# LYMPHOCYTE FUNCTIONAL CROSSTALK AND REGULATION

EDITED BY : Anil Shanker, Raghvendra Mohan Srivastava and Francesco Maria Marincola PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-414-9 DOI 10.3389/978-2-88963-414-9

#### 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

# LYMPHOCYTE FUNCTIONAL CROSSTALK AND REGULATION

Topic Editors: Anil Shanker, Meharry Medical College, United States Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States Francesco Maria Marincola, AbbVie (United States), United States

Citation: Shanker, A., Srivastava, R. M., Marincola, F. M., eds. (2020). Lymphocyte Functional Crosstalk and Regulation. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-414-9

# Table of Contents


Juhee Kim, Jun Young Lee, Kyungjin Cho, Sung-Wook Hong, Kwang Soon Kim, Jonathan Sprent, Sin-Hyeog Im, Charles D. Surh and Jae-Ho Cho


Tatyana Veremeyko, Amanda W. Y. Yung, Daniel C. Anthony, Tatyana Strekalova and Eugene D. Ponomarev


Dev Karan


Alka Dwivedi, Atharva Karulkar, Sarbari Ghosh, Afrin Rafiq and Rahul Purwar


*143 Adipose Tissue in Persons With HIV is Enriched for CD4+ T Effector Memory and T Effector Memory RA+ Cells, Which Show Higher CD69 Expression and CD57, CX3CR1, GPR56 Co-expression With Increasing Glucose Intolerance*

Celestine N. Wanjalla, Wyatt J. McDonnell, Louise Barnett, Joshua D. Simmons, Briana D. Furch, Morgan C. Lima, Beverly O. Woodward, Run Fan, Ye Fei, Paxton G. Baker, Ramesh Ram, Mark A. Pilkinton, Mona Mashayekhi, Nancy J. Brown, Simon A. Mallal, Spyros A. Kalams and John R. Koethe

*160 BTLA/HVEM Signaling: Milestones in Research and Role in Chronic Hepatitis B Virus Infection*

Xueping Yu, Yijuan Zheng, Richeng Mao, Zhijun Su and Jiming Zhang *168 NF-*k*B Signaling and IL-4 Signaling Regulate SATB1 Expression via Alternative Promoter Usage During Th2 Differentiation* Satyajeet P. Khare, Ankitha Shetty, Rahul Biradar, Indumathi Patta,

Zhi Jane Chen, Ameya V. Sathe, Puli Chandramouli Reddy, Riitta Lahesmaa and Sanjeev Galande

*181 Cytokines: Key Determinants of Resistance or Disease Progression in Visceral Leishmaniasis: Opportunities for Novel Diagnostics and Immunotherapy*

Alti Dayakar, Sambamurthy Chandrasekaran, Suresh V. Kuchipudi and Suresh K. Kalangi

*204 Tumor-Associated Disialylated Glycosphingolipid Antigen-Revealing Antibodies Found in Melanoma Patients' Immunoglobulin Repertoire Suggest a Two-Direction Regulation Mechanism Between Immune B Cells and the Tumor*

Beatrix Kotlan, Szabolcs Horvath, Klara Eles, Vanda K. Plotar, Gyorgy Naszados, Katalin Czirbesz, Miri Blank, Emil Farkas, Laszlo Toth, Jozsef Tovari, Andras Szekacs, Yehuda Shoenfeld, Maria Godeny, Miklos Kasler and Gabriella Liszkay

*218 Emerging Role of Lymphocyte Antigen-6 Family of Genes in Cancer and Immune Cells*

Geeta Upadhyay

*229 ATP Triggers Human Th9 Cell Differentiation via Nitric Oxide-Mediated mTOR-HIF1*a *Pathway*

Suyasha Roy and Amit Awasthi

*241 CTL-Derived Exosomes Enhance the Activation of CTLs Stimulated by Low-Affinity Peptides*

Shu-Wei Wu, Lei Li, Yan Wang and Zhengguo Xiao

*253 Targeting Adenosine in Cancer Immunotherapy to Enhance T-Cell Function*

Selena Vigano, Dimitrios Alatzoglou, Melita Irving, Christine Ménétrier-Caux, Christophe Caux, Pedro Romero and George Coukos

*283 Overexpression of PDE4A Acts as Checkpoint Inhibitor Against cAMP-Mediated Immunosuppression* in vitro

Klaus G. Schmetterer, Katrin Goldhahn, Liesa S. Ziegler, Marlene C. Gerner, Ralf L. J. Schmidt, Madeleine Themanns, Eva Zebedin-Brandl, Doris Trapin, Judith Leitner, Winfried F. Pickl, Peter Steinberger, Ilse Schwarzinger and Rodrig Marculescu

*298 CD8+ T Lymphocyte and NK Cell Network: Circuitry in the Cytotoxic Domain of Immunity*

Roman V. Uzhachenko and Anil Shanker


# Editorial: Lymphocyte Functional Crosstalk and Regulation

#### Raghvendra M. Srivastava<sup>1</sup> , Francesco M. Marincola<sup>2</sup> and Anil Shanker 3,4,5,6 \*

*1 Immunogenomics and Precision Oncology Platform, Memorial Sloan Kettering Cancer Center, New York, NY, United States, <sup>2</sup> Refuge Biotechnologies Inc., Menlo Park, CA, United States, <sup>3</sup> Department of Biochemistry, Cancer Biology, Neuroscience and Pharmacology, School of Medicine, Meharry Medical College, Nashville, TN, United States, <sup>4</sup> Host-Tumor Interactions Research Program, Vanderbilt-Ingram Comprehensive Cancer Center, Vanderbilt University School of Medicine, Nashville, TN, United States, <sup>5</sup> Vanderbilt Center for Immunobiology, Vanderbilt University School of Medicine, Nashville, TN, United States, <sup>6</sup> Vanderbilt Institute for Infection, Immunology and Inflammation, Vanderbilt University School of Medicine, Nashville, TN, United States*

#### Keywords: immune crosstalk, lymphocytes, T-cells, NK cells, dendritic cells, immunotherapy, cancer, infections

#### **Editorial on the Research Topic**

#### **Lymphocyte Functional Crosstalk and Regulation**

Lymphocyte effector responses constitute a key function of the vertebrate immunological defense. The responses are tightly controlled by a range of intracellular and intercellular mechanisms. A vast body of research demonstrates that the full potential of lymphocyte effector function is acquired following communications between the components of innate and adaptive immunity organized into an architecture of an interactive social network (1). For example, besides classical T-DC interactions (2–5), an interplay between dendritic cells (DC) and NK cells (6, 7), or T-cells and NK cells (8–18), plays important roles in resolving infections, solid cancers, or other pathologies. NK-T-regulatory (Treg) cell interplay maintains liver homeostasis (18). Furthermore, a tripartite crosstalk between T lymphocytes, DC, and NK cells is important for effector immune responses (19, 20). NK activity is also linked to T-cell and humoral responses (21). A cooperation between CD4<sup>+</sup> and CD8<sup>+</sup> T-cells during the effector phase has been suggested to eliminate bystander cancer cells (22). Thus, it is imperative to study immune compartments as an interdependent functional unit (23). A systematic understanding of the interaction between innate immune cells, innate lymphoid cells, including NK, and adaptive immune cells, such as T and B-cells can lead to improved immunotherapy approaches.

In pathologies of cancer, autoimmunity, chronic infections with prolonged inflammation, or tissue damage, lymphocytes are either compromised or they respond overtly. Immune dysfunction is linked to skewed immune cell distribution across several subtypes, including CD8+T, CD4+T, Treg, NK, innate lymphoid cells, DC, macrophages, neutrophils, or myeloid-derived suppressor cells (MDSC). Moreover, the efficiency of their crosstalk, and the frequency of intermediary players have a considerable role in determining the disease severity. Single cell and next generation sequencing technologies are revealing that numerous immune subtypes including previously uncharacterized subsets are affected in these diseases. Further, layers of underlying immune resistance and escape mechanisms interfere with the clinical outcomes, indicating our poor understanding of disease microenvironments and immune cell networkings.

This Research Topic was developed to gain an overview of these timely issues in lymphocyte biology with a particular emphasis on functional crosstalk and its regulation. In this topic, a series of articles ranging from basic science to translational and clinical reports, including a few very insightful reviews, provides meaningful insights toward this interesting field. These articles strongly support the premise that harnessing the immune cell crosstalk in immune disorders and cancer may uncover novel strategies to cure these diseases more effectively.

Edited and reviewed by: *Francesca Granucci, University of Milano Bicocca, Italy*

> \*Correspondence: *Anil Shanker ashanker@mmc.edu*

#### Specialty section:

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

Received: *16 November 2019* Accepted: *27 November 2019* Published: *10 December 2019*

#### Citation:

*Srivastava RM, Marincola FM and Shanker A (2019) Editorial: Lymphocyte Functional Crosstalk and Regulation. Front. Immunol. 10:2916. doi: 10.3389/fimmu.2019.02916*

**6**

To improve the outcome of T-cell immunotherapy in cancer patients, immune checkpoint inhibitors targeting PD-1/PD-L1, or CTLA-4 pathways received a lot of excitement in the past decade. Checkpoint antibodies were developed and approved by the Food and Drug Administration (FDA) in the USA and Europe. A blockade of T-cell exhaustion or contraction molecules, PD-1/PD-L1 and CTLA-4, respectively, should reactivate and expand cancer antigen-specific T-cells. However, clinical responses were heterogeneous to none, with several resistance mechanisms identified across cancer types. Surprisingly, a hyperprogression was also reported recently in multiple cancer types, including melanoma and nonsmall cell lung cancer, possibly mediated by antibody-Fc/FcR interaction and amplification of PD-1<sup>+</sup> Treg-cells (24–27). Seliger reviewed crucial factors pertaining to the limited efficacy of cancer immunotherapy, such as crosstalk between immune cells and gut microbiome, tumor-infiltrating regulatory myeloid cells, and the role of several immune cell subsets. The author also discussed how tumor mutational burden and neoantigen load regulates clinical outcomes (28). Limited success of negative immune regulators, i.e., PD-1 and CTLA-4, initiated the hunt for other promising negative checkpoint receptors and ligands. Several B7 molecules also regulate antitumor immunity (29). Included here, Cui et al. reported a novel B7-related molecule CD300c expressed on antigenpresenting cells, including B-cells, monocytes, macrophages, and DC. Its putative counter receptors were also identified on resting and activated CD8<sup>+</sup> and CD4<sup>+</sup> T-cells. They showed that CD300c blockade may reinvigorate T-cell responses. B- and T-lymphocyte attenuator (BTLA) is another crucial immunoregulatory receptor. Yu et al. discussed BTLA and its ligand HVEM (herpes virus entry mediator) signaling pathways, and highlighted chronological research needed in BTLA field. The widespread expression of BTLA across T, B, NKT, and DC indicates potential role of BTLA in immune crosstalk that could be harnessed to balance lymphocyte function and counteract immune disorders. The intracellular second messenger cyclic-AMP (cAMP) that acts in immunosuppressive signaling in T-cells is upregulated by multiple tumor-derived suppressive factors, such as prostaglandin E2 and adenosine. Schmetterer et al. show that ectopic overexpression of phosphodiesterase 4A (PDE4A) in T-cells leads to efficient degradation of cAMP. They suggest that PDE4A can be exploited as an immune checkpoint inhibitor against multiple suppressive factors.

Recently, the FDA approved chimeric antigen receptor (CAR) T-cell therapy for relapse and refractory B-cell acute lymphoblastic leukemia and diffuse large B-cell lymphoma following their success in multiple phase-I/II clinical trials. While CAR T-cells are considered as major breakthrough in the field of cancer immunotherapy, the regulation of CAR T-cells remains poorly understood. Dwivedi et al. reviewed the strategies that regulate CAR T-cell efficacy and persistence with focus on roles of different structural component of CAR construct.

Tumor microenvironment exhibits hypoxic and non-hypoxic areas heterogeneously. Hypoxic area is immune-deserted and appears to be resistant to immune cell attacks. Hypoxia is often associated with high amounts of adenosine, which is generated following ATP degradation. Adenosine restrains the effector function of T and NK cells. Several reports have indicated that interference with the adenosine signaling may improve anti-tumor immunity, especially in conjunction with immune checkpoint inhibitors. In a comprehensive review directed toward targeting adenosine in cancer immunotherapy, Vigano et al. presented an illustrious overview of several pre-clinical and clinical strategies to inhibit the negative effects of adenosine for superior cancer control. They described several mechanisms to intersect adenosine effects impacting T-cells, DC, macrophages, NK, neutrophils, MDSC, stromal cells, and tumor cells. Further enhancing the cross-talk among various immune cells in a hostile tumor microenvironment, Karan introduced the emerging concept of targeting the inflammasome to improve antitumor immunity. Since a multitude of cytokines play a pivotal role in the regulation of immune cell function, inflammasome targeting would likely modulate the profile of inflammatory cytokines reducing immunosuppression at tumor sites. It would be interesting to see how ongoing studies on inflammasomes would benefit the field of cancer immunotherapy.

Several groups have detected B-cell infiltration into solid cancers. Kotlan et al. investigated tumor-infiltrating Bcells. They performed antibody repertoire analysis at the genetic level and identified disialylated glycosphingolipid as tumor antigens in melanoma. They generated a novel tool, single-chain variable (ScFV) antibody, and performed several other tests to confirm that tumor-infiltrated Bcells can recognize tumor-associated glycosphingolipid on melanoma and other solid cancers. Since tumor-derived ganglioside can suppress B-cell antibody production, this B cell-cancer cell crosstalk could be utilized as a therapeutic target.

NK cells and T-cells recognize tumor cells with distinct types of receptors; however, they utilize common effector molecules, such as TNFα, perforin, and granzymes to exert their cytotoxic action. Uzhachenko and Shanker presented an emerging perspective that bidirectional crosstalk, and membranous reorganization between these CTL and NK cells may enable them to effectively eradicate cancer and infections. From human immune network proteome database, it is apparent that NK and CTL can profusely interact with DC to prime and optimize their effector function. Moreover, DC, being the professional antigen-presenting cells, have been tested as antigen delivery vehicle in several clinical trials. Vujanovic et al. interrogated the immunological effects of autologous DC transduced with MART-1, tyrosinase, and MAGE-A6 melanoma tumor antigens in a phase-I clinical trial. Interestingly, they identified that systemic high dose IFN-α2β after DC vaccination modulates a unique CD56dimCD16negative non-cytolytic NK subset in melanoma patients. This dominant immunoregulatory NK subset in tumor microenvironment appears to contribute to better clinical outcomes.

Exosomes mediate several immunoregulatory mechanisms. In an in vitro CTL activation model, Wu et al. report that CTL-derived exosomes can increase the effector function and proliferation of CTL activated by low affinity peptides. The exosomes appear to mediate crosstalk among high and lowaffinity CTLs at the beginning of CTL response. This may increase the repertoire of low affinity CTL and afford their long-term sustenance, thereby significantly improving T-cell mediated anti-cancer responses. In a mouse model, Kim et al. demonstrated that IL-7 and IL-15 cytokine-based crosstalk regulates the proportion and survival of naïve and memory Tcell populations. In the context of spontaneous proliferation of T-cells, Kim et al. showed an antigen-independent but IL-2 dependent crosstalk between CD4<sup>+</sup> and CD8+T-cells.

T-cells also play a central role in modulating adipose tissue inflammation. In people living with HIV infection, Wanjalla et al. correlated subcutaneous adipose tissue with significantly increased frequency of CD4<sup>+</sup> and CD8<sup>+</sup> Teffector-memory and effector-memory-RA<sup>+</sup> cells. Adipose tissue from HIV-infected individuals showed a higher expression of TLR2, TLR8, and multiple chemokines relevant to immune cell homing compared to HIV-negative controls with similar glucose tolerance.

The metabolic reprogramming of T cells by immunosuppressive drugs controls several inflammatory disorders (30). In humans infected with cytomegalovirus (CMV), Bak et al. showed significantly improved CMV-specific effectormemory T-cell function following inhibition of mammalian target of rapamycin (mTOR) with sirolimus. Monitoring of TCR-repertoire dynamics by next generation sequencing confirmed that the increased functionality was not related to sirolimus-resistant CTL-clones. Instead, environmental cues during CMV-CTL development via IL-2 receptor-driven signal transducer and activator of transcription-5 (STAT-5) signaling under mTOR inhibition allowed fine-tuning of T-cell programming for enhanced antiviral responses with stable TCRrepertoire dynamics. In a non-interventional prospective clinical trial in patients with multiple trauma, Hefele et al. assessed the role of platelets, Treg and Th17 cells in the post-traumatic immune response. They observed increased IL-17A expression in Th17 and Treg during the first 10 days following trauma. Moreover, despite a rising number of platelets, their analysis showed post-traumatic platelet dysfunction. Further studies are necessary to understand the underlying functional crosstalk between T-cells and platelets.

Recently, IL-9-producing Th9 cells have been identified in several disorders including cancer. Roy and Awasthi presented evidence that extracellular ATP promotes the differentiation of Th9 cells. They proposed a "feed-forward loop model" where nitric oxide production in Th9 cells enhances the mTOR-HIF-1-α pathway that further culminates in Th9 cell differentiation. This finding may have major implications in several cancer types. The role of Th9 cells in cancer is currently under investigation in several studies. Gaudino and Kumar described the crosstalk of T-cells and APC at multiple layers and presented a schematic of dysbiosis and gut microbiome. They highlighted how intestinal IL-17 receptor signaling and reciprocal crosstalk with gut microbiota can regulate autoimmunity. A comprehensive understanding of molecular interactions of immune cells and cytokines is important to refine appropriate immunotherapies. Dayakar et al. summarized the current understanding of a wide spectrum of cytokines and their interaction with immune cells that determine the clinical outcome of visceral leishmaniasis. They also highlighted opportunities for the development of novel diagnostics and intervention therapies for leishmaniasis.

Upadhyay describes the novel role of Ly6 gene family in cancer immune regulation. Ly6 genes are scattered in various chromosomes in human genome and have similarity to stem cell antigen-1 gene, a well-known cancer stem cell marker. The overexpression of this class of genes in solid cancers leads to poor survival. Some of these genes are expressed on both cancer cells and innate immune cells. Further research is necessary to understand if the Ly6 family genes could play a role in innate and adaptive immune crosstalk in the context of solid tumor microenvironments. Macrophage polarization is involved in many pathologies such as anti-cancer immunity and autoimmune diseases. Polarized macrophages exhibit plasticity when M2 macrophages are reprogrammed into an M1-like phenotype following treatment with IFNγ and/or LPS. At the same time, M1 macrophages are resistant to reprogramming in the presence of M2-like stimuli (IL-4). Veremeyko et al. explored the role of early growth response (Egr) family of transcriptional regulators in the induction and maintenance of M1 and M2 polarization. They demonstrated that a molecular crosstalk between Egr2 and CEBPβ transcription factors regulated macrophage polarization under distinct inflammatory conditions. An original study by Khare et al. investigated a switch of proximal and distal promoters fine-tuning the expression of genome organizer, special AT-rich sequence-binding protein (SATB1), which plays a crucial role in expression of multiple genes in a cell type-specific manner during the thymic development and peripheral differentiation and polarization of T-helper cells. Cytokine and TCR signaling crosstalk impacts SATB1 alternative promoter usage.

Collectively, the articles contained within the Research Topic highlight the leading concept of lymphocyte functional crosstalk and its underlying complexity that impacts disease pathology and outcomes. Further research into the functional dynamics of immune networks is essential and timely for advancing our understanding of the immunological basis of diseases and the design of preventive or therapeutic approaches. The knowledge acquired from the published articles will contribute to meaningful insights for the development of more refined and novel immune strategies.

#### AUTHOR CONTRIBUTIONS

RS and AS conceived, designed, and wrote the manuscript. FM provided substantial intellectual feedback. All authors read and approved the final manuscript for publication.

#### FUNDING

This work was supported by funds to AS by the following NIH grants: SC1CA182843 and U54 CA163069. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. There was no role of the funding

#### REFERENCES


bodies in the design or writing of the manuscript. No writing assistance was utilized in the production of this manuscript.

#### ACKNOWLEDGMENTS

We express our appreciation to all contributing authors, who participated in this Research Topic. Our gratitude is also due to all reviewers for agreeing to participate in the peer review process and providing their insightful comments and feedback on the manuscripts. We also thank Ms. Tonie Farris for critical reading of the manuscript.

is independent of CD4<sup>+</sup> T cells and dependent on natural killer cells. Br J Cancer. (2007) 96:1839–48. doi: 10.1038/sj.bjc.6603792


**Conflict of Interest:** FM is employed by the company Refuge Biotechnologies Inc., Menlo Park, CA. FM is also on the Advisory Boards of Calidi Biotherapeutics Inc., San Diego, CA and CHIPSATM Hospital, Playas Tijuana, Baja California, Mexico.

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

Copyright © 2019 Srivastava, Marincola and Shanker. 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.

# Spontaneous Proliferation of CD4<sup>+</sup> T Cells in RAG-Deficient Hosts Promotes Antigen-Independent but IL-2-Dependent Strong Proliferative Response of Naïve CD8<sup>+</sup> T Cells

#### Edited by:

Anil Shanker, Meharry Medical College, United States

#### Reviewed by:

Hyun Park, National Cancer Institute (NCI), United States Patrick Jerome Bertolino, Centenary Institute Australia, Australia

#### \*Correspondence:

Charles D. Surh csurh@ibs.re.kr Jae-Ho Cho jhcho90@ibs.re.kr

†Sin-Hyeog Im orcid.org/0000-0002-3173-1856

#### Specialty section:

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

Received: 07 April 2018 Accepted: 02 August 2018 Published: 23 August 2018

#### Citation:

Kim J, Lee JY, Cho K, Hong S-W, Kim KS, Sprent J, Im S-H, Surh CD and Cho J-H (2018) Spontaneous Proliferation of CD4<sup>+</sup> T Cells in RAG-Deficient Hosts Promotes Antigen-Independent but IL-2-Dependent Strong Proliferative Response of Naïve CD8<sup>+</sup> T Cells. Front. Immunol. 9:1907. doi: 10.3389/fimmu.2018.01907 Juhee Kim1,2, Jun Young Lee1,2, Kyungjin Cho1,2, Sung-Wook Hong1,2, Kwang Soon Kim1,2 , Jonathan Sprent 3,4, Sin-Hyeog Im1,2†, Charles D. Surh1,2 \* and Jae-Ho Cho1,2 \*

<sup>1</sup> Academy of Immunology and Microbiology, Institute for Basic Science, Pohang, South Korea, <sup>2</sup> Department of Integrative Biosciences and Biotechnology, Pohang University of Science and Technology, Pohang, South Korea, <sup>3</sup> Immunology Division, Garvan Institute of Medical Research, Darlinghurst, NSW, Australia, <sup>4</sup> University of New South Wales, Sydney, NSW, Australia

The fast and intense proliferative responses have been well documented for naïve T cells adoptively transferred into chronic lymphopenic hosts. This response known as spontaneous proliferation (SP), unlike antigen-independent lymphopenia-induced proliferation (LIP), is driven in a manner dependent on antigens derived from commensal microbiota. However, the precise nature of the SP response and its impact on homeostasis and function for T cells rapidly responding under this lymphopenic condition are still unclear. Here we demonstrate that, when naïve T cells were adoptively transferred into specific pathogen-free (SPF) but not germ-free (GF) RAG−/<sup>−</sup> hosts, the SP response of these cells substantially affects the intensity and tempo of the responding T cells undergoing LIP. Therefore, the resulting response of these cells in SPF RAG−/<sup>−</sup> hosts was faster and stronger than the typical LIP response observed in irradiated B6 hosts. Although the intensity and tempo of such augmented LIP in SPF RAG−/<sup>−</sup> hosts were analogous to those of antigen-dependent SP, the former was independent of antigenic stimulation but most importantly, dependent on IL-2. Similar observations were also apparent in other acute lymphopenic settings where antigen-dependent T cell activation can strongly occur and induce sufficient levels of IL-2 production. Consequently, the resulting T cells undergoing IL-2-driven strong proliferative responses showed the ability to differentiate into functional effector and memory cells that can control infectious pathogens. These findings therefore reveal previously unappreciated role of IL-2 in driving the intense form of T cell proliferative responses in chronic lymphopenic hosts.

Keywords: spontaneous proliferation, lymphopenia-induced homeostatic proliferation, naïve CD4 T<sup>+</sup> cells, naïve CD8<sup>+</sup> T cells, interleukin-2 (IL-2), specific pathogen-free (SPF), germ-free (GF)

# INTRODUCTION

Proliferation of naïve T cells under lymphopenic environments has long been accepted as a crucial homeostatic mechanism by which a diverse repertoire of these cells can be stably maintained at constant number during their peripheral life-time (1, 2). Therefore, this proliferative response, also known as lymphopenia-induced homeostatic proliferation (LIP), is considered as a common phenomenon for the most polyclonal and even monoclonal naïve T cell repertoire adoptively transferred into a recipient animal under various lymphodepletion settings either genetically (e.g., RAG−/−, TCRβ <sup>−</sup>/−, CD3ε <sup>−</sup>/−, and SCID mice) or conditionally (e.g., mice treated with irradiation and cytotoxic agents (3–7). The LIP response in these hosts is relatively slow with 2–3 rounds of cell division per week and is driven by a signal from self-ligands and elevated levels of IL-7 (3, 7–10). Despite the slow proliferation is unique for the LIP response in lymphopenic hosts, the relative strength of this response can be modulated variably depending on the cytokines being engaged (11). For instance, the LIP response driven by IL-2 and/or IL-15 is far stronger than those induced by IL-7 (12–14).

In sharp contrast to the LIP, however, totally different form of proliferative responses has also been reported for naïve T cells when adoptively transferred into the aforementioned chronic lymphopenic hosts, such as RAG−/<sup>−</sup> and TCRβ −/− (or CD3ε <sup>−</sup>/−) mice (15–17). The proliferative response in these hosts, known as spontaneous proliferation (SP), is much faster and stronger than those of typical LIP with more than ∼7 rounds of cell divisions per week and, unlike LIP, is driven in a manner independent of self-ligands and IL-7 (7, 15). Although the exact nature of the stimuli for the SP response is still incompletely understood, it has been demonstrated that this response is exclusively dependent on largely two key signals, namely from TCR engagements with its cognate foreign-peptide/major histocompatibility complex (foreign-pMHC) ligands and also from costimulatory interactions via CD28 (7, 15). The antigenic stimuli are thought to be derived from commensal microbiota because the SP response of naïve T cells is observed only in RAG−/<sup>−</sup> hosts raised under the specific pathogen-free (SPF), but not the germ-free (GF), condition (15). Precisely how the commensal antigens are presented to stimulate the SP response of T cells in these hosts and if so, why this phenomenon fails to occur in other lymphopenic hosts, such as irradiated C57BL/6 (B6) mice, remains to be addressed.

In this respect, recent studies have shown that the role of commensal microbial antigens in driving the SP response is not direct but rather indirect effect on T cells via a mechanism dependent on innate immune stimulation through toll-like receptor (TLR) on dendritic cells (DCs) (18–20). Therefore, it is possible that the SP response that occurs in chronic lymphopenic hosts is regulated at least in part, if not exclusively, by some forms of antigen-independent responses other than direct TCR engagements with foreign-pMHC ligands. In fact, it has been shown that IL-6 produced from DCs that are activated by bacterial ligands serves as a major driver for promoting the SP response of naïve T cells adoptively transferred into RAG−/<sup>−</sup>

hosts (18, 21). However, how these data supporting a role of antigen-independent components would be reconciled with the stringent requirement of antigen-dependent components of TCR stimulation for inducing the robust SP response in chronic lymphopenic hosts is largely unclear. In this study, we address these issues by investigating the mechanism of how the SP response of polyclonal or monoclonal T cells is regulated and influences their homeostasis and function during their recovery phase from various settings of lymphopenia. We show here that the SP response of naïve T cells observed in the lymphopenic hosts consists of at least two forms of intensive proliferative responses, namely an antigen-dependent "true" SP response and an antigen-independent but IL-2-dependent SP-like "bystander" response.

#### MATERIALS AND METHODS

#### Mice

C57BL/6 (B6), B6.PL (Thy1.1), B6.SJL (Ly5.1), Foxp3-eGFP (22), RAG−/−, L-2+/<sup>−</sup> mice, all on a B6 background, were purchased from The Jackson Laboratory. Sources of OT-I.RAG−/−, HY and SMARTA TCR Tg mice were previously described (4, 5, 23). Germ free (GF) mice are maintained sterilely at POSTECH Biotech Center (PBC, Korea) as described (24). OT-I.RAG−/−.Thy1.1, SMARTA.Thy1.1 and IL-2−/<sup>−</sup> mice were generated as described (4, 13). Unless it is described, 6–10 weeks old mice were used for the experiments according to the protocols approved by the Animal Experimental and Ethic Committee at the Institute for Basic Science (Korea).

#### Naïve T Cell Purification

Pooled (inguinal, axillary, cervical, and mesenteric) lymph node cells from SMARTA TCR Tg or Foxp3-eGFP mice were prepared for cell sorting as previously described (13, 25), with slight modifications. In brief, LN cells were first depleted of non-T cells by using the following biotinylated antibodies; CD11b, CD11c, CD24, CD19, B220, NK1.1, and IMag according to the manufacturer's protocol (BD biosciences). Enriched T cells were stained with fluorochrome conjugated antibodies to CD8α, CD4, CD44, CD62L, and/or CD5 and then either Foxp3-eGFP<sup>−</sup> CD4<sup>+</sup> CD44lo CD62L<sup>+</sup> (naïve CD4+), CD8α <sup>+</sup> CD62L<sup>+</sup> CD44 lo (naïve CD8+), CD8α <sup>+</sup> CD62L<sup>+</sup> CD44 lo CD5hi CD4<sup>−</sup> (naïve CD8<sup>+</sup> CD5hi), and CD8α <sup>+</sup> CD62L<sup>+</sup> CD44 lo CD5lo CD4<sup>−</sup> (naïve CD8<sup>+</sup> CD5lo) populations were sorted by using a Moflo XDP (Beckman Coulter, Brea, CA, USA) to >95% purity.

#### Adoptive Transfer

After purification, T cells were labeled with 5µM of either CFSE (Invitrogen) or CellTraceTM Violet (Molecular Probes), as previously described (26) and injected i.v. into hosts. For inducing lymphopenia, normal B6 mice were treated with anti-Thy1.2 mAb 30-H12 (anti-Thy1.2) (Bio X Cell, i.p. injection in a single dose of 200 µg/mouse, 2 days before cell transfer) or 600cGy of whole-body irradiation (1 day before cell transfer). For generating antigen-induced "SP-like" response, SMARTA CD4<sup>+</sup> T cells were transferred into the hosts, as indicated in the figures, followed by either LCMV Armstrong (2 × 10<sup>5</sup> PFU) or LCMV peptide GP61−<sup>80</sup> (20 µg/mouse) through i.p. injection 1 day post cell transfer. OT-I CD8<sup>+</sup> T cells were transferred into the hosts, as indicated in the figures, followed by immunization of OVA protein (Sigma Aldrich, 100 µg/mouse). HY.CD8<sup>+</sup> cells from female HY mice were transferred into female hosts.

# Reconstitution of IL-2−/<sup>−</sup> T Cells

Bone marrow (BM) cells were obtained from B6.SJL (Ly5.1) and IL-2−/<sup>−</sup> (Thy1.2) mice, mixed at 1:1 ratio. T cell-depletion was done in incubating BM cells with anti-CD4 (RL172), anti-CD8 (3.168), anti-CD24 (J11d) on ice for 10 minutes before adding complement (guinea pig). B6.PL (Thy1.1) mice were lethally irradiated (9.6Gy) before being injected i.v. with 4 × 10<sup>6</sup> T celldepleted BM cells. At 8 weeks after BM cell transfer, IL-2−/<sup>−</sup> naive CD4<sup>+</sup> T cells (CD90.2<sup>+</sup> CD44lo CD62Lhi CD4+) were obtained from these mixed chimeras by FACS sorting.

#### Tissue Preparation

Single-cell suspensions were prepared from mesenteric lymph nodes (MLNs), lamina propria (LP), epithelium, spleen (SPL), lung and liver as previously described (24, 27, 28). Briefly, MLNs, spleen and liver were pressed and filtered through cell strainers. Small intestine (SI) and large intestine (LI) were harvested and Peyer's patches removed prior to process. LP and lung were digested with collagenase D and DNase I. LP, IEL, and liver lymphocytes were enriched by 40:75% Percoll density gradient centrifugation.

### Flow Cytometry Analysis

For surface staining, isolated cells were stained for flow cytometry with the following mAbs from Biolegend, eBioscience and/or TONBO: CD3 (145-2C11), CD4 (GK1.5 and RM4–5), CD8α (53- 6.7), CD8? (YTS156.7.7), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), CD69 (H1.2F3), CD90.1 (HIS51 or OX-7), and TCR H-Y (T3.70) in a conjugation with FITC, PE, PE-Cy5, PE-Cy7, APC, APC-Cy7 or PB. Propidium iodide (PI) (Sigma Aldrich) was used at 500 ng/ml of final concentration for staining of 1–5 × 10<sup>6</sup> of cells to label dead cells. For intracellular staining, surface stained cells were fixed and permeabilized with BD cytofix/cytoperm according to manufacturer's protocol (BD Biosciences) and were stained with the following mAbs: IL-2 (JES6-5H4), IFN-γ (9D3.1C8), TNF-α (MP6-XT22) and Granzyme B (GB12). Flow cytometry samples were analyzed using a flow cytometer (LSR Fortessa and Canto-II; BD Biosciences) with DIVA software. Data were analyzed using Flowjo (Treestar).

# Immunohistochemistry

Small intestine was harvested and Peyer's patches were removed. Freshly collected tissues were "snap-frozen" in OCT (Leica) with liquid nitrogen. Tissue sections (6µm in thickness) were prepared, air-dried, fixed for 10 min at 4◦C in methanol (Merk). Cryosections were blocked for 30 min with biotin blocking solutions (Invitrogen), washed in PBS, and incubated overnight at 4◦C with anti-Thy1.1 (OX-7) and anti-CD8α (53– 6.7) antibodies (BioLegend). Sections were then washed with PBS and stained with DAPI. All slides were mounted with Prolong Antifade Reagent (Life technologies) and images were capture with Zeiss LSM 700 CLSM (confocal laser scanning microscope).

# Bacteria and Virus Infections

Listeria monocytogenes (LM) strain 10403s, carrying a recombinant internalin A (InIA) mutant, has been described in detail previously (29, 30). Briefly, B6 mice were infected with 5 × 10 <sup>10</sup> CFU Listeria monocytes (LM) InIA-OVA through oral gavage. For acute infections, B6 mice were infected i.p. with 2 × 10<sup>5</sup> PFU of LCMV Armstrong (31).

# Administration of Antibodies and/or Cytokines in vivo

IL-2/anti-IL-2 complexes were prepared as previously described (12, 13). In brief, 1 µg of recombinant mouse IL-2 (PeproTech) was mixed with 5 µg of anti-IL-2 (S4B6) (BD Biosciences) and injected i.p daily for three consecutive days after adoptive cell transfer in SPF RAG−/−. For the IL-2 blocking experiments, 100 µg of anti-IL-2 (JES6-1A12) (Bio X Cell) and anti-IL-2 (S4B6) (BD Biosciences) was injected i.p. every other day for 7 days after adoptive cell transfer in SPF RAG−/<sup>−</sup> and B6 hosts.

# Intravascular T Cell Staining

Following the previously published protocol (32), 3 µg of anti-Thy1.1 (OX-7) in 300 µl of PBS were injected i.v. into a mouse for intravascular staining of donor OT-I (Thy1.1) CD8<sup>+</sup> T cells. The mice were killed 10 min after injection and collected the tissues. For discrimination of vascular and tissue donor cells, the cells were stained with ex vivo Abs including anti-Thy1.1 (HIS51).

# Statistical Analysis

Results represent the mean ± SEM unless indicated otherwise. Statistical significance was determined by the unpaired Student's t test. Statistical analyses were performed using Prism GraphPad software v5.0. <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant).

# RESULTS

#### Spontaneous Proliferation of Polyclonal Naïve T Cells in RAG−/<sup>−</sup> Hosts

Given the well-known previous observations that polyclonal naïve CD4<sup>+</sup> or CD8<sup>+</sup> T cells undergo intense form of proliferative responses in a Rag-deficient host (15), which is referred to as spontaneous proliferation (SP), we sought to address whether and how this SP response of T cells influences their functional behavior and homeostasis during their reconstitution from lymphopenia. We thus first confirmed the prior notion that the SP occurs largely in an antigen-dependent manner with strong and fast rate of cell division kinetics.

For this, FACS-purified CTV-labeled polyclonal naïve CD4<sup>+</sup> T cells were adoptively transferred into three different lymphopenic hosts, namely C57BL/6 (B6) mice receiving sub-lethal doses (600 cGy) of irradiation and Rag1-deficient (RAG−/−) mice raised under the specific pathogen-free (SPF) or germ-free (GF) condition (**Figure 1A**, top). Donor cell division and recovery from the spleen (SPL) and mesenteric lymph nodes (MLN) were

analyzed on day 7 after adoptive transfer by flow cytometry. As shown in **Figure 1A**, donor CD4<sup>+</sup> T cells, as expected, exhibited only ∼2–3 rounds of slow rate of cell division (i.e., un-gated CTV<sup>+</sup> cells), referred to as lymphopenia-induced homeostatic proliferation (LIP) that is known to be dependent on TCR interaction with self-ligands and cytokine IL-7 (3, 7). In sharp contrast, cells transferred into SPF RAG−/<sup>−</sup> hosts showed robust proliferative responses, as evidenced by the full dilution of CTV dye (i.e., gated CTV<sup>−</sup> cells); however, these responses were abrogated substantially in GF RAG−/<sup>−</sup> hosts, confirming the previous findings showing stringent dependence of the SP responses of polyclonal naïve CD4<sup>+</sup> T cells on antigens derived from commensal microbiota (15). Unlike SP, the slower rate of LIP responses of donor cells was uninterrupted in the GF RAG−/<sup>−</sup> hosts, level of which was similar to that of irradiated B6 hosts (**Figure 1A**, left; compare ungated CTV<sup>+</sup> cells in the top and bottom histogram). Thus, the recovery of donor cells was ∼10-20-fold lower for the LIP responses in GF RAG−/<sup>−</sup> and irradiated B6 hosts than those for the SP responses observed in SPF RAG−/<sup>−</sup> hosts (**Figure 1A**, right). As for the SP of CD4<sup>+</sup> T cells, polyclonal naïve CD8<sup>+</sup> T cells from B6 mice also showed robust levels of SP, albeit at lower extent than CD4<sup>+</sup> T cell SP, in SPF RAG−/<sup>−</sup> hosts, but not in irradiated B6 hosts (**Figure S1**), which was also antigen-dependent because the SP response of CD8<sup>+</sup> T cells was abolished in GF RAG−/<sup>−</sup> hosts (data not shown).

Although the above data clearly confirmed the previously well-defined SP responses of either polyclonal naïve CD4<sup>+</sup> or CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts, interesting findings came from the experiments in which both naïve CD4<sup>+</sup> and CD8<sup>+</sup> T cells were co-transferred into these hosts. As shown in **Figure 1B**, in comparison to the proportion of SP responses after single transfer of either CD4<sup>+</sup> or CD8<sup>+</sup> T cells (63% ± 3.85 and 48% ± 2.95, respectively for MLN), we observed significantly enhanced SP responses of CD4<sup>+</sup> T cells and to a greater extent CD8<sup>+</sup> T cells on day 7 after co-transfer into SPF RAG−/<sup>−</sup> hosts (80% ± 1.92 and 78% ± 1.20, respectively for MLN). These data suggest that there is a positive cross-talk between two distinct T cell compartments by which the SP response of one compartment is facilitated by those of another compartment during reconstitution period after adoptive co-transfer into SPF RAG−/<sup>−</sup> hosts.

#### Influence of the SP Driven by Polyclonal CD4<sup>+</sup> T Cells on the LIP of TCR Tg CD8<sup>+</sup> T Cells

Based on the above findings of such potential cross-talk between polyclonal naïve CD4<sup>+</sup> and CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts, we then addressed the question of whether and how the SP responses of two distinct lineages of naïve T cell populations may affect each of their homeostasis in this lymphopenic condition. We thus next sought to investigate the potential impact of the SP response driven by polyclonal naïve CD4<sup>+</sup> T cell compartment on the typical slow rate of lymphopenia-induced homeostatic proliferation (LIP)—a response that is antigen-independent but IL-7-dependent of naïve CD8<sup>+</sup> T cells co-transferred into SPF RAG−/<sup>−</sup> hosts.

For this, we utilized monoclonal TCR transgenic (Tg) CD8<sup>+</sup> T cells in order to avoid antigen-dependent SP response, which was observed for polyclonal B6 CD8<sup>+</sup> T cells (**Figure 1B**), and validate the true influence of CD4<sup>+</sup> T cell-driven strong SP response on the weak LIP response of monoclonal CD8<sup>+</sup> T cells. Thus, naïve OT-I TCR Tg cells that are specific for H-2K<sup>b</sup> restricted ovalbumin (OVA) 257-264 peptide were labeled with CTV and adoptively transferred into irradiated B6 (600 cGy), SPF RAG−/−, or GF RAG−/<sup>−</sup> hosts and 7 days later, proliferative responses of donor OT-I cells were analyzed by flow cytometry (**Figure 2A**, top). As shown in **Figure 2A**, OT-I cells, unlike polyclonal CD8<sup>+</sup> T cells, did not show robust SP responses, but instead did show only ∼3–4 rounds of the typical slow-rate of LIP responses in SPF RAG−/<sup>−</sup> hosts, similar to those seen in irradiated B6 or GF RAG−/<sup>−</sup> hosts. Donor cell recoveries in the spleen were also similar in all three lymphopenic recipients, although there was a modest difference in the mLN (**Figure 2A**, bottom). These data thus confirmed the prior notion that the SP response observed in RAG−/<sup>−</sup> hosts is dependent on commensal microbial antigens and thus detectable only for polyclonal, but not monoclonal, CD8<sup>+</sup> T cells.

Based on the above results showing prevalent occurrence of the LIP but not the SP upon adoptive transfer with OT-I cells into SPF RAG−/<sup>−</sup> hosts, we then investigated the influence of the SP response driven by polyclonal CD4<sup>+</sup> T cells on the LIP response of OT-I cells. Thus, a mixture of FACS-purified, CTV-labeled naïve B6 CD4<sup>+</sup> T cells and OT-I CD8<sup>+</sup> T cells was transferred into SPF RAG−/<sup>−</sup> or GF RAG−/<sup>−</sup> hosts, and as a control, OT-I cells alone were also transferred into SPF RAG−/<sup>−</sup> hosts, and then analyzed on day 8 for their proliferation and cell recovery (**Figure 2B**, top). Here, the surprising finding was that, in marked contrast to the typical slow rate of LIP after single transfer of OT-I cells alone, these cells showed much faster and greater levels of proliferative responses when co-transferred with B6 CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts (**Figure 2B**, middle). More importantly, such intense responses of OT-I cells seen in SPF RAG−/<sup>−</sup> hosts were not observed in GF RAG−/<sup>−</sup> hosts even in the presence of B6 CD4<sup>+</sup> T cells being co-transferred, resulting in the poor donor cell recovery (**Figure 2B**, bottom). Careful analysis for the kinetics of these robust proliferative responses revealed that the appearance of the fast-dividing OT-I cells was apparent from day 4 after co-transfer with B6 CD4<sup>+</sup> T cells into SPF but not GF RAG−/<sup>−</sup> hosts, a time point of which CD4<sup>+</sup> T cells also began to show significant levels of SP responses (**Figure S2**), suggesting a role of B6 CD4<sup>+</sup> T cells undergoing SP.

Together, these findings strongly suggest that the SP response of polyclonal CD4<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts plays a role in promoting LIP response of co-transferred monoclonal (and also polyclonal) CD8<sup>+</sup> T cells, leading to the alteration of the speed and degree of their proliferation in a chronic lymphopenic environment.

### Influence of the Polyclonal CD8<sup>+</sup> T Cell-Derived SP on LIP of TCR TG CD8<sup>+</sup> T Cells

Given the above stimulatory effect of the SP response of polyclonal CD4<sup>+</sup> T cells on the LIP of monoclonal OT-I cells in SPF RAG−/<sup>−</sup> hosts, we next addressed whether the similar enhancing effect is also observed with the SP response of polyclonal CD8<sup>+</sup> T cells, because these cells also showed strong antigen-dependent SP in SPF RAG−/<sup>−</sup> hosts (**Figure 1B** and **Figure S1**). For this, a mixture of FACS-purified, CTV-labeled naïve B6 CD8<sup>+</sup> T cells and OT-I cells was transferred into SPF RAG−/<sup>−</sup> or GF RAG−/<sup>−</sup> hosts (**Figure 3A**, top). At day 7 after adoptive transfer, as expected, B6 CD8<sup>+</sup> T cells showed robust SP responses in SPF but not GF RAG−/<sup>−</sup> hosts (**Figure 3A**, middle and bottom left). Surprisingly, however, such SP responses of B6 CD8<sup>+</sup> T cells completely failed to induce strong LIP responses of the co-transferred OT-I cells in SPF RAG−/<sup>−</sup> hosts, as evidenced by the lack of CTV full-diluted, fast-dividing cells in these recipients, which was comparable to those seen in GF RAG−/<sup>−</sup> hosts (**Figure 3A**, middle and bottom right).

The rather unexpected results from these co-transfer experiments led us to speculate a possible involvement of clonal competition within the same lineage of donor CD8<sup>+</sup> T cell populations presumably for self-ligands, and suggest that SP response of polyclonal T cells in SPF RAG−/<sup>−</sup> hosts may result in a diverse degree of LIP differing from one clone to another.

#### Impact of TCR Affinity for Self-Ligands on the Strong LIP Response Mediated by the SP

The above clonal competition for self-antigens within the same lineage of CD8<sup>+</sup> T cell pools is of particular importance to determine the rate and magnitude of their LIP in lymphopenic hosts, as has been demonstrated by previous reports (6, 25, 33). It was thus intriguing for us to test whether the SP response of polyclonal B6 CD8<sup>+</sup> T cells has an enhancing effect on the LIP of monoclonal CD4+, if not CD8+, TCR Tg cells—in which clonal competition for self-pMHC ligands is avoided—similar to the effect observed with B6 CD4<sup>+</sup> T cells and OT-I CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts (**Figure 2B**).

For this, FACS-purified, CTV-labeled naïve B6 CD8<sup>+</sup> T cells were co-transferred with a mixture of AND TCR Tg (specific for I-E<sup>k</sup> -restricted pigeon cytochrome C 81-104 peptide) and OT-II TCR Tg (specific for I-A<sup>b</sup> -restricted Ova 323-339 peptide) naïve CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> or GF RAG−/<sup>−</sup> hosts and 7 days later, the mice were analyzed for donor cell proliferations by flow cytometry (**Figure 3B**, top). In marked contrast to the OT-I CD8<sup>+</sup> cells (**Figure 3A**), AND and to a lesser extent OT-II CD4<sup>+</sup> cells showed a significant increase in the proportion of their fast-rate of LIP by B6 CD8<sup>+</sup> T cells in SPF but not GF RAG−/<sup>−</sup> hosts (**Figure 3B**). Here, the greater effect on the LIP of AND cells than that of OT-II cells likely reflects the difference in their relative TCR affinity for self-ligands; thus, the LIP response was apparently faster and greater for T cells with a high affinity TCR (i.e., AND cells) than those with a low affinity TCR (i.e., OT-II cells) (34).

As for the above AND and OT-II CD4<sup>+</sup> cells, similar difference was observed with two distinct CD8<sup>+</sup> TCR Tg cells, namely OT-I vs. HY (specific for H-2D<sup>b</sup> -restricted male antigenderived peptide), in which their intrinsic TCR affinity for selfligands is much lower in HY cells than in OT-I cells (5, 6, 35). Thus, the results again clearly showed that the enhancing effect on the LIP was much greater for OT-I cells than for HY cells, when these cells were co-transferred with B6 CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts (**Figure S3A**). Likewise, the enhancing effect on the LIP of OT-I cells was also further confirmed together with co-transfer of P14 TCR Tg CD8<sup>+</sup> T cells (specific for H-2D<sup>b</sup> restricted lymphocytic choriomeningitis virus glycoprotein 33- 41 peptide) into SPF RAG−/<sup>−</sup> hosts (**Figure S3B**). These results thus suggest that the effect we observed is not OT-I-specific but broadly applicable for different monoclonal CD8<sup>+</sup> T cell populations with variable degrees depending on their relative TCR affinity for self-ligands.

Together, these findings indicate that the SP response of either polyclonal CD4<sup>+</sup> or CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts contributes to promoting the faster and greater LIP response of CD8<sup>+</sup> or CD4<sup>+</sup> T cells, respectively, in a TCR-self-MHC dependent manner.

### Effect of Ag-Specific CD4<sup>+</sup> T Cell Activation on the CD8<sup>+</sup> T Cell LIP in Acute Lymphopenic Conditions

The above data so far pointed out the unique ability of polyclonal T cells to promote the strong LIP of TCR Tg cells only when the former induces robust SP response in a chronic severe lymphopenic host such as SPF RAG−/<sup>−</sup> mice. Because the SP response in these hosts was commensal microbial antigen-dependent and thus failed to occur in a GF condition, it is possible that the above phenomenon is limited to a particular condition of SPF RAG−/<sup>−</sup> mice rather than a general occasion of typical lympho-depleted mice through irradiation, cytotoxic drugs, or T cell-depleting antibodies.

We thus sought to address this issue of whether antigeninduced strong T cell responses would also facilitate the rate

and intensity of LIP of naïve CD8<sup>+</sup> T cells (either from B6 and TCR Tg mice). For this, we generated two different lymphopenic settings derived from normal B6 mice by treatment of either a sub-lethal dose of irradiation (**Figure 4A**) or a monoclonal antibody against Thy1.2 for depleting host T cell compartment (anti-Thy1.2 mAb; clone 30H12; **Figure 4B**). In the first lymphopenic setting, normal B6 hosts receiving irradiation (600 cGy) were adoptively transferred with B6 CD8<sup>+</sup> T cells either alone or together with SMARTA TCR Tg CD4<sup>+</sup> T cells (specific for I-A<sup>b</sup> -restricted lymphocytic choriomeningitis virus glycoprotein 61-80 peptide; LCMV GP61) and 1 day later, were immunized with LCMV GP61 peptide antigen (**Figure 4A**, top); noted that the peptide antigen must be used here as a stimulus for SMARTA cells, because LCMV infection was lethal for the irradiated mice. Thus, while donor B6 CD8<sup>+</sup> T cells showed only a typical slow rate of LIP response when transferred alone with peptide injection, interesting finding was that these CD8<sup>+</sup> cells showed significantly elevated levels of rapidly dividing cells when SMARTA cells were co-transferred and activated with its cognate peptide antigen LCMV GP61 (**Figure 4A**, bottom). Here, the rather smaller increase of the fast-dividing donor B6 CD8<sup>+</sup> T cells in irradiated B6 hosts with activated SMARTA cells (∼37–52%; **Figure 4A**) relative to those of donor OT-I cells in SPF RAG−/<sup>−</sup> hosts with B6 CD4<sup>+</sup> T cells (∼78–80%; **Figure 2B**) might reflect a diverse range of heterogeneity in polyclonal CD8<sup>+</sup> T cell pools for their TCR affinity to self-ligands. In light of this view, we indeed found that the extent of the fast-rate of LIP driven by peptide-stimulated SMARTA cells in irradiated B6 hosts was much higher for CD5hi donor B6 CD8<sup>+</sup> T cells than for CD5lo counterparts co-transferred (∼76–79% vs. ∼13–14%, respectively; **Figure S4**). The data thus suggest a role of the activated CD4<sup>+</sup> T cells (here SMARTA cells) in promoting the rate and degree of antigen-independent LIP response of naïve CD8<sup>+</sup> T cells even in an irradiation-induced, acute lymphopenic condition.

To further confirm the above findings, we then utilized the second alternative approach in which normal B6 mice were injected with anti-Thy1.2 mAb (30H12) to acutely deplete host T cell compartment. The mAb-treated mice were then adoptively transferred with OT-I CD8<sup>+</sup> T cells either alone or along with SMARTA CD4<sup>+</sup> T cells, followed by either being un-infected or infected with LCMV to stimulate the latter SMARTA cells specifically (**Figure 4B**, top). Consistent with the results from the above irradiation-induced lymphopenic settings, the increased proportion of rapidly dividing OT-I cells was prominent with LCMV infection and subsequent activation of co-transferred SMARTA cells, whereas there was

were treated with a sub-lethal dose of irradiation (600 cGy) 1 day before cell transfer and then injected i.v. with CTV-labeled polyclonal naïve CD8<sup>+</sup> T cells (CD45.1; 5 × 10<sup>5</sup> cells) either alone or along with naïve SMARTA CD4<sup>+</sup> T cells from SMARTA TCR Tg mice (CD90.1; 5 × 10<sup>4</sup> cells) and 1 day later, immunized intraperitoneally (i.p.) with LCMV peptide GP61−80 (top). MLN and SPL were analyzed on day 7 by flow cytometry for CTV dilution (bottom left two panels) and percentages of the fast LIP of donor CD8<sup>+</sup> T cells (bottom right). (B) B6 mice (CD90.2) were injected i.p. with anti-Thy1.2 mAb (30H12) 2 days before cell transfer and then injected i.v. with CTV-labeled OT-I CD8<sup>+</sup> T cells (CD90.1; 5 × 10<sup>5</sup> cells) either alone or along with SMARTA CD4<sup>+</sup> T cells (CD90.1; 5 × 10<sup>4</sup> cells) and 1 day later, injected i.p. either with PBS or with LCMV Armstrong (2 × 10<sup>5</sup> PFU; top). MLN and SPL were analyzed on day 7 by flow cytometry for CTV dilution (bottom left two panels) and percentages of the fast LIP of donor OT-I cells (bottom right). Data shown are the mean ± SEM (n = 3-4 mice per group) and are representative of at least three independent experiments. \*p < 0.05; \*\*p < 0.01; ns, not significant.

no such increase of OT-I cell proliferation by SMARTA cells without LCMV infection (∼43% vs. ∼5–6%, respectively; **Figure 4B**, bottom two rows). As a control, there was only a slow rate of LIP of OT-I cells without SMARTA cells regardless of LCMV infection (**Figure 4B**, middle two rows); however, noted that, albeit at a smaller portion, LCMV infection appears to have a tendency of slight increase of OT-I cell LIP in this T-depleted condition, presumably due to some innate responses derived from residual host-derived, non-T cell populations.

The above effects on the OT-I cell LIP driven by the antigen-stimulated SMARTA cells were all observed in acute lymphopenic conditions (**Figures 4A,B**). Importantly, however, these effects were not detected in normal lympho-replete hosts; thus, upon LCMV infection without 30H12 mAb-induced lymphopenia, both the slow- and the fast-responding LIP of OT-I cells were severely decreased (**Figure S5**), therefore highlighting a stringent requirement of lymphopenia. Collectively, these findings strongly suggest that the above phenomenon seen in SPF RAG−/<sup>−</sup> hosts is not due to their unique environment of chronic lymphopenia. Instead, the elevated levels of the fast-rate LIP of naïve CD8<sup>+</sup> T cells can efficiently occur under various forms of acute lymphopenic conditions if two key requirements are provided, namely CD4<sup>+</sup> T cells and their specific cognate antigens capable of stimulating these cells.

mAb were injected i.v. with a mixture of CTV-labeled OT-I CD8<sup>+</sup> (CD90.1; 5 × 10<sup>5</sup> cells) and SMARTA CD4<sup>+</sup> T cells (CD90.1; 5 × 10<sup>4</sup> cells) and then infected i.p. with LCMV Armstrong (2 × 10<sup>5</sup> PFU; top). These mice were injected i.p. either with PBS or anti-IL-2 mAbs (two clones; JES6-1 and S4B6) at the indicated time points (top). CTV dilution (bottom left) and percentages of the fast LIP of donor OT-I cells (bottom right) were analyzed on day 8 by flow cytometry. Data shown are the mean ± SEM (n = 3 mice per group) and are representative of at least three independent experiments. \*p < 0.05; \*\*p < 0.01.

# Role of IL-2 in Promoting Strong LIP Response in Acute Lymphopenic Hosts

The above findings that the stimulated CD4<sup>+</sup> T cells under acute lymphopenic settings led to the faster and greater LIP of CD8<sup>+</sup> T cells prompted us to elucidate its underlying mechanisms and raise the question of how this is regulated and which factor is involved. Based on the fact that the LIP of naïve CD8<sup>+</sup> T cells is known to be antigen-independent and largely driven by cytokines, especially IL-7 (or both IL-7 and to a lesser extent IL-15 for naïve CD4<sup>+</sup> T cells) (7, 10), it is possible that the enhanced LIP response was just a mere reflection of greatly increased levels of these cytokines, presumably accompanied by the antigen-dependent activation of CD4<sup>+</sup> T cells in this condition.

We therefore tested this possibility of a role of the elevated amounts of IL-7 and/or IL-15 as a mechanism of promoting the antigen-independent strong LIP of CD8<sup>+</sup> T cells after LCMV infection. For this, a mixture of SMARTA CD4<sup>+</sup> and OT-I CD8<sup>+</sup> T cells was adoptively transferred into the aforementioned Tdepleted lymphopenic B6 mice (using 30H12 mAb treatment) of either wild-type (WT) or double knock-out (DKO) for both IL-7 and IL-15, and 1 day later, the mice were infected with LCMV and analyzed on day 7 by flow cytometry (**Figure 5A**, top). Here, the notable finding was that the enhancing effect on the LIP of OT-I cells was prominent with LCMV-activated SMARTA cells for both WT and DKO hosts compared to that of OT-I cells without SMARTA cells (**Figure 5A**, bottom), implying a role of different factor(s) other than IL-7 and IL-15.

In an attempt for searching the key factor(s), we then tested a possible role of IL-2 because this cytokine is mainly produced from T cells after antigenic stimulation. Moreover, especially for naïve CD8<sup>+</sup> T cells, IL-2 is known to induce an intense form of antigen-independent, rapid proliferative response in both lympho-deplete and even lympho-replete conditions (13). To address a role of IL-2, the 30H12-treated B6 mice were adoptively transferred with a mixture of SMARTA and OT-I cells and 1 day later, infected with LCMV in the presence or absence of anti-IL-2 mAbs (JES6-1 and S4B6 clones) for blocking in vivo IL-2, and then analyzed on day 8 by flow cytometry (**Figure 5B**, top). Here, the result was surprising; while the OT-I cells cotransferred with SMARTA cells showed the elevated proportion of the fast-dividing LIP, these cells failed to do so after LCMV infection along with IL-2 blockade (**Figure 5B**, bottom). The data thus suggest that the enhancing effect of the antigen-activated CD4<sup>+</sup> T cells on the LIP of CD8<sup>+</sup> T cells was associated at least in part with in vivo activity of IL-2.

#### Role of IL-2 in Enhancing Strong LIP Response in SPF RAG−/<sup>−</sup> Hosts

The implication of the above results with IL-2 blockade is that the IL-2 may act as a key factor and is produced by LCMV-activated SMARTA CD4<sup>+</sup> T cells, a level of which is perhaps sufficient to promote the rapid and robust form of antigen-independent LIP of co-transferred OT-I CD8<sup>+</sup> T cells in T-depleted hosts. Given the close similarity in the requirement of antigen-induced CD4<sup>+</sup> T cell activation, we reasoned that the enhancing effect on the LIP of OT-I cells observed in SPF RAG−/<sup>−</sup> hosts (**Figure 2B**) might also be mediated via a mechanism that is dependent on IL-2 presumably produced from activated CD4<sup>+</sup> T cells undergoing SP in these hosts.

We therefore tested again this possibility and the results were indeed the case with the following two observations: (1) a decrease and (2) an increase in the fast-dividing LIP by blocking or enhancing in vivo IL-2 activity, respectively (**Figure 6A** and **Figure S6A**). For this, SPF RAG−/<sup>−</sup> mice were adoptively transferred with either OT-I cells alone or a mixture of OT-I cells and B6 CD4<sup>+</sup> T cells, and then untreated or treated either with anti-IL-2 mAbs for blocking IL-2 (**Figure 6A**, top) or with IL-2 as a form of IL-2 and anti-IL-2 (S4B6) immune complex known to enhance IL-2 activity in vivo (**Figure S6A**) (12, 36). Here, the result was that OT-I cells co-transferred with B6 CD4<sup>+</sup> T cells failed to show the fast-dividing LIP after IL-2 blockade, while undergoing the slow-rate of LIP similar to those seen in OT-I cells transferred alone (**Figure 6A**, bottom). Conversely, however, the lack of the fast-rate LIP of OT-I cells transferred alone was completely restored after treatment with IL-2/anti-IL-2 complex, to the level comparable to those seen in OT-I cells co-transferred with B6 CD4<sup>+</sup> T cells (**Figure S6A**). Moreover, these IL-2/anti-IL-2 complexes also induced the faster and greater LIP of OT-I cells even when co-transferred with competing B6 CD8<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts (**Figure 3A** and **Figure S6B**). Besides these findings with OT-I cells, we also obtained similar data for polyclonal B6 CD8<sup>+</sup> and CD4<sup>+</sup> T cells co-transferred into SPF RAG−/<sup>−</sup> hosts with or without blocking IL-2 (**Figure 6B**); here, again, the in vivo IL-2 blockade led to ∼26-62% of significant reduction of the fast-dividing CD4<sup>+</sup> and CD8<sup>+</sup> T cells in these hosts.

We further confirmed the data from the above blocking experiments with anti-IL-2 mAb to validate the role of IL-2 produced by CD4<sup>+</sup> T cells undergoing SP in SPF RAG−/<sup>−</sup> hosts. For this, we first generated bone marrow (BM) chimera reconstituted with a 50:50 ratio of WT and IL-2−/<sup>−</sup> BM cells to avoid spontaneous T cell activation due to autoimmunity in IL-2 <sup>−</sup>/<sup>−</sup> mice (37) and 8 weeks later, naïve CD4<sup>+</sup> T cells derived from WT and IL-2−/<sup>−</sup> BM were isolated from the chimeric mice (**Figure S6C**). OT-I cells were then adoptively transferred alone or co-transferred with either the WT or the IL-2−/<sup>−</sup> CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts and analyzed on day 7 by flow cytometry (**Figure 6C**, top). While the fast-dividing LIP of OT-I cells was apparent with WT CD4<sup>+</sup> T cells, this response was totally abolished with IL-2−/<sup>−</sup> CD4<sup>+</sup> T cells despite their SP response was relatively intact (**Figure 6C**, bottom). Consistent with this finding, our data from additional transfer experiments also revealed that rapidly dividing activated CD4<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts indeed show a profound synthesis of intracellular IL-2 (and IFN-γ) after short-term in vitro restimulation, a level of which was much higher than that of OT-I cells co-transferred (**Figure S7**).

Together, all these findings strongly support the notion that IL-2 is produced from polyclonal CD4<sup>+</sup> (and CD8+) T cells that are activated as a result of antigen-dependent SP response in SPF RAG−/<sup>−</sup> hosts, which then acts as a key player for promoting the rapid and robust form of antigen-independent, but IL-2-dependent LIP of monoclonal CD8<sup>+</sup> T cells in these hosts.

#### Effector Activity of IL-2-Driven Strong LIP Response Against Bacterial Infection

Given the above data showing the strong proliferation of CD8<sup>+</sup> T cells driven by IL-2 that is produced from antigen-stimulated CD4<sup>+</sup> T cells co-transferred into either a chronic or an acute lymphopenic host, we next sought to address whether these robust form of antigen-independent, IL-2-driven response is functionally relevant for immune responses.

For this, SPF RAG−/<sup>−</sup> hosts were adoptively transferred with either a mixture of OT-I cells and B6 CD4<sup>+</sup> T cells or as a control OT-I cells alone, and for the latter control group, were either un-immunized or immunized with OVA antigen, and then analyzed on day 7 by flow cytometry (**Figure 7A**, top). Here, as expected, donor OT-I cells cotransferred with B6 CD4<sup>+</sup> T cells showed robust expansion and thus their recoveries from blood, SPL, lung and liver were

mice were then analyzed on day 7 by flow cytometry for CTV dilution (bottom left two panels) and percentages of the fast LIP of donor OT-I cells (bottom right). Data shown are the mean ± SEM (n = 3 mice per group) and are representative of three independent experiments. \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001, \*\*\*\*p < 0.0001.

i.p. either with LCMV Armstrong (2 × 10<sup>5</sup> PFU) or with OVA protein (100 µg/mouse; top). At 56 days after adoptive transfer, the mice were then challenged with OVA-expressing Listeria monocytogenes (LM-OVA) via oral gavage (5 × 10<sup>10</sup> CFU; top). The mice were analyzed on day 7 post-challenge for donor OT-I recovery from the SPL and SI (for intraepithelial lymphocytes; IEL) by flow cytometry (C), and were monitored at the indicated time points for body weight loss and survival (D), and also measured for bacterial counts in the SI and liver by plaque assay (E). Data shown are the mean ± SEM (n = 5 mice per group) and are representative of three independent experiments. \*p < 0.05; \*\*\*p < 0.001.

significantly greater than those of OT-I cells transferred alone without immunization (**Figure 7A**, bottom left). Importantly, the degree of expansion and recovery of OT-I cells were as near similar when co-transferred with B6 CD4<sup>+</sup> T cells as those of OT-I cells when transferred alone with OVA stimulation, and in particular, were also prominent in the gutassociated lymphoid tissues such as mLN, small, and large intestine (SI and LI, respectively) (**Figure 7A**, bottom right). Such preferential recovery of donor OT-I cells from the SI, especially intraepithelium (IEL) was also further confirmed by immunofluorescent tissue staining (**Figure 7B**). Furthermore, in line with the greater expansion, we found that the OT-I cells co-transferred with B6 CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts exhibited a high expression of granzyme B as well as an ability to produce IFN-γ and TFN-α upon short-term in vitro restimulation, characteristics of differentiation into functional effector cells (**Figure S8A**).

Based on the appearance of functional effector CD8<sup>+</sup> cells driven by IL-2 after co-transfer with B6 CD4<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts, we then investigated whether the similar events of effector differentiation would occur and also develop into functional memory cells in the aforementioned acute T celldepleted hosts with LCMV infection, a phenomenon that is also likely to be IL-2 dependent (**Figure 5B**). For this, 30H12 treated, T cell-depleted B6 mice were adoptively transferred with OT-I cells either alone or along with SMARTA cells, and then immunized with either LCMV or as a control OVA antigen (only for hosts receiving OT-I cells alone) (**Figure 7C**, top). At 2 months later, the mice were infected via oral gavage with internalin A (InlA)-mutated Listeria monocytogenes expressing OVA antigen (LM-OVA; **Figure 7C**, top), a mutant strain that is restricted to the intestinal infection through gut epithelial cells (29, 30). Here, at day 7 after LM-OVA challenge, the percentage and recovery of donor OT-I cells that had been primed with LCMV-stimulated SMARTA cells were as prominent in the SPL and to a much greater extent SI-IEL as those of OT-I cells that had been primed by OVA antigen, but much greater than those of OT-I cells that had been primed irrelevantly with LCMV in the absence of SMARTA cells (**Figure 7C**, bottom). Likewise, the OT-I cells from the former two groups of recipients also showed much higher ability to synthesize granzyme B and produce IFN-γ and TFN-α after in vitro restimulation (**Figure S8B**). Most importantly, in line with such prominent memory recall responses, OT-I cells that had been primed with LCMV-activated SMARTA cells resulted in the strong protective responses against lethal doses of LM-OVA challenges, comparable to those of OT-I cells that had been primed with OVA antigen, as evidenced by the smaller body weight reduction, higher survival rates, and lesser bacterial counts than those of control OT-I cells that had been un-primed (**Figures 7D,E**).

Collectively, these data indicate that the IL-2-driven, robust LIP response of CD8<sup>+</sup> T cells that is associated with antigendependent activation of CD4<sup>+</sup> T cells under lymphopenic conditions is accompanied by efficient generation of and differentiation into effector and memory cells that are functional for protecting hosts from pathogenic infections.

#### DISCUSSION

Unlike the slow rate of lymphopenia-induced homeostatic proliferation (LIP), the rapid and robust form of proliferative responses has been documented for naïve T cells particularly in a chronic lymphopenic host such as RAG−/<sup>−</sup> and TCRβ −/− (and also CD3ε <sup>−</sup>/−) mice (15, 17). Although this response (called as spontaneous proliferation; SP) is known to be antigendependent—an antigen derived from commensal microbiota, precise nature of this response and its impact on homeostasis and function for the responding T cells during their recovery from lymphopenia remain incompletely understood. In the present study, we addressed these issues and demonstrated that, upon adoptive transfer of polyclonal B6 naïve CD4<sup>+</sup> and CD8<sup>+</sup> T cells into SPF but not GF RAG−/<sup>−</sup> hosts, strong SP response of these cells affects the intensity and the tempo of the responding T cells, especially for those of undergoing antigen-independent LIP. Thus, the resulting LIP response in SPF RAG−/<sup>−</sup> hosts was rapid and intense, and was influenced by TCR affinity for self-ligands and most importantly, heavily dependent on IL-2 that is produced from activated T cells undergoing antigendependent SP. Notably, these observations were not limited to a unique environment of chronic lymphopenia but rather broadly applicable for various other acute lymphopenic conditions with two crucial requirements, namely antigen-dependent T cell activation and availability of relatively high amounts of IL-2. As a consequence, T cells undergoing IL-2-driven strong LIP showed a full capacity to differentiate into functional effector and memory cells that can provide a protective response against pathogenic bacterial infection.

The exact nature of stimuli for driving the strong SP response in RAG−/<sup>−</sup> hosts is unclear but a number of evidence revealed that this response mainly depends on a strong TCR signal via its engagement with a cognate antigen—presumably from commensal microbial components—and also costimulatory signal through CD28 (15, 18, 38, 39). Such dependency explains why the SP response occurs only with polyclonal but not monoclonal T cell populations, and is undetectable in GF RAG−/<sup>−</sup> hosts or severely reduced even in SPF RAG−/<sup>−</sup> hosts after treatment with antibiotics (15, 18). This notion therefore seems to fit well with the idea that the SP response by polyclonal naïve CD4<sup>+</sup> or CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts is a reflection of strong responsiveness of a few, albeit rare, clones that have specificity to a variety of commensal-derived peptide antigens. In fact, it has been shown that for CD4<sup>+</sup> T cells, the SP response is impaired in chronic lymphopenic mice lacking MHC-II expression (15, 17) but unimpaired in mice either being treated with mAb for blocking IL-7 or lacking IL-7 expression (7, 40, 41). Given this notion, our data showing enhanced SP response of B6 CD4<sup>+</sup> and CD8<sup>+</sup> T cells co-transferred into SPF RAG−/<sup>−</sup> hosts compared to those of either CD4<sup>+</sup> or CD8<sup>+</sup> T cells transferred separately were rather unexpected (**Figure 1B**), because these cells were unlikely to share TCR specificity for cognate antigens and they were subjected to strong activation independently in an antigen-specific manner.

Based on the findings from the above co-transfer experiments, it was tempting to speculate that the SP response is perhaps influenced at least in part by a factor other than TCR engagement with cognate antigens. This prediction was indeed true and supported by the following two surprising observations: the rapid and robust "SP-like" proliferative responses of (1) OT-I TCR Tg CD8<sup>+</sup> T cells co-transferred with B6 CD4<sup>+</sup> T cells and (2) AND TCR Tg CD4<sup>+</sup> T cells co-transferred with B6 CD8<sup>+</sup> T cells into SPF RAG−/<sup>−</sup> hosts (**Figures 2B**, **3B**). Because there are no cognate antigens specific for OT-I and AND cells, such robust proliferations of these cells we observed in SPF RAG−/<sup>−</sup> hosts appear to rule out a role of antigen-dependent signals through OT-I or AND TCR per se. However, despite the lesser importance of antigen-specific TCR engagement, the robust responses of TCR Tg cells in SPF RAG−/<sup>−</sup> hosts still depend on a covert TCR signal derived from its contacts with self-ligands. Thus, the reduced proliferative responses were apparent with monoclonal or polyclonal T cells of lower TCR affinity for self-ligands, e.g., HY or OT-II cells and B6 CD5lo cells transferred into SPF RAG−/<sup>−</sup> hosts (compared to those of OT-I or AND cells and B6 CD5hi cells, respectively; **Figure 3B** and **Figures S3A**, **4**). In this respect, the observed self-dependence of these TCR Tg cells and relative difference in their proliferative response in SPF RAG−/<sup>−</sup> hosts closely resemble those seen with these cells in acute lymphopenic hosts (7, 8), although the intensity and tempo of the proliferative responses are distinctly different. Hence, the logical explanation for the proliferative response of monoclonal TCR Tg cells driven by polyclonal T cells in SPF RAG−/<sup>−</sup> hosts is that this phenomenon falls into the same category as the typical IL-7-driven LIP response in that the response occurs (1) only in lymphopenic conditions; depends (2) on TCR engagement with self-ligands; and is (3) independent of cognate antigenic stimulation. But the difference is that the former is much faster and stronger than the latter LIP and is driven by a combination of two important stimuli, namely a tonic TCR self-reactivity and presumably a much more potent cytokine than IL-7.

In an attempt to search for the latter cytokine factor that has a stimulatory activity, IL-2 is the most suitable candidate. This prediction is based on our findings that the robust LIP of TCR Tg cells in SPF RAG−/<sup>−</sup> hosts is stringently dependent on the presence of polyclonal B6 T cells and their strong activation, likely inducing effector cells capable of producing IL-2. In fact, we and others have previously shown that IL-2 acts as a potent stimulator for naïve CD8<sup>+</sup> T cells by itself even in the absence of antigenic stimulation both in vitro and in vivo (35, 42). Moreover, this cytokine—either alone or along with IL-15—was shown to induce much intense form of proliferative responses for naïve T cells in lympho-deplete or even lympho-replete hosts (12, 35, 43), the intensity and tempo of which are similar to those of the strong LIP response we observed in SPF RAG−/<sup>−</sup> hosts. Because these previous studies, however, utilized an experimental setting in which naïve T cells are exposed to a supraphysiological level of IL-2 in a rather unphysiological condition, a role of IL-2 was needed to be validated in our system. In this regard, two critical questions arise: (1) does IL-2 indeed act as a key stimulatory factor? and (2) are polyclonal B6 T cells that are activated and proliferated in SPF RAG−/<sup>−</sup> hosts a major source for in vivo IL-2 production? These were indeed the case and clearly supported by a series of our in vivo data showing a decrease of the robust LIP response of OT-I cells by IL-2 blockade (**Figure 6A**), an increased response by IL-2 administration (**Figures S6A,B**), and finally an impaired response when co-transferred with IL-2−/<sup>−</sup> CD4<sup>+</sup> T cells (**Figure 6C**). Although a stimulatory role of IL-2 was better highlighted with TCR Tg cells (here OT-I cells), it should be noted, however, that the effect by IL-2 was also prominent for polyclonal B6 T cells (**Figure 6B**). Thus, the reduction of the strong proliferative responses by IL-2 blockade was apparent for both B6 CD8<sup>+</sup> and to a lesser extent CD4<sup>+</sup> T cells. Why the B6 CD4<sup>+</sup> T cells were less effective for the IL-2 blockade is unclear. Whether this is a reflection of less contribution of IL-2-driven LIP yet more reliance on antigen-driven SP for CD4<sup>+</sup> T cells than for CD8<sup>+</sup> T cells remains to be addressed.

Based on the above effect of IL-2, it seems conceivable that the typical SP response of polyclonal B6 T cells in SPF RAG−/<sup>−</sup> hosts largely consists of at least two different forms of proliferation, namely an antigen-dependent "true" SP as well as an antigen-independent and IL-2-dependent "bystander" LIP, with its strength and rate of proliferation akin to those of the SP. In light of this notion, it should be taken into caution that the SP response reported in some previous studies utilizing co-transfer experiments of polyclonal CD4<sup>+</sup> and CD8<sup>+</sup> T cells into the same chronic lymphopenic hosts might include such a bystander component of IL-2-driven rapid LIP. Because this response in these hosts occurs in a manner independent of TCR engagement with cognate antigens, T cells (especially for CD8<sup>+</sup> T cells) would be strongly responding and proliferating even in a situation where MHC-I expression is limited or absent if MHC-II expression is intact for CD4<sup>+</sup> T cells to drive their SP and IL-2 production. In fact, a previous study has shown that, when transferred with an unseparated mixture of naïve CD4<sup>+</sup> and CD8<sup>+</sup> T cells, the strong proliferative response of the latter population is found even in MHC-I-deficient TCRβ <sup>−</sup>/<sup>−</sup> hosts and concluded that this phenomenon is MHC-II-, but not MHC-I-, dependent (17). In this situation, however, determining whether the observed response of CD8<sup>+</sup> T cells in an MHC-I-lacking environment would reflect the effect of IL-2 produced from the co-transferred antigen-activated CD4<sup>+</sup> T cells will be interesting.

How IL-2 can induce such a robust SP-like bystander response is unclear. Previously we have addressed this issue and demonstrated that IL-2 (or IL-15) has a unique ability to drive activation and proliferation of naïve T cells, particularly CD8<sup>+</sup> T cells, via a mechanism dependent on the density of lipid rafts on the T cell membrane (35). In this study, a relatively high concentration of IL-2 in vitro could induce a rapid membrane relocalization and clustering of IL-2Rβ chain (CD122) into the membrane micro-domains of lipid rafts, leading to the enhanced activation and amplification of IL-2R and its downstream signal transduction pathways, including activation of JAK/STAT, ERK, and PI3K/AKT signaling pathways (35). Hence, the implication from these studies is that naïve CD8<sup>+</sup> T cells (and to a lesser extent CD4<sup>+</sup> T cells) are able to undergo rapid SP-like responses only when these cells are exposed to IL-2 at high concentrations and that this response may occur in any situation where IL-2 is increased at sufficiently high levels in vivo. Inducing IL-2 production may be easily achieved by strong antigenic stimulation of naïve T cells, yet reaching an effective concentration at levels sufficient for inducing robust SP-like response seems less easy largely because of shorter half-life of in vivo IL-2 and its rapid consumption by overwhelming numbers of T cells being expanded during antigen-specific immune responses. Indeed, we found that, upon adoptive transfer of OT-I cells into normal lympho-replete B6 hosts, potent antigenic stimulation of either polyclonal B6 (hostderived) or monoclonal SMARTA CD4<sup>+</sup> T cells (co-transferred) triggered by LCMV infection fails to induce IL-2-driven strong proliferative responses of OT-I cells (**Figure S5**). The result, however, was totally different when such antigenic stimulation

occurs in a lympho-deplete condition, resulting in the robust IL-2-dependent SP-like response of OT-I cells (**Figures 4**, **5**). Our findings therefore provide strong support for the notion that such robust proliferative responses of naïve T cells can be easily driven by relatively increased levels of in vivo IL-2 produced from antigen-stimulated T cells in both situations of chronic and acute lymphopenia.

Our study presented here does not rule out the possibility that the IL-2 may function indirectly through CD4<sup>+</sup> T cells or DCs rather than direct action on CD8<sup>+</sup> T cells (19). Here, the possible scenario is that IL-2 may modulate DCs to induce co-stimulatory molecules such as B7.1 (CD80) and B7.2 (CD86) that can enhance a stimulatory activity of DCs or, alternatively, would help to promote CD4<sup>+</sup> T cell activation and subsequent upregulation of co-stimulatory molecule, such as CD40L, engagement of which then results in maturation and stimulation of otherwise immature DCs. As a consequence of such IL-2 conditioning, the resulting DCs would then facilitate the robust proliferative responses of bystander CD8<sup>+</sup> T cells. In fact, despite some controversial results (44), the upregulation of CD80 and CD86 on DCs has been documented in RAG−/<sup>−</sup> mice (19); here, the expression of these molecules on DCs was higher in RAG−/<sup>−</sup> mice than in normal B6 mice, and was largely attributed to the lack of CD4<sup>+</sup> T regulatory (Treg) cells, which are known to restrain upregulation of costimulatory molecules and maturation/activation of steady-state DCs via CTLA4 (45–47). More importantly, DC costimulation was even further upregulated in RAG−/<sup>−</sup> mice when adoptively transferred with naïve CD4<sup>+</sup> T cells. Although IL-2 has not been considered as a modulatory factor for DCs in this study, the enhanced DC expression of co-stimulatory molecules observed with CD4<sup>+</sup> T cell transfer into RAG−/<sup>−</sup> hosts did not seem to require the activity of IL-2—presumably produced from CD4<sup>+</sup> T cells transferred—on DCs with mainly two reasons: First, the enhancement of CD80 and CD86 expression on DCs in RAG−/<sup>−</sup> mice was not detected with in vivo administration of IL-2 delivered as a form of IL-2 and anti-IL-2 immune complexes (19), for which we also confirmed in our study (data not shown). Second, similar to the results observed with CD4<sup>+</sup> T cell transfer into RAG−/<sup>−</sup> hosts, the enhanced expression of DC costimulatory molecules has also been reported in mice lacking IL-2 or its receptors (CD25 and CD122), which is known to be accompanied by spontaneous activation of conventional CD4<sup>+</sup> CD25<sup>−</sup> T cells resulting from the lack of Treg suppression in these mice (48). Hence, although we favor the view that IL-2 is dispensable for inducing the enhanced DC expression of co-stimulatory molecules, the additional experiments—e.g., using adoptive transfer with CD25- or CD122-deficient OT-I cells into RAG−/<sup>−</sup> hosts—will be necessary to provide strong evidence for a direct stimulatory role of IL-2 on CD8<sup>+</sup> T cells.

Whether these observations with increased levels of IL-2 have any particular relevance in a normal physiological condition remains to be addressed. In this respect, it has been shown that, similar to responses in RAG−/<sup>−</sup> hosts, naïve CD4<sup>+</sup> T cells also undergo the rapid and intense form of SP response upon adoptive transfer into neonatal hosts (neonates at 1–3 days of age) (49). Therefore, it will be of interesting to investigate whether some, if not all, of the SP response of naïve T cells (especially for CD8<sup>+</sup> T cells) that occur during a neonatal period is perhaps IL-2-dependent and how these cells would behave for their homeostasis and function throughout an adult life. With regard to the functional aspect of IL-2-driven T cell responses, we showed that these cells after the intense proliferative responses (both in RAG−/<sup>−</sup> hosts and in lymphodepleted B6 hosts with LCMV infections) are able to fully differentiate into effector and memory cells that are functional for controlling pathogenic infections at levels equivalent to those of antigen-induced effector/memory cells (**Figure 7**). Therefore, it seems clear that these findings, together with additional future studies, would provide better understanding of the precise nature of the strong SP response of polyclonal naïve T cells in chronic lymphopenic hosts and of previously unappreciated role of IL-2 in regulating their homeostasis and functional responses.

In summary, besides a role of enhanced DC costimulation (19), a clear stimulatory role of IL-2 is apparent for naïve CD8<sup>+</sup> T cells in SPF RAG−/<sup>−</sup> hosts. We show here in this study that this IL-2-driven stimulatory effect occurs only in a particular situation where CD4<sup>+</sup> T cells co-exist and their strong antigendependent activation is induced under various lymphopenic environments. Although the phenomenon observed in these particular conditions seems less physiological, our findings nevertheless would have an implication for the development of therapeutic interventions against cancer, especially for those of using pre-conditioning regime of lymph-depletion prior to adoptive T cell transfer-based immunotherapy (50–54).

# ETHICS STATEMENT

This research was approved by the Institutional Animal Care and Use Committees (IACUC) of the Pohang University of Science and Technology (2013–01–0012). Mouse care and experimental procedures were performed in accordance with all institutional guidelines for the ethical use of non-human animals in research and protocols from IACUC of the Pohang University of Science and Technology.

#### AUTHOR CONTRIBUTIONS

JK, JL, and KC performed experiments. JK, CS, and J-HC designed experiments and analyzed and interpreted the data. S-WH, KK, JS, and S-HI contributed to this study with valuable discussion and critical comments. JK and J-HC wrote the manuscript.

#### ACKNOWLEDGMENTS

We thank all the AIM/IBS laboratory members for both direct and indirect supporting; Haejin Jung and Me Ok Lee for assistance with cell sorting by flow cytometry; POSTECH Biotech Center animal facility for animal breeding, housing and supply; and all AIM/IBS administrative staffs. We thank Naomi Taylor and Valerie Dardalhon for helpful comments on this study. We thank Brian S. Sheridan for Listeria monocytogenes. This work was supported by project IBS-R005-D1 from the Institute for Basic Science, Korean Ministry of Science and Information/Communication/Technology.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.01907/full#supplementary-material


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

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

# Early Growth Response Gene-2 Is Essential for M1 and M2 Macrophage Activation and Plasticity by Modulation of the Transcription Factor CEBPβ

Tatyana Veremeyko<sup>1</sup> , Amanda W. Y. Yung<sup>1</sup> , Daniel C. Anthony <sup>2</sup> , Tatyana Strekalova3,4,5 and Eugene D. Ponomarev 1,6 \*

*<sup>1</sup> Faculty of Medicine, School of Biomedical Sciences, The Chinese University of Hong Kong, Shatin, Hong Kong, <sup>2</sup> Department of Pharmacology, University of Oxford, Oxford, United Kingdom, <sup>3</sup> Department of Neuroscience, Maastricht University, Maastricht, Netherlands, <sup>4</sup> Institute of General Pathology and Pathophysiology, Moscow, Russia, <sup>5</sup> Laboratory of Psychiatric Neurobiology, Institute of Molecular Medicine and Department of Normal Physiology, Sechenov First Moscow State Medical University, Moscow, Russia, <sup>6</sup> Kunming Institute of Zoology-Chinese University of Hong Kong Joint Laboratory of Bioresources and Molecular Research of Common Diseases, Kunming, China*

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Rohit Mathur, University of Texas MD Anderson Cancer Center, United States Geeta Upadhyay, Uniformed Services University of the Health Sciences, United States*

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

#### Specialty section:

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

Received: *10 May 2018* Accepted: *11 October 2018* Published: *01 November 2018*

#### Citation:

*Veremeyko T, Yung AWY, Anthony DC, Strekalova T and Ponomarev ED (2018) Early Growth Response Gene-2 Is Essential for M1 and M2 Macrophage Activation and Plasticity by Modulation of the Transcription Factor CEBP*β*. Front. Immunol. 9:2515. doi: 10.3389/fimmu.2018.02515* The process of macrophage polarization is involved in many pathologies such as anti-cancer immunity and autoimmune diseases. Polarized macrophages exhibit various levels of plasticity when M2/M(IL-4) macrophages are reprogrammed into an M1-like phenotype following treatment with IFNγ and/or LPS. At the same time, M1 macrophages are resistant to reprogramming in the presence of M2-like stimuli. The molecular mechanisms responsible for the macrophages polarization, plasticity of M2 macrophages, and lack of plasticity in M1 macrophages remain unknown. Here, we explored the role of Egr2 in the induction and maintenance of macrophage M1 and M2 polarization in the mouse *in vitro* and *in vivo* models of inflammation. Egr2 knockdown with siRNA treatment fail to upregulate either M1 or M2 markers upon stimulation, and the overexpression of Egr2 potentiated M1 or M2 marker expression following polarization. Polarisation with M2-like stimuli (IL-4 or IL-13) results in increased Egr2 expression, but macrophages stimulated with M1-like stimuli (IFNγ, LPS, IL-6, or TNF) exhibit a decrease in Egr2 expression. Egr2 was critical for the expression of transcription factors CEBPβ and PPARγ in M2 macrophages, and CEBPβ was highly expressed in M1-polarized macrophages. In siRNA knockdown studies the transcription factor CEBPβ was found to negatively regulate Egr2 expression and is likely to be responsible for the maintenance of the M1-like phenotype and lack plasticity. During thioglycolate-induced peritonitis, adoptively transferred macrophages with Egr2 knockdown failed to become activated as determined by upregulation of MHC class II and CD86. Thus, our study indicates that Egr2 expression is associated with the ability of unstimulated or M2 macrophages to respond to stimulation with inflammatory stimuli, while low levels of Egr2 expression is associated with non-responsiveness of macrophages to their activation.

Keywords: monocytes/macrophages, Egr2, CEBPβ, M1/M2 balance, activation, plasticity, inflammation

# INTRODUCTION

Currently, it is accepted that at least two (among others) pathways for macrophage activation exist leading to two distinct or polarized states: the M1- and M2-like phenotypes (1, 2). Changes in M1 vs. M2 balance is a hallmark of many autoimmune diseases (2). Common autoimmune diseases such as multiple sclerosis and rheumatoid arthritis are associated with the presence of M1, M2 or mixed M1/M2 subsets with the predominant pathological role of an M1 subset (3, 4). The M1 like phenotype is mediated by Th1-associated cytokine IFNγ and microbial products such as LPS. M1-like macrophages express a high level of MHC class II and CD86 and effectively stimulate CD4 T cells to drive strong pro-inflammatory properties (2, 5, 6). These include the manufacture of NO (produced by enzyme NOS2) and the secretion of a number of potent proinflammatory cytokines such as IL-1β, IL-6, and TNF (2, 6). The M2-like macrophage phenotype is induced by the Th2 associated cytokines IL-4 or IL-13. M2 macrophages express low levels of CD86 and are poor stimulators of CD4 T cells, and seem to promote tissue repair during the resolution phase of inflammation (1, 7). Under normal conditions, most tissueresident macrophages (e.g., peritoneal macrophages, microglia) exhibit M2-like characteristics. M2-like macrophages express a number of specific markers that include Arg1 (arginase), Fizz1 (cysteine-rich secreted protein), and Ym1 (extracellular matrixbinding lectin) (1, 8, 9). M2-like macrophages are known to produce more anti-inflammatory cytokines including IL-10 and TGFβ1; however, IL-10 is not considerate to be a specific M2 marker but is probably more indicative of the presence of deactivated macrophages (2, 10). The transcription factor PPARγ was recently shown to be involved in M2 polarization (6, 11), while its upstream transcription factor CEBPβ was shown to be necessary for the induction of both M1 and M2 markers upon stimulation with either M1 or M2 directing stimuli (2, 12–15).

Early growth response (Egr) proteins are a family of transcriptional regulators that mediate expression of multiple genes involved in cell growth and differentiation (16–18). There are four Egr proteins in the family and three (Egr1, Egr2, and Egr3) are expressed in macrophages (16, 19). The exact role of Egr1, Egr2, and Egr3 in the development and function of myeloid cells, such as monocytes and macrophages, remain unclear. It has been suggested that Egr1 plays an important role in the regulation of monocyte and macrophage differentiation. However, Egr1 deficient mice have normal numbers of macrophages, which suggests other functions for this protein must exist (19). Egr1, Egr2, and Egr3 expression was upregulated in myeloid progenitors when these cells are differentiated in vitro into macrophages in the presence of M-CSF (19). The role of Egr proteins was recently investigated in lymphoid cells (16); it was reported in this study that a conditional knockout for Egr2 and Egr3 resulted in a lethal autoimmune syndrome that was associated with excessive systemic levels of pro-inflammatory cytokines (20). The knockout also exhibited impaired antigen receptor-induced proliferation of B and T cells. It was suggested that Egr2 negatively regulates T and B cell activation and production of pro-inflammatory cytokines by the induction of suppressor of the cytokine signaling (SOCS) molecules SOCS1 and SOCS3 (20). In macrophages, Egr1 was shown to induce expression of SOCS1 in LPS-stimulated M1 macrophages (21). Similarly, Egr2 was found to be the positive regulator for SOCS1 and STAT5 in dendritic cells (22). In Tregs, Egr2 has been shown to regulate the expression of the anti-inflammatory cytokine TGFβ3 (23). In non-immune biology, Egr2 was found to be critical for hindbrain development and peripheral myelination and led to the perinatal death of Egr2-deficient mice (24, 25). However, the downstream action of Egr molecules in macrophages is still not well understood. Bone marrowderived myeloid precursors from Egr1/Egr3-double knockout mice exhibited macrophage differentiation that is identical to that of wild-type mice (19). Moreover, fetal liver-derived myeloid precursors from Egr1/Egr2-double-deficient mice did not show abnormalities in macrophage differentiation (19). These data indicate that other functions of Egr-family proteins in myeloid cells exist beyond the development and differentiation of myeloid progenitors into monocytic cells. Drawing upon published results on both lymphoid and myeloid cells, we hypothesized, that Egr family proteins are likely to be involved in the control of macrophage activation and/or polarization.

Although polarized M1 and M2 macrophages exhibit distinct phenotypes in vitro, the picture is not so clear in vivo during inflammation, where macrophages often exhibit dually activated or a mixed M1/M2-like phenotype (7–9, 26). The mixed M1/M2-like phenotype was associated with a high level of macrophage plasticity: polarized macrophages seem to change their phenotype at a whole population-based level over the time (27). However, to date, it has not been clear how this process is regulated at a molecular level. There are several studies, including some of our own, that demonstrate how M2 macrophages could be switched to the M1-like phenotype in vitro and in vivo (28, 29). Much less is known about the potential to switch M1-like macrophages toward an M2-like phenotype. There is one study regarding the switch of human M1-like macrophages to M2, but his study was very restricted in terms of the choice of M1 and M2 markers and the M1-polarizing conditions were modest (30). It was concluded based on extensive literature analysis that it would not be common for M1 macrophages to be switched to M2 in vivo and that it would only happen in a mild inflammatory reaction (1). In our previous study, we found that IFNγ/LPS-treated M1-like macrophages were unresponsive to IL-4 and would not upregulate M2-associated molecules such as Mrc1 (CD206) and microRNA-124 (miR-124) and did not downregulate M1 associated miR-155, MHC class II and CD86 (28). The reason for such non-responsiveness of M1-like macrophages remained unexplained at a molecular level. One of the possible mechanisms is mitochondrial dysfunction of M1-like macrophages, but exact molecular pathways that lead to this state remained unknown (31).

Only recently Egr2 was proposed as a new M2 marker for macrophage alternative activation (32); however, exact functions of transcriptional regulator Egr2 in M1-like [M(IFNγ) and M(LPS)] and M2-like [M(IL-4) and M(IL-13)] macrophage polarization was not investigated in details. Since Egr2 was not involved in macrophage development, it was important to investigate the role of Egr2 in the ability of macrophages to become activated toward M1- or M2-like phenotypes and maintain their M1- or M2-like phenotype during their reprogramming. In contrast to myeloid dendritic cells, macrophages are quite a heterogeneous population and their mode of activation depends on local microenvironment leading to classic (M1-like) and alternative (M2-like) states. While Egr2 was found to be a negative regulator of activation of dendritic cells targeting SOCS1 (22), the role of Egr2 in macrophage activation is not clear. Therefore, it is very important to understand the role of Egr2 in macrophage activation.

In this study, we found that the Egr family proteins, Egr3 and Egr2, were differentially expressed in M1 and M2 macrophages. We established that although Egr2 was upregulated in M2 macrophages, the expression of this protein was required for upregulation of M1 or M2 markers in response to M1-like (IFNγ, LPS, TNF, IL-6) or M2-like (IL-4, IL-13) stimuli respectively. Treatment of macrophages with IFNγ and/or LPS resulted in long-term downregulation of Egr2, which lasted for more than 72 h. Low levels of Egr2 was correlated with non-responsiveness of M1 macrophages to further stimulation. Knockdown of Egr2 by siRNA decreased expression of the M2 markers Arg1, Ym1, Fizz1 and upregulated IL-10 in IL-4-treated macrophages. The introduction of Egr2 siRNA also made macrophages less responsive to IFNγ and/or LPS resulting in the downregulation of the M1-associated markers NOS2, TNF, IL-6, IL-1β and Ptgs2 (Cox-2) and upregulation of IL-10. Alternatively, overexpression of Egr2 increased expression of a number of M1- and M2 associated markers in response to IFNγ and IL-4 respectively. We also demonstrated that Egr2 positively regulated the expression of CEBPβ, while, in turn, Egr2 was negatively regulated by CEBPβ. Expression of Egr2 in macrophages appeared to be critical for upregulation of MHC class II and CD86 under inflammatory conditions. These data indicate a new role for Egr2 as a factor that positively regulates macrophage activation and plasticity during M1 and M2 polarization.

#### MATERIALS AND METHODS

#### Mice

C57BL/6 (B6) and C57BL/6-Tg(ACTB-bgeo/DsRed.MST (DsRed transgenic) mice were originally purchased from the Jackson Laboratory (Bar Harbor, ME) and bred locally at the Laboratory Animal Services Center (LASEC) at the Chinese University of Hong Kong. All animal procedures were conducted under individual licenses from the Hong Kong government and were approved by the animal ethics committee at the Chinese University of Hong Kong.

#### Cells

Bone-marrow-derived macrophages (BMDMs) were grown in the presence of M-CSF as described earlier in our studies (28, 33). Bone marrow from 2 to 3 of 6 to 8 week-old B6 or DsRed transgenic mice was isolated by flushing from femur bones with PBS and a single cell suspension was obtained by passing the cell suspension through a 70µm Cell Strainer (Falcon). Macrophages were grown in DMEM media (Gibco) supplemented with 10% FBS (Gibco) and 10 ng/ml M-CSF (R&D) for 5 days in 24 well plate (0.5 ml media/well). Media was replaced after every 2–3 days. After 5 days, the cells were used in experiments at the density of 300,000–400,000 cells per well, and the purity of the cells was more than 95% as determined by two-color flow cytometry analyzing expression of CD11b and F4/80. For macrophage polarization, the cells were treated with IL-4 or IL-13 (both from R&D, 50 ng/ml), or IFNγ (R&D, 100 ng/ml), or LPS (100 ng/ml; Sigma), or TNF (100 ng/ml), or IL-1β (100 ng/ml), or IL-6 (100 ng/ml), or GM-CSF (50 ng/ml) (all from R&D) for the periods indicated ranging from 2 to 24 h. The mouse macrophage cell line RAW264.7 was purchased from ATCC and maintained in DMEM media supplemented with 10% FBS.

#### Flow Cytometry

For analysis of cell surface markers, the cells were stained with anti-CD86-FITC or anti-CD86-PE, anti-MHC class II-PE-Cy5 (all from BD Biosciences), and F4/80-FITC (Biolegend) or F4/80-APC (eBioscience). Transfection plasmids and siRNA had fluorescent reporters (GFP for plasmid and Cy3-labeled RNA for siRNA), which were used to gate on transfected cells. Macrophages were analyzed for expression of surface markers (MHC class II, CD86, F4/80) or intracellular markers (Egr2, IL-6, TNF). FcRs were blocked with mAb specific for mouse FcR (2.4G2; BD Biosciences). Intracellular staining for Egr2, IL-6 or TNF was performed similarly as described earlier (32, 33) using fixation/permeabilization agent (eBioscience) and anti-Egr2-APC (eBioscience), anti-IL-6-PE, or anti-TNF-PE mAbs (both from BD Biosciences). The cells were analyzed using LSRFortressaTM cytometer (BD Biosciences) and FlowJo software (Tree Star Inc.) as we described earlier (34).

### Quantitative Real-Time RT PCR

For quantitation by real-time RT PCR, total RNA was isolated by Qiagen or MirVana (Applied Biosystems) kits from BMDMs or RAW264.7 cell line. Real-time RT-PCR analyses were performed using TaqMan miRNA assays (Applied Biosystems). For analysis of mRNA expression of M1 and M2 associated molecules, we used primers that are described in **Table 1**. Relative expression levels were calculated using the 11C<sup>T</sup> method and normalized to the expression of GADPH housekeeping gene and then to the expression of a control sample that was defined as 1 (34). For the analysis of miR-155 expression, we used specific primers from Applied Biosystems and analyzed data by normalizing to the expression of short non-coding housekeeping snoRNA-55 and then to the reference sample using 11C<sup>T</sup> method as described previously (33, 35).

#### Western Blotting

Western blot analysis was performed according to a standard protocol as previously reported (33, 34). Antibodies for β-Actin (cat#4967), NOS2 (cat#9819), and Arg1 (cat#2982) were purchased from Cell Signaling. Antibodies for Egr2 (cat#692002) were purchased from Biolegend. Antibodies for CEBPβ (cat#606202) were purchased from Biolegend.

TABLE 1 | Primer sequence for mRNA expression analysis.


#### Macrophage Plasticity Assay

Plasticity assay was performed as described earlier (28). BMDMs were polarized toward M1 with IFNγ (100 ng/ml) and LPS (100 ng/ml) or, toward M2 with IL-4 (50 ng/ml) for 24 h, before washing and culture in media alone (DMEM with 10% FBS and 10 ng/ml M-CSF) or they were further polarized in M1 (IFNγ, 100 ng/ml and LPS, 100 ng/ml) or M2 (IL-4, 50 ng/ml) for another 24-h period.

#### Knockdown Experiments Using Small Interfering RNA

Egr2 interfering RNAs (siRNAs) (siGenome smart pool, cat# M-040303-01), and CEBPβ siRNA (siGenome smart pool, Cat # M-043110-02), and CEBPα siRNA [(siGenome smart pool, Cat # M-040561-01) and control siRNA (siGenome nontargeting control pool), cat# D-0011206-13-05] were purchased from Dharmacon (Lafayette, CO, USA) and used similarly as described earlier in our studies (36). BMDMs were transfected with siRNAs using TransIT-X2 System as a transfecting agent (Madison, WI, USA) according to the manufacturer's instructions. The efficiency of transfection was 80–90% as determined by the expression of a fluorescent reporter (Cy3 labeled RNA) in F4/80-positive cells as determined by flow cytometry.

#### Overexpression Plasmids and Experiments

Expression plasmid pMIG-Egr2 were generated as follows. The first fragment of Egr2 gene was obtained from plasmid mEgr2/LZRS (Addgene; plasmid #27784) by using enzyme digestion via two restriction sites XhoI and BglII. Another fragment of Egr2 was amplified from mouse DNA library using primers 5′ -CCTGAACTGGACCACCTCTACTC-3′ and 5 ′ -CAGCAGATCTCACGGTGTCCTGGTTC-3′ . Full Egr2 gene was amplified using primers 5′ -TGCTCTCGAGATGATGAC CGCCAAGGCCG-3′ and 5′ -CGTCTGAATTCTCACGGTGTC CTGGTTCGAGAG-3′ and introduced into plasmid pMIG via restriction sites XhoI and EcoRI and subsequent ligation. Expression plasmid pMIG-Cebpb was kindly provided by Prof. Thomas Graf (Centre for Genomic Regulation, Barcelona, Spain). The efficiency of transfection was 20–50% as determined by expression of GFP reporter in F4/80-positive RAW264.7 cells as determined by flow cytometry.

#### Nitric Oxide Synthase Activity Assay

NOS activity in BMDM cell lysate was measured using a colorimetric assay kit from Abcam (cat#ab211083) according to manufacturer's instructions.

#### Thioglycollate-Induced in vivo Inflammation

Thioglycollate-induced peritonitis was initiated in the group of 4–5 B6 mice by i.p. injection of 2 ml of 4% thioglycollate broth media (Sigma) in PBS. The cells were isolated via peritoneal lavage 4 days after injection of thioglycollate media. Adoptive transfer of BMDMs transfected with Egr2 siRNA or Control miRNA was performed 30 min prior to administration of thioglycolate medium. BMDMs were grown in the presence of M-CSF from bone marrow of DsRed transgenic mice, transfected as described above, and 4–5 × 10<sup>6</sup> cells per recipient B6 mouse were injected i.p. For FACS analysis, the cells were stained for CD86-FITC, MHC class II-PE-Cy5 and F4/80-APC, and DsRed+F4/80<sup>+</sup> gated cells were analyzed for the expression of MHC class II and CD86.

#### Statistical Analysis

The results are presented as a mean ± standard error (S.E.). Unpaired Student's t-tests were used to determine significance between two independent groups. P-values of less than 0.05 were considered to be significant. SigmaPlot software was used for the creation of the graphs and statistical analysis.

# RESULTS

#### EGR2 Is Upregulated Only in M2 but Not M1 Macrophages

In this study, we hypothesized that Egr1, Egr2, and Egr3 play an important role in the regulation of macrophage activation toward M1- and/or M2-like phenotypes. In the first series of experiments, we aimed to verify our hypothesis by investigating the kinetics of expression of these three genes in M2 and M1 polarized conditions. We used IL-4 or IFNγ to polarize macrophages toward M2 or M1, respectively. When we monitored expression of the M2 marker Arg1 in IL-4 treated macrophages, we found it was upregulated as early as 3 h after treatment and peaked at 24 h. Conversely, IFNγ treatment did not upregulate Arg1 with the 3–24 h period, which indicated successful M2 polarization (**Figure 1A**). The M1 marker NOS2 was not upregulated by IL-4, but it was upregulated by IFNγ as early as 3 h, peaked at 8 h, and was slightly decreased by 24 h (**Figure 1B**).

Next, we investigated the level of expression of Egr1 and found that it was upregulated by both IL-4, and IFNγ (**Figure 1C**). Egr2 was upregulated by IL-4, which is remained high for 24 h. By contrast, IFNγ treatment did not upregulate Egr2 (**Figure 1D**). Egr3 was only transiently upregulated by IFNγ that peaked at 8 h (**Figure 1E**). Thus, we found that all three members of Egr family were differentially regulated in macrophages following M1 or M2 polarization; Egr2 is associated with M2, Egr3 with M1-like macrophages, and Egr1 is a polarization marker (M0 to M1/M2).

### Egr2 Is Upregulated by IL-4 and IL-13 and Downregulated by IFNγ, LPS, IL-6 and TNF

We investigated whether Egr1, Egr2, and Egr3 were differentially regulated by M2-like stimuli (IL-4 and IL-13) and an expended set of M1-like stimuli (IFNγ, LPS, IL-6, IL-1, TNF, and GM-CSF). Egr1 was upregulated by both M2-like stimuli IL-4 and IL-13 (**Figure 2A**) and a number of the M1-like stimuli (IFNγ, IL-6, and GM-CSF) (**Figure 2B**). However, Egr1 was downregulated by LPS and TNF (**Figure 2B**). At the same time cell viability was not decreased in IFNγ, LPS, IL-6, or TNF treated macrophages as determined by vital dye staining (not shown). When we investigated Egr2, it was upregulated by IL-4 and IL-13 (**Figure 2C**) and it was downregulated by IFNγ, LPS, IL-6, and TNF, but not GM-CSF (**Figure 2D**). Egr3 was transiently downregulated IL-4 and IL-13 (**Figure 2E**), while it was transiently upregulated by two M1-like stimuli IFNγ and IL-6 (**Figure 2F**). These data indicated that only Egr2 was differentially regulated by M2- and most of the M1-like stimuli and mRNA expression of this molecule was stably high in M2 and stably low in M1 cells. Expression of Egr2 was also stably upregulated on protein level by IL-4 (**Figures S1A,B**), indicating a great importance of Egr2 vs. Egr1/2 in control of macrophages activation toward M2. Thus, we established that Egr2 was upregulated by M2 stimuli IL-4 and IL-13 and downregulated by a number of M1 stimuli IFNγ, LPS, IL-6, and TNF.

FIGURE 1 | Kinetics of expression of Egr1, Egr2, and Egr3, M2 marker Arg1, and M1 marker NOS2 in macrophages polarized toward M2 (IL-4), or M1 (IFNγ). Bone-marrow-derived macrophages (BMDMs) were analyzed as untreated (C) or treated with IL-4 (IL4) or IFNγ (IFN) as described in *Materials and Methods*. The cells were analyzed after 3, 5, 8, and 24 h of incubation with indicated cytokines. For analysis, the cells were washed, mRNA was isolated and the expressions of *Arg1* (A), *Nos2* (B), *Egr1* (C), *Egr2* (D), and *Egr3* (E) were analyzed by real-time RT PCR as described in *Materials and Methods.* Mean ± S.E. of 12 separate culture plate wells is shown [\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001 when compared to untreated (C) cells].

FIGURE 2 | Kinetics of expression of Egr1, Egr2, and Egr3 in macrophages activated with various M2 (IL-4 and IL-13) and M1 (IFNγ, LPS, TNF, IL-6, and GM-CSF) stimuli. (A–F) Bone-marrow-derived macrophages (BMDMs) were treated with M2 stimuli IL-4 or IL-13 (A,C,E) or M1 stimuli IFNγ, LPS, TNF, IL-6, and GM-CSF (B,D,F) as described in *Materials and Methods.* The cells were analyzed immediately (0 h), or after 8 or 24 h of incubation with indicated M2 or M1 stimuli. For analysis, the cells were washed, mRNA was isolated and the expressions of *Egr1* (A,B), *Egr2* (C,D), and *Egr3* (E,F) were analyzed by real-time RT PCR as described in *Materials and Methods.* Mean ± S.E. of six separate culture plate wells is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001 when compared to 0 h cells). (G) BMDMs were treated with LPS or IFNγ as described in *Materials and Methods* and the cells were analyzed as untreated (C) or after 24 h, or 48 h, or 72 h of incubation with indicated M1 stimuli. For analysis, the cells were washed, mRNA was isolated and the expression of was analyzed by real-time RT PCR as described in *Materials and Methods.* Mean ± S.E. of six separate culture plate wells is shown (\*\**p* < 0.01; \*\*\**p* < 0.001). (H) BMDMs were treated with LPS as described in *Materials and Methods* and the cells were analyzed as untreated (C) or 24 h, 48 h, and 72 h of incubation with LPS (100 ng/ml). For analysis, the cells were washed, and the expression of Egr2 was analyzed on a protein level by western blot as described in *Material and Methods*. β-Actin has used a loading control. Representative western blot is shown in Figure S1C. Mean ± S.E. of 3–4 separate culture plate wells is shown (\*\**p* < 0.01).

### LPS and IFNγ Cause Long-Term Downregulation of Egr2 in M1 Macrophages

Since Egr2 was differentially expressed in M1- and M2-like macrophages we focused on this factor and looked at the long-term kinetics of the expression of Egr2 in M1-stimulated macrophages. We found that IFNγ- or LPS- treated macrophages displayed long-term downregulation of Egr2, which lasted for more than 72 h (**Figure 2G**). Downregulation of Egr2 on mRNA level was also confirmed on a protein level indicating ∼3 fold decrease in expression of this protein in M1 macrophages (**Figure 2H**, **Figure S1C**). Thus, M1 polarized macrophages downregulate Egr2 mRNA and protein levels, while Erg2 is upregulated in M2 macrophages.

#### M1 Macrophages Exhibiting Low Levels of Egr2 Fail to Downregulate M1 Markers and Weakly Upregulate M2 Markers in Response to Treatment With IL-4

In this study, we investigated the possibility to reprogram the M1- and M2-polarized macrophages. M1-like macrophages treated with IL-4 retained the capacity to express high levels M1 markers NOS2 (**Figure 3A**, LPS/IFNγ and LPS/IFNγ Second stimulus: IL-4), IL-6 (**Figure 3B**, LPS/IFNγ and LPS/IFNγ Second stimulus: IL-4), TNF (**Figure 3C**, LPS/IFNγ and LPS/IFNγ Second stimulus: IL-4), and displayed limited Arg1 induction (**Figure 3D**, IL-4 and LPS/IFNγ Second stimulus: IL-4) or Ym1 induction (**Figure 3E,** IL-4 and LPS/IFNγ Second stimulus: IL-4). Importantly the level of expression of Egr2 remained low/negative in M1 macrophages treated with IL-4 (**Figure 3F**, LPS/IFNγ , and LPS/IFNγ Second stimulus: IL-4). Conversely, M2 macrophages, which exhibited a high level of Egr2 expression (**Figure 3F**, IL-4), could switch and upregulated M1 markers NOS2, TNF, and IL-6 (**Figures 3A–C**, IL-4 Second stimulus: LPS/IFNγ ) and downregulated M2 markers Arg1, Ym1, and Egr2 (**Figures 3D–F**, IL-4 Second stimulus: LPS/IFNγ ). These data were consistent with our previous study where we investigated surface markers MHC class II, CD86, and CD206 (28). Thus, our data demonstrate that M1 macrophages retained a low level of Egr2, which was associated with retention of M1 markers and weak upregulation of M2 markers.

#### Knockdown of Egr2 Resulted in a Decrease in Expression of M2 Markers Arg1, Fizz1 and PPARγ and Upregulation of IL-10 in IL-4-Treated Macrophages

We hypothesized that Egr2 may be essential for M2 polarization. To test this we knocked down Egr2 using siRNA technology. We found that the siRNA treatment markedly decreased expression of Egr2 in unmanipulated and in IL-4-stimulated macrophages on mRNA (**Figure 4A**) and protein (**Figures S2**, **S3**) levels. We found that knockdown of Egr2 decreased the expression of Arg1 (**Figure 4B**), Fizz1 (**Figure 4C**), Ym1 (**Figure 4D**), and PPARγ (**Figure 4E**). The expression of IL-10 was significantly increased (**Figure 4F**). Thus, these results indicate that Egr2 promotes expression of the M2 markers Arg1, Fizz1, Ym1, and PPARγ and inhibits the expression of IL-10.

#### Knockdown of Egr2 Resulted in a Decrease in the Expression of M1 Markers NOS2, Cox-2, TNF, Il-1β in IFNγ-Treated Macrophages

We further examined the role of Egr2 in M1 polarization. Again, we found that siRNA for Egr2 substantially decreased expression of mRNA for Egr2 in IFNγ-stimulated M1 macrophages (**Figure 5A**). We found that knockdown of Egr2 significantly decreased expression of M1 markers NOS2 (**Figure 5B**), IL-1β (**Figure 5C**), TNF (**Figure 5D**) and the activation marker Cox-2(Ptgs2) (**Figure 5E**). Surprisingly, as in M2-like macrophages, Egr2 knockdown in M1-like macrophages (**Figure 4F**) resulted in an increase in the expression of IL-10 (**Figure 5F**). The expression of M1-associated microRNA miR-155 was also inhibited in IFNγstimulated macrophages (**Figure S4**), further demonstrating the importance of Erg2 in the induction of M1-associated regulatory RNAs. Thus, these results indicate that Egr2 promotes expression of RNA of M1-associated molecules NOS2, Cox-2, TNF, IL-1β, and miR-155.

#### Knockdown of Egr2 Resulted in a Decrease in the Expression of M1 Markers NOS2, TNF, IL-6, CD86, MHC Class II, and M2 Marker Arg1 on a Protein Level

We further validated the decrease in expression of Egr2, IL-6, and TNF in macrophages on a protein level using intracellular staining and quantitative FACS analysis. We found that IL-4-treated macrophages expressed a high level of Egr2, which was significantly decreased by Egr2 siRNA (**Figures 6A,B**). When compared to IL-4-treated cells, IFNγ-treated macrophages expressed a ∼10-fold lower level of Egr2, which was further decreased by Egr2 siRNA (**Figures 6A,B**). Knockdown of Egr2 in IFNγ-treated macrophages significantly decreased the production of M1 cytokines IL-6 and TNF on a protein level (**Figures 6C–F**).

Next, we validated downregulation of Arg1 in M2-like and NOS2 in M1-like macrophages on a protein level. We found that expression of Arg1 was not detected in IL-4 treated macrophages on a protein level on day 1, but appeared on day 4 post-IL-4 treatment (not shown). Knockdown of Egr2 significantly decreased the expression of Arg1 on a protein level on day 4 (**Figure 7A**). Expression of NOS2 was detected in M1 macrophages on day 1 post-IFNγ treatment and expression of this protein was significantly decreased by Egr2 siRNA (**Figure 7B**). NOS activity was also decreased in M1-like macrophages with an Egr2 knockdown, as was determined by actual NO production (**Figure 7C**).

Finally, we performed analysis of expression of classical macrophage activation marker MHC class II and M1-associated marker CD86 and found that both

FIGURE 3 | Analysis of expression of M1 markers (NOS, IL-6, and TNF), M2 markers (Arg1 and Ym1), and Egr2 in M1 macrophages that were subsequently stimulated toward M2 and vice versa. Bone-marrow-derived macrophages (BMDMs) were treated with M2 stimulus IL-4 or M1 stimulus IFNγ/LPS for 24 h, washed and then subjected to opposite M1 (IL-4) and M2 (IFNγ/LPS) stimuli for another 24 h-period as described in *Materials and Methods*. The cells were analyzed as unstimulated (C), or after 24 h of incubation with IL-4 only (IL4), or after incubation with IL-4 for 24 h, and then with IFNγ/LPS for 24 h (IL4 Second stimulus IFN/LPS), or after 24 h of incubation with IFNγ/LPS only (IFNγ/LPS), or after incubation with IFNγ/LPS for 24 h, and then with IL-4 for 24 h (IFN/LPS Second stimulus IL4). For analysis, the cells were washed, mRNA was isolated and the expressions of *NOS2* (A), *TNF* (B), *IL-6* (C), *Arg1* (D), *Ym1* (E), and *Egr2* (F) were analyzed by real-time RT PCR as described in *Materials and Methods.* In (A–F), mean ± S.E. of 4–6 separate culture plate wells is shown [\*\*\**p* < 0.001; \*\*\*\**p* < 0.0001 when compared to unstimulated (C) cells].

markers were downregulated by siRNA for Egr2 in IFNγtreated M1 macrophages (**Figure 8A**). Quantification is shown in **Table 2**. At the same time, siRNA for Egr2 did not significantly affect the expression of these markers in M2-like macrophages (**Table 2**). Thus, we found that although Egr2 is upregulated in M2 macrophages, expression of Egr2 is also important for M1 polarization.

# Knockdown of Egr2 Resulted in a Decrease in the Expression of MHC Class II and CD86 on Macrophages During Thioglycolate-Induced Inflammation of the Peritoneum

We then tested whether Egr2 is important during inflammation in vivo. To test this, we used adoptively transferred bone-marrow

< 0.0001; \*\*\*\*\**p* < 0.00001).

derived macrophages (BMDMs) that expressed genetic markers DsRed under the actin promoter and were transfected with siRNA for Egr2 or Control siRNA. On day 4 of thioglycollateinduced peritonitis, adoptively transferred DsRed+F4/80<sup>+</sup> macrophages transfected with control siRNA upregulated MHC class II and CD86, while macrophages with siRNA for Egr2 had a significantly lower level of expression of these molecules (**Figure 8B**). Quantification is shown in **Table 2.** Thus, we found that expression of Egr2 was important for macrophages activation in vivo during inflammation.

### Overexpression of Egr2 in IFNγ- and/or LPS-Treated Macrophages Resulted in Upregulation of M1 Markers TNF, NOS2, IL-6, IL-1β, and Cox-2

To confirm that expression of Egr2 is important for upregulation of M1-associated markers we overexpressed Egr2 in mouse macrophage cell line RAW264.7 as BMDMs are resistant to transfection with plasmids. We found that overexpression of Erg2 enhanced expression of NOS2 (**Figure 9A**), IL-1β

FIGURE 5 | Analysis of expression of Egr2 and M1-associated markers (NOS2, TNF, IL-1β, Cox-2) and IL-10 in IFNγ-activated macrophages with knockdown of Egr2. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which the cells were used as unstimulated [C(si) and Egr2(si)] or activated with IFNγ for another 24 h-time period [C(si)+IFN and Egr2(si)+IFN] as in Figure 3. The cells were washed, mRNA was isolated and the expressions of *Egr2* (A), *NOS2* (B), *TNF* (C), *IL-1*β (D), *Cox-2* (E) and *IL-10* (F) were analyzed by real-time PCR as described in *Materials and Methods* similar as for Figure 2. In (A–F), mean ± S.E. of six separate culture plate wells is shown (\*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\*\**p* < 0.00001).

(**Figure 9B**), IL-6 (**Figure 9C**) TNF (**Figure 9D**), and Cox-2 (**Figure 9E**) in the macrophage cell line stimulated with IFNγ. Similarly, overexpression of Erg2 enhanced expression of NOS2 (**Figure 10A**), IL-1β (**Figure 10B**), IL-6 (**Figure 10C**), and Cox-2 (**Figure 10E**) in the RAW264.7 cell line stimulated with LPS. However, the expression of TNF was downregulated in LPS-stimulated RAW264.7 cells with overexpression of Egr2 (**Figure 10D**). Thus, we further confirmed that although Egr2 was upregulated in M2 macrophages, expression of this protein is also important for the polarization of M1-like macrophages and regulation of expression of M1 markers.

#### Overexpression of Egr2 in IFNγ-Treated Macrophages Resulted in Upregulation of Egr2, IL-6, and TNF on a Protein Level

We validated that transfection of RAW264.7 cells with Egr2 plasmid resulted in upregulation of Egr2 on a protein level

siRNA (solid line) vs. Control siRNA (dotted line) are shown on representative histogram graphs (A,C,D). Staining with isotype-matched control mAbs (Control ISO) is shown by shaded histograms. Quantitative analysis (mean fluorescence intensity level for Egr2 and percentage of positive cells for IL-6 and TNF) is shown in (B,D,F). In (B,D,E), mean ± S.E. of six separate culture plate wells is shown (\*\**p* < 0.01; \*\*\**p* < 0.001).

using multi-color flow cytometry. We found that expression of Egr2 was significantly increased in F4/80+GFP<sup>+</sup> transfected cells treated with IL-4 or IFNγ (**Figures 11A,B**). Moreover, the expressions of M1 markers IL-6 (**Figures 11C,D**) and TNF (**Figures 11E,F**) were also increased on a protein level in F4/80+GFP<sup>+</sup> transfected cells treated with IFNγ. Thus, we confirmed that expression of Egr2 resulted in upregulation of M1 markers on a protein level.

macrophages transfected with Egr2 siRNA vs. Control siRNA (C). In (A–C), mean ± S.E. of quadruplicate is shown. Statistically significant differences in the expression levels with are shown on the figures (\*\*\**p* < 0.001; \*\*\*\**p* < 0.001).

#### Knockdown of Egr2 Did Not Downregulate Expressions of SOCS1, SOCS2, and SOCS3 in IFNγ-Stimulated Macrophages

It has been previously reported that Egr2 directly regulates the expression of SOCS1 and/or SOCS3, but not SOCS2 proteins, in lymphoid and dendritic cells, which are important regulators of pro-inflammatory (IL-6, TNF) and anti-inflammatory (e.g., IL-10) cytokines (20, 21). Here, we verified whether Egr2 regulates the expression of SOCS1, SOCS2, and SOCS3 in activated macrophages. First, we investigated the kinetics of expression of SOCS1, SOCS2, and SOCS3 in M2 and M1 macrophages stimulated with IL-4 or IFNγ, respectively. Similar to the pattern of expression of Egr1, Egr2, and Egr3 proteins, SOCS1 was upregulated by both IL-4, and, to a higher degree, by IFNγ (**Figure S5A**). SOCS2 was upregulated by IL-4 (**Figure S5B**), and SOCS3 was upregulated by IFNγ, but not by IL-4 (**Figure S5C**).

thioglycollate-induced inflammation as described in *Materials and Methods.* (A) After *in vitro* incubation in media, IL-4 or IFNγ, the cells were washed, stained for surface markers F4/80, MHC class II and CD86 and F4/80<sup>+</sup> gated macrophages were analyzed for the expression of MHC class II and CD86 by three-color follow cytometry as described in *Materials and Methods*. The expressions for MHC class II (top histograms) and CD86 (bottom histograms) of untreated (left histograms) or activated with IL-4 (middle histograms) or IFNγ (right histograms) macrophages transfected with Egr2 siRNA (solid line) vs. control siRNA (dotted line) are shown on representative histogram graphs. Staining with isotype-matched control mAbs is shown by shaded histograms. (B) The transfected DsRed-positive macrophages were injected i.p. into a group of 4-5 mice and peritoneal inflammation was induced by injection of thioglycollate medium as described in *Materials and Methods*. On day 4 after induction of inflammation, the cells were isolated by peritoneal lavage and cells were washed, stained for surface markers F4/80, MHC class II and CD86. F4/80+DsRed<sup>+</sup> gated macrophages were analyzed for the expression of MHC class II and CD86 by four-color follow cytometry as described in *Materials and Methods.* The expressions for MHC class II (left histograms) and CD86 (right histograms) on F4/80+DsRed<sup>+</sup> macrophages transfected with Egr2 siRNA (solid line) vs. control siRNA (dotted line) are shown on representative histogram graphs. Staining with isotype-matched control mAbs is shown by shaded histograms. (C) In (A,B), quantifications and statistics are shown in Table 2.

However, in contrast to expected results that Egr2 positively regulate SOCSs proteins, knockdown of Egr2 in IFNγ-treated BMDMs resulted in upregulation (not downregulation) of SOCS1, SOCS2, and SOCS3 (**Figures S6A–C**). Thus, we found that SOCS1, SOCS2, and SOCS3 were not positively regulated by Egr2 in M1-like macrophages.

TABLE 2 | Effect of Egr2 knockdown on expression of macrophage activation markers MHC class II and CD86 during inflammatory conditions<sup>a</sup> .


*<sup>a</sup>BMDMs were grown from bone marrow of B6 or DsRed transgenic mice in the presence of M-CSF for 5 days and transfected with Control siRNA or Egr2 siRNA. The cells were analyzed in vitro or injected i.p. into the group of 4–5 mice with thioglycollate-induced inflammation. For in vitro analysis, the cells were incubated in media alone (unstimulated) or in the presence of IL-4 (50 ng/ml) or IFN*γ *(100 ng/ml) as described in Materials and Methods. After 24 of incubation in vitro or on day 4 after induction of thioglycollate-elicited inflammation, the cells were stained for MHC class II, CD86, and F4/80. The F4/80*<sup>+</sup> *gated (in vitro) or F4/80*+*DsRed*<sup>+</sup> *(ex-vivo isolated) gated cells were analyzed for expression of MHC class II and CD86 by 3–4-color flow cytometry and mean fluorescence intensity (MFI) levels for expression of MHC class II and CD86 were measured. Mean* ± *S.E. of three separate experiments or 4–5 individual animals is shown.*

*<sup>b</sup>P* < *0.001 when compared to control siRNA.*

*<sup>c</sup>P* < *0.01 when compared to control siRNA.*

*<sup>d</sup>P* < *0.01 when compared to control siRNA.*

#### Knockdown of Egr2 Downregulate Expressions of CEBPβ in M0 and M2 Macrophages, While Overexpression of Egr2 in M1 Macrophages Upregulate CEBPβ

We further investigated mechanisms by which Egr2 is involved in the upregulated expression of M1 and M2 markers in macrophages. We found that SOCSs molecules are not the target for Egr2 in macrophages (**Figures S5**, **S6**). Therefore, we investigated other possible direct targets for Egr2. It has been demonstrated that the Egr family proteins Egr1, Egr2, and Egr3 directly upregulate CEBPβ, one of the major transcription factors in macrophages (37). Many studies have previously shown that CEBPβ is important for macrophage activation and for the induction of expression of both M1 and M2 markers (12– 14). Here we investigated whether Egr2 affected the expression of CEBPβ. We found that siRNA for Egr2 downregulated this transcription factor in unstimulated (M0), IL-4 treated (M2), and IFNγ-treated macrophages (**Figure 12A**). When we upregulated Egr2 in an IFNγ-stimulated M1 macrophage line that exhibited a low level of baseline Egr2 expression, we found that CEBPβ was upregulated (**Figure 12B**). Thus, we found that Egr2 positively regulates the expression of CEBPβ in unstimulated and in M1- and M2-stimulated macrophages.

#### CEBPβ Downregulates the Expression of Egr2 in M1-Like Macrophages

It was previously reported that CEBPβ could negatively regulate the expression of Egr2 (38). We proposed that similar mechanism play a role in M1-like macrophages where Egr2 is significantly downregulated. To test this hypothesis, we investigated whether the expression of Egr2 would be upregulated in unstimulated M0 or stimulated M1 and M2 macrophages where CEBPβ knockdown by siRNA treatment. Expression of Egr2 was upregulated in M0, M2 (IL-4), and M1 (IFNγ) macrophages treated with siRNA for CEBPβ [**Figure 12C**, Cebpb(si)]. When siRNA for CEBPα (as an additional negative control) was employed, Egr2 expression was unaffected [**Figure 12C**, Cebpa(si)]. We also found that siRNA for CEBPβ resulted in a decrease in M1 (NOS2, TNF) and M2 (Arg1, Ym1) markers in IFNγ- or IL-4- stimulated macrophages (not shown), which confirms previous reports. Thus, we found that CEBPβ, but not CEBPα, inhibits the expression of Egr2.

Next, we investigated the level of expression of CEBPβ in unstimulated or M2 (IL-4) or M1 (IFNγ/LPS)-stimulated macrophages. CEBPβ was upregulated 1.7- and 1.4-fold in M2 macrophages at 8 and 24 h. In M1 macrophages, upregulation of CEBPβ was quite marked (5-fold) at 8 h, but only transient returning to control level at 24 h (**Figure 12D**). When we compared the expression of CEBPβ at a protein level, we found that this transcription factor was not detectable in untreated M0 cells. However, CEBPβ was detected in M1- and M2-like macrophages after 24 h of stimulation with IFNγ/LPS or IL-4 (**Figure 12E**, **Figure S7**). CEBPβ was expressed at a much higher level in M1-like macrophages compared to M2-like macrophages (**Figure 12E**, **Figure S7**), which was correlated with the low level of Egr2 in M1-like macrophages or macrophages with CEBPβ knockdown (**Figure 12C**). Thus, these data suggest that CEBPβ negatively regulate Egr2 in M1-like macrophages.

#### DISCUSSION

In this study, we demonstrated an important role of Egrfamily proteins in macrophage activation driven by M1- or M2-like stimuli. We found that Egr2 is expressed in nonactivated M0 macrophages, upregulated in M2 macrophages, and significantly downregulated in M1 macrophages where Egr2 expression remained at a low level for an extended period. We also discovered that Egr2 expression was important for the upregulation of both M1 and M2 markers. Our data also indicate that Egr2 expression is important for M0 or M2 macrophages to make the transition to M1, while M1 macrophages exhibited a low level of Egr2 24 h post-stimulation, and respond poorly to M1**-** or M2-like stimuli, and exhibit an Egr2low deactivated phenotype, which is characterized by upregulation of IL-10 (**Figure S8**).

Mechanistically we found that Egr2 was important for upregulation of CEBPβ in M2 and to less extent of degree M1

macrophages, which in turn, inhibited Egr2 expression in M1 macrophages. These data are in good agreement with the fact that Egr2 is significantly downregulated in M1 but not in M2 cells (**Figures 2C,D**). Moreover, the peak of expression of Egr2 was at 8 h in M2 macrophages, and after that, the expression of Egr2 was slightly declined at 24 h (**Figure 1D**), which was consistent with the highest level of CEBPβ at 8 h and a slight decline at 24 h (**Figure 12D**). Factor Egr1 could also contribute to the upregulation of CEBPβ in M2 macrophages at 24 h since Egr1 reached the highest level at that time (**Figure 1A**). In case of M1 cells, CEBPβ could be transiently induced by Egr1 and Egr3 both of which are expressed at the highest level in

IFNγ-stimulated macrophages at 8 h and decline by 24 h poststimulation (**Figures 2C,E**). Egr2 is also still present in M1 macrophages at 8 h contributing to a high level of mRNA for CEBPβ at that time (**Figure 2D**). Base on all these data we proposed the model that Egr2 positively regulate CEBPβ that, in turn, promote expression of M1 and M2 markers, while CEBPβ provides a negative regulatory feedback loop downregulating Egr2 (**Figure 12F**). Thus, taken together our study demonstrate that high level of CEBPβ negatively regulated Egr2 in M1 macrophages.

Although distinct pathways of macrophages activation are recognized, little is known about the regulation of such activation on a molecular and transcriptional level (2). In this study, we demonstrated that Egr2 is involved in the process of activation, polarization, and plasticity of macrophages. Several studies suggest that M1 macrophages exhibit a low level of plasticity in vivo (28). Moreover, it was clearly demonstrated that high doses of LPS induce tolerance in macrophages when they decrease the production of proinflammatory cytokines IL-1β, IL-6, and TNF and upregulate

fluorescence intensity level for Egr2 and percentage of positive cells for IL-6 and TNF) is shown in (B,D,F). In (B,D,E), mean ± S.E. of six separate culture plate wells is shown. Statistically significant differences in the expression levels are shown on each figure (\**p* < 0.05; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001).

IL-10 (39). In our study, we found that Egr2low macrophages also downregulated these markers and up-regulated IL-10. Thus, Egr2 may play an important role in LPS tolerance and other types of tolerance in M1-polarized macrophages in vivo.

An important question is how Egr2 and CEBPβ mediate macrophages activation in mixed M1/M2 activation when IL-4 and IFNγ/LPS are present at the same time. Since M1 stimuli downregulate Egr2 relatively late (24 h post-activation) and M2 stimuli upregulate Egr2 as early as 3–8 h after stimulation (**Figure 1D**), we expect that Egr2 and CEBPβ would be first upregulated at 3–8 h and induce expression of M1 and M2 markers at these time-points. At 24 h CEBPβ would downregulate Egr2 in a way similar to that of M1 macrophages (**Figure 12**). Since both M1 and M2 stimuli upregulate CEBPβ (**Figures 12D,E**), we expect to see a high level of expression

FIGURE 12 | Analysis of reciprocal regulation of CEBPβ and Egr2 in M1 and M2 macrophages. (A) Effect of inhibition of Egr2 on CEBPβ expression in M0, M2, and M1 macrophages. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which the cells were used as unstimulated [C(si) and Egr2(si)] or activated with IL-4 [C(si) + IL4 and Egr2(si) + IL4] or IFNγ [C(si)+IFN and Egr2(si)+IFN] for another 24 h-time period as in Figure 3. The cells were washed, mRNA was isolated and the expressions of *CEBP*β was analyzed by real-time PCR as described in *Materials and Methods*. (B) Effect of overexpression of Egr2 on CEBPβ expression in M1 macrophages. Macrophage cell line RAW264.7 was transfected with pMIG expression vector plasmid or empty plasmid vector [V(C) or vector with Egr2 (V(Egr2)] for 24 h and after which the cells were used as unstimulated [V(C) and V(Egr2)] or activated with LPS for another 24 h-time period [V(C)+LPS and V(Egr2)+LPS] as in Figure 8. The cells were washed, mRNA was isolated and the expressions of *CEBP*β were analyzed by real-time PCR as described in *Materials and Methods.* (C) Effect of inhibition of CEBPβ on Egr2 expression in unstimulated (M0), M2, and M1 macrophages. BMDMs were transfected with siRNA cocktail for CEBPα [CEBPA(si)], CEBPβ [CEBPB(si)], or control siRNA [C(si)] for 24 h as described in *Materials and Methods*. After which, the cells were used as unstimulated [C(si), CEBPA(si), and CEBPB(si)] or activated with IL-4 [C(si)+IL4, CEBPA(si)+IL4, and CEBPB(si)+IL4] or IFNγ [C(si)+IFN, CEBPA(si)+IFN and CEBPB(si)+IFN] for another 24 h-time period as in Figure 3. The cells were washed, mRNA was isolated and the expressions of *Egr2* was analyzed by real-time PCR as described in *Materials and Methods*. (D) Kinetics of expression of *CEBP*β on mRNA level in M2 (IL-4) and M1 (IFNγ/LPS) macrophages. BMDMs were stimulated with IL-4 or IFNγ/LPS and the levels of *CEBP*β expression was *(Continued)*

FIGURE 12 | determined by real-time RT PCR immediately (0 h), 8, and 24 h after activation similarly as for Figure 2. Mean ± S.D. of triplicate is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\*\**p* < 0.0001 when compared to untreated (0 h) cells). (E) Expression of CEBPβ on protein level in M2 (IL-4) and M1 (IFNγ/LPS) macrophages. BMDMs were used unstimulated (Cont) or stimulated with IL-4, or IFNγ with LPS for 24 h as for Figure 3 and the level of expression of CEBPβ was analyzed by western blot as described in *Materials and Methods*. A representative blot (upper image) and quantitative analysis (bottom graph) are shown here. The whole blots are shown in Figure S7. (F) Model of reciprocal regulation of CEBPβ and Egr2 in M1 and M2 macrophages. In both M1 and M2 macrophages Egr1 upregulate CEBPβ, which in turn induce expression of M1 or M2 markers, respectively. In M1 macrophages, high level of CEBPβ inhibits Egr2 leading to non-responsiveness of these cells to further simulation. In (A–C,E), mean ± S.E. of 4–6 separate culture plate wells is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001; \*\*\*\*\**p* < 0.00001).

of CEBPβ at 24 h post-activation in M1/M2 mixed activation condition. High level of CEBPβ would downregulate Egr2 in M1/M2-activated macrophages. Thus, we assume that under mixed M1/M2 activation conditions, activated macrophages would express both M1 and M2 markers and would be resistant to further stimulation due to downregulation of Egr2**.**

Egr2 is involved in the expression of multiple genes in various cells types. In lymphoid cells, it was shown to upregulate expression of SOCS1 and SOCS3 (20). In dendritic cells, conditional knockout of Egr2 abolished expression of SOCS1 (22). In our studies, we did not find downregulation of SOCS1, SOCS2, and SOCS3 in macrophages with knockdown of Egr2. Thus, we believe that it is unlikely that Egr2 mediated this effect through SOCS molecules in activated macrophages. It was demonstrated that Egr2 directly regulated expression of the transcription factors CEBPβ (37). Here, CEBPβ was shown to be important for M1 and M2 polarization. In case of M2 polarization, it is suggested that CEBPβ act through downstream transcription factor PPARγ (2). Our study demonstrates that PPARγ and CEBPβ are regulated by Egr2. Therefore, CEBPβ-PPARγ axis is most likely the mechanism of action of Egr2 in macrophages M2 polarization. For M1 polarization, it was demonstrated that CEBPβ is upregulated in macrophages in response to IFNγ, IL-1, IL-6, TNF, and LPS and induce expression of M1 markers (40). CEBPβ is also required for induction of Th1/Th17-mediated autoimmune neuroinflammation associated with M1/M2 mixed macrophages activation (9, 41). Our study suggests that Egr2 upregulate expression of M1 and M2 markers through activation of CEBPβ and its downstream targets. On the other hand, CEBPβ inhibited Egr2. Since we found that CEBPβ protein is expressed in M1 macrophages at a very high level than in M2 macrophages, this transcription factor appears to inhibit Egr2 in the M1 macrophages causing long-term downregulation of Egr2 and loss of plasticity.

We believe that Egr2 is important for the induction of expression of both M1 and M2 markers (**Figures 4**,**5**) since it is expressed in M0 cells, upregulated early in M2 macrophages, and downregulated late in M1 macrophages. Particularly, our study demonstrated that Egr2 was upregulated 3–8 h post-activation in M2 conditions and downregulated only at 24 h post-activation in M1 conditions (**Figure 1D**). Therefore, we believe that Egr2 is important for both M1 and M2 conditions since Egr2 is expressed at a relatively high level at 3–8 h post-activation in both M1 and M2 macrophages (**Figure 1D**). We found that most of the M1 markers were upregulated early. For example, M1 marker NOS2 was upregulated by IFNγ as early as 3 h, peaked at 8 h, and then was slightly decreased by 24 h (**Figure 1B**). This is consistent with the expression of Egr2 at 3–8 h and downregulation at 24 h in M1 macrophages (**Figure 1D**).

We believe that Egr2 directly regulate the expression of CEBPβ by binding to its promoter and activating transcription. Our in silico analysis revealed that mouse Cebpb promoter area contains ten Egr-binding sites (**Figure S9A**). In support of this analysis, it was reported that Egr proteins directly induced CEBPβ expression (37). It was also shown that CEBPβ is important for expression of both M1 and M2 markers (12, 14, 34), while we found that CEBPβ is expressed at a very high level in M1 macrophages (**Figures 12D,E**). Thus, the knockdown of Egr2 in M1 macrophages result in low level of CEBPβ (**Figure 12A**) leading to low level of M1 markers (**Figure 5**).

Our study also demonstrated that CEBPβ inhibits Egr2 (**Figure 12C**). The mechanism for this phenomenon is not as straightforward as positive regulation of CEBPβ by Egr2. We believe that CEBPβ could act alone or together with other co-factors such as Nrf1 to inhibit Egr2 expression by binding to promoter area and repressing Egr2 transcription. It was demonstrated that CEBPβ and Nrf1 efficiently inhibited expression of DSPP gene during odontoblast differentiation by binding to the promoter sequence of this gene (42). In the case of the DSPP gene, both CEBPβ and Nrf1 had the ability to bind to the promoter and repress the transcription acting individually or synergistically by forming complexes with each other. We analyzed Egr2 promoter area and found three potential binding sites for CEBPβ and ten binding sites for Nrf1 (**Figure S9B**) suggesting the possible involvement of CEBPβ and Nrf1 in Egr2 downregulation. In support of our hypothesis, it was shown that Nrf1 is expressed in macrophages during inflammation in vivo and it is upregulated in LPStreated M1 macrophages in vitro at 24–36 h post-treatment (43–45). Moreover, it was shown that Nrf1 in complex with other co-factors (e.g., c-Jun, ATF2) positively regulated TNF expression, a known M1-associated cytokine (46). Thus, the negative transcriptional regulation of Egr2 is likely mediated by CEBPβ alone and/or with the help of other co-factors such as Nrf1.

Our study indicated an important role of Egr2 in macrophages activation in vivo during inflammation. The decrease in Egr2 expression inhibited upregulation of MHC class II and CD86, which is very important for antigen presentation and stimulation of CD4 T cells. Our in vitro studies also clearly demonstrated that Egr2 is critical for both M1 and M2 types of activation. Thus, in contrast to studies of the role of Egr2 in T/B cells and dendritic cells where Egr2 was found to be a negative regulator of activation via targeting SOCS molecules, Egr2 serve as a positive regulator of macrophage activation via targeting CEBPβ transcription factor but not SOCS molecules. This suggests the potential possibility to target Egr2 in macrophages to inhibit their activation during inflammatory diseases such as autoimmunity. Our study also suggests an importance of having a high level of Egr2 during macrophages reprogramming.

Egr2 belongs to early-immediate response genes, which are induced by multiple stimuli including growth factors, cytokines, hypoxia, and cell stress. The most studied gene from Egr family is Egr1, while the function of Egr2 is less studied (47, 48). There are several conditions such as fibrotic process and hypoxia that affect in vivo (49, 50). Therefore it is important to know the level of Egr2 expression in macrophages in such conditions. Indeed, it was demonstrated that Egr2 is upregulated by TGF-β1 during fibrosis, suggesting likely involvement of Egr2 in M2 polarization during the fibrotic process and/or scar formation (51). Hypoxia was shown to upregulate Egr1 and Egr3 and downregulate Egr2 by 45% in unstimulated human monocytes (52). At the same time, hypoxia upregulated both M1 (TNF, IL-6, CD86) and M2 (Arg1) markers in unstimulated human monocytes (52). These data indicate that for unstimulated macrophages hypoxia condition could be viewed as M1/M2 mixed activation stimulus leading to upregulation of M1 and M2 markers and downregulation of Egr2. Another study demonstrated that hypoxia decreased expression of M1 (CD80, CD86, MHC class II, TNF) and M2 (CD206, TREM) markers in LPS- and IL-4- treated human macrophages, respectively (53). This is in line with our data showing that knockdown of Egr2 decreased expression of both M1 and M2 markers (**Figures 4**, **5**).

In our study, we used in vitro knockdown/overexpression strategy combined with the adoptive transfer of macrophages during the development of inflammation in vivo. Usage of Egr2 knockout mice appeared to be problematic for direct investigation of the role of Egr2 in macrophage activation in vivo since Egr2 knockout mice die within a 2-week period after birth (24). It is possible to use irradiation chimeras by doing transplantation of newborn liver hemopoietic stem cells from donor Egr2-deficient mice into lethally irradiated wildtype recipient mice to overcome this problem as was reported earlier (19). However, even in this case in vivo experiments investigating macrophage activation during inflammation would be problematic to interpret since Egr2 is also involved in activation and/or development of T/B cells and dendritic cells (20). There is a theoretical possibility to make a conditional knockout of Egr2 only in macrophages using macrophage-specific myeloid-Cre strains crossed with Egr2-loxp animals. However, all current "macrophage-specific" myeloid-Cre systems (CD11b-Cre, F4/80-Cre, LysM-Cre etc.) target also dendritic cells, granulocytes and only a subpopulation of monocytes/macrophages (54). Therefore these systems would be also very problematic to use for in vivo experiment with thioglycollate-elicited peritonitis without doing in vitro experiments along with adoptive transfer, which we did in the current study. Thus, the direct modulation of Egr2 expression in macrophages in vivo is still very challenging, which pose limitations for macrophage reprogramming during pathological conditions in mouse models.

Despite methodological problems, reprogramming of macrophages from M2 to M1 and from M1 to M2 is an important therapeutic ambition in many pathologies. For anti-cancer therapy, it is considered important to reprogram M2 tumor-infiltrating macrophages to M1 (55). Conversely, for many types of inflammatory and autoimmune diseases such as autoimmune neuroinflammation, it is important to reprogram M1 macrophages toward M2 (9). Our study indicates that the second task is likely to be more problematic as it will require the overexpression of Egr2 in addition to stimulation with M2 stimuli to enable the transition of M1 cells to M2. This strategy may be important for future gene therapies of treatment of inflammatory diseases associated with high level of M1 activation such as multiples sclerosis, rheumatoid arthritis, atherosclerosis, sepsis, and others.

# AUTHOR CONTRIBUTIONS

TV and EP conceived the study, planned experiments. TV, AY, and EP conducted experiments. TV, TS, DA, and EP analyzed the data and prepared manuscript.

# FUNDING

The work was supported by the Research Grant Council-Early Career Scheme grant from the Government of Hong Kong, reference no. 24100314 (TV, AY, and EP) and by "5–100" Russian Research Excellence program (TS and DA).

# ACKNOWLEDGMENTS

We thank Marina Dukhinova and Ekaterina Kopeikina (School of Biomedical Sciences, The Chinese University of Hong Kong) for technical assistance with animals.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02515/full#supplementary-material

Figure S1 | Analysis of expression of Egr2 in IL-4- and LPS-treated macrophages on a protein level. Bone-marrow-derived macrophages (BMDMs) were treated with IL-4 for 24 h (A,B) or LPS (C) for 72 h as described in *Materials and Methods* and the cells were analyzed as untreated (Control) or after 24 h, 48 h, and 72 h of incubation with LPS (100 ng/ml). For analysis, the cells were washed, and the expression of Egr2 was analyzed on a protein level by western blot as described in *Material and Methods*. A representative images are shown in (A,C). Quantitative analysis of relative expression levels of Egr2 normalized to β-Actin is shown in (B). Mean ± S.E. of three experiments is shown (∗∗*p* < 0.01).

Figure S2 | Analysis of expression of Egr2 protein in unstimulated and IL-4-treated macrophages with knockdown of Egr2. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which the cells were used as unstimulated [C(si) and Egr2(si)] or activated with IL-4 for another 24 h-time period [C(si)+IL4 and Egr2(si)+IL4] as in Figure 3. The expression of Egr2 was analyzed by western blot as described in *Materials and Methods*. Quantitative analysis of relative expression levels of Egr2 normalized to β-Actin is shown. Representative images are shown in Figure S3. Mean ± S.E. of three separate experiments is shown (∗∗*p* < 0.01).

Figure S3 | Analysis of expression of Egr2 in unstimulated and IL-4-treated macrophages with knockdown of Egr2. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which the cells were used as unstimulated [C(si) and Egr2(si)] or activated with IL-4 for another 24 h-time period [C(si)+IL4 and Egr2(si)+IL4] as in Figure 3. The expression of Egr2 was analyzed by western blot as described in *Materials and Methods*. A representative image of whole blots for Egr2 and β-Actin are shown with marked relevant samples.

Figure S4 | Analysis of expression of miR-155 in IFNγ-treated macrophages with knockdown of Egr2. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which the cells were used as unstimulated [C(si) and Egr2(si)] or activated with IFNγ for another 24 h-time period [C(si)+IFN and Egr2(si)+IFN] as in Figure 3. The cells were washed, mRNA was isolated and the expression of miR-155 was analyzed by real-time PCR as described in *Materials and Methods*. Mean ± S.E. of six separate culture plate wells is shown (∗∗∗∗*p* < 0.0001).

Figure S5 | Kinetics of expression of SOCS1, SOCS2, and SOCS in macrophages polarized toward M2 with IL-4 or M1 with IFNγ. Bone-marrow-derived macrophages (BMDMs) were analyzed untreated (Control) or treated with IL-4 (IL4) or IFNγ (IFN-γ) as described in *Materials and Methods* and the cells were analyzed after 3, 5, 8, and 24 h of incubation. For analysis, the cells were washed, mRNA was isolated and the expressions of *SOCS1* (A), *SOCS2* (B), and *SOCS3* (C) were analyzed by real-time RT PCR as described in *Materials and Methods.* Mean ± S.E. of six separate culture plate wells is shown.

Figure S6 | Effect of inhibition of Egr2 on the regulation of SOCS1, SOCS2 and SOCS3 expressions in M2- and M-like macrophages. Bone-marrow-derived macrophages (BMDMs) were transfected with siRNA cocktail for Egr2 [Egr2(si)] or control siRNA [C(si)] for 24 h as described in *Materials and Methods*, and after which, the cells were used as unstimulated [C(si) and Egr2(si)] or activated with

#### REFERENCES


IL-4 [C(si)+IL4 and Egr2(si)+IL4] or IFNγ [C(si)+IFN and Egr2(si)+IFN] for another 24 h-time period as in Figure 3. The cells were washed, mRNA was isolated and the expressions of *SOCS1* (A), *SOCS2* (B), and *SCOS3* (C) were analyzed by real-time PCR as described in *Materials and Methods*. In (A–C), mean ± S.E. of six separate culture plate wells is shown (∗*p* < 0.05; ∗∗ , *p* < 0.01; ∗∗∗*p* < 0.001).

Figure S7 | Analysis of the expression of the CEBPβ protein in M2/M(IL-4) and M1/M(FNγ/LPS) macrophages. Bone-marrow-derived macrophages (BMDMs) were used as unstimulated (M0) or stimulated with IFNγ and LPS (M1), or IL-4 (M2) for 24 h as for Figure 12E and the level of expression of CEBPβ was analyzed by western blot as described in *Materials and Methods*. Representative whole blots for CEBPβ and β-Actin are shown.

Figure S8 | Model of regulation of macrophages plasticity by Egr2. Unstimulated M0 macrophages express Egr2 and do not express M1 and M2 markers such as NOS2 and Agr1 exhibiting Egr2+NOS2−Arg1<sup>−</sup> phenotype. After stimulation with IFNγ and/or LPS macrophages downregulate Egr2 and become M1 with Egr2−NOS2+Arg1<sup>−</sup> phenotype. Further stimulation of M1 macrophages with IL-4 or IFNγ/LPS result in deactivated (M-dea) Egr2−IL-10<sup>+</sup> phenotype with a low level of M1 and M2 markers. On the other hand, M0 macrophages stimulated with IL-4 toward M2 become Egr2+++NOS2−Arg1+. Further stimulation of M2 macrophages with IFNγ/LPS result in M1 phenotype with upregulation of M1 and downregulation of M2 markers.

Figure S9 | *In silico* analysis of Cebpb gene promoter area for the presence of Egr2-binding sites, and the analysis of Egr2 promoter area for the presence of CEBPβ- and Nrf1- binding sites. (A) Mapping of 3,000 bp promoter region upstream of mouse Cebpb gene (chromosome 2) using MULAN software (https:// mulan.dcode.org/). Egr-binding sites upstream of Cebpb gene are shown by red boxes. (B) Mapping of 3,000 bp promoter region upstream of mouse Egr2 gene (chromosome 10) using MULAN software (https://mulan.dcode.org/). CEBPβ- and Nrf1- binding sites upstream of Egr2 are shown by yellow and green boxes, respectively.


and the triggering receptor expressed on myeloid cells-1. Front Immunol. (2017) 8:1097. doi: 10.3389/fimmu.2017.01097


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

Copyright © 2018 Veremeyko, Yung, Anthony, Strekalova and Ponomarev. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) 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.

# Corrigendum: Early Growth Response Gene-2 Is Essential for M1 and M2 Macrophage Activation and Plasticity by Modulation of the Transcription Factor CEBPβ

Tatyana Veremeyko<sup>1</sup> , Amanda W. Y. Yung<sup>1</sup> , Daniel C. Anthony <sup>2</sup> , Tatyana Strekalova3,4,5 and Eugene D. Ponomarev 1,6 \*

*<sup>1</sup> Faculty of Medicine, School of Biomedical Sciences, The Chinese University of Hong Kong, Shatin, Hong Kong, <sup>2</sup> Department of Pharmacology, University of Oxford, Oxford, United Kingdom, <sup>3</sup> Department of Neuroscience, Maastricht University, Maastricht, Netherlands, <sup>4</sup> Institute of General Pathology and Pathophysiology, Moscow, Russia, <sup>5</sup> Laboratory of Psychiatric Neurobiology, Institute of Molecular Medicine and Department of Normal Physiology, Sechenov First Moscow State Medical University, Moscow, Russia, <sup>6</sup> Kunming Institute of Zoology-Chinese University of Hong Kong Joint Laboratory of Bioresources and Molecular Research of Common Diseases, Kunming, China*

#### Approved by:

*Frontiers in Immunology Editorial Office, Frontiers Media SA, Switzerland*

#### \*Correspondence:

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

#### Specialty section:

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

Received: *27 November 2018* Accepted: *28 November 2018* Published: *13 December 2018*

#### Citation:

*Veremeyko T, Yung AWY, Anthony DC, Strekalova T and Ponomarev ED (2018) Corrigendum: Early Growth Response Gene-2 Is Essential for M1 and M2 Macrophage Activation and Plasticity by Modulation of the Transcription Factor CEBP*β*. Front. Immunol. 9:2923. doi: 10.3389/fimmu.2018.02923* Keywords: CEBPβ, Egr2, M1/M2 balance, activation, inflammation, monocytes/macrophages, plasticity

#### **A Corrigendum on**

#### **Early Growth Response Gene-2 Is Essential for M1 and M2 Macrophage Activation and Plasticity by Modulation of the Transcription Factor CEBP**β

by Veremeyko, T., Yung, A. W. Y., Anthony, D. C., Strekalova, T., and Ponomarev, E. D. (2018). Front. Immunol. 9:2515. doi: 10.3389/fimmu.2018.02515

In the published article, there was an error regarding the affiliation for Tatyana Strekalova. As well as having affiliations 3 and 4, she should also have "Laboratory of Psychiatric Neurobiology, Institute of Molecular Medicine and Department of Normal Physiology, Sechenov First Moscow State Medical University, Moscow, Russia".

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2018 Veremeyko, Yung, Anthony, Strekalova and Ponomarev. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) 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.

Frontiers in Immunology | www.frontiersin.org 1 December 2018 | Volume 9 | Article 2923

# A CD300c-Fc Fusion Protein Inhibits T Cell Immunity

Cheng Cui 1,2, Min Su<sup>1</sup> , Yujun Lin<sup>1</sup> and Laijun Lai 1,3 \*

<sup>1</sup> Department of Allied Health Sciences, University of Connecticut, Storrs, CT, United States, <sup>2</sup> Department of Physiology, College of Basic Medical Science, China Medical University, Shenyang, China, <sup>3</sup> University of Connecticut Stem Cell Institute, University of Connecticut, Storrs, CT, United States

T cell responses are fine-tuned by co-stimulatory and co-inhibitory molecules. Among the T cell regulators, the B7 family members are of central importance. The recent success in targeting the B7 family molecules for the treatment of immune-related diseases has attracted intense interest in identifying additional B7-related molecules. In this study, we describe CD300c as a novel T cell co-inhibitory molecule that shares significant sequence homology with existing B7 family members. CD300c protein is expressed on professional antigen-presenting cells (APC), including B cells, monocytes, macrophages, and dendritic cells (DCs). The putative CD300c counter-receptor is expressed on CD4 and CD8 T cells, and the expression levels are upregulated upon activation. Soluble human and mouse CD300c-Fc fusion proteins significantly inhibit the proliferation, activation, and cytokine production by CD4 and CD8 T cells in vitro. Administration of CD300c-Fc protein attenuates graft-vs.-host disease (GVHD) in mice. Our results suggest that therapeutic interaction with the CD300c inhibitory pathway may represent a new strategy to modulate T cell-mediated immunity for the treatment of GVHD and autoimmune disease.

#### Edited by:

Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States

#### Reviewed by:

Christoph Wülfing, University of Bristol, United Kingdom Wei Yang, Memorial Sloan Kettering Cancer Center, United States

#### \*Correspondence:

Laijun Lai laijun.lai@uconn.edu

#### Specialty section:

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

Received: 11 August 2018 Accepted: 29 October 2018 Published: 15 November 2018

#### Citation:

Cui C, Su M, Lin Y and Lai L (2018) A CD300c-Fc Fusion Protein Inhibits T Cell Immunity. Front. Immunol. 9:2657. doi: 10.3389/fimmu.2018.02657 Keywords: T cell co-inhibitory molecule, B7 family, T cell proliferation, T cell activation, GVHD

# INTRODUCTION

T cell immune responses are tightly controlled by co-stimulatory and co-inhibitory molecules. The co-stimulatory molecules contribute to the development of immune responses against cancers and foreign pathogens, while the co-inhibitory molecules are critical for peripheral tolerance to avoid autoimmunity, GVHD and transplant rejection.

A number of T cell co-stimulatory and co-inhibitory ligands and receptors have been identified. Among them, B7 family members are of central importance. The B7 family ligands include B7-1 (1), B7-2 (2), PD-L1 (B7-H1) (3, 4), PD-L2 (B7-DC) (5, 6), inducible T cell co-stimulator ligand (ICOSL) (also known as, B7-H2, B7h, B7RP-1, GL50, LICOS) (7–10), B7-H3 (11), B7-H4 (B7x, B7S1) (12–14), B7-H5 (HHLA2) (15, 16), and B7-H6 (17), etc. The importance of these molecules has been highlighted by the FDA approval of several drugs for the treatment of cancer and autoimmune disease by targeting the ligands or their receptors. For example, recombinant fusion protein CTLA-4-Fc (Abatacept or Balatacept), an inhibitory receptor for B7-1 and B7-2, has been used in the treatment of rheumatoid arthritis and kidney transplantation rejection. In contrast, antibodies against PD-L1/PD-1 or CTLA-4, such as Pembrolizumab (Keytruda), Nivolumab (Opdivo), Atezolizumab (Tecentriq), Avelumab (Bavencio), Durvalumab (Imfinzi), Ipilimumab (Yervoy), can significantly enhance antitumor immunity and the survival of cancer patients.

**52**

The unprecedented success by targeting the ligands or receptors of the B7 family members for the treatment of cancer and autoimmune disease has attracted considerable interest in identifying additional T cell regulators. In this study, we identify CD300c as a novel T cell inhibitory molecule. Human CD300c (hCD300c), also known as CMRF-35A or CMRF-35, was originally isolated from human leukocytes (18–20). It has been reported that CD300c has effects on NK cells, monocytes, macrophages, B cells, DCs and basophils (21–30). CD300c can bind to plasma membrane lipids phosphatidylethanolamine (PE) and phosphatidylserine (PS) (23, 30, 31). However, the functional significance of CD300c on T cells has not been characterized.

We show here that hCD300c is homologous to known B7 family members in amino acid sequence. CD300c is expressed on APCs, and its putative counter-receptor is expressed on CD4 and CD8 T cells. Both soluble human and mouse CD300c-IgG2a Fc (CD300c-Ig) fusion proteins significantly inhibit the proliferation and activation, and cytokine production of CD4 and CD8 T cells in vitro. Administration of hCD300c-Ig protein ameliorates GVHD in mice. Therefore, this unique T cell inhibitory pathway may provide a new strategy to modulate T cell-mediated immunity to treat immune-related diseases.

#### MATERIALS AND METHODS

#### Bioinformatics Analysis of CD300c

Sequence alignments of CD300c and known B7 family members, and CD300c orthologous proteins were analyzed via the Clustal W program in MacVector 16.0.5 (MacVector, Inc.). The leader peptide, transmembrane, and Ig-like domain were predicted with SignalP 4.0 (http://www.cbs.dtu.dk/services/SignalP), TMHMM server version 2.0 (http://www.cbs.dtu.dk/services/TMHMM/), and InterPro (https://www.ebi.ac.uk/interpro).

#### Cloning and Purification of hCD300c and mCD300c2

The extracellular domains of hCD300c (aa29-183) and mCD300c2 (aa22-193) were cloned and fused into a pCMV6-AC-FC-S expression vector containing the constant region of mouse IgG2a (ORIGENE). The vectors were transfected into HEK293F cells. The fusion proteins were purified for supernatant using Protein G Sepharose 4 Fast Flow according to the manufacturer's instructions (GE Healthcare). Purified proteins were verified by SDS-PAGE, Coomassie Staining and Western blot. Protein was quantified using the PierceTM BCA Protein Assay Kit (Pierce, Rockford, IL). Control Ig (recombinant mouse IgG2a Fc protein) was purchased from BXCell (West Lebanon, NH).

### SDS-PAGE and Western Blot

Purified CD300c-Ig was loaded on a 12% SDS-PAGE, and stained with Coomassie blue or transferred to a polyvinylidene fluoride membrane. The protein containing membrane was incubated with HRP conjugated anti-mouse IgG2 antibody, or antihCD300c antibody (Novus Biologicals, Littleton, CO) followed by HRP conjugated second antibody, and then developed with Super Signal <sup>R</sup> West Pico chemiluminescent Substrate (Thermo Scientific).

#### Flow Cytometry Analysis

Single cell suspensions of organs were stained with the fluorochrome-conjugated antibodies protein as described (32– 35). For intracellular staining, the cells were first permeabilized with a BD Cytofix/Cytoperm solution for 20 min at 4◦C. Direct or indirect staining of fluorochrome-conjugated antibodies included: CD4, CD8, CD19, B220, CD11c, CD11b, F4/80, H2<sup>b</sup> , Annexin V, Ki67, CD44, CD62L, CD69, CTLA-4, CD28, PD-1, BTLA, and ICOS and mCD300c2 (BioLegend or BD Biosciences, San Jose, CA, San Diego, CA). mCD300c2-Ig and hCD300c-Ig were biotinylated with sulfo-NHS-LC-Biotin (Pierce). The samples were analyzed on a FACSCalibur or LSRFortessa X-20 Cell Analyzer (BD Biosciences). Data analysis was done using FlowJo software (Ashland, OR).

#### Limulus Amebocyte Lysate (LAL) Assay

The endotoxin level in the purified proteins was determined by the endpoint chromogenic LAL test according to the manufacturer's instructions (Lonza, Walkersville, MD) (36).

#### In vitro T Cell Proliferation Assays

Normal human peripheral blood CD3<sup>+</sup> Pan T Cells that were negatively isolated from mononuclear cells using an indirect immunomagnetic Pan-T labeling system were purchased from ALLCELLS, LLC (Alameda, CA). Murine CD3<sup>+</sup> T cells were purified from C57BL/6 mice by an immunomagnetic system (Miltenyi, Auburn, CA), and the purity of the cells was usually >95%. T cells were stimulated with anti-CD3 and/or anti-CD28 antibodies (Biolegend) in the presence of CD300c-Ig or control Ig. Proliferative response was assessed by pulsing the culture with 1 µCi of [3H] thymidine (PerkinElmer, Inc., Downers Grove, IL) 12 h before harvest. Incorporation of [3H] thymidine was measured by liquid scintillation spectroscopy (PerkinElmer, Inc.). For carboxyfluorescein diacetate succinimidyl ester (CFSE) assay, splenocytes were labeled with CFSE (ThermoFisher Scientific), and stimulated with anti-CD3 in the presence of CD300c-Ig or control Ig. The cells were analyzed by flow cytometry.

#### Mice

Four-week-old female C57BL/6 and BALB/c mice were purchased from Jackson Laboratory. The mice were used in accordance with a protocol approved by the Institutional Animal Care and Use Committee of the University of Connecticut.

#### GVHD Model

BALB/c recipients received 900 cGy total body irradiation from a 137Cs source (Gammator-50 Gamma Irradiator; Radiation

**Abbreviations:** APC, antigen-presenting cells; DCs, dendritic cells; GVHD, graftvs. -host disease; ICOSL, T cell co-stimulator ligand; Ig, immunoglobulin; CD300c-Ig, CD300c-IgG2a Fc; CFSE, carboxyfluorescein diacetate succinimidyl ester; hCD300c, human CD300c; mCD300c, mouse CD300c; CLM-6, CMRF-35-like molecule-6; LMIR2, leukocyte mono-Ig-like receptor 2, DIgR1, dendritic cellderived Ig-like receptor 1; MAIR-II, myeloid-associated Ig-like receptor II; MW, molecular weight; BM, bone marrow; BMT, BM transplantation; SI, small intestine.

Machinery Corporation, Parsippany, NJ). Two to four hours later, the mice were injected intravenously (i.v.) with BM and spleen cells from C57BL/6 mice. The recipients were injected i.p. with hCD300c-Ig, or control Ig. The severity of GVHD was evaluated with a clinical GVHD scoring system. In brief, GVHD recipients in coded cages were individually scored every week for five clinical parameters on a scale from 0 to 2: weight loss, posture, activity, fur texture and skin integrity. A clinical GVHD index was generated by summation of the five criteria scores (maximum index = 10).

GVHD target organs were harvested for histopathological analysis. The organs were formalin-preserved, paraffinembedded, sectioned and hematoxylin/eosin (H&E)-stained. Assessment of tissue damage was performed based on scoring systems previously described (37). Briefly, liver GVHD was scored on the number of involved tracts and severity of liver cell necrosis; the maximum score is 10. Gut GVHD was scored on the basis of crypt apoptosis and lamina propria inflammation; the maximum score is 8. Lung GVHD was scored on the periluminal infiltrates, pneumonitis, and the severity of lung tissues involved; the maximum score is 9.

#### Statistical Analysis

P-values were based on the two-sided Student's t-test. A confidence level above 95% (p < 0.05) was determined to be significant.

#### RESULTS

#### CD300c Shares Sequence and Structural Homology With the B7 Family Molecules

Recognizing the importance of the B7 family in controlling immune responses, we performed a series of genome-wide database searches to find molecules that are homologous to known B7 family members. We discovered that hCD300c shares varying levels of amino acid identity and similarity with B7-1 (17 and 13%), B7-H2 (16 and 12%), B7-H3 (13 and 12%), B7-H4 (12 and 15%), PD-L1 (14 and 19%), and PD-L2 (13 and 15%) (**Figure 1A**). It has been reported that human B7-1 shares 13– 21% of amino acid identity with other B7 family members (15). The levels of amino acid identity of hCD300c with the known B7 family members suggest that CD300c is a B7 family-related molecule.

It has been reported that the mouse orthologs of hCD300c are mouse CD300c (mCD300c) [also called CMRF-35-like molecule-6 (CLM-6)] and mCD300c2 [also known as leukocyte mono-Iglike receptor 2 (LMIR2), dendritic cell-derived Ig-like receptor 1 (DIgR1), myeloid-associated Ig-like receptor II (MAIR-II), or CLM-4] (24–28). hCD300c shares 51 and 48% identity, and 6 and 8% similarity with mCD300c and mCD300c2, respectively (**Figure 1B**). mCD300c and mCD300c2 also share 8–10% amino acid identity and 9–14% amino acid similarity with mouse B7-1, B7-H2, B7-H3, B7-H4, PD-L1, and PD-L2 (**Supplemental Figure 1**). hCD300c, mCD300c, and mCD300c2 belong to the immunoglobulin (Ig) superfamily and are type I transmembrane proteins that contain an extracellular region with

a single Ig-V like domain, a transmembrane segment, and a short cytoplasmic tail (**Figure 1B**) (18–22, 38, 39).

### hCD300c Inhibits the Proliferation and Activation of Mouse and Human T Cells in vitro

To investigate whether like the known B7 family members, CD300c protein can affect T cell function, we produced an hCD300c-Ig fusion protein by cloning the extracellular domain of the hCD300c gene into an expression vector containing the constant region of the mouse IgG2a. The expression vector was then transfected into human HEK-293 cells to produce hCD300c-Ig fusion protein that was then purified from the supernatant of the cells. A relatively high purity of hCD300c-Ig protein was obtained, as determined by Coomassie blue-stained SDS-PAGE (**Supplemental Figure 2**). The identity of the fusion protein was verified by Western blot using anti-IgG2a antibody or anti-hCD300c antibody (**Supplemental Figure 2**). The actual molecular weight (MW) of the hCD300c-Ig was higher than the predicted MW, suggesting that the recombinant protein was glycosylated. The endotoxin level was <0.01 EU/ml of 1 µg of purified protein.

We then determined whether hCD300c-Ig protein affected T cell proliferation. To do this, CD3<sup>+</sup> T cells were purified from splenocytes of C57BL/c mice, and cultured on plates precoated with anti-CD3 antibody in the presence of graded doses of hCD300-Ig (800, 1,600, and 3,200 ng/ml) for 3 days. Since the molecular weight of hCD300-Ig fusion protein is ∼1.5 fold higher than that of control Ig protein, we used equimolar amounts of the control Ig as a control. T cell proliferation was measured by [3H] thymidine incorporation. As shown in **Figure 2A**, hCD300c-Ig inhibited anti-CD3-activated T cell proliferation in a dose-responsive manner, with ∼32, 72, and 78% inhibition by 800, 1,600, and 3,200 ng/ml hCD300c-Ig, respectively, as compared to equimolar amounts of control Ig. We also determined whether hCD300c could inhibit anti-CD3 and anti-CD28 antibody-activated T cell proliferation. Similarly, hCD300c-Ig reduced anti-CD3 and anti-CD28-activated T cell proliferation in a dose-dependent manner, although to a lesser extent than that with anti-CD3 activation only (**Figure 2B**).

To confirm the effect on T cell proliferation and to determine whether hCD300c affects CD4 and/or CD8 T cells, we performed a carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution assay. Murine splenocytes were labeled with CFSE, and then cultured with anti-CD3 antibody in the presence of graded doses of hCD300c-Ig or control Ig. T cell proliferation was measured by CFSE fluorescent dilution in CD4 and CD8 T cells. As shown in **Figures 2C–F**, hCD300c-Ig inhibited anti-CD3 activated proliferation of both CD4 and CD8 T cells in a dosedependent manner.

We next determined whether hCD300c-Ig affects the activation of T cells in vitro. CD69 is an early activation marker. After splenocytes were cultured with anti-CD3 antibody and hCD300c-Ig or control Ig, the expression of the CD69 on CD4 and CD8 T cells was analyzed 24 h later. As shown in **Figures 2G–J**, hCD300c-Ig at the dose of 3,200 ng/ml


FIGURE 1 | CD300c is a B7 family-related molecule. (A) Alignment of hCD300c with some known B7 family members. Identical amino acids are shaded black. Amino acids with strong homologies are shaded in gray. Conserved cysteine residues are labeled with an asterisk (\*). (B) Alignment of hCD300c with mCD300c and mCD300c2. Predicted signal peptide, IgV-like, and transmembrane (TM) domains for hCD300 are marked.

FIGURE 2 | incorporation. (C–F) Mouse splenic cells were labeled with CFSE and cultured with anti-CD3 antibody (1µg/ml) and graded doses of hCD300c-Ig protein (1,600 and 3,200 ng/ml) or equimolar amounts of control Ig protein for 3 days. The cells were stained with anti-CD4 and CD8 antibodies, and analyzed for CFSE levels by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. (C,E) Representative flow cytometric profiles and (D,F) statistical analysis of CFSElo proliferating CD4<sup>+</sup> or CD8<sup>+</sup> T cells. (G–J) Mouse splenic cells were cultured with anti-CD3 antibody and graded doses of hCD300c-Ig protein or control Ig protein as (C) for 24 h. The cells were then analyzed for CD69 expression by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. (G,I) Representative flow cytometric profiles, and (H,J) statistical analysis of CD69<sup>+</sup> cells in CD4<sup>+</sup> or CD8<sup>+</sup> T cells. (K) Purified human CD3<sup>+</sup> T cells were cultured with plate-bound anti-human CD3 antibody (1µg/ml) in the presence of graded doses of hCD300c-Ig protein (1,500 and 3,000 ng/ml) or control Ig protein for 3 days. Cell proliferation was measured by [3H] thymidine incorporation. (L,M) Purified human CD3<sup>+</sup> T cells were cultured with plate-bound anti-human CD3 antibody (1µg/ml) in the presence of graded doses of hCD300c-Ig protein (1,600 and 3,200 ng/ml) or control Ig protein for 1 days. The cells were then analyzed for human CD69 expression by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. The data were pooled from 3 independent experiments and expressed as mean ± SD. \*P < 0.05 compared with control Ig.

significantly reduced the expression of CD69 on both CD4 and CD8 T cells. The results suggest that hCD300c also inhibits the activation of CD4 and CD8 T cells.

Having demonstrated that hCD300c-Ig inhibited murine T cells proliferation in vitro, we examined whether hCD300c-Ig affected human T cells. Purified human T cells were cultured with anti-CD3 antibody in the presence of graded doses of hCD300-Ig or control Ig, and T cell proliferation was measured by [3H] thymidine incorporation. Similarly, hCD300c-Ig markedly inhibited human T cell proliferation with ∼53 and 77% inhibition by 1,500, and 3,000 ng/ml hCD300c-Ig, respectively (**Figure 2K**). Furthermore, hCD300c-Ig at both 1,600 and 3,200 ng/ml doses significantly reduced the expression of CD69 on both human CD4 and CD8 T cells (**Figures 2L,M**).

Taken together, our results suggest that hCD300c-Ig inhibits TCR-mediated proliferation and/or activation of both mouse and human T cells in vitro. hCD300c has similar inhibitory effects in both human and mouse primary T cells, suggesting that its binding partner and its conferred function on T cells may be conserved across species.

#### mCD300c2 Inhibits the Proliferation and Activation of Mouse T Cells in vitro

We also produced a mCD300c2-Ig protein by fusing the extracellular domain of mCD300c2 to the mouse IgG2a constant region, and analyzed the effects of purified mCD300c2-Ig fusion protein on mouse T cell proliferation and activation in vitro. We found that mCD300c2-Ig markedly inhibited anti-CD3-induced T cell proliferation, with more than 90% inhibition by 800, 1,600, or 3,200 ng/ml of mCD300c2-Ig (**Figure 3A**). mCD300c2- Ig also inhibited anti-CD3 and anti-CD28 antibody-induced T cell proliferation, with ∼81, 85, and 94% inhibition by 800, 1,600, and 3,200 ng/ml hCD300c-Ig, respectively (**Figure 3B**). CFSE dilution assay confirmed that mCD300c2-Ig inhibited anti-CD3-induced proliferation of both CD4<sup>+</sup> and CD8<sup>+</sup> T cells (**Figures 3C–F**).

Like hCD300c-Ig, mCD300c2-Ig significantly reduced the expression of CD69 on both CD4<sup>+</sup> and CD8<sup>+</sup> T cells induced by either anti-CD3 antibody, or anti-CD3 plus anti-CD28 antibodies, and the reduction was also in dose-dependent manner (**Figures 3G–L**). To further confirm that mCD300c2-Ig inhibits T cell activation, we analyzed the expression of CD44 and CD62L by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. It has been reported that naïve T cells are CD44lowCD62Lhi, while effective memory T cells are CD44hiCD62Llow. As shown in **Figures 3M–Q**, mCD300c2- Ig significantly increased the percentages of CD44lowCD62Lhi naïve cells in anti-CD3-activated CD4<sup>+</sup> and CD8<sup>+</sup> T cells, but decreased the percentages of CD44hiCD62Llow effective memory T cells. The results further suggest that mCD300c2-Ig inhibits the activation of CD4<sup>+</sup> and CD8<sup>+</sup> T cells.

Collectively, our results suggest that both hCD300c-Ig and mCD300c2-Ig inhibits TCR-mediated proliferation and activation of both CD4 and CD8 T cellsin vitro, providing further evidence that CD300c is a novel B7 family-related molecule with T cell co-inhibitory properties.

#### mCD300c2 Inhibits Cytokine Production From T Cells

We then determined the effect of mCD300c2 on cytokine production from T cells in vitro. CD3<sup>+</sup> T cells were purified from the spleens of C57BL/6 mice and stimulated with anti-CD3 antibody in the presence of graded doses of mCD300c2-Ig or control Ig protein for 3 days. The contents of cytokines in the supernatants were measured by ELISA. As shown in **Figure 4**, mCD300c2-Ig inhibited the production of IFNγ, IL-2, IL-17A, and IL-10, but not TNFα. The results suggest that mCD300c2-Ig suppresses certain Th1/Th2/Th17 cytokine production by T cells induced by TCR signaling.

# The Expression Pattern of CD300c on Murine Immune Cells

We analyzed cell surface expression of mCD300c2 protein on murine immune cells by flow cytometry using a monoclonal antibody against mCD300c2 (clone TX52). Although the manufacture's manual indicates that this antibody recognizes mCD300c and mCD300d, this antibody was generated by using mouse MAIR-II (mCD300c2)-transfected cells as an immunogen. As indicated in the Discussion Section, the nomenclature for mCD300c is still confusing and MAIR-II is also known as mCD300c2/CD300d. Therefore, we believe that this antibody reacts to mCD300c2 protein. We found that resting splenic CD4<sup>+</sup> and CD8<sup>+</sup> T cells scarcely expressed mCD300c2 protein (**Figure 5**). After activation by anti-CD3 and anti-CD28 antibodies, a small percentage of activated CD4<sup>+</sup> and CD8<sup>+</sup> T cells expressed mCD300c2. We then examined the expression of mCD300c2 protein on other immune cells, and found that resting and activated B220<sup>+</sup> B cells, CD11b<sup>+</sup> monocytes and F4/80<sup>+</sup> macrophages expressed various levels of mCD300c2 (**Figure 5**). The expression level of mCD300c2 on macrophages

(Continued)

FIGURE 3 | incorporation. (C–F) Mouse splenic cells were labeled with CFSE and cultured with anti-CD3 antibody (1µg/ml) and graded doses of mCD300c2-Ig protein or control Ig protein as (A). The cells were then stained with anti-CD4 and CD8 antibodies and analyzed for CFSE levels by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. (C,E) Representative flow cytometric analysis of CFSE distribution of CD4<sup>+</sup> or CD8<sup>+</sup> T cells. (D,F) Statistical analysis of CFSElo proliferating CD4<sup>+</sup> or CD8<sup>+</sup> T cells. (G–Q) Mouse splenic cells were cultured with (G–Q) anti-CD3 antibody and (I,L) anti-CD28 antibody in the presence of graded doses of mCD300c2-Ig or control Ig protein as (A,B). The cells were analyzed for the expression of (G–L) CD69 24 h later, (M–Q) CD44, and CD62L 72 h later. (G,J,M) Representative flow cytometric profiles, and (H,I,K,L,N–Q) statistical analyses of the percentages of CD69+, CD44lowCD62Lhi naïve, and CD44hiCD62Llow effective memory CD4 and CD8 T cells. The data were pooled from 3 independent experiments and expressed as mean ± SD. \*P < 0.05 compared with control Ig.

was upregulated upon activation. In addition, although CD300c protein was scarcely expressed on resting CD11c<sup>+</sup> DCs, it was induced upon activation by LPS (**Figure 5**). These results suggest that CD300c is expressed on a variety of APCs.

### The Expression of the Putative CD300c Counter-Receptor

To determine the expression pattern of the CD300c counterreceptor, purified mCD300c2-Ig and control Ig proteins were biotinylated. Splenocytes from C57BL/c mice were stained with the biotinylated proteins, followed by streptavidin-PE. Flow cytometric analysis showed that mCD300c2-Ig bound to resting CD4<sup>+</sup> and CD8<sup>+</sup> T cells, and the binding was increased when CD4<sup>+</sup> and CD8<sup>+</sup> T cells were activated by anti-CD3 and anti-CD28 antibodies (**Figures 6A,B**).

We also analyzed the expression of the CD300c counterreceptor on other immune cells. We found that mCD300c2-Ig bound to both resting and activated B220<sup>+</sup> B cells, CD11c<sup>+</sup> DCs, CD11b<sup>+</sup> monocytes and G4/80<sup>+</sup> macrophages (**Figures 6A,B**). The expression levels of the putative mCD300c counter-receptor on these immune cells were not significantly changed upon activation by LPS.

To determine whether mCD300c binds to molecules previously identified as receptors of the known B7 family members, HEK-293 cells were transfected with an expression vector containing the mouse CD28, CTLA-4, PD-1, BTLA, or ICOS gene. The expression of these receptors on the transfected 293 cells was confirmed by flow cytometric analysis with the antibodies against the respective receptors (**Figure 6C**). The binding of mCD300c to the transfected HEK-293 cells was then analyzed. As shown in **Figure 6D**, mCD300c2 did not bind to the CD28, CTLA-4, PD-1, BTLA, or ICOS transfected cells.

Taken together, our results suggest that the mCD300c2 counter-receptor is expressed on resting and activated CD4 and CD8 T cells, B cells, DCs, monocytes, and macrophages. The expression levels of the receptor on activated CD4 and CD8 T cells is upregulated. The mCD300c counter-receptor seems to be distinct from CD28, PD-1, CTLA-4, PD-1, or BTLA.

### hCD300c-Ig Protein Ameliorates GVHD in Mice

Although bone marrow (BM) transplantation (BMT) has been widely used in the treatment of many diseases, GVHD remains a major complication after allogeneic BMT. Acute GVDH is primarily caused by T cells in donor transplants attacking recipient's tissues. We used a well-defined MHC-mismatched [C57BL/6 (H2<sup>b</sup> ) → BALB/c (H2<sup>d</sup> )] GVHD mouse model to validate the effect of CD300c on T cells in vivo. BALB/c mice were lethally irradiated and injected i.v. with BM and splenic cells from allogeneic C57BL/6 mice. The recipients were then injected i.p. with hCD300c-Ig, or control Ig. The development of GVHD was monitored over time. As shown in **Figures 7A–C**, control

compared with isotype Ab. \*\*P < 0.05 compared with resting cells.

Ig-treated GVHD recipients revealed gradual body weight loss and all succumbed by day 35 after BMT. hCD300c-Ig treatment significantly reduced the mortality and morbidity of GVHD, with 45% of the mice still surviving at day 40 post-transplantation (**Figures 7A–C**). GVHD severity was confirmed by pathologic analysis, showing that pathology scores of the liver, small intestine (SI), and lung in hCD300c-Ig-treated recipients were significantly lower than those in control Ig-treated recipients (**Figures 7D,E**). These data suggest that hCD300c-Ig treatment attenuates GVHD in vivo.

We then analyzed T cell proliferation, survival, and activation in hCD300c-Ig- or control Ig-treated GVHD mice. Lethally irradiated BALB/c recipients were injected i.v. with BM and splenic cells from C57BL/6 mice. The mice were injected i.v. on day 0 and i.p. on day 2 with 20 µg hCD300c-Ig or control Ig protein. The recipients were euthanized and the spleens were harvested on day 4. We analyzed for the expression of Ki67, a cell marker of proliferation. As shown in **Figures 7F–I**, the percentages of Ki67<sup>+</sup> cells in donor CD4<sup>+</sup> and CD8<sup>+</sup> T cells of hCD300c-Ig-treated recipients were significantly lower than those in control Ig-treated mice. We also analyzed the survival of donor CD4<sup>+</sup> and CD8<sup>+</sup> T cells, and found that the percentages of annexin V<sup>+</sup> 7-ADD<sup>−</sup> apoptotic CD4<sup>+</sup> or CD8<sup>+</sup> T cells were not significantly different between hCD300c-Ig- and control Ig-treated groups (data not shown). We next examined the expression of activation markers by CD4<sup>+</sup> and CD8<sup>+</sup> T cells. Although the percentages of CD69<sup>+</sup> cells in donor CD4<sup>+</sup> and CD8<sup>+</sup> T cells were not significantly different between hCD300c-Ig- and control Ig-treated groups (data not shown), the percentages of CD44hi cells and CD62Llo cells in donor CD4<sup>+</sup> and CD8<sup>+</sup> T cells were significantly reduced in hCD300c-Igtreated GVHD mice (**Figures 7J–M**).

FIGURE 6 | mCD300c2-Ig vs. control Ig. The data were pooled from 3 independent experiments. (A,B) \*P < 0.05 compared with control Ig. \*\*P < 0.05 compared with resting cells. (C,D) HEK-293 cells were transfected with an expression vector containing the mouse CD28, CTLA-4, PD-1, BTLA, or ICOS gene. The transfected cells were stained with (C) antibodies against the respective CD28, CTLA-4, PD-1, BTLA, or ICOS protein (open histograms) or isotype Ab (shaded histograms), or (D) biotinylated mCD300c2-Ig (open histograms) or control Ig protein (shaded histograms), and analyzed by flow cytometry. Representative flow cytometric profiles showing the binding of (A,D) mCD300c-Ig or control Ig, or (C) indicated antibodies to (A) resting and activated immune cells, or (C,D) transfected HEK-293 cells.

Taken together, our data suggest that hCD300c-Ig treatment attenuates GVHD, likely by inhibition of the proliferation and activation of donor T cells in response to alloantigen stimulation.

#### DISCUSSION

In an effort to identify additional immune regulators, the present study describes CD300c as a novel T cell co-inhibitory molecule. CD300c has a significant amino acid sequence and structural homology with the known B7 family members. CD300c protein is expressed on APCs and its counter-receptor is expressed on T cells. Functionally, CD300c-Ig protein inhibits the proliferation, activation and cytokine production of T cells. Therefore, CD300c contains typical features of B7 family members, suggesting that it is a B7 family-related molecule.

The nomenclature for mCD300c is still confusing (21). mCD300c is also called CLM-2, and mCD300c2 is also known as LMIR2/DIgR1/MAIR-II/CLM-4/CD300d (21, 22, 38–40). It has been reported that the mCD300c2 gene is located on mouse chromosome 11 (22, 38) and hCD300c is located near to human chromosome 17 (18), the syntenic region of mouse chromosome 11. The data further suggest that mCD300c2 is the mouse homolog of hCD300c (22, 38).

The B7 family is a member of the Ig superfamily. The extracellular region of the known B7 family members typically contain IgV and IgC domains. However, both human and mouse CD300c have only one IgV domain in the extracellular region. It has been reported that the interaction site in the Ig superfamily members is often mapped to the distal Ig domain, which would be the IgV domain in the CD300c. Therefore, the absence of the IgC domain in the CD300c would not significantly reduce its ability to inhibit T cell functions.

We have shown that the mCD300c2 protein is expressed on the cell surface of a variety of APCs, including B cells, monocytes, macrophages and DCs. It has been reported that that mCD300c2 mRNA was abundant in the spleen (21, 22, 38, 39) and that mCD300c2 protein was expressed on cell surface of DCs, monocytes, macrophages and B cells (21, 22, 38, 39). High hCD300c transcript levels have also been detected in the human spleen and thymus (28) and hCD300c protein has been shown on human monocytes, macrophages, granulocytes, and DCs, as well as, in a subpopulation of B and T cells (19, 24, 28). Our results are consistent with these reports.

Our results demonstrate that the mCD300c3 counter-receptor is expressed on resting and activated CD4 and CD8 T cells, B cells, DCs, monocytes, and macrophages. The expression levels of the counter-receptor on activated CD4 and CD8 T cells is upregulated upon activation, while the expression levels of the mCD300c counter-receptor on resting and activated B cells, DCs, monocytes, and macrophages were not significantly different. mCD300c2 protein did not bind to CD28, PD-1, CTLA-4, PD-1, or BTLA-expressing cells, indicating that the mCD300c2 counter-receptor is distinct from known members of the CD28 receptor family. CD300c is also considered as a receptor, and lipids such as phosphatidylserine and phosphatidylethanolamine can act as ligands for CD300c (23, 31, 41). It remains to be determined whether hCD300c and mCD300c2-Ig affects T cell functions through these lipids or a protein counter-receptor.

The expression of CD300c protein on APCs and its counterreceptor on T cells suggests that CD300c affects T cells. Indeed, we have demonstrated that both hCD300c and mCD300c2 significantly inhibit the proliferation, activation, and/or cytokine production of CD4 and CD8 T cells in vitro. It seems that the inhibition of the proliferation of murine T cells by hCD300c-Ig was slightly greater than that of human T cells (**Figure 2A** vs. **Figure 2K**), whereas the inhibition of the expression of CD69 on human T cells was greater than that on murine T cells (**Figures 2H**,**J** vs. **Figures 2L,M**). The phenomena may be related to the expression of the CD300c counter-receptor on T cells because we found that murine and human CD4 and CD8 T cells expressed different levels of CD300c counter-receptor (**Figure 6B** vs. **Supplemental Figure 3**).

We have also shown that hCD300c-Ig treatment attenuates acute GVHD in mice. To the best of our knowledge, this is the first report that CD300c is able to inhibit T cell function and treat GVHD. The effect of CD300c on GVHD is associated with the inhibition of T cell function in vivo. In agreement with the in vitro data, hCD300c-Ig inhibits T cell proliferation and activation in the GVHD model. However, although both mCD300c2-Ig and hCD300c-Ig inhibit the expression of CD69 in T cells in vitro, we did not observe that hCD300c-Ig treatment reduced CD69 expression by donor T cells in vivo. This inconsistency is most likely caused by time differences in analyzing this marker. CD69 is an early activation marker. We analyzed the expression of this marker 1 day after activation by anti-CD3 antibody or anti-CD3 and anti-CD28 antibodies in vitro, but 4 days after activation by allogeneic antigens in the GVHD model. hCD300c-Ig may inhibit the expression of CD69 in vivo at early time points, but this inhibition was not in effect 4 days later. This notion is supported by our results that hCD300c-Ig reduced the percentages of two other T cell activation markers CD44hi cells and CD62Llo cells, in CD4<sup>+</sup> and CD8<sup>+</sup> T cells in vitro 3 days (**Figures 3M–Q**) and in vivo 4 days (**Figures 7J–M**) after activation.

It has been reported that CD300c has effects on other types of immune cells (21–30). Crosslinking of CD300c on NK cells by its antibody induce cytokine secretion and degranulation (23). CD300c is also involved in the modulation of IgE-mediated basophil activation (30). mCD300c2 non-covalently associates

C57BL/6 mice at day 0 and i.p. with 20 µg hCD300c-Ig, or control Ig every 3 days for 6 times. (A–C) Recipients were monitored for (A) survival, (B) weight change, and (C) clinical GVHD. (D,E) In separate experiments, recipients given 20 µg hCD300c-Ig or control Ig at 3-day intervals from days 0 to 12 were euthanized 2 weeks (Continued) FIGURE 7 | after BMT. (D,E) The liver, SI and lung were analyzed for histologic damage. (D) Representative photomicrographs (the magnification was X200), and (E) mean ± SD of histopathology scores. (F–M) hCD300c-Ig inhibits T-cell proliferation and activation in response to alloantigens in vivo. Lethally irradiated BALB/c mice were injected i.v. with 5 × 10<sup>6</sup> BM 10 × 10<sup>6</sup> splenic cells from C57BL/6 mice. The recipients were injected i.v. on day 0 and i.p. on day 2 with 20 µg hCD300c-Ig, or control Ig. On Day 4 post-transplant, the percentage of (F–I) Ki67+, (J,K) CD44hi, and (L,M) CD62Llo cells in donor T cells (H2b+CD4+, or H2b+CD8+) of the spleens were examined by flow cytometry. (F,H,J,L) Representative flow cytometric profiles and (G,I,K,M) statistical data are shown. (J,L) Dash lines: control Ig; solid lines: hCD300c. Pooled data from 2 separate experiments are represented; with 5–6 mice per group in each experiment. \*P < 0.05 compared with control Ig-treated mice.

with the signaling adaptor DAP10 in macrophages and B cells (27, 34). mCD300c negatively regulates adaptive immune responses by B cells in a DAP12-dependent manner. mCD300c2– deficient B cells had an enhanced proliferation in response to BCR and CpG stimulation. In contrast, expression of mCD300c2 into mCD300c2−/<sup>−</sup> B cells was able to suppress BCR- and CpG-mediated proliferation (24, 35). In addition, mice deficient in MAIR-II (LMIR-2) are more susceptible to caecal ligation and puncture-induced peritonitis than wild-type mice (36). Furthermore, cross-linking of hCD300c by its antibody on DCs resulted in decreased expression of MHC-II and reduced production of TNFα and IL-6, but increased production of type I of IFN (29, 37). In contrast, mCD300c2 stimulates proinflammatory cytokines and chemokine secretions from macrophages. Cross-linking of mCD300c2 on macrophages resulted in increased production of TNFα, IL-6, and MCP-1 (27). It is unclear whether the effects of CD300c on these immune cells are caused by a direct effect. Similarly, although we used purified T cells for the [3H] thymidine incorporation and cytokine production assays, we used splenocytes that include T cells and other immune cells for the T cell activation studies. The possibility that the inhibition of T cell activation by CD300c-Ig may be caused by its indirect effect on other immune cells cannot be excluded at this stage.

In summary, we have identified CD300c as a B7 familyrelated molecule. CD300c protein is expressed on APCs, and

#### REFERENCES


its counter-receptor is expressed on T cells. Soluble human or mouse CD300c-Ig fusion proteins significantly inhibit T cell proliferation, activation, and cytokine production in vitro. Administration of hCD300c-Ig protein attenuates GVHD in mice. Therefore, CD300c protein has the potential to be used in the treatment of GVHD, autoimmune disease, and transplant rejection.

#### AUTHOR CONTRIBUTIONS

CC performed experiments and analyzed data. MS and YL performed experiments. LL designed experiments, analyzed data, supervised the study, and wrote the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by grants from NIH (1R01AI123131- 01), Connecticut Regenerative Medicine Research Fund (16- RMB-UCONN-02).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02657/full#supplementary-material


expression is increased in cow s milk allergy. J Allergy Clin Immunol. (2018). doi: 10.1016/j.jaci.2018.05.022. [Epub ahead of print].


**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 Cui, Su, Lin and Lai. 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.

# Selective Effects of mTOR Inhibitor Sirolimus on Naïve and CMV-Specific T Cells Extending Its Applicable Range Beyond Immunosuppression

Szilvia Bak <sup>1</sup> , Sabine Tischer <sup>1</sup> , Anna Dragon<sup>1</sup> , Sarina Ravens <sup>2</sup> , Lars Pape<sup>3</sup> , Christian Koenecke<sup>4</sup> , Mathias Oelke5,6, Rainer Blasczyk <sup>1</sup> , Britta Maecker-Kolhoff 7† and Britta Eiz-Vesper <sup>1</sup> \* †

<sup>1</sup> Hannover Medical School, Institute for Transfusion Medicine, Hannover, Germany, <sup>2</sup> Hannover Medical School, Institute of Immunology, Hannover, Germany, <sup>3</sup> Department of Pediatric Nephrology, Hannover Medical School, Hannover, Germany, <sup>4</sup> Department of Hematology, Hemostasis, Oncology and Stem Cell Transplantation, Hannover Medical School, Hannover, Germany, <sup>5</sup> Department of Pathology, John Hopkins School of Medicine, Baltimore, MD, United States, <sup>6</sup> NexImmune Inc., Gaithersburg, MD, United States, <sup>7</sup> Department of Pediatric Hematology and Oncology, Hannover Medical School, Hannover, Germany

#### Edited by:

Anil Shanker, Meharry Medical College, United States

#### Reviewed by:

Krishna Beer Singh, University of Pittsburgh, United States Dev Karan, Medical College of Wisconsin, United States

> \*Correspondence: Britta Eiz-Vesper eiz-vesper.britta@mh-hannover.de

†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: 13 August 2018 Accepted: 30 November 2018 Published: 17 December 2018

#### Citation:

Bak S, Tischer S, Dragon A, Ravens S, Pape L, Koenecke C, Oelke M, Blasczyk R, Maecker-Kolhoff B and Eiz-Vesper B (2018) Selective Effects of mTOR Inhibitor Sirolimus on Naïve and CMV-Specific T Cells Extending Its Applicable Range Beyond Immunosuppression. Front. Immunol. 9:2953. doi: 10.3389/fimmu.2018.02953 Cytomegalovirus (CMV) infection/reactivation remains among the most important complications of immunosuppression after transplantation. However, recent clinical observations indicate that mammalian target of rapamycin (mTOR) inhibition with sirolimus may improve the outcome of CMV complications. Underlying mechanisms of this observation, particularly the effect of sirolimus on naïve- and CMV-specific cytotoxic CD8<sup>+</sup> T-cell (CMV-CTL) functionality is still undiscovered. Here, the influence of sirolimus on naïve and memory CMV-CTLs was determined by CD3/CD28 crosslinking and alloreactivity assays. After stimulating CMV-CTL with HLA-A∗02:01-restricted CMVpp65-peptide loaded artificial antigen-presenting cells (aAPCs), we measured the effect of sirolimus on T-cell proliferation, phenotype, and functionality. Sirolimus significantly improved CMV-specific effector memory T-cell function and negatively influenced naïve T cells. This unique mechanism of action was further characterized by increased secretion of interferon-gamma (IFN-γ), granzyme B (GzB) and enhanced target-cell-dependent cytotoxic capacity of activated CMV-CTLs. Next-generation-sequencing (NGS) was applied to monitor T-cell receptor (TCR)-repertoire dynamics and to verify, that the increased functionality was not related to sirolimus-resistant CTL-clones. Instead, modulation of environmental cues during CMV-CTL development via IL-2 receptor (IL-2R)-driven signal transducer and activator of transcription-5 (STAT-5) signaling under mTOR inhibition allowed fine-tuning of T-cell programming for enhanced antiviral response with stable TCR-repertoire dynamics. We show for the first time that sirolimus acts selectively on human naïve and memory T cells and improves CMV-specific T-cell function via modulation of the environmental milieu. The data emphasize the importance to extend immune monitoring including cytokine levels and T-cell functionality which will help to identify patients who may benefit from individually tailored immunosuppression.

Keywords: HCMV, antiviral T cells, mTOR inhibitor, personalized immunosuppression, transplantation, sirolimus

# INTRODUCTION

Immunosuppressive therapy to deplete T cells, redirect T-cell trafficking, or terminate T-cell response pathways after solid organ (SOT) and hematopoietic stem cell transplantation (HSCT) is mainly used to prevent graft rejection or severe graftvs.-host disease (GvHD) (1–3). Immunocompromised patients are highly susceptible to viral infection and reactivation by endogenous herpes viruses such as cytomegalovirus (CMV) and Epstein-Barr virus (EBV), which are associated with high morbidity and mortality (4, 5). Current treatment strategies involve administering effective antiviral drug therapies, reducing the degree of immunosuppression, or changing the individual immunosuppressive drug regimen in order to restore virusspecific T cell-mediated immune responses (4, 6–10).

The immunosuppressive drug sirolimus was first discovered as an antifungal metabolite in Streptomyces hygroscopicus in 1975 (11), and was later found to potently inhibit the proliferation of immune cells such as T cells and dendritic cells (DCs) (12). Its target is the cellular kinase called mammalian target of rapamycin (mTOR), which is present in two functionally district complexes: complex 1 (mTORC1, sirolimus-sensitive) and complex 2 (mTORC2). Similar to other mTOR inhibitors (so-called rapalogs) such as everolimus, sirolimus prevents the translation of proteins that promote cell survival and proliferation by engaging with FK506-binding protein (FKBP). The sirolimus-FKBP complex binds to the sirolimus-sensitive mTORC1 protein complex and thus inhibits downstream phosphorylation activities, resulting in the blockade of G1/S cell cycle progression (13–17). The drug further mediates immunosuppressive function by attenuating signaling through the interleukin-2 receptor (IL-2R) and other cytokine receptors (12).

In 2005, Ozaki et al. were the first to report that sirolimus monotherapy results in better outcomes in renal transplant patients with CMV disease than standard calcineurin inhibitor-based immunosuppression (18). This observation was strengthened by accumulating evidence of better control of CMV viremia in sirolimus-treated patients following HSCT and SOT (18–22). Initially, it was speculated that by targeting the mTOR complex during the lytic phase of CMV infection, sirolimus abrogates the infection, and inhibits reactivation since CMV utilizes the mTORC1 pathway for viral replication (18). However, recent studies have shown that the favorable outcomes after transplantation are not associated with the direct molecular blockade of CMV reactivation, but can be attributed to indirect effects on the immune system (19). In 2009, two independent groups reported that sirolimus exerts dose-dependent immunostimulatory effects on CD8<sup>+</sup> memory T cells in mice and rhesus macaques exposed to viral pathogens (12, 23, 24). High-dose sirolimus suppressed CD8<sup>+</sup> T-cell expansion, whereas the quality and quantity of T-cell response was dependent on the duration and timing of treatment. When studying the immunostimulatory effects of sirolimus on bacterial-induced CD8<sup>+</sup> T-cell responses against skin transplants in a transgenic mouse system, Ferrer et al. (25) observed that sirolimus boosted antigen-specific T-cell responses to the pathogen, but not to the transplant. These effects seem to be intrinsic to T cells and influenced by the environment in which the antigen is presented.

Further studies demonstrated the link between the unique metabolic requirements of T cells and the ability of mTORC1 to integrate environmental cues involved in direct T-cell differentiation and function during sirolimus treatment (26–28). These results indicate that the drug functions as a signaling component downstream of T-cell receptor (TCR)/CD3-mediated activation. In addition to TCR-stimulation, co-stimulation, and IL-2R signaling also appear to play an important role in the effects of sirolimus on T-cell functionality (26, 29). Despite sirolimussensitive mTORC1, IL-2 signaling in T cells is also mediated by the signal transducer and activator of transcription 5 (STAT-5) (30–32). Although many reports focus on the role of mTORC1 signaling, cross-talk between these key regulators and the signal that drives T-cell function in the presence of sirolimus have not been defined yet.

In this study, diligent characterization of the effects mediated by sirolimus and its interactions with TCR, IL-2R, mTORC1, and STAT-5 on the functionality of CMV-specific CD8<sup>+</sup> cytotoxic T lymphocytes (CTLs) and naïve T cells was assessed. To exclude the influence of sirolimus on other cells besides T cells, artificial antigen-presenting cells (aAPCs) loaded with HLA-A<sup>∗</sup> 02:01 restricted CMVpp65 peptide (A02pp65p) was used (33, 34).

We found that naïve T cells showed no significant response to treatment with sirolimus. In contrast on memory T cells sirolimus had differential effects on key elements of T-cell activation and function such as (1) the dynamics of the TCR repertoire, (2) the phosphorylation of proteins involved in TCR/mTORC1/IL-2R signaling, and (3) the expression of micro-RNAs (miRNAs, e.g., miRNA-21) and effector genes like granzyme B (GzB) and interferon-gamma (IFN-γ). The modulation of environmental cues during antiviral memory Tcell development through the activation of IL-2R driven STAT-5 signaling under the cover of mTORC1 inhibition allows the fine-tuning of antiviral T-cell programming for improved CMVspecific T-cell response.

These results suggest a need to optimize the monitoring of immunosuppressed patients with an elevated risk of pathogen infection or reactivation by determining serum IL-2 or IL-2R subunit-sharing cytokine levels and antigen-specific T-cell functionality for further individualization of immunosuppressive therapy.

#### MATERIALS AND METHODS

#### Isolation of PBMCs and T Cells

Experiments were performed with residual blood samples from platelet (PLT) apheresis disposables used for routine PLT collection of regular anonymous healthy donors of the Hannover Medical School (MHH) Institute for Transfusion Medicine. Informed consent was obtained from all donors following approval by the Ethics Committee of MHH (ethical number: 3639-2017, 2744-2015), and trial subject data were treated as confidential information protected by medical confidentiality. Peripheral blood mononuclear cells (PBMCs) were isolated from HLA-A<sup>∗</sup> 02:01-positive CMV-seropositive donors by discontinuous-gradient centrifugation. Untouched CD3<sup>+</sup> and CD8<sup>+</sup> T cells were enriched by magnetic cell sorting (MACS) using negative selection kits (Miltenyi Biotec, Bergisch, Gladbach, Germany), according to the manufacturer's instructions. The purity was routinely higher than 90%, as determined by flow cytometry. Magnetically labeled non-T cells were collected from CD3<sup>+</sup> T-cell isolation and were used as target cell population in alloreactivity approach.

#### Alloreactivity and CD3/CD28 Crosslinking Approach

To investigate the effects of sirolimus on human naïve and memory CD8<sup>+</sup> T-cell populations, alloreactivity assay was performed using 1 × 10<sup>5</sup> CD3<sup>+</sup> T cells stimulated for 2 days with 5 × 10<sup>5</sup> allogeneic CD3<sup>−</sup> cells which were simultaneously collected during CD3<sup>+</sup> T-cell isolation and then irradiated by high dose gamma irradiation (30Gy). For antigen-independent stimulation 5 × 10<sup>5</sup> isolated CD8<sup>+</sup> T cells were stimulated on anti-CD28-coated (CD28.2, Becton Dickinson (BD, Heidelberg, Germany) 48-well plates or with human T activator CD3/CD28 Dynabeads (Thermo Fisher Scientific, Waltham, MA) for 3 days according to the manufacturer's instructions. Cell culture media (RPMI 1640, Lonza, Basel, Switzerland) was supplemented with 10% heat-inactivated human AB serum (c.c.pro Oberdorla, Germany), IL-2 (50 U/ml, PeproTech, Hamburg, Germany) in the presence or absence of sirolimus (Sigma-Aldrich by Merck, Darmstadt, Germany) at the recommended therapeutic concentration (10 ng/ml) (13, 35). T cells were analyzed for their phenotype and expression of activation markers by flow cytometry.

# Generation of CMV-Specific CD8<sup>+</sup> T Cells Using aAPC Beads

To examine the direct effects of sirolimus on CMV-specific CTLs, aAPC beads were used. These aAPCs were generated by coupling HLA-A<sup>∗</sup> 02:01 molecules (DimerX, BD), loaded with anti-CD28 mAbs (BD) and HLA<sup>∗</sup> 02:01-restricted CMVpp65495−<sup>503</sup> peptide (NLVPMVATV, A02pp65p, ProImmune, Oxford, UK) onto M-450 Epoxy beads (Thermo Fisher Scientific, Waltham, MA, USA), as previously described (33). The beads were ready to use and were stored at 4◦C up to 6 months. Isolated CD8<sup>+</sup> T cells were cultured at the recommended 1 to 1 cells to aAPC beads density for 7 days in aAPC medium [(RPMI 1640 (Lonza) supplemented with 1% sodium pyruvate (c.c.pro), 5 or 10% heatinactivated human AB serum (c.c.pro), 0.4% MEM vitamins and 1% non-essential amino acids (Thermo Fisher Scientific)] in the presence or absence of sirolimus (0.5–1000 ng/ml, Sigma-Aldrich by Merck). Sirolimus and aAPCs were added at the same time to the isolated CD8<sup>+</sup> T cells on day 0. The aAPC medium was supplemented with IL-2 (50 U/ml) on days 0 and 3 for the generation of CMV-specific CD8<sup>+</sup> T cells. IL-7/IL-12/IL-15 or IL-21 cytokines (each 10 ng/ml; all PeproTech) were added independently as indicated in the result section and were used to replace IL-2 for the evaluation of IL-2 and sirolimus-related expansion and functionality of CMV-specific T cells. The frequency of A02pp65p-positive (CMV-specific multimer+) CD8<sup>+</sup> T cells was assessed on day 7 using peptide major histocompatibility complex (pMHC) multimer staining and further analyses were performed as described in detail below.

### Flow Cytometry Analysis, Multimer Staining, and Phenotyping

Phenotypic characterization of T cells was carried out after alloreactivity, CD3/CD28 crosslinking approaches and aAPC stimulation, using the following antibodies: anti-CD3-peridininchlorophyll (PerCp) (SK7), anti-CD8-AlexaFluor700 (AF-700) (SK1), anti-CD25- allophycocyanin(APC)/phycoerythrin (PE)/Cy7/BV421 (BC96), anti-CD45RA-PE/Cy7/BV510 (HI100), anti-CD62L-APC/Cy7 (DREG-56), anti-CD69- APC/Cy7/PE/Cy7 (FN50), anti-137-APC (4B4-1), anti-CD366- APC/Cy7 (F38-2E2), anti-CD223-fluorescein-isothiocyanate (FITC) (11C3C65), anti-CD152-PE/Cy7 (L3D10) (all BioLegend, San Diego, CA, USA) and anti-CD279-PE (EH12.1), anti-CD154-PE (TRAP1) (BD). All flow cytometry analyses were performed using the FACS Canto 6c and 10c systems (BD) and BD FACSDiva Software version 8.0.1. At least 10,000 events were acquired in the CD3<sup>+</sup> or CD8<sup>+</sup> gate. Gates were set based on the light-scatter properties of lymphocytes.

Multimer staining was performed to monitor the frequency of A02pp65p-positive (CMV-specific) CD8<sup>+</sup> T cells. It was assessed before and after aAPC stimulation using PE or APC-conjugated HLA-A<sup>∗</sup> 02:01/CMVpp65p-specific (Immudex, Copenhagen, Denmark) dextramers. To be considered positive (multimer+), the sample had to (1) be a well-defined cell population and/or (2) contain ≥0.5% multimer+CD8<sup>+</sup> T cells.

### Cell Counting, T-Cell Proliferation, and Cell Death

Trypan blue dye (Thermo Fisher Scientific) exclusion technique for counting of living cells manually was performed using bifocal light microscope before and after T-cell stimulation assays. Cell proliferation was monitored by carboxyfluorescein succinimidyl ester (CFSE) labeling (Thermo Fisher Scientific) at final concentration 1µM on day 0 and CFSE-labeled cells were stimulated 7 days according to aAPC approach. Dead cells were excluded by 7-amino-actinomycin (7AAD) (BioLegend) staining in combination with the following cell surface antibodies: anti-CD8-PE/Cy7 (SK1) (BioLegend), anti-CD3-APC (SK7) (BD). CFSE dilution and 7AAD were analyzed by flow cytometry.

# Intracellular Cytokine Staining

After 7 days of aAPC stimulation and sirolimus treatment, cells were re-stimulated with 10µg/ml A02pp65<sup>p</sup> at a density of 1– 2 × 10<sup>5</sup> cell/well for 1 h at 37◦C and incubated with Brefeldin A (1:1,000, BioLegend) for additional 4 h at 37◦C. Expression of intracellular cytokines were assessed by pMHC multimer and surface staining for CD3 and CD8 following intracellular staining with anti-Granzyme B-PacificBlue (GB11), anti-TNFα-PE/Cy7 (Mab11) (BioLegend), anti-IFNγ-PE (45.15) (Beckmann Coulter, Brea, CA, USA) using IntraPrep Kit (Beckmann Coulter) according to the manufacturer's instructions. Briefly, following 5 h peptide re-stimulation, multimer, and surface antibody staining were performed and then the cells were permeabilized by addition of IntraPrep Reagent 1 and 2, subsequently. Cells were washed afterwards, stained with intracellular antibodies, and analyzed by flow cytometry.

### IFN-γ ELISpot Assay

Antigen-specific IFN-γ-producing CD8<sup>+</sup> T cells were determined after 7 days of aAPC stimulation and treatment by IFN-γ Enzyme Linked Immuno Spot Assay (ELISpot) as previously described (36), using pre-coated IFN-γ EliSpot plates (Lophius Biosciences, Regensburg, Germany). Briefly, 2.5 × 10<sup>3</sup> CD8<sup>+</sup> T cells were plated in 125 µl aAPC media/well and incubated overnight with 10µg/ml A02pp65<sup>p</sup> or left unstimulated (negative control). Spots were developed based on the manufacturer's recommendation and data were acquired on an "AID iSpot Reader System" with "AID EliSpot Software Version 7.0" and spot counting was performed with "AID EliSpot Software Version 8.0." All spot counts are mean values from duplicates and expressed as spot-forming unit (SFU) or SFU per 1,000 multimer+CD8<sup>+</sup> T cells, respectively.

#### Multiplex Cytokine Profiling

The secretion levels of effector molecules in the T-cell supernatants after culture of 7 days aAPC stimulation with or without sirolimus treatment in the presence of IL-2 were determined by LEGENDplexTM bead-based immunoassay (BioLegend) following overnight A02pp65<sup>p</sup> re-stimulation. The LEGENDplexTM Human CD8/NK Panel was used to quantify simultaneously 13 human cytokines, including IL-2, IL-4, IL-10, IL-6, IL-17A, tumor necrosis factor alpha (TNF-α), soluble Fas (sFas), sFas ligand (sFasL), IFN-γ, granzyme A (GzA), GzB, perforin, and granulosyn according to manufacturer's protocol.

# CD107a Degranulation Assay

Cytotoxicity of CMV-specific CD8<sup>+</sup> T cells was assessed by detecting cell surface expression of CD107a. On day 7, 2.5 × 10<sup>5</sup> aAPC stimulated and sirolimus treated cells (as previously described) were re-stimulated with A02pp65p. Anti-CD107a-PE/Cy7 (H4A3) and Monensin (1:1,000, both from BioLegend) were added and cells were incubated for 4 h at 37◦C before cell surface staining with anti-CD3-FITC (UCHT1) and anti-CD8- APC (SK1) (both BioLegend) was performed. pMHC multimer staining was assessed before re-stimulation for 10 min at 37◦C. Data were acquired by flow cytometry.

#### Evaluation of Cytotoxicity in Response to Target Cell Recognition

Cytotoxicity of the 7 day aAPC stimulated and sirolimus treated CMV-specific CD8<sup>+</sup> effector T cells was evaluated in the presence of HLA-A<sup>∗</sup> 02:01 transduced and CFSE (final concentration 1µM) labeled K562 target cells (1.5 × 10<sup>7</sup> cells) For peptideloading, target cells were re-suspended in aAPC media and plated into 24-well plates at a cell density of 2.5 × 10<sup>6</sup> cells/well. Peptide (A02pp65) was added at a concentration of 10µg/ml and incubated overnight at 37◦C. On day 7, following aAPC stimulation and sirolimus treatment in the presence of IL-2 or IL-15, CMV-specific, effector and peptide loaded K562 target cells were cocultured for 5 h at 37◦C in 96 well-plates in fresh aAPC media containing IL-2 or IL-15, respectively. Effector to target (E:T) ratios of 1:1, 5:1, and 10:1 were obtained by setting target cell number constant (2.5 × 10<sup>4</sup> cells/well). Specific lysis of target cells was detected by 7AAD staining and data were acquired using flow cytometry.

# Phosphorylation Analysis

Following 7 days of aAPC stimulation and sirolimus treatment CD8<sup>+</sup> T cells were re-stimulated with 10µg/ml peptide (A02pp65) for 1 h and phosphorylation of extracellular regulated kinase 1/2 (pERK1/2), protein kinase B on Ser473 (pAktSer473), on Thr308 (pAktThr308), ribosomal protein 6 (pS6), STAT-5 (pSTAT-5) was evaluated by phospho-flow cytometry. Phosphorylation was determined by surface staining with anti-CD3-FITC (UCHT1) (BioLegend), anti-CD8-PerCP/Cy5.5 (RPA-T8) (BD), followed by fixation (Fix Buffer I, BD), permeabilization with Perm Buffer III (BD), and intracellular staining with anti-pS6-AlexaFlour647 (N7-548), anti-pERK1/2- AlexaFlour647 (20A), anti-pSTAT5-AlexaFlour647 (47/STAT5), or immunoglobulin (Ig)G1 AlexFlour647 isotype control (MOPC-21), anti-pAkt(T308)-PE (J1-223.371), anti-pAkt(S473)- PE (M89-61), or IgG1 PE isotype control (MOPC-21) (all BD).

# STAT-5 Inhibition

To investigate the mode of action of STAT-5, CMV-specific CD8<sup>+</sup> T cells were first stimulated for 7 days with aAPCs and treated with or without sirolimus in the presence of IL-2 or IL-15. Stimulated CD8<sup>+</sup> T cells were incubated with STAT-5 inhibitor (STAT-5i, Merck Millipore) at a concentration of 50µg/ml for ∼20 h at 37◦C. In addition to intracellular cytokine staining (ICS), T-cell STAT-5 and S6 phosphorylation was determined by phospho-flow cytometry and fluorescence microscopy as described below.

#### Immunofluorescence Microscopy

A total of 1.5 × 10<sup>5</sup> CD8<sup>+</sup> T cells were re-stimulated with A02pp65<sup>p</sup> for 1 h or left unstimulated. Thereafter, cells were stained with PE-conjugated pMHC multimers for 30 min at RT. Each sample was fixed then with Fix Buffer I and permeabilized with Perm Buffer III (both from BD). Cells were stained with anti-pSTAT5/b (5G4) primary antibody (at a final concentration of 1µg/ml) and IgG-FITC secondary antibody (both from Santa Cruz Biotechnology, Dallas, TX, USA) followed by staining with anti-pS6-AlexaFlour647 (N7-548) (BioLegend). Staining was performed for 30 min at 4◦C. Between each individual steps, cells were subsequently washed twice with PBS+1%BSA. Cells were placed onto microscope slides and after mounting of the coverslip with mounting medium (Dianova, Hamburg, Germany)**,**samples were analyzed by using Olympus IX81 fluorescent microscope (Olympus, Shinjuku, Tokyo, Japan) at magnification 60×.

#### Gene and miRNA Expression Analysis

Total RNA from CD8<sup>+</sup> T cells after aAPC stimulation and treatment (as previously described: after re-stimulation with 10µg/ml CMVpp65<sup>p</sup> for overnight) was isolated using mirVana RNA isolation Kit (Thermo Fisher Scientific). cDNA was reverse-transcribed by either the microRNA Transcriptions Kit or the High-capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific), according to the manufacturer's instructions. Expression of miR-21, miR-155, miR-181a, perforin, cyclin D1 (Bcl-1), suppressor of cytokine signaling 1 (SOCS1), phosphatidylinositol-3,4,5-trisphosphate 5-phosphatase-1 (SHIP-1), T-cell associated transcription factor (T-bet), mitogen activated protein kinase 1 (MAPK1)/ERK, eomesodermin

(EOMES), and Ki-67 were quantified by inventoried mixes (Thermo Fisher Scientific). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and miR-191 served as internal control.

Global gene expression analysis was performed on multimersorted CMV-specific CD8<sup>+</sup> T cells (expanded and treated with sirolimus for 7 days in the presence of IL-2 as described above) following overnight aAPC re-stimulation. CMV-specific CD8<sup>+</sup> cells were sorted based on their multimer-specificity by high speed flow cytometry sorters at the Research Facility Cell Sorting of MHH. The Microarray utilized in this study represents a refined version of the Whole Human Genome Oligo Microarray 4 × 44K v2 (Design ID 026652, Agilent Technologies), called "026652QM\_RCUG\_HomoSapiens" (Design ID 084555) developed by the Research Core Unit Genomics (RCUG) of MHH. Microarray design was created at Agilent's eArray portal using a 1 × 1 M design format for mRNA expression as template. All non-control probes of design ID 026655 have been printed four times within a region comprising a total of 181560 Features (170 columns × 1,068 rows). Four of such regions were placed within one 1M region giving rise to four microarray fields per slide to be hybridized individually (Customer Specified Feature Layout). Control probes required for proper Feature Extraction software operation were determined and placed automatically by eArray using recommended default settings. Measurements of on-chip replicates were averaged and normalized by quantile normalization approach. Then clustering and heat map were created using the Morpheus web-based tool. GeneCards <sup>R</sup> database was used to receive genomic and proteomic information about the particular genes.

# TCR Sequencing

For mRNA-isolation of flow cytometry sorted cells the Qiagen Micro Kit was used, following rapid amplification of cDNA ends using the Smarter 5'RACE cDNA amplification kit (Clontech 634923) according to the recommended protocol. Per sample 5 µl RNA was used for cDNA synthesis. Next, complementarity-determining region 3 (CDR3) regions of the human TCR beta chain were amplified through gene-specific primers for the constant region of the beta (β)-chain (GCACACCAGTGTGGCCTTTTGGG) and the introduced SMARTER oligonucleotide (CTAATACGACTCACTATAGGGC) using the Advantage 2 PCR kit (Clontech 639206) in a 50 µl reaction. Primer sequences contain 16 S Illumina overhang adapter sequences. Cycling conditions were as following: 120 s 95◦C; 30 times 30 s 95◦C, 45 s 64◦C, 60 s 72◦C; 60 s 72◦C. Generated PCR amplicons were agarose gel purified. Next, samples were labeled with Nextera Illumina Index reads within 10 additional PCR cycles using the Advantage 2 PCR kit (CLontech) and purified with Agencourt AMPpure XP beads (Beckman Coulter) and. subjected to Illumina MiSeq analysis using V2 500 cycles or V3 600 cycle paired-end sequencing reagent. Obtained Fastq files were annotated according to IMGTHighV/quest. For downstream bioinformatics analysis only productive reads were taken into consideration. Individual CDR3 sequences were ranked according to their abundance within the respective samples. For multisample comparison obtained reads of CDR3 sequences were normalized to all productive reads per sample. Shannon diversity indices were calculated using the R library "vegan" prior to normalization to 5,000 productive sequences.

### Patients and Treatment Regimen

PBMCs from in vivo sirolimus-treated kidney- (n = 3) and stem cell-transplanted (n = 2) patients or from healthy individuals (n = 6) with or without sirolimus treatment (10 ng/ml) were rested overnight at 37◦C. Patient information is summarized in **Table S1**. Thereafter, analysis of STAT-5 and S6 phosphorylation were assessed as described previously following 15 min CD3/CD28 Dynabeads stimulation in the presence or absence of IL-2. Additionally, intracellular cytokine staining was performed following 5 h incubation/stimulation under the aforementioned conditions. Phosphorylation of S6 and STAT-5 and cytokine expression was assessed as described above on gated CD8<sup>+</sup> T cells by flow cytometry.

#### Statistics

Statistical analyses were performed in GraphPad Prism version 7.0 software (GraphPad Software, San Diego, CA, USA) using two-paired Student's t-test or two-way analysis of variance. Levels of significance were expressed as p-values [∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, not significant (n.s.)].

# RESULTS

#### Expression of Activation Marker on Sirolimus-Treated Memory T Cells Is Induced by TCR Activation and IL-2 Supplementation

As the purpose of immunosuppression is to prevent complications like graft rejection and GvHD, which are mainly caused by alloreactive naïve T cells, the selective effect of sirolimus was investigated on human naïve and memory CD8<sup>+</sup> T-cell populations. T-cell activation marker expression was analyzed with an alloreactivity assay using CD3<sup>+</sup> T cells stimulated with irradiated allogeneic CD3<sup>−</sup> cells (**Figure 1A**) after co-stimulation with anti-CD28-coated plates (**Figure 1B**) and stimulation by anti-CD3/CD28 crosslinking (**Figure 1C**) in the presence of IL-2 and sirolimus at the recommended therapeutic concentration (10 ng/ml). On naïve CD8+CD45RA+CD62L<sup>+</sup> T cells, the mean normalized percentage of cells (sum) expressing CD25, CD69, CD137, and CD154 relative to that in untreated controls (equalized to a sum of 400%) was not affected by sirolimus alone

the presence (+) or absence (–) of IL-2 (50 U/ml). Shown are mean normalized percentages of activation markers relative to untreated controls (indicated as dashed lines). (A) Stimulation on allogeneic irradiated CD3<sup>−</sup> cells (n = 6). (B) Stimulation on anti-CD28-coated plates (n = 5). (C) Stimulation with anti-CD3/CD28 beads (n = 3).

(405.6%) and was moderately downregulated by supplemental IL-2 alone (388.1%) and by sirolimus and IL-2 combined (379.4%) (**Figure 1A**). Sirolimus alone led to a decrease in memory CD8+CD45RA<sup>−</sup> T-cell activation (349.8%). Interestingly, this suppressive effect was overcome by combining sirolimus with IL-2 (531.8%). These results imply that the immunosuppressive effects of sirolimus on memory T cells in the allogeneic TCR-dependent alloreactivity assay were overcome by IL-2.

Following CD28 co-stimulation, activation marker expression decreased slightly after treatment with sirolimus alone (naïve, 340.3%; memory, 245.6%), increased on both T-cell populations after IL-2 supplementation (naïve, 474.3%; memory, 792%), and was highest after treatment with IL-2 alone (naïve, 777%; memory, 1,477%) (**Figure 1B**).

Antigen-independent stimulation via the TCR using anti-CD3/CD28 beads was further assessed in order to determine if the positive effect of sirolimus depends not only on IL-2, but also on TCR-activation (**Figure 1C**). Sirolimus treatment did not influence the overall activation marker expression on naïve T cells. On memory T cells, the addition of IL-2 resulted in slight upregulation (408.5%), which was further increased in the presence of sirolimus (419.3%). Thus, in the absence of TCR signaling (**Figure 1B**), sirolimus had a negative effect on naïve and memory T cells which was compensated by IL-2 supplementation, but was still lower than with IL-2 alone. Overall, the positive effect of sirolimus plus IL-2 was highest on memory T cells. These results indicate that memory T cells are more susceptible to the immunostimulatory effect of sirolimus, and that this effect strongly depends on activation

of the TCR via either allogeneic target cells (**Figure 1A**) or anti-CD3/CD28 (**Figure 1C**) as well as on the presence of IL-2.

#### Sirolimus Suppresses CMV-Specific T-Cell Expansion in a Dose-Dependent Manner

To investigate the effects of sirolimus on CMV-specific CD8<sup>+</sup> T cells, the optimal concentration range of sirolimus for in vitro experiments was determined, which was set to be 0.5–1,000 ng/ml (**Figure 2**). As expected, sirolimus had a dose-dependently negative effect on the expansion of CMV-specific T cells. Normalized values for the generation of A02pp65p-specific (multimer+) CD8<sup>+</sup> T cells (**Figure 2A**), 5, 10, and 40 ng/ml were determined to be the respective inhibitory concentration (IC) for 25, 50, and 75% inhibition of generation of CMV-specific multimer+CD8<sup>+</sup> T cells relative to numbers in untreated controls (100%).

**Figure 2B** shows one representative result. Overall and relative to untreated controls (mean of 62.2% multimer+CD8<sup>+</sup> T cells), 47.5, 35.1, and 25.5% multimer+CD8<sup>+</sup> T cells were generated at IC25, IC50, and IC75 (**Figure 2C**). IC50 (10 ng/ml) was preferably used in subsequent experiments as it reflects the therapeutic concentration (13, 35); moreover, the number of

Data are shown as means ± SD. The two-paired Student's t-test was used to test for statistically significant differences [\*p < 0.05, \*\*\*\*p < 0.0001, non-significant (ns)].

CMV-specific T cells were generated at IC50 to perform T-cell functional assays.

To investigate whether the observed suppression of T-cell expansion was due to decreased proliferation and/or increased cell death, cell proliferation was monitored by CFSE dilution on day 7 (**Figure 2D**, **Figure S1A**), trypan blue exclusion (**Figure S1B**), and 7AAD staining (**Figure S1C**). Only a slight effect of sirolimus on the proliferation capacity of CMV-specific T cells was observed (mean IC25: 89.4%, mean IC50: 92.3%, and mean IC75: 86.6% vs. mean control value of 95.6%, **Figure 2D**). As expected, treatment significantly inhibited the proliferation (total number) of CD8<sup>+</sup> T cells in a dose-dependent manner (**Figures S1A,B**), without increasing cell death (**Figure S1C**). Thus, sirolimus inhibits the generation frequency of CMV specific T-cells without influencing their proliferation capacity.

#### Sirolimus Has No Effect on Effector Memory Phenotype but Upregulates Activation Marker Expression on CMV-Specific T Cells

The phenotype of CMV-specific CD8<sup>+</sup> T cells generated in response to sirolimus treatment was determined based on CD45RA and CD62L expression before and after 7 days of treatment. As expected, CMV-specific (**Figure 3A**), and total CD8<sup>+</sup> T cells (**Figure S2A**) were mainly effector memory (EM) T cells (CD45RA−CD62L−, multimer+: 79% and total CD8+: 64%, respectively), and their frequencies were only slightly higher than those in untreated controls (multimer+: 77%, total CD8+: 68%).

Upregulation of classical activation (CD25, CD69) and exhaustion marker (PD-1, Lag-3) expression was markedly increased after sirolimus treatment in CMV-specific T cells

FIGURE 4 | Improved functionality of CMV-specific T cells after sirolimus treatment. (A) Relative quantities (RQ) of IFN-γ and granzyme B (GzB) secretion were assessed using RT-qPCR following overnight A02pp65<sup>p</sup> re-stimulation on aAPC stimulated and sirolimus treated and untreated CD8<sup>+</sup> T cells. (B) IFN-<sup>γ</sup> expression levels determined by IFN-γ ELISpot assay and expressed as the number of spot-forming units (SFU) per 1,000 CMV-specific multimer+CD8<sup>+</sup> T cells. (C) Percentages (%) of intracellular IFN-γ, GzB, and TNF-α on CMV-specific multimer+CD8<sup>+</sup> T cells, as determined by intracellular staining using multicolor flow cytometry. (D) Target-cell recognition assay was performed on day 7. Total CD8<sup>+</sup> T cells were co-cultured for 5 h with A02pp65p-loaded and CFSE-labeled A\*02-transduced K562 target cells (squares) at effector to target ratios of 1:1 (n = 8), 5:1 (n = 8), and 10:1 (n = 6). Unloaded K562 cells served as controls (spheres). Percentages of dead cells were detected by 7AAD staining and multicolor flow cytometry. (E) Degranulation was determined as the percentage and median fluorescence intensity (MFI) of CD107a expression on CMV-specific multimer+CD8<sup>+</sup> T cells by multicolor flow cytometry following 4 h of re-stimulation with A02pp65<sup>p</sup> (<sup>n</sup> <sup>=</sup> 5). Values are displayed as mean (±) SD. Statistical analysis: (A–C) two-paired Student's t-test and (D,E) two-way analysis of variance [\*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001, non-significant (ns)].

(**Figure 3B**) compared to total CD8<sup>+</sup> T cells (**Figure S2B**). These results were not unexpected since the transient expression of exhaustion markers is also used to describe T-cell activation (37, 38). Similar to the lack of effect on the expression of CTLA-4 and Tim-3 on multimer<sup>+</sup> T cells (**Figure 3B**), treatment had no significant effect on the overall expression of activation and exhaustion markers on total CD8<sup>+</sup> T cells (**Figure S2B**). Interestingly, Tim-3 and Lag-3 expression on CD8<sup>+</sup> T cells was significantly downregulated in response to sirolimus treatment. Taken together, these data suggest that CMV-specific T cells expand on aAPCs in the presence of IL-2 and sirolimus exhibit an effector memory-like phenotype characterized by strong expression of activation markers.

These results are in line with those shown in **Figure 1** and highlight the strong immunostimulatory effects of sirolimus on memory T cells in the presence of IL-2 and TCR signaling.

#### CMV-Specific T Cells Show Potent Increase in Functionality and Target Cell Recognition After Sirolimus Treatment

To evaluate whether sirolimus has an influence on antigenspecific effector function (**Figure 4**), mRNA expression of effector molecules such as IFN-γ and GzB was measured following overnight antigen re-stimulation by reversetranscription quantitative PCR (RT-qPCR). Sirolimus-treated cells showed significantly higher levels of IFN-γ (relative quantity (RQ) mean IC25: 1.9, IC50: 4.3, IC75: 3.6) and GzB (RQ mean IC25: 2.1, IC50: 2.2, IC75: 2) mRNA expression than untreated controls (**Figure 4A**).

IFN-γ ELISpot assay and combined ICS with multimer staining were done to confirm these results and further characterize the responsiveness of CMV-specific T-cell responses (**Figures 4B,C**, **Figures S3A,B**). Analysis of the total number of SFU showed that sirolimus treatment led to a slight reduction of IFN-γ secretion compared to the controls (control: mean of 509 SFU, IC25: 411, IC50: 473, and IC75: 347; **Figure S3A**). However, analysis of SFU per 1,000 CMV-specific CD8<sup>+</sup> T cells showed increased IFN-γ expression (control: mean of 308 SFU, IC25: 427, IC50: 548, and IC75: 654; **Figure 4B**). Our evaluation of the effector function of CMV-specific CD8<sup>+</sup> T cells by ICS (**Figure 4C**, **Figure S3B**) showed significant increases in the frequency (**Figure 4C**) and in median fluorescence intensity (MFI) (**Figure S3B**) of IFN-γ, GzB, and TNF-α secretion in sirolimus-treated virus-specific T cells. LEGENDplexTM Human CD8/NK Panel Detection Antibodies were used for further quantification of these and other effector cytokines from cell culture supernatant (**Figure S3C**). Overall expression of effector cytokines (e.g., IL-2, perforin etc.) was increased in sirolimustreated cells compared to untreated controls. These results showed clear evidence of a higher functionality of sirolimustreated CMV-specific T cells, which was further strengthened by proof of the capacity of sirolimus-treated CD8<sup>+</sup> T cells to lyse A02pp65p-expressing K562 target cells (**Figure 4D**). CD8<sup>+</sup> T cells generated over 7 days with and without IC50-level sirolimus treatment were co-cultured with peptide-unloaded or -loaded target cells for 5 h at the following three different effector:target (E:T) ratios: 1:1, 5:1, and 10:1. Compared to untreated cells, the capacity of sirolimus-treated CD8<sup>+</sup> T cells to recognize and lyse target cells (CFSE+7AAD+) was higher at every E:T ratio and was the highest at 10:1 (**Figure 4D**).

Surface expression of CD107a on CMV-specific CD8<sup>+</sup> T cells upon peptide re-stimulation was also measured to further analyze cytotoxicity. Compared to controls, treated cells showed increased CD107a expression in terms of both frequency and MFI (**Figure 4E**), with a significant difference in MFI. Taken together, these data indicate that sirolimus improves the functional quality of CMV-specific CD8<sup>+</sup> T cells (**Figure 4**) while suppressing the expansion of CMV-specific T cells (**Figure 2**).

#### Sirolimus Does Not Affect the Dynamics of CDR3 Repertoires in CMV-Specific T Cells

In order to answer the question of whether sirolimus promotes the expansion of sirolimus-resistant T-cell clones, an RNAbased next-generation sequencing (NGS) approach was applied to monitor the dynamics of TCR β-chain repertoires in multimersorted CMV-specific CD8<sup>+</sup> T cells on day 7 (**Figure 5**). As expected, the pre- and post-expansion TCR repertoires were highly clonal independently of treatment, and they consisted of a very small number of expanded clones as displayed in stacked area graphs in **Figure 5A**. Likewise the clonal sizes between repertoires were highly similar (**Figure 5B**).

Interestingly, none of the most frequent CDR3 sequences were shared by the analyzed healthy donors (as indicated by the color code and displayed CDR3 sequences in **Figure 5A**). Overall, diversity was low and did not change after expansion and treatment (**Figure 5C**). Taken together, RNA-based NGS and CDR3 based analysis reflected no change in the T-cell repertoire post-sirolimus treatment, which significantly argue against effect of sirolimus on any T-cell clone. These results suggest that the αβ TCR repertoire reflects the immunological history of an individual rather than the selective pressure of immunosuppression on a healthy individual.

### Signaling Pathways Involved in T-Cell Activation and Function Are Differently Regulated by Moderate mTORC1 Inhibition

The mechanism of sirolimus to improve the functionality of CMV-specific CD8<sup>+</sup> T cells was evaluated by analyzing the

phosphorylation of kinases such as Akt and ERK1/2, respectively, and of the proteins (S6 and STAT-5) which are involved in T-cell activation and signaling (**Figure 6**; **Figure S4**). Antigenspecific phosphorylation of S6, a downstream target of mTORC1 was evaluated to confirm the inhibitory effect of sirolimus on mTORC1. Since, the IC50 was used in order to generate a sufficient number of CMV-specific T cells for further functional analysis, incomplete inhibition of mTORC1 was expected. The frequency of phosphorylated S6 was reduced in sirolimus-treated cells (mean 52.3%, **Figure 6A**), but MFI analysis (mean 7092.5, **Figure S4A**) showed that phosphorylation was higher in treated cells compared to untreated controls (62.1%, 4785.5). Treated CD8<sup>+</sup> T cells showed a TCR responsiveness gain compared to controls, as reflected by increased frequency and MFI values for phosphorylation of distal signaling molecules such as ERK1/2, AktSer473, and AktThr308 (**Figures 6B–D**, **Figures S4B–D**). These results are in line with the detected increase in CD25 (also known as IL-2R alpha chain) expression on sirolimus-treated cells (**Figure 1**, **Figure 3B**), and they might indicate a shift toward IL-2-dependent regulation of CMV-specific CD8<sup>+</sup> T-cell development. Indeed, sirolimus treatment resulted in significant increases in STAT-5 phosphorylation in terms of frequency (**Figure 6E**, 30.6 vs. 12.9%) and MFI values (**Figure S4E**, 2098.3 vs. 1,424), and functional improvement of CMV-specific T-cells.

# Sirolimus-Induced Functional Improvement Correlates With IL-2R Activation on CMV-Specific T Cells

Next, to determine if IL-2, IL-7, IL-12, IL-15, and IL-21 can differentially affect the sirolimus-related expansion and functionality of CMV-specific T cells, these immune regulatory cytokines were added to the culture media independently as a different sets of cell cultures (**Figures 7A,B**, **Figure S5A**). The results showed that IL-7, IL-12, and IL-21 are not as essential as IL-2 and IL-15 for CMV-specific T-cell expansion (**Figure 7A**). In particular, the expansion of multimer+CD8<sup>+</sup> T cells was barely affected by these cytokines (frequencies<10%). Only the addition of IL-2 and IL-15 resulted in T-cell expansion, which was further impaired by sirolimus treatment (IL-2 control: 82.5% and IC50: 54.8%, IL-15 control: 70.4%, and IC50: 40.2%). Upon evaluating the effector function of CMV-specific CD8<sup>+</sup> T cells by ICS, we observed an overall increase in IFN-γ expression (**Figure 7B**, **Figure S5A**) after sirolimus treatment in the presence of IL-15. The tendency observed with IL-2 and IL-21 was comparable to that in untreated controls. However, the responses observed with IL-2 or IL-15 were more robust in terms of the total number of cells (data not shown) and CMV-specific CD8<sup>+</sup> T-cell responses determined by multimer and intracellular staining (**Figures 7A,B**, **Figure S5A**).

(±) SD. Statistical analysis: (A,B) Student's t-test and (C,D,F) two-way analysis of variance (\*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001).

Increased expression of IFN-γ corresponded to an overall increase in capacity of sirolimus-treated CMV-specific CD8<sup>+</sup> T cells to recognize and lyse A02pp65p-loaded K562 target cells and to the expression of CD107a in the presence of IL-15 (**Figures S5B,C**), although these responses were stronger with IL-2 and sirolimus (**Figures 4D,E**). These results indicate that IL-2R subunit-sharing cytokines, particularly IL-2 and IL-15, support antiviral T-cell responses during mTORC1 inhibition by sirolimus.

# Sirolimus Treatment and STAT-5 Inhibition Impair the Quality and Quantity of CMV-Specific CD8<sup>+</sup> T cells

Sirolimus-treated CMV-specific CD8<sup>+</sup> T cells were further tested for differences in functionality and S6 and STAT-5 phosphorylation after overnight STAT-5 inhibition to prove the beneficial effect of STAT-5 on functionality (**Figures 7C–F**). Following short STAT-5 inhibition, CD8<sup>+</sup> T cells displayed a decreased frequency of expanded CMV-specific multimer<sup>+</sup> T cells in comparison with cells treated with and without sirolimus alone or STAT-5 inhibitor alone in the presence of IL-2 (**Figure 7C**) or IL-15 (**Figure S5D**). The frequency of pS6 (**Figure 7D**) on CD8<sup>+</sup> T cells and its localization on multimer+CD8<sup>+</sup> T cells was lower after combined treatment as compared to other conditions (**Figure 7E**). No differences could be seen in phosphorylation of S6 in cells treated with IL-15 (**Figure S5E**).

Reduced antigen-specific expression of IFN-γ was observed following inhibition of STAT-5 alone or with sirolimus in combination with either IL-2 (**Figure 7F**) or IL-15 (**Figure S5F**). This indicated a decrease in STAT-5 function, which was further confirmed by decreased phosphorylation of STAT-5 in cells treated with either IL-2 (**Figures 7D,E**) or IL-15 (**Figure S5E**).

cells on day 7 obtained from untreated (Control 1 and 2) and sirolimus-treated cells (IC50 1 and 2) of two donors (n = 2). CD8<sup>+</sup> T cells were sorted based on their multimer specificity using high-speed flow cytometry sorters. Following overnight A02pp65p re-stimulation, total RNA was isolated and investigated by microarray analysis. Clustering and heat map analyses were performed using the Morpheus web-based tool. The data are means ± SD. The two-paired Student's t-test was used to test for statistically significant differences [non-significant (ns)].

However, IFN-γ and pSTAT-5 expression remained higher in cells treated with the combination of sirolimus and STAT-5 inhibitor than in untreated controls; expression levels of both were moderately inhibited and showed the same tendency as in cells treated with sirolimus alone. Collectively, these data suggest that IL-2R-driven STAT-5 signaling plays a major role in the improvement of antiviral T-cell responses during mTORC1 inhibition.

#### Sirolimus Differently Affects Expression of miRNAs and Effector/Target Genes

Expression of miR-21, miR-155, and miR-181a was studied since these miRNAs are known to be involved in the regulation of Tcell activation and function as well as in the expression of several target and effector molecules, such as perforin, Bcl-1, SOCS1, SHIP-1, T-bet, MAPK1/ERK, EOMES, and Ki-67. Expression levels were analyzed by RT-qPCR following overnight antigen re-stimulation on sirolimus treated or untreated CD8<sup>+</sup> T cells (**Figure 8A**).

Sirolimus reduced miR-21 expression but did not affect miR-155 (**Figure 8A**). Interestingly, while both miR-155 and miR-181a target SHIP-1 and SOCS1, miR-155 was not affected. Sirolimus treatment resulted in increased expression of miR-181a, perforin, MAPK1, Bcl-1 (one of the main targets of mTORC1 and STAT-5 signaling), Ki-67, T-bet and EOMES on CD8<sup>+</sup> T cells (**Figure 8A**). RNA-based microarray analysis was subsequently performed to confirm the RT-qPCR results and to investigate how these and other key genes are regulated in multimer-sorted CMV-specific T cells. Similar gene expression profiles were observed on treated and untreated multimer-sorted antigen-specific CD8<sup>+</sup> T cells following antigen re-stimulation (**Figure 8B**). Genes coding proteins in mTORC1 were clearly downregulated by sirolimus (FKBP5, DEPTOR), whereas those involved in IL-2R pathways (IL2R, JAK1, STAT5B, IFNG) were upregulated. Although these results support our hypothesis that the improved functionality of CMV-specific T cells strongly depends on the presence of TCR activation, co-stimulation and IL-2, further investigation is necessary to better understand and obtain more insight into this highly complex network regulating CD8<sup>+</sup> T-cell biology.

#### Improvement of Functional Responses of CD8<sup>+</sup> T-Cells Treated With Sirolimus in vivo Strongly Depends on the Presence of TCR Activation, Co-stimulation and IL-2

To further investigate the impact of sirolimus and IL-2, we analyzed peripheral blood mononuclear cells (PBMCs) from sirolimus-treated patients for the phosphorylation of S6 and STAT-5 and for the expression of effector cytokines such as IFNγ, GzB, and TNF-α following short-term polyclonal CD3/CD28 stimulation in the presence or absence of IL-2 and compared to controls (PBMCs from healthy donors). As expected, sirolimus treatment downregulated S6 phosphorylation in patients and healthy controls compared to levels in untreated healthy controls (**Figure 9A**). Although, stimulation with IL-2 alone led to an only slight increase in pS6 frequencies, the same tendency was observed in all groups. In contrast, stimulation with CD3/CD28 had no effect. Overall, pS6 was lower in patients and only slightly impaired in healthy donors treated with sirolimus than in untreated controls. Phosphorylated STAT-5 (**Figure 9A**) was slightly higher in healthy controls with or without sirolimus than in patients treated with sirolimus. IL-2 alone or in combination with CD3/CD28 cross-linking always resulted in the upregulation of pSTAT-5 frequencies; the highest in in vivo treated sirolimus patients with IL-2 alone was observed. Although no differences in IFN-γ, GzB, or TNF-α expression were measured between the in vitro treated and untreated groups, IL-2-specific production of these cytokines was slightly higher (**Figure 9B**), and was increased further by CD3/CD28 stimulation compared to unstimulated controls and to in vivo sirolimus treated patients. Taken together, these results indicate that IL-2 increases the functionality of CD8<sup>+</sup> cells treated with sirolimus in vivo, and that this effect can be further enhanced via TCR activation.

# DISCUSSION

The protein kinase mTOR acts as a multichannel processor in a cellular-nutrient-sensing network and plays a major role in integrating diverse environmental signals that affect immune cell growth, proliferation and function. The inhibition of mTOR signaling by the immunosuppressive drug sirolimus is an established therapeutic strategy in transplantation medicine (1, 39). On the one hand, only moderate effects of sirolimus monotherapy in preventing graft rejection and GvHD have been observed in the transplant setting. On the other hand, various investigators have shown that transplant patients treated with this mTOR inhibitor or got everolimus-based immunosuppressive regimen have better control of pathogen infections or reactivations and, therefore, have better clinical outcomes (1, 12, 18–25, 40).

This comprehensive study provides insight into the paradoxical effect of sirolimus on naïve and CMVpp65-specific CD8<sup>+</sup> memory T cells generated by a unique aAPC-based assay. We investigated the effects of TCR signaling and co-stimulatory signals and the role of mTORC1 signaling on memory T cells, focusing on fundamental elements of T-cell function and on the diversity of TCR repertoire, activation of signaling pathways, and expression of target and effector molecules. Surprisingly, we found that sirolimus–treated cells are not only functional but also have significantly better function than untreated controls. The paradoxical effect of sirolimus strongly depends on the presence of TCR activation with co-stimulation and IL-2.

As expected and in contrast to memory T-cell activation, the activation of naïve T cells was lower in response to allogeneic TCR-dependent stimulation, antigen-independent TCR stimulation via CD3/CD28 crosslinking, and co-stimulatory signals alone and in the presence of IL-2. These results underline the rationale behind using immunosuppressive agents such as sirolimus to prevent graft rejection and GvHD, as these transplant complications are caused by alloreactive cells residing within the naive CD45RA<sup>+</sup> T-cell compartment (41, 42).

According to recent findings, the immunostimulatory effect of sirolimus on memory antigen-experienced T cells is strongly dependent on the presence of TCR and co-stimulatory signals and can be enhanced by IL-2. To further investigate the mode of action of sirolimus on memory T cells, a CMV-specific Tcell generation model was chosen that allowed to investigate

phospho-flow cytometry. (B) Percentages of intracellular expression of IFN-γ, GzB and TNF-α in CD8<sup>+</sup> T cells were measured by multicolor flow cytometry after 5 h of stimulation. The data are shown as mean ± SD. Statistical analysis was done by two-way analysis of variance. No significant differences were determined.

direct effects on those cells. Bead-based aAPCs allowed for very robust expansion of antigen-specific memory T cells by providing signals for TCR recognition via HLA-A<sup>∗</sup> 02:01-CMVpp65 peptide complexes and co-stimulation through anti-CD28 monoclonal antibodies (mAbs) (33). Despite a decrease in the overall Tcell proliferation rate, we observed a strong increase in the frequency of CMV-specific multimer<sup>+</sup> T cells, which was of course lower in the sirolimus-treated cells than in the untreated controls. This finding highlights the selective direct positive effect of sirolimus on antigen-specific T cells and the negative effects on those cells which were not specifically activated. The dosedependent negative impact of sirolimus on CMV-specific Tcell expansion was counterbalanced by significant improvement in effector cell functionality, as determined by antigen-specific cytokine release (IFN-γ, GzB, and TNF-α) and cytotoxicity assays. Increased expression levels of classical activation markers such as CD25 and CD69 and transient expression of the classical exhaustion markers PD-1 and Lag-3 on CMV-specific T cells further strengthen the hypothesis that sirolimus has a positive immunostimulatory effect on antiviral T-cell functionality (37, 38). As expected, mTORC1 inhibition had no effect on the effector memory phenotype (CD45RA−CD62L−).

Although our results strongly underline earlier in vitro and in vivo observations (12, 18, 19, 23), this is the first work that shows that sirolimus has a direct immunostimulatory effect on the functionality of human CMV-specific CD8<sup>+</sup> T cells. Surprisingly, we found that sirolimus not only left the treated cells functional, but also significantly improved their functionality compared to that of untreated controls. Activation that allows T cells to proliferate and develop into effective antiviral CTLs relies on four essential signals from the following sources: (1) TCR stimulation, (2) co-stimulation, (3) cytokines, and (4) chemokines (1, 43). Previous studies have shown that stimulation of TCR and CD28 in resting T cells results in IL-2-driven proliferation through the activation of phosphatidylinositide 3-kinase (PI3K) and mTOR (14). The mTORC1 inhibition with sirolimus blocks not only signals 1 and 2 but also signal 3 by attenuating IL-2R signaling, thereby preventing the full activation of T cells and the optimal expression of cyclins. Conversely, recent studies revealed that the immunosuppressive effect of sirolimus on Tcell proliferation can be diminished if those signals occur (25, 26, 29). A study of sirolimus-treated T cells by Colombetti et al. nicely showed that CD3/CD28 and IL-2/IL-2R pathways for antigen-independent TCR recognition independently regulate Tcell proliferation in sirolimus-treated T cells, that both pathways are controlled to a different extent by PI3K and mTOR, and that the CD3/CD28-driven activation of T cells is abolished in the absence of IL-2. In contrast, IL-2 induced T-cell proliferation was independently regulated by these signaling molecules (44). The present study showed that T-cell activation is influenced in the same way because the immunosuppressive effect of sirolimus can only be overcome by antigen-independent TCR stimulation via CD3/CD28 co-crosslinking and IL-2.

In 2004, Slavik et al. postulated that the functional outcome of antigen-specific memory T cells is associated with strong signals due to high affinity TCR and CD28 (26, 29). Their findings are substantiated by our results suggesting the antigenic stimulation of high-affinity HLA-A<sup>∗</sup> 02:01-restricted CMV-specific TCR and co-stimulation signals from aAPCs. In addition, IL-2 converted the sirolimus-resistant functionality of T cells in our tests using an antigen-specific and CD3/CD28 crosslinking approach. To our knowledge, this study is the first showing that sirolimus has a direct selective immunosuppressive effect on naïve and immunostimulatory effect on antigen-specific T cells.

In addition, our results showed that mTORC1 inhibition with sirolimus during aAPC stimulation results in unexpectedly higher levels of CD25 (IL-2R) expression. Therefore, it is likely that the antigen-specific T cells became more susceptible to IL-2 or IL-2R subunit-sharing cytokines (e.g., IL-15), which are known to be fundamental in the maintenance and differentiation of effector T cells (45, 46). Moreover, the addition of IL-15 resulted in the same positive effect on T-cell functionality in the presence of sirolimus, whereas supplementation with IL-7, IL-12, or IL-21—cytokines known to improve T-cell function—did not overcome the immunosuppressive effect of sirolimus in our study. Antigen-independent analysis of the T-cell functionality of samples from patients treated with sirolimus and healthy control confirmed this positive effect of IL-2.

In order to generate sufficient numbers of CMV-specific T cells to perform functional assays, we opted for incomplete inhibition of T-cell function at the IC50 level (10 ng/ml), which is approximately equivalent to the typical circulating level of the drug in immunosuppressed patients. Therefore, we expected the phosphorylation of S6, a downstream target of sirolimussensitive mTORC1, to occur following stimulation. Given that the drug induced moderate phosphorylation of pS6 in addition to significantly higher phosphorylation of STAT-5, a main downstream target of IL-2, it is tempting to speculate that sirolimus activates at least one more downstream pathway that enables T cells to escape cell cycle arrest and thus maintain their effector and cytotoxic functions. This hypothesis was supported by findings of increased expression of Bcl-1 and higher phosphorylation of ERK1/2 and Akt, and was further strengthened by the findings observed after inhibition of STAT-5. In conclusion, we found that the interactions of mTORC1 and IL-2R-driven STAT-5 signaling influenced the immune balance by modulating the expansion and functionality of CMV-specific T cells. To our knowledge, this is the first study investigating the interplay between mTORC1 and IL-2R-driven STAT-5 signaling in antiviral human CMVpp65p-specific CD8<sup>+</sup> T cells.

To gain deeper insight into the mechanism of action of this mTOR inihibitor, we further investigated the mTOR inhibitor on miRNA-mediated regulation of CMV-specific T-cell function in cells treated with sirolimus. Stimulation via TCR and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling normally induces the expression of miR-21, miR-155, and miR-181, all of which are known to promote T-cell proliferation, survival and effector function (47). As sirolimus did not affect miR-155 expression in this study, the increase in the expression of the target genes SOCS1 and SHIP1 might be related to the slight downregulation of miRNA-21. Considering these results, we would expect sirolimustreated cells to show less functionality. On the other hand, increased miR-181a expression might be responsible for the observed increase in T-cell functionality. The expression of markers such as MAPK1, CD28, and CD40LG as determined by microarray analysis and the increased secretion of effector molecules such as IFN-γ and GzB support this idea. Further studies using anti-miRNA oligonucleotides to neutralize miRNA function will help to gain more insight into the highly complex network regulating CD8<sup>+</sup> T-cell biology via miRNAs and their counterparts (48).

Recent findings demonstrated the ability of mTOR to interpret signals in the immune microenvironment and to program the generation of effector vs. memory CD8<sup>+</sup> T cells via the direct line between metabolism and function (14). Our microarray analysis of multimer-sorted CMV-specific CD8<sup>+</sup> T cells confirmed this, with results showing that sirolimus induces the upregulation of genes involved in glycolysis, such as CD28, AKT, and HIF1A, which promote increases in glucose uptake, PDK1, which increases the conversion of pyruvate into lactate, and upstream glycolytic enzymes such as LDH and MYC. These results are consistent with the findings of another study demonstrating that the functionality of effector CD8<sup>+</sup> T cells is relies on glycolysis (49). In addition, it underlines the immunostimulatory effect of sirolimus on antiviral T-cell responses by modulating multiple environmental cues during antiviral T-cell expansion under the influence of mTORC1 inhibition.

Sequencing of the TCR β-chain repertoire of sorted CMVspecific T cells was performed to answer the question whether sirolimus–resistant clones may serve as a reservoir of shared and functional T cells and expand during immunosuppression. The ability of the adaptive immune system to respond to a wide variety of pathogens depends on the presence of a unique TCR repertoire reflecting the initial V(D)J recombination events shaped by the selection of self and foreign antigens presented by HLA molecules on APCs (50). We observed no selective effect of sirolimus on the highly unique clonal HLA-A<sup>∗</sup> 02:01 restricted CMV-specific TCR β-chain repertoires. Interestingly, no sharing of the TCRs between the analyzed healthy individuals was observed. We hypothesize that the αβ T cell repertoire reflects the individual history of immunological exposure, which is driven by the environmental milieu and not by the selective pressure of immunosuppression on a healthy individual. Further independent analysis on the TCR α-chain would be essential to complete these study findings and strengthen this hypothesis. Although we did not observe any effect of sirolimus on the TCR repertoires examined in this study, next-generation sequencing offers the possibility to identify drug-resistant TCR clones in patients who develop primary infection or reactivation during immunosuppressive treatment. Furthermore, such analyses will help to identify the dynamics of both αβ and γδ TCR repertoires in immunosuppressed recipients with viral complications treated with various immunosuppressive (51, 52).

The effect of sirolimus on human cells is even wider than previously expected. In this study, we show for the first time that sirolimus acts selectively on naive and memory T cells and directly on CMV-specific CD8<sup>+</sup> T cells to promote their responses to antigens. These results clearly demonstrate the benefits of sirolimus treatment for transplant patients. On the one hand, sirolimus suppresses alloreactive naïve T cells and thereby prevents the development of GvHD. On other hand, it increases the functionality of antigen-specific T cells and might therefore promote graft-vs.-infection (GvI) and graft-vs.-tumor (GvT) responses. A better understanding of the complex effects of immunosuppressive drugs that regulate effector cell functions should provide new opportunities to further individualize immunosuppressive therapy in patients with an increased risk of viral infection and/or reactivation (53). In addition, mTOR inhibition in combination with the modulation of environmental cues using agents such as IL-2 might lead to new strategies for the treatment of infectious diseases or immunosuppressive tumors.

#### AUTHOR CONTRIBUTIONS

SB helped to design the study, performed the analyses, carried out the T-cell stimulation experiments and functional assays for healthy donors and patients, did the data generation and statistical analysis, and wrote the manuscript. ST assisted with assessing the immunofluorescence staining and microscopy, contributed helpful discussions and aided to draft the manuscript. AD helped with performing the alloreactivity and CD3/CD28 crosslinking approaches and aided to draft the manuscript. SR carried out NGS experiments, contributed helpful discussion with respect to the NGS. LP provided patient material and aided to draft the manuscript. CK helped by contributing clinical requirements and discussions and aided

#### REFERENCES


to draft the manuscript. MO contributed helpful and critical discussions with respect aAPC approach and helped to draft the manuscript. RB contributed helpful and critical discussions, helped to draft the manuscript, and approved the final version of the manuscript for publication. BM-K helped to design the study, provided patient material, helped by contributing critical and valuable discussions about clinical background issues and by drafting the manuscript. BE-V conceived the study, participated in its design and coordination, designed the T-cell stimulation assays, immunoassays, and data analysis procedures, and co-wrote the manuscript.

#### FUNDING

This work was supported by the German Federal Ministry of Education and Research (reference numbers: 01EO0802, 01EO13029), the Hannover Biomedical Research School (HBRS), and the PhD program Molecular Medicine.

#### ACKNOWLEDGMENTS

We would like to thank Sarina Lukis and Marina Kramer for their excellent technical assistance. We would like to acknowledge the assistance of the Cell Sorting Core Facility of the MHH. Microarray raw data used in this publication were generated by the Research Core Unit Transcriptomics of Hannover Medical School.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02953/full#supplementary-material


cells. Immunity (1999) 11:225–30. doi: 10.1016/S1074-7613(00) 80097-7


**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 Bak, Tischer, Dragon, Ravens, Pape, Koenecke, Oelke, Blasczyk, Maecker-Kolhoff and Eiz-Vesper. 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.

# Inflammasomes: Emerging Central Players in Cancer Immunology and Immunotherapy

#### Dev Karan\*

Department of Pathology, MCW Cancer Center and Prostate Cancer Center of Excellence, Medical College of Wisconsin, Milwaukee, WI, United States

Inflammation has an established role in cancer development and progression and is a key player in regulating the entry and exit of immune cells in the tumor microenvironment, mounting a significant impact on anti-tumor immunity. Recent studies have shed light on the role of inflammasomes in the regulation of inflammation with a focus on the subsequent effects on the immunobiology of tumors. To generate strong anti-tumor immunity, cross-talk between innate, and adaptive immune cells is necessary. Interestingly, inflammasome bridges both arms of the immune system representing a unique opportunity to manipulate the role of inflammation in favor of tumor suppression. In this review, we discuss the impact of inflammasomes on the regulation of the levels of inflammatory cytokines-chemokines and the efficacy of immunotherapy response in cancer treatment.

Keywords: inflammation, inflammasome, immunotherapy, cancer, tumor microenvironment

#### INTRODUCTION

Tumor development is a complex, multistep process wherein early-stage tumor cells undergo a series of genetic/epigenetic changes and modify the surrounding environment by secreting a plethora of cytokines-chemokines for their advantage. This progressively transformed tumor microenvironment (TME) is infiltrated by different immune cell types that are likely to kill early-stage tumor cells. However, the success or failure of the immune cells to kill tumor cells and to inhibit abnormal cell growth depends on a variety of events occurring at the tumor site. Investigating such interactions both at the cellular and molecular levels in the TME would provide rationale for the development of new treatment approaches.

The role of inflammation is well-accepted in the growth and development of tumors, and it constitutes a leading hallmark in cancer development (1, 2). Inflammation-associated cytokines-chemokines within the TME are secreted by various cell types including tumor and immune cells (residual and infiltrated). Some of the cytokines such as IL-10 and TGF-β suppress the immune cell function and help tumor cells evade the immune system, while cytokines like IL-12, IFN-γ, and TNF-α support immunological functions by enhancing anti-tumor immunity. However, an imbalance in the cytokines-chemokines influx and increased inflammation in the TME results in a tumor-promoting environment. In this review, we provide an update on the potential of targeting inflammasomes to manipulate the role of inflammation in the TME, thereby suppressing tumor promotion activities.

#### Edited by:

Anil Shanker, Meharry Medical College, United States

#### Reviewed by:

Hrishikesh Pandit, National Cancer Institute at Frederick, United States Dalil Hannani, UMR5525 Techniques de l'Ingénierie Médicale et de la Complexité Informatique, Mathématiques et Applications, Grenoble (TIMC-IMAG), France

> \*Correspondence: Dev Karan dkaran@mcw.edu

#### Specialty section:

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

Received: 29 September 2018 Accepted: 07 December 2018 Published: 20 December 2018

#### Citation:

Karan D (2018) Inflammasomes: Emerging Central Players in Cancer Immunology and Immunotherapy. Front. Immunol. 9:3028. doi: 10.3389/fimmu.2018.03028

**85**

# INFLAMMASOMES AND CANCER

Inflammasomes are multi-protein complexes consisting of nucleotide-binding and oligomerization domain (NOD)-like receptor (NLR), the adaptor protein (ASC: apoptosis-associated speck-like protein containing CARD), and procaspase-1. Among NLRs, NLRP1, NLRP2, NLRP3, NLRP4, NLRP6, NLRP12, and NLRP14 are known to form multi-protein complexes and are grouped into a canonical inflammasome pathway (3, 4). Absent in Melanoma 2 (AIM2), also a member of the canonical inflammasome pathway, recognizes aberrant cytoplasmic dsDNA and induces cytokine maturation, release, and pyroptosis (5). The non-canonical inflammasomes include a complex that consists of CARD9, Malt1, Bcl-10, caspase-8, and ASC (6). During non-canonical inflammasome signaling, caspase-1 is known to cleave gasdermin D (GSDMD), resulting in GSDMDp30 pores in macrophages, releasing IL-1β and IL-18, and inducing pyroptosis. The non-canonical pathway involving NLRP3 inflammasome leading to murine caspase-11 or human caspase-4,-5 maturation is also reported (7). However, later studies confirmed lipopolysaccharides (LPS)-induced toll-like receptor (TLR4)-mediated signaling for activation of caspase-4, -5, and -11 (8–11).

The canonical pathway of inflammasome activation is regulated at both the transcriptional and post-translational levels following two signals (4). The first signal, referred to as the priming signal, is associated with the upregulation of NLRP3 induced by the TLR/nuclear factor (NF)-κB pathway. Alternatively, mitochondrial-derived reactive oxygen species (ROS) can also prime NLRP3 inflammasome activation utilizing TLR4/MyD88 signaling (12). During this priming process, the precursor forms of interleukin (IL) IL-1β, (pro-IL-1β), and IL-18 (pro-IL-18) are also upregulated. The second signal is exerted by various stimuli including pathogen-associated molecular patterns (PAMPs), damage-associated molecular patterns (DAMPs), adenosine triphosphate (ATP), uric acid crystals, or other toxins facilitating the functional assembly of the NLRP3 associated multi-protein complex consisting of ASC and pro-caspase-1 (13, 14). Upon activation, inflammasome assembly regulates the activity of caspase-1, responsible for the proteolytic cleavage of pro-IL-1β and pro-IL-18 to its mature and bioactive forms, hence inducing a variety of biological effects.

Current studies have highlighted the role of inflammasomemediated inflammation in cancer (2, 15). The most studied and best characterized inflammasome, NLRP3, is an emerging, key player in the development and progression of cancer, and an increased expression of NLRP3 has been associated with multiple cancer types (16–19). Activation of NLRP3 inflammasome has also been shown to promote inflammationinduced tumor growth and metastasis in head and neck cancer and oral squamous carcinoma (20, 21). Contrary to the effects of NLRP3 in the promotion of cancer, studies in colorectal cancer demonstrated that increased NLRP3 inhibits colorectal metastasis. It was demonstrated that IL-18 secretion downstream of NLRP3 inflammasome is associated with increased interferongamma (IFN-γ) production and activation of signal transducer and activator of transcription (STAT1) involved in protection against colorectal tumorigenesis (22). Additional mechanisms showed that IL-18 primed natural killer (NK) cells trigger FasL-induced apoptosis in the tumor (23). In a pre-clinical model of colitis-associated cancer (CAC), mice lacking the inflammasome adaptor proteins PYCARD (ASC) and caspase-1 demonstrated increased disease outcome, morbidity, and polyp formation. Expression of inflammasome component NLRP3 was also negatively associated with the progression of hepatocellular carcinoma (24). Immunohistochemical analysis in prostate biopsies showed almost a uniform expression of NLRP3 and did not reveal any association with prostate cancer progression (15).

Besides NLRP3, studies describing the role of other inflammasome components are limited. Melanoma studies demonstrated the role of NLRP1 in tumor promotion by increasing inflammasome activity and suppressing apoptosis in metastatic melanoma (25). Due to limited studies in the literature, the role of NLRC4 seems inconsistent. One study demonstrated a negative correlation between NLRC4 and colitisassociated tumors. Utilizing azoxymethane (AOM)/dextran sodium sulfate (DSS) model, it was shown that NLRC4 knockout mice developed increased tumor volume (26). However, in another study, yet with the same model system, no difference in tumor growth was found between NLRC4-deficient and wild-type mice (26, 27). Similarly, the absence of NLRP6 accelerated colitis-associated tumors in mice, while its presence was demonstrated to suppress inflammation and carcinogenesis. It was found that NLRP6 helps to preserve the integrity of epithelial barriers and hence prevents adenoma formation (28). Immunohistochemical analysis of archival prostate biopsy specimens showed a significant increase in NLRP12 expression in prostate cancer tissues as compared to that in benign tissues (15). On the contrary, it was demonstrated that NLRP12 suppresses colon inflammation and tumorigenesis through negative regulation of NF-κB signaling (29, 30). Increased activation of AIM2 inflammasome is associated with the early course of acute pancreatitis (31). AIM2 inflammasome is also reported to play a critical role in the development of human prostatic diseases (32). AIM2, an interferon (IFN) inducible protein, is constitutively down-regulated in prostate cancer, however, IFN-induced AIM2 inflammasome activation leads to increased production of IL-1β and IL-18 in prostate cancer cell lines.

Emerging studies on the expression of inflammasome components ASC and caspase-1 also support the critical role of these molecules in tumor growth and development. In many cancer types, including prostate, breast, lung, and glioblastoma, ASC is downregulated due to hypermethylation (33–36). ASC contributes to the process of apoptosis and functions as a Bax adaptor by translocating Bax to mitochondria (37). In fact, ASC is essential in bridging the activity of NLRP inflammasomes and pro-caspase-1. ASC exerts its functional activity by translocating from the nucleus to the cytoplasm and localizes with the inflammasome components NLRP and pro-caspase-1 (38). Taken together, the multi-protein complex of inflammasomes drives a cascade of pro-inflammatory cytokines regulating various cellular activities.



HNCC, Head and neck cell cancer; OSCC, Oral squamous cell carcinoma; HCC, Hepatocellular carcinoma; GBM, Glioblastoma.

IL-1 is the most extensively characterized inflammasomerelated cytokine promoting cancer, whereas IL-1β is a pleiotropic inflammatory cytokine associated with cell proliferation, differentiation, tissue regeneration, and immune cell regulation. Additionally, the role of IL-1β in angiogenesis, tumor promotion, metastasis, resistance to chemotherapy, and immunosuppression are well-described (39, 40). Several other members of the IL-1β family, including IL-18 and IL-33, are processed and activated by caspase-1, a driving component of inflammasome activity (41). Upregulation of IL-33 has been reported in tumor growth and metastasis in lung, colorectal, and gastric cancer (42). An increased serum level of IL-33 is associated with poor prognosis in breast and lung cancer (43–45). Thus, the inflammasomemediated production of pro-inflammatory cytokines has been identified as a critical modulator of disease outcome. Multiple studies have explored the role of inflammasomes in carcinogenesis and anti-tumor activities with conflicting observations (46–50). Pro-tumor and anti-tumor role of NLRP inflammasome is summarized in **Table 1**. We observed that in a majority of tumor types, increased level of NLRP contributes to tumor progression while in colitis-associated tumors, NLRP showed anti-tumor activity. Such a dissociated role for the NLRP inflammasome in colitis-associated tumors could be attributed to microbiota, which orchestrate the colonic microenvironment in association with inflammasomes. However, further studies are needed to determine their tissue-specific functional activation of inflammasome components.

#### INFLAMMASOMES IN MYELOID CELLS

Initially, NLRs were thought to be expressed in innate immune cells of monocyte lineage. However, NLRP3 is central to inflammasome research and is expressed in multiple cell types, including monocytes, macrophages, granulocytes, dendritic cells (DCs), epithelial cells, and osteoblasts (51–54). NLRP3 inflammasome is activated by a number of DAMPs and PAMPs recognized by TLRs, which signal via myeloid differentiation marker (MyD88) and Toll-IL-1 receptor (TIR) domain-containing adaptor-inducing interferons (TRIF), and pro-inflammatory cytokines that direct the adaptive immune response (55). Inflammasomes are also an integral part of adaptive Th1 cell response, where assembly of NLRP3 inflammasome is shown in CD4<sup>+</sup> T cells leading to caspase-1 dependent IL-1β secretion (56). In fact, DCs naturally express TLRs, a family of pattern recognition receptors (PRR), which recognizes PAMPs and DAMPs on the cell surface, whereas NLRs serve as cytosolic sensors (57). In the event of danger signals, NLR and TLR synergize to expand the maturation of DCs, migration, antigen presenting function, and adaptive immune system activation. However, in the absence of TLR involvement, inflammasome activation stimulates an immunosuppressive behavior in DCs (57).

The information on the molecular profiling of other inflammasome components is sporadic. NLRP6 is expressed in myeloid cells such as granulocytes, dendritic cells and macrophages and is a potential regulator of innate immunity, wherein lack of NLRP6 protects mice against bacterial pathogens, which is attributed to an increased number of monocytes and neutrophils (58, 59). In response to microbial infection, assembly of NLRP7 has been shown in human macrophages (60, 61). Lack of NLRP10 has been associated with impairment in DC functions and initiation of adaptive immunity (62, 63). Bacterial pathogens also activate AIM2 inflammasome in macrophages and DCs, resulting in caspase-1 activation and inducing pyroptotic cell death to control bacterial infection (64, 65). These studies highlight the emerging importance of various inflammasomes in the cells of myeloid origin which are helpful in maintaining homeostasis and regulating inflammation, infection, and immunity.

### EMERGING CONCEPT OF INFLAMMASOMES IN CANCER IMMUNOTHERAPY

Cancer immunotherapy has evolved considerably from approaches such as cell-specific targeting, TME manipulation, systemic inhibition of immune suppressor cells to immune check-point inhibitors, and their subsequent combinations. However, T cells and natural killer (NK) cells remain the primary source of ammunition for targeting cancer. T cells are viewed as most suitable for antigen-specific tumor targeting while NK cells kill both tumor cells and virally infected cells. In the development of effector T cells and NK cells, antigen presenting cells (APCs) remain central in augmenting strong anti-tumor response. To generate antigen-specific effector T cell response, APCs such as dendritic cells (DCs) present the antigen to the cytotoxic T cells (CTL), known as signal-I, which is accompanied by signal-II associated with co-stimulatory molecules. Combination of these two signals leads to fully activated T cells with anti-tumor potential. However, an additional signal-III is also necessary for the proliferation and differentiation of effector cells with enhanced anti-tumor immunity (66). DCs also perform a critical task with regard to translating the interplay between innate and adaptive immunity, and these interactions are being utilized to improve current anti-cancer immunotherapies. While previous studies were more focused on T cell-based immune targeted therapies, the potential of NK cell-based therapies are well-recognized and are advancing the field of immunotherapy in conjunction with T cells. Recent studies showed that NK cells are also regulatory cells engaged in reciprocal interactions with dendritic cells, macrophages, T cells, and endothelial cells (67, 68). There is a strong triangular relationship, directly or indirectly between APCs, NK cells, and T cells. In fact, an optimum modulation of APCs (e.g., DCs) is critical to generate highly tumor-reactive NK or T cells.

A divergent process of immune cell (including DCs) mobilization, which is secretion of various cytokines, plays a significant role in the generation of anti-tumor immunity. Cytokines and chemokines expressed by DCs have a significant impact on the development as well as recruitment and priming of T helper cells. Additionally, the presence of IFN-α activates NK cells and the Th1 cytokines such as IFN-γ and IL-12, which helps in augmenting the antigen-specific immunity emphasizing the critical role of cytokines in immune cell regulations and tumor immunity. IFN-γ supports anti-tumor activities in multiple ways by upregulating major histocompatibility complex-I (MHC-I) in tumor cells, inhibiting the process of angiogenesis and tumor cell proliferation, and at the same time augmenting cytotoxic effector cell functions of CTLs, NK cells, macrophages, and CD4<sup>+</sup> T cells polarization (69). On the contrary, IFN-γ has also been associated with tumor promotion by supporting the resistance of tumor cells to immune cell-mediated killing (70). Similarly, TNF-α mediates anti-tumor immunity through simultaneous recruitment and activation of macrophages and DCs. However, dysregulated TNF-α signaling is also associated with the promotion of tumor cell growth via the mediation of epithelial-mesenchymal transition (EMT) (71). Thus, the regulation of pro-inflammatory cytokines following inflammasome activation will have a potential impact on immune cell interactions and the process of differentiation.

The central role of IL-33, a member of the IL-1 family, in inducing tumor-promoting type 2 responses has recently gained attention. Treatment of IL-33 in tumor-bearing mice impairs the functional activity of NK cells and dendritic cells and influences macrophages to M2 polarization, thus, suppressing innate and adaptive anti-tumor immunity (42). Administration of IL-33 in a murine model of breast cancer resulted in increased tumor growth and development of metastases, which correlated with increased intra-tumoral numbers of IL-13-producing innate lymphoid cells (ILCs), IL-13 receptor 1-expressing myeloidderived suppressor cells (MDSCs), and regulatory T cells (Tregs) (72). IL-13 has been shown to activate tumor-promoting MDSCs and their production of anti-inflammatory transforming growth factor-beta (TGF-β) (73). In addition, IL-13 can polarize macrophages toward a pro-tumorigenic M2 phenotype and actively participate in immune evasion (74). Therefore, a tight regulation of cytokines-chemokines will help in the maintenance of homeostatic balance and is potentially regulated via inflammasomes.

Since inflammasomes influence the production of cytokines and immune cells differentiation by regulating the functions of APCs, it seems obvious that the role of inflammasomes could directly be associated with events at the site of the tumor. The role of immune suppressor cells such as T-regs, MDSCs, and tumor-associated macrophages (TAMs) is wellestablished in the promotion of tumor growth and metastasis. These cells suppress the immune effector cell functions in multiple ways and hamper the clinical impact of immunotherapy. Therefore, chemotherapeutic drugs are used for selective killing of immune suppressor cells to enhance the clinical impact of immunotherapy approaches. Further substantiating the role of inflammasomes in the immunobiology of cancer is the significantly higher activation of NLRP3 inflammasome observed during chemotherapy treatment. It is suggested that dying tumor cells release ATP that is sensed by the P2X7 receptor of DCs leading to NLRP3 activation and is associated with chemo-resistant tumor growth. Additionally, chemotherapy triggers cathepsin B release in myeloid-derived suppressor cells, activating the NLRP3 inflammasome leading to MDSCderived IL-1β and angiogenesis and promoting tumor growth and metastasis (75, 76). IL-1β is also known to enhance the production of IL-17 by CD4<sup>+</sup> T cells, which in turn favors angiogenesis and tumor growth (77).

It is interesting to note that the inflammasome component NLRP3 also impairs the impact of anti-tumor vaccine. Dendritic cell-based vaccination of NLRP3-deficient mice bearing melanoma tumors showed a significant increase in survival as compared to their respective controls. This improved survival was attributed to low numbers of MDSCs in NLRP3-deficient tumors suggesting the role of NLRP3 in facilitating the migration of MDSCs to the TME (78). In a model of chemically-induced carcinogenesis, mice lacking NLRP3 showed low tumor burden and suppression of metastasis (79). NLRP3-deficiency induces NK cell infiltration and is associated with increased production of chemokines CCL5 and CXCL9 promoting the anti-metastatic activity of NK cells. Similarly, in a breast cancer model, deficiency of inflammasome components (NLRP3 or caspase-1 knockout) reduced tumor growth and metastasis and was correlated with reduced infiltration of MDSCs within the tumors (80). In a mouse model of pancreatic ductal adenocarcinoma (PDA), it was observed that NLRP3 promotes the expansion of immune-suppressive macrophages. NLRP3 signaling in macrophages drives the differentiation of CD4<sup>+</sup> T cells into tumor-promoting Th2, Th17, and T-regulatory cell types, while suppressing Th1 cell polarization and cytotoxic CD8<sup>+</sup> T cell activation. Inflammasome signaling also modulates IL-12 secretion, hence affecting T helper cell polarization. Subsequent inhibition of NLRP3 signaling or mice deficient in inflammasome components (NLRP3, ASC, or caspase-1) showed immunogenic reprogramming of innate and adaptive immunity within the TME and were protected against PDA (81). These preclinical studies provide a proof-of-concept that targeting the inflammasome reduces the quality and quantity of immune suppressor cells and hence inhibits tumor growth and metastasis. Therefore, targeted inhibition of inflammasome activation will help to generate a balancing act of pro-inflammatory cytokinechemokine in the TME and is likely to restore the immune surveillance, augmenting anti-tumor immunity.

# INFLAMMASOME INHIBITORS AND THERAPEUTIC INTERVENTION

So far, we described that the absence of inflammasomes helped in protecting mice against tumor growth and was associated with reduced immune suppressor cells in the circulation as well as in the tumor microenvironment. However, the use of synthetic compounds or small molecule inhibitors for in vivo targeting of NLRP3 and examining the infiltrating lymphocytes in the tumor bed are just at the initial stage. Several inhibitors have been proposed to target NLRP3 inflammasome activation in vitro and in vivo, resulting in reduced levels of IL-1β, and inflammasome components associated with anti-inflammatory diseases (82, 83). Using in vitro studies in cell lines, we also showed that a dietary agent withaferin-A disintegrates the inflammasome complex and modulates multiple cytokines and chemokines associated with inflammation and cancer (13). Interestingly, intraperitoneal administration of MCC950 to block NLRP3 inflammasome activity significantly reduces the number of T-regs, MDSCs, and TAMs, and increases the numbers of CD4<sup>+</sup> and CD8<sup>+</sup> T cells in mice. MCC950 inhibited NLRP3 activation and caspase-1 dependent IL-1β processing, and hence improved anti-tumor immune response in head and neck squamous cell carcinoma (84). Another small molecule inhibitor, andrographolide, protected mice against colitis-associated cancer (CAC) through mitophagy-mediated NLRP3 inhibition (85). Andrographolide was administered intragastrically daily following AOM/DSS-induced CAC in C57BL/6 mice and significantly attenuated colitis progression and tumor burden. Although regulation of immune cells was not determined, the protection of mice against CAC was associated with the disruption of inflammasome assembly and reduced IL-1β secretion. This is in contrast with the CAC studies in NRLP3- or caspase-1-deficient mice, which are susceptible to aggressive AOM/DSS-induced CAC. Since inflammasomes are involved in host defense mechanism against infection and autoinflammatory diseases, a complete knockout of inflammasome components (e.g., NLRP3, ASC, or caspase-1) may cause an imbalance in intestinal microbiota and/or cytokines-chemokines profile promoting an aggressive colitis progression. This clearly indicates that the inhibition of inflammasome activity using small molecules has preventive and

(A), and a summary of events involved in tumor growth and metastasis and a proposed mechanism to block inflammasome activity (B). IL-1β and IL-18 are well-known primary targets; however, multiple cytokines-chemokines could be simultaneously reduced in the tumor microenvironment with subsequent inhibition of immune suppressor cells, providing a space to enhance anti-tumor immunity. WFA: Withaferin-A.

therapeutic potential in reducing pro-inflammatory events at the tumor site and providing an opportunity to boost the efficacy of immunological manipulations in cancer treatment.

#### CONCLUSIONS

Manipulation of the body's immune system has emerged as one of the most promising approaches leading the way for successful immunotherapies targeting various cancers. In addition to classical approaches of vaccination-induced antitumor immunity, development of programmed T cells, transfer of apoptosis-resistant T cells, and metabolically more active or cancer-specific chimeric antigen-receptor (CAR) T cells are being tested in clinics. However, the cytokines-chemokines profile is critical in maintaining the anti-tumor immunity as well as the proliferation and interaction of immune cells. Multiple studies have analyzed the impact of NLRP3 inflammasomes in autoimmune and auto-inflammatory diseases using NLRP3 genespecific knockout mice. However, studies are emerging that aim to determine the role of NLRP3 in the regulation of immune cells in a tumor setting. As described above, inflammasome components regulate the cytokines-chemokines profile, bridge the innate, and adaptive immune responses, and hence, impact immune cell functions. While inflammasome components revealed an anti-tumor effect in colon and CAC, it is evident

#### REFERENCES


that in most of the solid tumors, the NLRP inflammasome is associated with tumor promotion and metastasis (**Table 1**). It is likely that small molecule based targeting of specific inflammasome components (NLRP1, NLRP3, NLRC4, NLRP6, NLRP12, or caspase-1), rather than using gene-specific knockout strains, might divulge useful information about the role of inflammasomes in mice protected against colon and CAC and hence strengthening the role of inflammasomes in tumorigenesis. Therefore, targeted inhibition of inflammasomes might provide a novel opportunity to manipulate the immunobiology of cancer to augment the efficacy of immunotherapeutic approaches. A schematic representation of inflammasome-centralized crosstalk between APC, NK, and T cells and the potential events in cytokine-chemokine modulations following inhibition of NLRP3 inflammasome components, is summarized in **Figure 1**.

#### AUTHOR CONTRIBUTIONS

DK conceived the concept, searched the literature and wrote the manuscript.

#### ACKNOWLEDGMENTS

We gratefully acknowledge the funding support by the National Cancer Institute (CA204786).


against colorectal tumor formation. J Immunol. (2010) 185:4912–20. doi: 10.4049/jimmunol.1002046


regulation of dendritic cell-derived IL-12 release. Front Immunol. (2017) 8:1462. doi: 10.3389/fimmu.2017.01462


**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 Karan. 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.

# Lymphocytes in Cellular Therapy: Functional Regulation of CAR T Cells

Alka Dwivedi † , Atharva Karulkar † , Sarbari Ghosh† , Afrin Rafiq† and Rahul Purwar\*

*Department of Biosciences and Bioengineering, Indian Institute of Technology Bombay, Mumbai, India*

#### Edited by:

*Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States*

#### Reviewed by:

*Avery Dexter Posey Jr., University of Pennsylvania, United States Rajshekhar Alli, St. Jude Children's Research Hospital, United States*

> \*Correspondence: *Rahul Purwar purwarrahul@iitb.ac.in*

*†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: *12 September 2018* Accepted: *27 December 2018* Published: *18 January 2019*

#### Citation:

*Dwivedi A, Karulkar A, Ghosh S, Rafiq A and Purwar R (2019) Lymphocytes in Cellular Therapy: Functional Regulation of CAR T Cells. Front. Immunol. 9:3180. doi: 10.3389/fimmu.2018.03180* Lymphocytes especially autologous T cells have been used for the treatment of numerous indications including cancers, autoimmune disorders and infectious diseases. Very recently, FDA approved Chimeric Antigen Receptor T cells (CAR T cells) therapy for relapse and refractory CD19+ B cell acute lymphoblastic leukemia (r/r B-ALL) and r/r diffuse large B cell lymphoma (r/r DLBCL) upon their remarkable success in multiple Phase I-II clinical trials. While CAR T cells are considered as major breakthrough in the field of cancer immunotherapy, the regulation of CAR T cells remains poorly understood. In this review we will discuss the strategies that regulate the CAR T cells efficacy and persistence with focus on roles of different structural component of CAR construct. Different domains of CAR construct, for example, antigen binding domain, hinge, transmembrane, and signaling domain as well as immune-regulatory cytokines have significant impact on CAR T cell efficacy. Finally, this review will highlight the strategies that will promote CAR T cells efficacy and will reduce the toxicity.

Keywords: chimeric antigen receptor, cancer immunotherapy, immunoregulation, anti-tumor efficacy, cytokines

# INTRODUCTION

Chimeric antigen receptor T cells (CAR T cells) have achieved remarkable success in the field of cancer tumor immunotherapy since last decade (1). Promising clinical outcomes were observed in case of hematological malignancies leading to FDA-approval of KymriahTM and YescartaTM for r/r B-ALL in pediatric and young adults and adult patients with large B cell lymphoma, respectively. This has led to a paradigm shift in the field of cancer immunotherapy especially in treatment of some hematological cancers across the globe. The enormous success of CAR T cells in hematological malignancies is attributed to numerous factors, most important being the choice of CD19 expression on all B cells (2, 3). Other factors are easy sampling of the tumor and trouble free homing of the T cells to hematologic organs such as blood, bone marrow and the lymph nodes (4). On the contrary, efficacy of CAR T cells targeting solid tumors is still in its infancy due to multiple challenges such as lack of unique tumor associated antigen (TAAs), inefficient T cell homing to the tumor bed and due to immunosuppressive tumor microenvironment. These factors lead to limited persistence and sub-optimal efficacy of the CAR T cells in solid tumor settings (5).

CARs consist of four main domains, namely ectodomain for specific target antigen recognition and endodomain that provides costimulatory and activation signals. These two domains are connected by hinge and transmembrane domain (**Figure 1**). The major advantages of CAR T cells over other cell based therapies are (1) killing of tumor targets in a MHC independent manner and thereby overcoming certain tumor escape mechanisms such as MHC-I down regulation and faulty antigen processing, (2) engineering of multiple anti-tumor immuno-modulators, and (3) targeting wide array of antigens (protein, carbohydrate, and glycolipid).

While CAR T cells are effective in achieving long-term remission in certain types of malignancies, the major challenge remains in controlling CAR T cells in case of dysregulated activation. Hence, our major focus in this review is to elucidate the structure-function relationship of different components of CAR construct on CAR T cells persistence and efficacy.

# Impact of ScFv Ectodomain on CAR T Cell Functions

The scFv ectodomain, which plays important role in the efficacy and safety of CAR T cells, is a smallest synthetic functional module containing variable heavy (VH) as well as variable light (VL) chain portion of an antibody linked with a long flexible linker. Most commonly used linker in several CAR construct is (Gly4Ser)3. Glycine residues provide flexibility and serine residues provide solubility. This renders a properly folded scFv thereby maintaining antigen-binding capability of the parental IgG. Differences in orientation of the VH, Linker and VL may affect the scFv's affinity and specificity (6). scFvs are critical for antigen specific CAR T cell activation. Non-specific or crossreactive CARs can augment the off-target toxicity and inefficient CAR T cell activation. Hence, considerable efforts are needed for scFv designing and its characterization.

Two approaches especially animal immunization and surface display methods have been widely used for designing and screening of scFv. Although the most effective method is immunization, in case of well-conserved antigen there has been reports of generation of neutralizing antibodies due to cross reactivity (7).

There are various antibody display technologies such as phage display, cell display, ribosomal display, mRNA display and DNA display, out of which, phage display is one of the commonly used screening method (8–10). Display methodologies such as phage display overcomes the limitation of immune tolerance since they are based on in vitro selection from naïve or immune libraries (11). Yeast surface display emerged as an alternative technology to phage display, generating 10<sup>8</sup> − 10<sup>9</sup> library members. These antibodies have better affinity and specificity profiles through combination of library screening by flowcytometry and affinity maturation by in vitro codon variation or mating mediated chain shuffling (12, 13). In recent years high throughput eukaryotic cell display technologies have been successfully utilized. The advantage of this technology is real time analysis and characterization of library along with machineries for proper folding before being displayed on the surface of the cell. High throughput display technologies creates antibody libraries from which antibody fragments or domains can be selected for better effector function, tissue penetration and pharmacokinetics (14). Therefore, in order to cater the screening of antigen binding of scFv domains in CAR, either of the above methods have been utilized and have a significant role in deciding the CAR T cell efficacy.

The four important characteristics of scFv are immunogenicity, affinity, specificity, and its binding epitope. The monoclonal antibodies (mAbs) obtained from murine hybridomas were found to be immunonogenic in humans which resulted in low efficacy and immediate elimination from circulation (15, 16). They also showed systemic inflammatory responses resulting in serious physiological complications. Hence humanization of scFv can help to enhance safety and therapeutic potential of a CAR. Anti-folate receptor α (FRα) CAR T cells were developed against metastatic ovarian cancer using MOv18scFv which is a murine mAb for FRα. But, the CAR T cells showed poor persistence and anti-tumor efficacy (15, 17). In another study involving mesothelin-targeted CAR T cells containing SS1 (murine scFv), anaphylactic shock was observed in a patient. This was probably promoted by IgE antibodies specific for murine scFvs. This further indicates potential immunogenicity of murine scFv containing CARs (16). These CARs showed less in vivo persistence along with poor anti-tumor efficacy. Less immunogenicity was observed due to humanization resulting in enhanced persistence and safety of CAR T cells. A low affinity but highly specific CAR for epidermal growth factor receptor variant III (EGFRvIII) was humanized and included in the second-generation CAR T cells containing EGFRvIII scFv, 4-1BB and CD3ζ domains. Patients infused with this CAR showed minimum off-target toxicity and decreased cytokine release syndrome (18). The above humanized CARs showed better persistence and functionality but they still pose a risk of off-tumor toxicity owing to the 5% residual mouse sequences. This leads to the necessity of developing fully humanized scFvs, either from phage display or transgenic mouse models. In this connection, M28z CAR, consisting of m912 scFv (fully human anti-mesothelin mAb) was generated to resolve the immunogenicity issue which resulted in long term complete remission as reported in in vivo tumor models (19). Few other humanized CARs such as anti-FRα CAR for ovarian cancer and anti-CD22 CAR derived from m971 are in clinical trials (2, 3, 20, 21). With these advantages of using humanized scFv derived CARs, a case report of anti-HER2 CAR T cells containing scFv from trastuzumab (humanized mAb-herceptin) showed exceptional fatality with dosage of 1 × 10<sup>10</sup> cells/infusion (22). In contrast to this, the patients receiving a low dose (1 × 10<sup>8</sup> cells/m<sup>2</sup> ) of anti-HER2 CAR T cells derived from murine clone FRP5 showed increased tolerance along with minimum toxicity (23). In response to this observation, the change in epitope binding affinity and avidity might have an impact on the efficacy and toxicity of the anti-HER2 CAR T cells. The epitope of HER-2 recognition is distinct for trastuzumab (derived from 4D5 clone) in comparison to murine FRP5 clone. Other factors which might have role in reduced toxicity are the T cells dosage, lack of lymphodepletion regime and less persistence of murine FRP5

anti-HER-2 CAR over trastuzumab containing CAR T cells. However, due to involvement of multiple factors in treatment, it remains unclear to identify the exact reason of the exceptional fatality using humanized scFv. Tumor associated antigens (TAAs) are the prominent targets for immunotherapy that are highly expressed on tumor tissue and also expressed at lower level in healthy tissues. This leads to unwanted recognition and sometimes life threatening toxicity. Therefore, the scFv selection is crucial while designing CARs in order to discriminate between tumor cells and normal tissues. To avoid these complications, the interactions between CAR scFvs and target antigens should be carefully understood and a cross reactivity study should be performed before finalization of scFv. One approach is to increase the scFv affinity toward target antigens. For example, CARs containing high affinity scFvs for receptor tyrosine kinase like orphan receptor I (ROR1) and folate receptor β (FRβ) have shown superior effector functions than CARs with low affinity scFvs (24, 25).

In addition, a fine balance between specificity and affinity for the target antigen also plays a determining role in CAR T cell function. Affinity tuned CARs having more specificity and less affinity showed high therapeutic index (26). Further, epitope positioning also determines the efficiency of CAR T cell activation. For example, scFv recognizing a CD22 epitope, lying proximal to B cell plasma membrane showed enhanced anti-tumor function as compared to one recognizing membranedistal epitope (27). Therefore, the approach of incorporating a flexible linker in the CAR construct can modulate the antigen binding.

Upon the long term follow-up of the r/r B-ALL patients, it was observed that the tumors have acquired resistance to anti-CD19 CAR T cells (28–30). Possible reasons for the resistance toward anti-CD19 CAR T cell could be: (1) loss of target antigen**, (**2) exhaustion and lesser persistence of anti-CD19 CAR T cells (3) immunosuppressive tumor microenvironment (31, 32). To overcome the loss of antigen target due to mono-antigen specific CARs, bi-specific CARs have been designed to recognize two antigens in a true Boolean OR-gate fashion (i.e., either of the two antigens binding should be sufficient to trigger robust T-cell output). The studies have shown that in a high-disease burden setting, the bispecific CD19-CD20 CARs CAR proved both effective and less toxic than single CARs in pre-clinical settings(33, 34) Similarly, bispecific CAR has been designed to target both human epidermal growth receptor 2 (HER2) and IL13Rα2 and it showed enhanced potency and anti-tumor activity in vivo compared to two separate CARs (35). There are multiple on-going clinical trials in children and adults using bispecific CAR T cells (NCT03241940, NCT03233854, and NCT03448393). As reported in Phase 1 trial of anti-CD19/CD22 CAR T cells, anti-leukemic activity was observed with complete remission in 5/5 patients with CD19 dim/neg B-ALL for a duration of 6 months (20). Long term follow-up of these clinical trials will give the insights on superiority of bispecific CAR strategies over current monospecific CAR T cells.

Hence, these four aspects of scFv (immunogenicity, specificity, affinity, target positioning) are vital in determining the safety and functional efficacy of CAR T cells.

# Regulation by Hinge Region

Hinge region connects the ectodomain and the transmembrane domain of a CAR. Amino acid fragments from CD8α, IgG1, and IgG4 are the most common hinge used in majority of the CARs. Hinge region can affect the CAR function by providing flexibility and length to a CAR. Function of hinge region may vary depending on the targeted antigen. For example, hinge-less CARs against different antigens like CD19, carcinoembryonic antigen (CEA), neural cell adhesion molecule (NCAM) and 5T4 showed different effector functions compared to CARs having CH2CH3 hinge adapted from IgG1. CAR T cells against CD19 and 5T4 showed enhanced functional efficacy after adding CH2CH3 hinge in contrast to CEA and NCAM CAR T cells (36). As described in scFv section accessibility of a target antigen can modulate the efficacy of a CAR provided by a flexible linker as well as hinge region. Investigators have seen that in case of anti-CD22 CAR having a spacer derived from IgG1 Fc receptor impacts the positions of the targeted antigen with respect to cell surface and leads to increased efficacy (27). Moreover, length of hinge region is also critical for CAR T cells efficacy, shorter hinge region showed enhanced antitumor efficacy against ROR1 antigen compared with longer hinge (37). Very recently mesothelin CAR containing IgG4 hinge have shown higher efficacy and CAR T cell proliferation than the CAR without hinge region by bringing the mesothelin antigen proximal to membrane and thereby reducing the steric inhibitory effects between scFv and its target epitope (38). Recently investigators have found that replacing IgG1 hinge to IgG2 enhanced the efficacy further and reduced off target effect of a CAR against an antigen prostate stem cell antigen (PSCA) which is overexpressed on many solid tumors (39). These findings suggest that hinge region provides the flexibility to overcome the steric hindrances as well as reduces the distance between CAR scFv and antigen. Further studies are ongoing to understand the effect of hinge domain on flexibility, distance and elimination of the off target effects in relation to CARs.

#### Roles of Transmembrane Domain of CAR Construct in CAR T Cell Function

Transmembrane region is a linker between hinge region and endodomain of a CAR. Type I proteins such as CD3ζ, CD28, and CD8α have been used as transmembrane domains in CAR constructs. Earlier it was thought that transmembrane domain does not have much impact on CAR T cell efficacy except anchoring CAR molecule to the membrane. But recent investigations have shown that transmembrane domain impacts the efficacy as well as stability of CARs (40). For example, in first generation CARs, CD3ζ, containing transmembrane domain dimerizes with endogenous TCR and enhances the CAR T cell function as compared to the mutated CD3ζ transmembrane domain (41). In addition to this Savoldo et al. have shown that CARs containing CD3ζtransmembrane domain are less stable on the cell surface as compared to the CD28 transmembrane domain in second generation CARs (42). According to Guedan et al. transmembrane domains helps in persistence as well as enhance antitumor efficacy in third generation ICOS based CAR. It was observed that the ICOS TM domain is required for the optimal in vivo phenotype of third generation ICOS based CAR T cells. However, ICOS transmembrane domain did not show any effect on CAR cell-surface expression or tonic signaling (43). Very recently investigators have demonstrated critical role of hinge and transmembrane domain in context of release of effector cytokines which is one of the crucial challenges of CAR T cell therapy known as cytokine release syndrome (CRS) that involves release of excess amounts of cytokines causing toxicity. CAR T cells having CD8α hinge and transmembrane domain release less IFN-γ and TNF-α as compared to those having CD28 hinge and transmembrane domain with no significant differences in CAR T cell efficacy and proliferation (44). Therefore, designing of a robust CAR should involve the best combination of hinge and transmembrane domain to regulate the functionality of CAR T cells.

#### Regulation by Intracellular Endodomains

Different generations of CARs vary in their respective intracellular/co-stimulatory signaling domains. While first generation CARs contain only CD3ζ intracellular domain, the second and third generation CARs contain CD3ζ intracellular domain along with either single co-stimulatory domain or two co-stimulatory domains like CD28 and 4-1BB (4).Upon antigen binding phosphorylation cascade of immunoreceptor tyrosinebased activtion motif ITAMs present in CD3ζ intracellular domain is initiated leading to activation and priming of CAR T cells (45). However, second generation CAR modification over the first generation CARs by inserting the CD28 co-stimulatory domain demonstrated increased expansion and persistence of CAR T cells as compared to first generation counterparts (42).

On the contrary, 4-1BB (CD137), another co-stimulatory receptor is responsible for enhanced T cell survival. In a clinical study, the second generation CD19 CAR (CTL019) containing anti-CD19 scFv, CD3ζ domain along with 4-1BB co-stimulatory domain revealed robust expansion and long persistence of CTL019 cells along with sustained remissions in patients with relapsed/refractory chronic lymphocytic leukemia CLL (46). Another study using umbilical cord blood T cells (UCB T cells) revealed that UCB-19BBz and UCB-1928BBz CAR T cells were more cytotoxic toward CD19+ leukemic cell lines as compared to UCB-19z and UCB-1928z CAR T cells (47). Few studies compared the utility of CD28 and 4-1BB in various CAR constructs and data revealed the almost similar early response rates in ALL patients when treated with either CD28 or 4-1BB CAR (48, 49). However, in case of CLL, the 4-1BB CARs exhibited superior efficacy than the CD28 CARs, probably due to increased persistence of 4-1BB (CD137) CAR T cells and exhaustion of CD28 CAR T cells driven by CD28 endodomain signaling (46, 50, 51). The study done by Brentjens et al. in 8 CLL patients showed that infusion of 19-28z T cells resulted in complete reduction in lymphadenopathy in 1/8 patients (12.5%), progressive stable disease in 3/8 patients and no objective response in 4 patients. However, according to Porter et al., overall response rate in CLL patients when treated with CTL019 (CD19 scFv+ 4-1BB costimulatory domain) was found to be 8/14 (57%), where 4 patients showed complete remissions (CR) and 4 showed partial remissions (PR). The CAR T cells persisted and remained functional for more than 4 years in case of two patients who achieved CR and showed no relapse symptoms.

In another study, 4-1BB CARs (GD2-BBz CAR T cells) have been demonstrated to ameliorate CAR T cell exhaustion by decreasing the expression of exhaustion related molecules and by upregulation of three critical pathways such as hypoxia inducible signaling, cellular metabolism and negative apoptosis regulation (51). However, it remains poorly understood how these pathways contribute in ameliorating CAR T cell exhaustion in 4-1BB CARs.

In addition, CAR signaling impact metabolic reprogramming in T cells by modulating bioenergetics and mitochondrial biogenesis. The CD28z CAR T signaling facilitates differentiation of TEM and increased aerobic glycolysis in T cells. On the contrary, 4-1BBz CAR T cells display differentiation to TCM cells along with increased mitochondrial biogenesis and oxidative metabolism (52). The high rate of mitochondrial respiratory capacity as observed in case of 4-1BBz CAR T cells promotes CD8+ T cell memory differentiation and the oxidative phosphorylation acts a major source of energy thereby supporting increased CAR T cell proliferation (53). Therefore, in circumstances where long term CAR T cell persistence can cause severe off tumor toxicity, short-lived CARs can be designed by incorporating CD28 co-stimulatory domains.

Although many studies demonstrated that 4-1BB CAR is safe with increased persistence, in case of CD5+ tumor targeting, CD5-4-1BB CAR showed reduced efficacy compared with CD5- CD28z CAR due to enhanced T cell fratricide. To understand the mechanism of these observations, it has been described that the tumor necrosis factor receptor associated factor (TRAF) signaling induced by 4-1BB co-stimulatory domains upregulates expression of intracellular adhesion molecule 1 (ICAM-1) (usually not expressed on T cells) which in turn stabilizes the immunological synapse between CD5 CAR T cells (54). In order to circumvent this CAR (either 4-1BB or others) induced toxicity, a regulated CAR expression system may be developed.

While majority of available CARs either used 4-1BB or CD28 CARs, another CD4-related co-stimulatory receptor OX40 (CD134) emerged as a prominent approach in CAR signaling. Investigators have shown that a third generation CAR CD28 z-OX40 helped CCR7 (-) T cells avoid apoptosis and show potent anti-tumor functional efficacy (55, 56). Another report by the same group revealed that in comparison to a CD28-z CAR, the 3rd generation CD28-z-OX40 CAR decreased secretion of repressive cytokine like IL-10 without altering secretion of pro-inflammatory cytokines, T cell proliferation and cytotoxic potential. OX40 signaling further repressed Treg mediated IL10 secretion (56). This aspect of OX40 signaling can be harnessed in controlling Treg to effector T cell ratios in adoptive immunotherapy.

The ICOS expressing CARs induce ICOS signaling in the T cells and thereby increase the expression of IL-17A, IL-17F, and IL-22 (57). ICOS mediated co-stimulatory CAR signaling established Th17 characteristics such as increased expression of RORC, CD161, IL-1R1, and NCS-1. ICOS signaling also fostered Th17/Th1 polarization by enhancing IFN-γ and Tbet expression. In vivo animal studies further revealed enhanced persistence and anti-tumor responses of CD4+ CAR T cells in case of ICOS based CARs in contrast to the CD28 or 4-1BB containing CARs (43, 57). The extended persistence of ICOS-based CARs can be translated to non-lymphoid tumors where CAR T cell persistence is poorly understood. However, effect of ICOS based CARs is yet to be tested clinically in patients.

Hence, the co-stimulatory domains play a key regulatory role in determining the anti-tumor efficacy, functionality and persistence of CAR T cells both in vitro and in vivo. A thorough understanding of the signaling cascades associated with various co-stimulatory molecules will help in designing more effective CAR with better clinical implications.

#### Regulating CAR T Cell Efficacy With Homeostatic Cytokines and Other Genes

Efficient T cell activation requires three signals, T cell receptor (TCR) signaling (Signal 1), and activation by co-stimulatory molecules (Signal 2) and immune-stimulatory cytokines (Signal 3). So far majority of the CARs designed and discussed possess signal 1 and signal 2, however, signal 3 generally provided by homeostatic cytokines is absent in the conventional CAR T cells and also less abundant in the tumor microenvironment (15) Therefore, next generation of CAR T cells would require an additional cytokine signaling to satisfy the need of signal 3 for optimal CAR T cell activation (58). The major cytokines involved in T cell activation belong to γc class like IL-2, IL-7, IL-15, IL-21, and IL-9. These cytokines control T cell survival and proliferation, which ultimately has significant roles in CAR T cell persistence and efficacy (59). These cytokines are currently employed in ex vivo expansion of CAR T cells prior to therapy in combinations or alone (60).

Starting with the well-studied cytokine in T cell activation and regulation, IL-2 has ranked first upon receiving a US FDA approval for employment in immunotherapy of melanoma in 1998 (61). However, along with induction of potent anti-tumor T cells, IL-2 is also known to cause Activation Induced Cell Death (AICD) and differentiation into immunosuppressive regulatory T cells (62). Hence, low dose IL-2 in CAR T cell therapy has been suggested for achieving better anti-tumor responses and to overcome the immunosuppressive effects of IL-2 (63).

Another approach to lessen the immunosuppressive effect of IL-2 on T lymphocyte is to make way for other cytokines like IL-7 and IL-15 in adoptive immunotherapy. With the discovery of T stem cell like memory cells (Tscm) it has been described that IL-7 and IL-15 are known to generate and maintain less differentiated TSCM population in the T lymphocyte pool thereby improving the antitumor responses significantly (64–66). In the absence of these regulatory cytokines the differentiation status of CAR T cells is skewed toward terminal differentiation thereby reducing functional potency of the CAR T cells. Owing to this fact, incorporation of IL-7 and IL-15 in the in vivo expansion of CAR T cells has been a new addition in the adoptive immunotherapy field (60).

Attempts have been made for administration of cytokines like IL-2, IL-7, IL-15, and IL-12 in clinical trials against various malignancies. However, the anti-tumor responses are greatly masked by toxicities generated due to the cytokine administration alone or along with CAR T cells (67–70). The toxic effects of recombinant cytokines are due to higher doses of i.v. administration in patients leading to off target toxicity. Therefore, attempts are been made to engineer CAR T cells to secrete or express these cytokines on the T cell surface giving rise to next generation of CAR T cells. A variant of IL-7R has been engineered to be expressed along with CAR T cells against solid tumors and has shown a good response in improving the anti-tumor efficacy in vivo models (71). CAR T cells secreting IL-15 have been tested in hematological malignancies along with anti-CD19 CAR expression. The IL-15 expression has shown to regulate differentiation of these modified CAR T cells by inhibiting apoptosis and showing reduced Treg induction unlike IL-2 (72). Along with the secretory form, the membrane bound expression of IL-15 has also shown to be a good strategy in achieving less differentiated CAR T cells and in generating potent anti-tumor responses in leukemia models (73).

In addition to this, IL-12 armored CAR T cells has shown to eliminate the need of pre-treatment regime before the onset of CAR T therapy in a preclinical setting. IL-12 armored CAR T cells also possess an intrinsic resistance to regulatory T cells mediated inhibition (74). However, another group of researchers conducting clinical trials with engineered TILs to secrete IL-12 showed a highly toxic response owing to the uncontrolled cytokine release (75). In addition, a report suggested leukemic transformation and clonal expansion of CD8+ T cells transduced to express IL-15 gene. This further helped in realizing the need for control of signal 3 generated by immune regulatory cytokines (76).

In order to regulate the adverse effects mediated by cytokine secretion as well as to regulate toxicities due to CAR T cells as observed in clinical trials, the scientists are designing strategies to tune these cells in a ligand dependent manner (77). One such "safety-switch" is the inducible caspase 9 molecule integrated with CAR T cells which can be controlled by an external ligand AP1903. Administration of the ligand induces caspase 9 mediated apoptotic pathway thereby clearing the CAR T cells from body and reducing the off-target cytotoxicity. These switches have been combined with cytokines IL-7 and IL-15 in the preclinical studies to provide a hope for managing toxicities due to the uncontrolled cytokine secretion (71, 78).

With such newer modifications in the CAR T cell designs, the focus of the field is gradually shifting toward regulation of actions of CAR T cells with presence of tunable signals and use of inhibitory receptors. Very recently, a new generation of CAR T cells, the SUPRA CAR T cells is been demonstrated (79). SUPRA stands for Split, Universal and Programmable CAR T cells, which can respond to multiple antigens and can be inducible controlled to manage T cell activation mediated toxicity. Other such CAR T cells such as iCARs (inhibitory CARs) are analyzed in preclinical studies wherein use of CTLA-4 and PD-1 based self-regulating inhibitory receptors have been employed to control off-target toxicities (80).

The current advances in the employment of cytokines and receptors thereof in regulating the CAR cell response is more on the preclinical front. The clinical outputs of these engineered CAR T cells along with newly designed switches would provide a useful insight into planning strategies for better regulation of CAR T cells in immunotherapy. Until then, the management of signal 3 for the regulation of CAR T cells stills remains elusive and opens a new field of research in adoptive immunotherapy.

#### CONCLUSION AND FUTURE PROSPECTIVE

Chimeric antigen receptors emerged as new genre of drugs with huge therapeutic potential for hematological malignancies as well as solid tumors. Different components of a CAR strongly impact the anti-tumor efficacy, potency and safety of the genetically engineered CAR T cells after infusion into the patients. This review highlights the importance of scFv designing in determining flexibility, affinity and specificity of the CAR along with strategies to minimize the off target toxicity. The regulatory role of hinge domain and transmembrane domain in providing proper positioning of the CAR, imparting stability and reducing cytokine release toxicity has also been discussed. Moreover, this review conveys the key role of intracellular domain in relation to long term survival and persistence of CAR T cells along with the regulatory aspects of various cytokines like IL-2, IL-15, IL-7, and IL-21. Fine tuning of the crucial components and parameters as discussed in this review will pave the way for developing more safe and efficacious CAR T cells for a wide array of malignancies.

Despite the successful achievements of CAR T cell therapy in case of hematological malignancies it displays to serious side effects such as cytokine release syndrome (CRS) that causes systemic inflammatory responses, multi-organ failure and death as well as neuro-toxic effects like aphasia, hallucinations and seizures (67). Another major concern is the "on target off-tumor" toxicity. The efficacy and safety of CAR T cells is monitored in preclinical animal models prior to its clinical testing, which includes syngeneic, human xenograft, immunocompetent transgenic, and humanized transgenic mice. These pre-clinical models neither reflect the obstacles in clinical efficacy nor do they predict potentially life-threatening safety concerns and hence lack to display the complications which might arise in the clinical setting. Moreover, the lack of host immune system does not allow testing of the tumor microenvironment, the tumor's metastatic potential, or the host response to CAR T cells. Each murine model has its own advantages and shortcomings which highlights the absence of a single competent model that can evaluate CAR T cell efficacy as well as toxicity issues. Recently, many primate models such as macaque or dog are also being studied in relation to CAR T side effects (81, 82). Hence, the evolution and refinement of preclinical models will lead to improved prediction of CAR T safety and efficacy in the clinic.

With all the evidences available with varied designs of the CAR in the field, so far it remains elusive to understand the use of correct combination of these domains which regulate the CAR T cell efficacy. In order to generate a predictive model to understand the efficacy and toxicity of the designed CAR T cells in in vitro experiments, arises the need to design computational biology tools. The development of computational approaches with a combination of clinical outcomes from the currently developed CAR T cells, may have a major impact in the designing superior CAR T cells with minimal toxicity and improved efficacy in future.

# 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 intramural grants of IIT Bombay (15IRSGHC002), funding from Wadhwani research foundation (DO/2017-WRCB002-016/PRE\_APPRVL\_REQ/188581), Tata center (DGDON422/PRE\_APPRVL\_REQ/147655) at IIT Bombay and Tata Trust (RD/0117-TATAE00-001), to RP.

# REFERENCES


T cells for ovarian cancer. Clin Cancer Res. (2006) 12:6106–15. doi: 10.1158/1078-0432.CCR-06-1183


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

#### Approved by:

*Frontiers in Immunology Editorial Office, Frontiers Media SA, Switzerland*

#### \*Correspondence:

*Rahul Purwar purwarrahul@iitb.ac.in*

*†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: *12 February 2019* Accepted: *15 February 2019* Published: *08 March 2019*

#### Citation:

*Dwivedi A, Karulkar A, Ghosh S, Rafiq A and Purwar R (2019) Corrigendum: Lymphocytes in Cellular Therapy: Functional Regulation of CAR T Cells. Front. Immunol. 10:401. doi: 10.3389/fimmu.2019.00401*

# Corrigendum: Lymphocytes in Cellular Therapy: Functional Regulation of CAR T Cells

Alka Dwivedi † , Atharva Karulkar † , Sarbari Ghosh† , Afrin Rafiq† and Rahul Purwar\*

*Department of Biosciences and Bioengineering, Indian Institute of Technology Bombay, Mumbai, India*

Keywords: chimeric antigen receptor, cancer immunotherapy, immunoregulation, anti-tumor efficacy, cytokines

#### **A Corrigendum on**

#### **Lymphocytes in Cellular Therapy: Functional Regulation of CAR T Cells**

by Dwivedi, A., Karulkar, A., Ghosh, S., Rafiq, A., and Purwar, R. (2019). Front. Immunol. 9:3180. doi: 10.3389/fimmu.2018.03180

In the original article, author names were incorrectly published with the first and last names reversed. The correct authors names are: "Alka Dwivedi," "Atharva Karulkar," "Sarbari Ghosh," "Afrin Rafiq," and "Rahul Purwar."

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2019 Dwivedi, Karulkar, Ghosh, Rafiq and Purwar. 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.

# CD56dim CD16<sup>−</sup> Natural Killer Cell Profiling in Melanoma Patients Receiving a Cancer Vaccine and Interferon-α

Lazar Vujanovic1,2, Christopher Chuckran<sup>3</sup> , Yan Lin1,4, Fei Ding1,4, Cindy A. Sander 1,2 , Patricia M. Santos 1,2, Joel Lohr <sup>3</sup> , Afshin Mashadi-Hossein<sup>5</sup> , Sarah Warren<sup>5</sup> , Andy White<sup>5</sup> , Alan Huang<sup>5</sup> , John M. Kirkwood1,2 and Lisa H. Butterfield1,2,3,6 \*

*<sup>1</sup> University of Pittsburgh Hillman Cancer Center, Pittsburgh, PA, United States, <sup>2</sup> Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States, <sup>3</sup> Department of Immunology, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States, <sup>4</sup> Department of Biostatistics, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States, <sup>5</sup> NanoString Technologies, Seattle, WA, United States, <sup>6</sup> Department of Surgery, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States*

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*R. Keith Reeves, Harvard Medical School, United States Nathalie Jacobs, University of Liege, Belgium*

\*Correspondence: *Lisa H. Butterfield lbutterfield@parkerici.org*

#### Specialty section:

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

Received: *24 August 2018* Accepted: *04 January 2019* Published: *29 January 2019*

#### Citation:

*Vujanovic L, Chuckran C, Lin Y, Ding F, Sander CA, Santos PM, Lohr J, Mashadi-Hossein A, Warren S, White A, Huang A, Kirkwood JM and Butterfield LH (2019) CD56dim CD16*<sup>−</sup> *Natural Killer Cell Profiling in Melanoma Patients Receiving a Cancer Vaccine and Interferon-*α*. Front. Immunol. 10:14. doi: 10.3389/fimmu.2019.00014* Natural killer (NK) cells are innate cytotoxic and immunoregulatory lymphocytes that have a central role in anti-tumor immunity and play a critical role in mediating cellular immunity in advanced cancer immunotherapies, such as dendritic cell (DC) vaccines. Our group recently tested a novel recombinant adenovirus-transduced autologous DC-based vaccine that simultaneously induces T cell responses against three melanoma-associated antigens for advanced melanoma patients. Here, we examine the impact of this vaccine as well as the subsequent systemic delivery of high-dose interferon-α2b (HDI) on the circulatory NK cell profile in melanoma patients. At baseline, patient NK cells, particularly those isolated from high-risk patients with no measurable disease, showed altered distribution of CD56dim CD16<sup>+</sup> and CD56dim CD16<sup>−</sup> NK cell subsets, as well as elevated serum levels of immune suppressive MICA, TN5E/CD73 and tactile/CD96, and perforin. Surprisingly, patient NK cells displayed a higher level of activation than those from healthy donors as measured by elevated CD69, NKp44 and CCR7 levels, and enhanced K562 killing. Elevated cytolytic ability strongly correlated with increased representation of CD56dim CD16<sup>+</sup> NK cells and amplified CD69 expression on CD56dim CD16<sup>+</sup> NK cells. While intradermal DC immunizations did not significantly impact circulatory NK cell activation and distribution profiles, subsequent HDI injections enhanced CD56bright CD16<sup>−</sup> NK cell numbers when compared to patients that did not receive HDI. Phenotypic analysis of tumor-infiltrating NK cells showed that CD56dim CD16<sup>−</sup> NK cells are the dominant subset in melanoma tumors. NanoString transcriptomic analysis of melanomas resected at baseline indicated that there was a trend of increased CD56dim NK cell gene signature expression in patients with better clinical response. These data indicate that melanoma patient blood NK cells display elevated activation levels, that intra-dermal DC immunizations did not effectively promote systemic NK cell responses, that systemic HDI administration can modulate NK cell subset distributions and suggest that CD56dim CD16<sup>−</sup> NK cells are a unique non-cytolytic subset in melanoma patients that may associate with better patient outcome.

Keywords: melanoma, CD56dim CD16−, natural killer cell subsets, dendritic cells, recombinant adenoviral vector, interferon-α

#### INTRODUCTION

Malignant melanoma is the most lethal cutaneous cancer that causes the majority of skin cancer-related deaths (80%). It is a neoplasm of melanocytes that commonly initiates in the skin and eye. Early-stage, localized lesions are normally treatable with surgical excision (5-year survival rate of 90–100% for stage 0/I). Unfortunately, due to its invasive properties, the median survival for individuals with metastatic disease is only 6–10 months (1). Recent advances in molecular biology and immunology have led to the identification of a number of promising therapeutic targets, with therapies that engage these targets having become a focus of drug development for melanoma. Among these, therapies targeting mitogen-activated protein kinase (MAPK) signaling pathway in lesions that harbor mutated BRAF kinase and negative regulatory checkpoint molecules of the immune system [cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed death receptor-1 (PD-1)] became standards of care for metastatic melanoma (1–3). Clinical benefit of these therapies has been limited to a subset of patients, stressing the need for novel therapeutic modalities to treat melanoma. Cancer vaccines, such as dendritic cell (DC)-based vaccines could be one such strategy.

DC are professional antigen presenting cells that can effectively activate naïve CD4+ and CD8+ T cell responses as well as Natural Killer (NK) cells, and play a key role in host immune responses against infections and cancers (4). DC-based vaccines have been shown to be safe, immunogenic and capable of promoting long-term tumor-specific immunity, survival and durable objective clinical responses in a minority of patients [4.2–7.1%; (5–8)]. Our lab has developed a novel autologous DC-based vaccine capable of inducing adaptive and innate antitumor immunity in vitro. We engineered lipopolysaccharide (LPS) and interferon-γ (IFNγ)-matured, type-1-skewing DCs to express three melanoma tumor antigens (tyrosinase, MART-1 and MAGE-A6) through replication-defective adenovirus (AdV) transduction (9–11). The vaccines were administered to patients at three time-points, and patients were subsequently randomized to observation or to receive boost of 1 month of high-dose systemic interferon-α2b (IFNα; HDI). IFNα is a type I interferon which is a potent modulator of immune responses. It can promote strong type-1 anti-tumor T cell responses and is a major regulator of NK cell-mediated cytotoxicity (12–15). Because IFNα has been shown to have a significant benefit on the overall survival and prolonged relapse-free survival, it is commonly used as an adjuvant immunotherapy for melanoma (16). The clinical trial results revealed a clinical response rate of 6% and increased vaccine antigen recognition by T cells in the majority of vaccinated patients. HDI adjuvant therapy did not enhance T cell or clinical responses (Butterfield et al., under review).

Because DC have been shown to be potent inducers of NK cell activation by elevated expression of activation markers as well as enhanced IFNγ secretion, lytic activity and proliferation, we investigated whether the DC ± IFNα strategy also enhanced NK cell activity in melanoma patients (17, 18). NK cells are an important class of innate cytotoxic lymphocytes which mediate killing of virally-infected or transformed cells. Their function is regulated by an elegant interplay of numerous activating and inhibitory surface receptors that are either specific for self-peptides expressed on healthy cells or for molecules that are upregulated during stress (19, 20). Consequently, missing self-antigens, marked overexpression of normal self-peptides, or the expression of altered self-peptides can stimulate NK cell cytotoxicity (21). The integration of multiple simultaneous receptor-ligand interactions and cytokine signals determines NK cell activation as well as functional response (19, 20). Depending on context, NK cells have the ability to act as either immunoregulators through cytokine production or direct effector cells (20).

In human cancer, NK cells play a significant role in both killing of transformed cells as well as immune regulation through the production of cytokines (20, 22, 23). Decreased patient NK cell function has been reported in hepatocellular carcinoma, head and neck cancer, breast cancer, and melanoma (24–27). A seminal 11-year observational study suggested that increased peripheral blood NK cell lytic activity correlated with decreased cancer incidence (28). In the melanoma setting, it is generally believed that NK cell dysfunction correlates with reduced tumor immunosurveillance and increased cancer susceptibility (22, 23, 29–31). However, a recent study has indicated that tumor-draining lymph nodes generate and/or recruit highly cytotoxic CD56dim NK cells, indicating that NK cell functions are not as suppressed in melanoma patients as previously believed (32).

We have shown that AdV-transduced DC (AdV.DC) recruit CD56dim CD16<sup>+</sup> and CD56bright CD16<sup>−</sup> NK cell subsets via CXCL8/IL-8 and CXCL10/IP-10, respectively, and activate NK cell Th1 functions through cell contact-mediated interactions of membrane-bound tumor necrosis factor (TNF) and IL-15 (18, 33). We have also shown that combination of IFNα with AdV.DC further enhances NK cell activation and cytotoxicity in vitro (11). Based on these data, we examined the impact of intradermal AdV.DC ± systemic HDI administration on peripheral blood NK cell profiles in melanoma patients. We characterized differences in immunosuppressive serum factors, NK cell cytotoxicity, phenotype, and subpopulation distribution between patients with and without measurable disease and healthy donor controls in blood, and profiled subpopulation distributions of tumor-infiltrating NK cells (TINKs).

# MATERIALS AND METHODS

### Antibodies

NK cell phenotype of melanoma patients enrolled in the trial was examined using fluorochrome-conjugated antibodies against the following cell-surface markers: CD56-FITC, CD3-PC7, CD16- APC, CD69-BV421, NKp30-BV711, CXCR3-BV421, CCR3- BV510 (BD Biosciences; San Diego, CA), NKp44-PerCP eFluor 710 (eBioscience; San Diego, CA), CXCR1-PE (R&D Systems; Minneapolis, MN), CCR7-BV711 (BioLegend; San Diego, CA), and matching IgG isotype controls from the same vendors. The immune checkpoint and NK cell activation receptor panel included the following markers: Zombie NIR Fixable Viability Dye (BioLegend; San Diego, CA), CD3-PE-Vio770 (Miltenyi Biotec; San Diego, CA), ANK-1-PE (Santa Cruz Biotechnology; Dallas, TX), TIGIT-PerCP eFluor 710 (eBioscience), CD45- BUV395, CD56-BV510, CD16-BUV737, NKG2D-APC, NKp46- BV711, CD69-BV421, and PD-1-BV650 (BD Biosciences).

#### Patients and Their Treatments

This was a Phase I, single site study to evaluate the immunological effects of autologous DC transduced with the MART-1, tyrosinase and MAGE-A6 genes in 35 subjects with recurrent, unresectable stage III or IV melanoma (M1a, b, or c), or resected stage IIIB-C or IV melanoma (**Supplemental Table 1**). 5 × 10<sup>6</sup> - 10<sup>7</sup> AdV.DC were given intradermally every 2 weeks for a total of 3 vaccines. After the AdV.DC immunizations, subjects were randomized to either receive a boost of HDI or no boost. Subjects randomized to receive the IFNα boost received Interferonα2b (Intron A, Schering-Plow), 20 MU/m2/d (rounded to the nearest 1 million units) administered intravenously for 5 consecutive days (Monday through Friday) every week for 4 weeks. Administration began approximately 30 days (±7 days) after the 3rd vaccine (Butterfield et al., under review).

#### Patient Sample Acquisition and Storage

With informed consent, peripheral blood and tumor biopsies were obtained from healthy donor (HD) and melanoma patients (HCC #04-001, #09-021 and #96-099). Patient characteristics are described in **Supplemental Tables 1**, **2**. Peripheral blood mononuclear cells (PBMCs) were separated from HD blood using Ficoll Hypaque gradient centrifugation (Corning, Manassas, VA) as previously described (34) and cryopreserved as aforementioned. Monocytes and lymphocytes isolated by elutriation from the baseline, day 43 and day 89/101 leukaphereses were cryopreserved in 50% RPMI, 40% HuAB serum (Gibco; Fisher Scientific; Waltham, MA) and 10% DMSO (Sigma). A red top tube (no anticoagulant) was also drawn at each time point for serum to test for cytokine/chemokine/growth factor/immunosuppressive factor levels. Patient samples were acquired in parallel with a HD control sample. Serum was clotted at room temperature, aliquoted, and frozen at −80◦C. Serum was kept in a monitored freezer and tested after a single thaw. Bulk melanoma single cell suspensions were collected and cryopreserved as previously reported (35). All patient specimens were processed by competency-trained technologists under standard operating protocols in the Immunologic Monitoring Laboratory.

# NK Cell Isolation and Culture

Cryopreserved patient lymphocyte and HD PBMC samples were thawed using RPMI + 10% FBS media supplemented with 0.5% DNAse (Sigma) and immediately prepared for analysis and testing. Thawed cell viabilities were between 65 and 92% as measured by trypan blue exclusion (Gibcol Fisher Scientific), with the mean viability of 81%. One portion of cells was used for multi-color flow cytometric analysis of NK cells. NK cells were purified from the remaining cells by negative magnetic cell sorting selection using the Human NK cell Isolation Kit (Miltenyi Biotec). NK cells were cultured in AIM-V medium supplemented with 5% human AB serum. NK cells were resuspended in AIM-V media (Gibco; Fisher Scientific) at a concentration of 5.0 × 10<sup>6</sup> cells/mL.

#### Cell Line

K562 erythroleukemia cell line was obtained from ATCC in 2000, and was authenticated by flow cytometry. K562 was cultured in RPMI 1640 medium, supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin, and 1% Lglutamine (Life Technologies), at 37◦C, in a humidified 5% CO<sup>2</sup> atmosphere. It was passaged bi-weekly, and was used in the described experiments after every other passage. The cell line was negative for Mycoplasma contamination as shown by GEN-PROBE Mycoplasma Tissue Culture Non-Isotopic Rapid Detection System (Gen-Probe, Inc.; San Diego, CA).

# NK-Target Cell Visualization Assay (NK-TVA)

Target cells (K562) were cytosolically labeled with TVA dye (Cell Technologies Ltd.; Shaker Heights, OH) and co-cultured with isolated patient NK cells for 3 h at 37◦C and 5% CO2. NK:K562 ratios used were 100:1, 50:1, 25:1. 12.5:1, 6.25:1, 3:1, 1.5:1, and 0:1. Killing was measured by counting the remaining target cells after co-culture using an ImmunoSpot S6 ULTIMATE analyzer (Cell Technologies Ltd.). The percentage of cytotoxic activity and lytic units (LU)20/10<sup>7</sup> effector cells were calculated as previously reported (36).

# Absolute Cell Counts

One heparinized tube of whole blood was drawn at each time point from each patient for fresh whole blood flow cytometry to obtain absolute counts and percentages of PBMC subsets (Beckman TQ-prep, Beckman Coulter, Brea, CA). Cells were stained for CD3, CD16, and CD56. Flow cytometry was performed on a Coulter FC500 and analyzed by CXP software (Beckman Coulter) (Butterfield et al., under review).

#### Flow Cytometry

One-step staining of cell-surface antigens was performed using fluorochrome-conjugated primary antibodies as previously described (10, 18, 33). For the analysis of blood NK cells two antibody panels were constructed around CD56-FITC, CD16-APC, and CD3-PE-Cy7 (BD Bioscience, San Jose, CA) antibodies. NK cell activation receptors were evaluated with CD69-BV421, NKp30-BV711 (BD Bioscience), and NKp44- PerCP eFluor 710 (eBioscience, San Diego, CA) antibodies. Chemokine expression levels were tested using CXCR1-PE (R&D Systems, Minneapolis, MN), CXCR3-BV421, and CCR7- BV510 (BD Bioscience) antibodies. For the TINK analysis, Zombie NIR (BioLegend), CD45 BUV395, CD56 BV510, CD16 BUV737 (BD Bioscience), and CD3 PE-Vio770 (Miltenyi Biotec) antibodies were used. Suitable IgG controls were acquired from the same vendors. FACS analyses were performed using the BD LSRFortessaTM cell analyzer, and analyzed using FlowJo v10 (FlowJo, LLC; Ashland, OR) software.

#### Serum Testing

Patient and HD sera were assessed for multiple cytokines, chemokines, and immunosuppressive factors using the Cytokine/Chemokine/Growth Factor 45-Plex Human Panel 1 and Immuno-Oncology Checkpoint 14-Plex Human Panel 2 Luminex kits (ThermoFisher, Pittsburgh, PA).

#### Multiplex Gene Expression Analysis

Total RNA isolated from patient tumor biopsies (QIAGEN; Valencia, CA) was profiled on the nCounter <sup>R</sup> MAX platform with the PanCancer Immune Profiling Panel (NanoString Technologies, Inc.; Seattle, WA). Single molecule imaging was performed on the nCounter SPRINT Profiler (NanoString Technologies) and gene expression data were analyzed using the R statistical software. Gene expression data from nCounter platform were log2 transformed and normalized using stable expression of housekeeping genes. Genes and the weights involved in calculations of metagenes were developed a priori by NanoString. Following calculation of the metagene values for each sample, scores were used to evaluate association of each score to patient outcome. The outcome here was defined either as a binary value based on "better" [partial responder (PR), stable disease (SD) and patients with resected tumors and no evidence of disease (NED)] and "worse" (progressive disease) prognosis. Each outcome was regressed against the scores calculated. In the case of the binary outcome, a univariate linear model regressing each score on the binary outcome generated expected log2 ratios of scores for the patients with better prognosis relative to those with worse prognosis. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (37) and are accessible through GEO Series accession number GSE124574 (https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE124574).

#### Statistical Analysis

Two-tailed non-parametric tests were used for all analyses. To explore the relationship between two tests groups [e.g., HD vs. melanoma patients; Day 43 vs. Baseline differences to determine the impact of DC immunizations on the profile of NK cells; and IFNα vs. observation groups], the Wilcoxon rank-sum test was performed. When comparing three groups [e.g., HD vs. patients with NED vs. patients with measurable disease (MD)], Kruskal-Wallis test was used to get the overall p-value, followed by the Wilcoxon rank-sum to determine pairwise p-values. For correlation plots, statistical significance was estimated by linear correlation. To explore the relationship between NK cell killing ability, NK cell subsets distributions and expression of activation markers on these subsets, Spearman's correlation was used. Graphs were generated using GraphPad Prism v6 (GraphPad Software, Inc., La Jolla, CA). P values < 0.05 were considered to be statistically significant.

#### RESULTS

#### Late-Stage Melanoma Patient Sera Contain Elevated Levels of Multiple Factors That Suppress NK Cell Function

NK cell dysfunction has generally been associated with late-stage melanoma. Multiple mechanisms of NK cell suppression in the cancer setting have been described, including those mediated by tumor- or immune cell-derived soluble factors (30, 38). HD and melanoma patient sera were tested for immunosuppressive factors MICA, NT5E/CD73 and tactile/CD96, as well as perforin, which is released by cytotoxic NK and T cells and is a surrogate biomarker for in vivo cytolytic activity of these effectors (39). MICA, NT5E/CD73 and tactile/CD96 showed statistically significant increase in melanoma patients vs. HDs (**Figure 1A**). These analytes were tumor burdendependent as NED had similar serum levels of these factors to HDs (**Figure 1B**). Surprisingly, elevated perforin presence was concurrently observed in melanoma patient sera irrespective of their disease burden possibly indicating aberrant lymphocytic cytolytic activity (**Figures 1A,B**) indicating enhanced cytolytic activity of NK and T cell effectors (40, 41). None of these analytes were affected by DC immunizations or HDI (**Figures 1C,D**).

#### Circulatory NK Cells Isolated From Late-Stage Melanoma Patients Display an Elevated Level of Activation

Human NK cells can be segregated into at least three major subsets based on their CD56 (NCAM) and CD16 (FcγRIII) expression levels: CD56bright CD16−, CD56dim CD16+, and CD56dim CD16<sup>−</sup> NK cells. These subsets have unique phenotypic and functional profiles. NKp30 and CD69 expression levels are upregulated on all three subsets upon activation, but are also present at low levels on resting NK cells. In contrast, NKp44 is exclusively expressed on activated NK cells (33, 42). These three subsets also have specific chemokine receptor profiles which dictate their specific chemokine preferences. We have previously shown that CCR7 is expressed on all three NK cell subsets, CXCR3 is the dominant receptor expressed on CD56bright CD16<sup>−</sup> NK cells, and CXCR1 is the central receptor presented on CD56dim CD16<sup>+</sup> NK cells. CD56dim CD16<sup>−</sup> NK cells express a variety of chemokine receptors, including CXCR1 and CXCR3 (33). NK cell chemokine receptors have been reported to be upregulated following activation (43). We

unresectable (measurable disease; MD) tumors were compared for their serum concentrations of immunosuppressive factors to HDs. (C) Mel sera isolated prior to all the treatments (Baseline) and after DC immunizations (Day 43) are compared. (D) Mel sera isolated post DC immunizations are tested in patients in the observation arm of the study and those that received IFNα therapy (IFNα). Intersecting lines and whiskers represent mean and standard error of mean values, respectively. Black circles (A–C) represent patients that were PR.

performed a phenotypic analysis of fresh whole-blood NK cells, as well as a more detailed multicolor analysis of cryopreserved lymphocytes (**Supplemental Figure 1**).

Statistically significant differences in absolute counts of HD and melanoma patient NK cell subsets were not observed, nor were counts significantly affected by DC and HDI treatments (**Supplemental Figure 2**). A more robust phenotypic evaluation of cryopreserved systemic lymphocytes showed a distribution imbalance of CD56dim CD16<sup>+</sup> and CD56dim CD16<sup>−</sup> NK cells in a cohort of melanoma patients, particularly NED patients (**Figure 2A**, **Supplemental Figure 3A**). Patient NK cells also displayed elevated level of activation in comparison to HDs. Average expression levels of CD69, NKp44, CXCR1, CXCR3, and CCR7 were elevated on multiple patient NK cell subsets. Of

those, CD69, NKp44, and CCR7 showed trends toward increased expression on one or more NK cell subsets (**Figures 2B**, **3**). NKp30 was the only surface protein that showed little to no modulation on any of the melanoma patient NK cell subsets (**Figure 2B**, **Supplemental Figure 3B**). Expression of NKp44 was particularly affected by tumor burden as NED patients showed elevated expression on all of the NK cell subsets (**Supplemental Figures 3B**, **4**).

# HDI, but Not DC Vaccines Impacted Melanoma Patient NK Cell Subsets

Intradermal DC immunizations had little effect on distribution and activation of the three NK cell subsets (**Figures 4**, **5**). They may have had a modest suppressive ("normalizing") impact on CXCR1 expression levels on CD56bright CD16<sup>−</sup> and CD56dim

CD16<sup>−</sup> NK cells (**Figure 5**; not significant). All patients tested who subsequently received HDI had increased presence of circulating CD56bright CD16<sup>−</sup> NK cell frequencies (**Figure 6A**). HDI therapy did not significantly impact expression levels of activation and chemokine receptors evaluated on any of the NK cell subsets (**Figures 6B**, **7**).

#### Cytokine and Chemokine Levels Measured in Patient Sera Prior to and Post DC and HDI Therapies

Sera were evaluated for the presence of multiple cytokines (e.g., IL-12p70, IL-18, IL-15, TNF) and chemokines (e.g., IL-8/CXCL8, IP-10/CXCL10) associated with NK cell activation and recruitment. None of the cytokines evaluated were detectable on a consistent basis in either HD or patient sera (data not shown). In contrast, measurable levels of MCP-1/CCL2, MIP-1α/CCL3, MIP-1β/CCL4, RANTES/CCL5, eotaxin-1/CCL11, IP-10/CXCL10, and SDF-1α/CXCL12 were found in the majority of donor sera tested. Patients and HDs displayed similar serum levels of all detectable chemokines with the exception of eotaxin-1/CCL11 which was slightly decreased in patient sera, particularly in patients with measurable disease (**Supplemental Figure 5**). DC therapy led to a slight increase in MIP-1α/CCL3 and RANTES/CCL5, while HDI had no effect on chemokine levels in patient sera (**Supplemental Figure 6**).

FIGURE 4 | DC immunization impact on NK cell subset distributions and expression of activation markers. (A) Distribution of the three major NK cell subsets based CD56 and CD16 expression levels; and (B) CD69, NKp30, and NKp44 expression levels measured on the three NK cell subsets are shown for melanoma patients prior to the treatments (Baseline) and after three rounds of DC immunizations (Day 43). Intersecting lines and whiskers represent mean and standard error of mean values, respectively. Black circles represent patients that were PR.

### NK Cells Isolated From Melanoma Patients With Measurable Disease Display Elevated Lytic Ability

Lytic capability of patient peripheral blood NK cells was also examined because this metric has previously been correlated with disease incidence and prognosis (28). Patient and HD systemic NK cells were evaluated on per-cell basis. Purified NK cells were tested at different effector-to-target (E:T) ratios against K562 cells using the NK-TVA assay (**Supplemental Figure 7**). At baseline, melanoma patient NK cells showed elevated mean lytic ability vs. HDs (189 vs. 131 LU; **Figure 8A**). There were two groups of patients: those that had lower lytic ability than HDs and those that had higher. Closer evaluation of patient NK

cell responses indicated that 5/7 NED patients showed decreased (120 LU on average), while 7/9 patients with measurable disease displayed increased killing (300 LU on average) ability (**Figure 8B**). Interestingly, the two NED patients that showed elevated killing ability (Mel13 and Mel21; **Supplemental Table 1**) are those that clinically progressed following the completion of the trial (**Figure 8B**; Mel13 and Mel 21 had progression-free survival of 18.4 and 19.2 months, respectively; Butterfield et al., under review). DC immunizations and systemic HDI therapy did not significantly impact the NK cell lytic ability significantly (**Figures 8C**,**D**).

#### Enhanced NK Cell Lytic Ability Strongly Correlates With Enhanced CD56dim CD16<sup>+</sup> NK Cell Presence

Lytic unit values calculated for each donor sample and collection time-point were cumulatively correlated with paired percentage distributions of each major NK cell subset. There was a highly significant positive correlation between LUs and CD56dim CD16<sup>+</sup> cell prevalence, as expected based on their known cytotoxic phenotype (**Figure 9A**) (20, 44–47). Reciprocally, there was a significant negative correlation between LU and CD56dim CD16<sup>−</sup> cells, whereas there was no correlation for CD56bright CD16<sup>−</sup> cell percentage (**Figure 9A**). In agreement, CD56dim CD16<sup>+</sup> NK cell percentage values inversely correlated to paired CD56dim CD16<sup>−</sup> percentages (**Supplemental Figure 8**). These

data indicate that CD56dim CD16<sup>−</sup> cells likely have a different function. The relationship between LU and expression levels of tested markers was explored and expression of CXCR3 on CD56dim CD16<sup>+</sup> and, surprisingly, CD56dim CD16<sup>−</sup> positively correlates with LU (**Figure 9B**). Increased CD56dim CD16<sup>−</sup> NK cell incidence did not correlate with NKp30, NKp46, CXCR1 and CCR7 expression levels (data not shown). However, enhanced expression levels of CD69 and CXCR3 did positively and negatively correlate with CD56dim CD16<sup>−</sup> NK cell presence, respectively (**Supplemental Figure 9**), suggesting that elevated CD56dim CD16<sup>−</sup> NK cell presence in some patients may be due to their dysregulated retention in circulation because of altered expression levels of CD69 (48) and CXCR3.

# CD56dim CD16<sup>−</sup> NK Cells Are the Dominant Subset in the Melanoma Microenvironment

Since the two circulating CD56dim subsets appear to be dysregulated in a cohort of melanoma patients and mediate

distinct functions it was important to evaluate which of these subsets infiltrates melanoma. As viable tumor biopsies from the patients enrolled in our clinical trial were limited, we analyzed PBMC and melanoma cells from patients not associated with the trial by multi-color flow cytometry (**Supplemental Table 2**, **Supplemental Figures 10**, **11**, and **Figures 10A–C**). CD56<sup>+</sup> CD3<sup>−</sup> NK cells were found in higher frequencies among circulating vs. tumor-infiltrating leukocytes (mean 9.37 vs. 0.63% among PBMC and TIL, respectively; **Figure 10A**). The CD56dim CD16<sup>+</sup> subset was the dominant population among circulating NK cells (mean 63.32 vs. 10.474% in blood NK cells and TINK, respectively). In sharp contrast, CD56dim CD16<sup>−</sup> subset was the dominant population among TINK (mean 24.9 vs. 86% in blood NK cells and TINK, respectively). Interestingly, CD56bright CD16<sup>−</sup> subset was not detected in most of the melanoma lesions evaluated (**Figure 10B**). All of the gated TINK subsets expressed NKG2D along with moderate levels of NKp46, which confirmed their NK cell lineage (**Supplemental Figure 10B**). Expression of CD69, a marker of activation and tissue residency, was increased on all the TINK subsets (48). In contrast, expression of ANK-1, a marker expressed on CD56dim CD16<sup>−</sup> adherent NK cell precursors [pre-A-NK; 49] was, on average, 3 fold decreased in tumors vs. blood (**Supplemental Figures 10C**, **11**). We also explored whether PD-1 and TIGIT were coexpressed on CD56dim CD16<sup>−</sup> TINKs as a surrogate measure for functional exhaustion (49). Unlike tumor-infiltrating T cells (**Supplemental Figure 10D**), almost no co-expression of PD-1 and TIGIT was observed on any of the TINK subsets (**Figure 10C**). Interestingly, PD-1 expression was elevated on

standard error of mean values, respectively.

FIGURE 8 | Evaluation of the impact of tumor load, DC vaccine and systemic IFNα therapies on the circulating NK cell lytic ability. Purified circulating NK cells from HD and melanoma patients (Mel) were tested by NK-TVA assay (Supplemental Figure 6) for their ability to lyse K562 targets as depicted by lytic units (LU). NK cell subsets isolated from HD and melanoma patients (Mel) was measured. (A,B) Melanoma patient NK cells isolated prior to treatments (Baseline) were evaluated. (B) NED and MD patients were compared for their NK cell subset distributions to HDs. (C) NK cell subset distributions prior to all the treatments (Baseline) and after DC immunizations (Day 43) are compared. (D) NK cell subset distributions post DC immunizations are compared for patients in the observation arm of the study and those that received HDI (IFNα). Intersecting lines and whiskers represent mean and standard error of mean values, respectively. Black circles (A–C) represent patients that were PR.

a cohort of CD56dim CD16<sup>−</sup> TINKs (mean 0.7 vs. 4.5% in blood NK cells and TINK, respectively), while TIGIT was downregulated on a separate cohort of CD56dim CD16<sup>−</sup> (mean 16.2 vs. 8.1% in blood NK cells and TINK, respectively), as well as on CD56bright CD16<sup>−</sup> TINKs (mean 9.5 vs. 0.5% in blood NK cells and TINK, respectively). CD56dim CD16<sup>+</sup> NK cells consistently presented highest levels of TIGIT, with TINKs displaying increased average expression levels (mean 17.8 vs. 24.6% in blood NK cells and TINK, respectively), however this observation did not achieve statistical significance.

### CD56dim NK Cell Gene Signature in Melanomas Is Higher in Patients With Better Clinical Response

To relate aforementioned TINK observations to patients enrolled in our clinical trial, total RNA samples isolated from 16 patient biopsies prior to all treatments were evaluated by NanoString. Relative cell population abundance was evaluated as previously described (50). The CD56dim NK cell gene signature was higher in patients with better clinical outcome (PR, SD and NED; **Figure 10D**), but the trend did not achieve statistical significance (p = 0.1525). Cumulatively, these data suggest that increased melanoma infiltration by CD56dim CD16<sup>−</sup> NK cells may be beneficial to patient outcome.

#### DISCUSSION

Our group has performed a phase I trial where a novel, type 1-skewing AdV.DC vaccine was used to promote T cell responses against tyrosinase, MART-1 and MAGE-A6. Twentyfour patients with measurable disease and 11 surgically treated NED patients were enrolled. Two out of Twenty-Four patients with measurable disease achieved a PR and 8/24 patients had stabilization of the disease. Of 11 NED patients, 4 remain NED at a median follow-up of 3 years (Butterfield et al., under review). While the vaccine induced T cell responses against multiple antigens, it was also important to characterize whether AdV.DC ± HDI affected peripheral blood NK cells as predicted by our pre-clinical studies in terms of subset distribution, activation and chemokine receptor expression, and functionality.

NK cells have traditionally been studied as predominantly two separate populations, CD56bright CD16<sup>−</sup> and CD56dim CD16+, more attention has recently been given to the CD56dim CD16<sup>−</sup> subset. CD56bright CD16−cells exhibit superior cytokine production, whereas CD56dim CD16<sup>+</sup> cells primarily demonstrate enhanced cytotoxicity (20, 44–47). CD56bright CD16<sup>−</sup> NK cells preferentially localize to secondary lymphoid tissues and CD56dim CD16<sup>+</sup> cells occupy peripheral blood, lungs, and sites of inflammation (45, 51–54). Such tropisms are determined by unique chemokine receptor expression profiles which include CCR7, and CXCR3 for CD56bright CD16<sup>−</sup> cells and CXCR1 and CX3CR1 for CD56dim CD16<sup>+</sup> NK cells (33, 53, 55).

The CD56dim CD16<sup>−</sup> NK subset may be a highly heterogenous population consisting of both maturing and target cell-activated cells. Precursors of adherent NK cells (pre-A-NK) that express ANK-1, a polysialylated 230-kD isoform of NCAM, have been identified within the CD56dim CD16<sup>−</sup> NK cell population (56). We show that ANK-1<sup>+</sup> CD56dim CD16<sup>−</sup> NK cells are underrepresented within melanoma lesions as compared to circulation, likely indicating altered cytolytic state of CD56dim CD16<sup>−</sup> TINKs. More recent studies have shown that there are both poorly and highly cytotoxic populations found within this subset (57, 58). In concert with these reports, CD16 expression on CD56dim CD16<sup>+</sup> cells has been shown to be downregulated following target cell-induced activation of matrix metalloproteases, specifically ADAM17. CD16 shedding strongly correlates with increased CD107a expression, indicating that degranulation coincides with CD16 downregulation (59, 60).

Enhanced understanding of NK responses as well as the mechanisms by which NK cells promote immunity in these therapies could benefit patient outcomes by: (1) identifying suitable treatment candidates based on NK cell distributions and marker expression levels; (2) provide a system to evaluate patient disease progression or regression in response to treatment; (3) determine the potential efficacy of IFN-α adjuvant for specific patients. Though it was expected that NK cells from patients with measurable disease would show decreased expression of

activating receptors and extravasation-associated receptors, as well as lower lytic ability (61), the opposite was found. While a number of soluble immunosuppressive factors were found to be elevated in the sera of metastatic disease patients, their NK cells displayed increased expression of multiple activation markers and chemokine receptors, and enhanced lytic ability that directly correlated with higher representation of CD56dim CD16<sup>+</sup> NK cells. A previous study has shown that CD56dim CD57lo CD69<sup>+</sup> CCR7<sup>+</sup> KIR<sup>+</sup> NK cells expand in tumorinfiltrated lymph nodes and display enhanced cytotoxic activity against autologous melanoma cells. The same study has shown that the frequency of circulating NK cells expressing the receptors for IL-8/CXCL8 is increased in metastatic melanoma patients compared with healthy subjects, agreeing with our results (32). The observation that systemic NK cells in patients with metastatic melanoma display an enhanced degree of activation may be the consequence of lack of extravasation by activated NK cells into tumors, thereby increasing the presence of systemic NK cells with elevated lytic ability. It has been shown that stage I melanoma patients with high systemic NK cell activity have less lymphocyte infiltrate at the base of the tumor than those with low NK cell activity (62). To address this hypothesis, tumor infiltration will have to be analyzed for activated NK cells. Our exploratory transcriptomic analysis indicates that the CD56dim NK cell gene signature in bulk tumor biopsies is associated with better clinical response. Phenotypic evaluation of TINKs suggests that the CD56dim CD16<sup>−</sup> subset is the dominant NK cell population in melanomas. These results are not surprising as we have previously shown that the same subset in blood expresses the broadest repertoire of chemokine receptors, including CCR2, CCR3, CCR4, and CCR5 which are not expressed on other subsets, indicating that these NK cells are highly capable of recruitment into inflamed tissues (33). We also show that TINKs likely do not have the commonly-used "exhaustion" (co-expression of multiple checkpoints, like PD-1 and TIGIT). Future studies need to explore in greater detail the function and functional state of tumor-infiltrating CD56dim CD16<sup>−</sup> NK cells subsets in blood and in the tumor microenvironment. Based on our cumulative data, as well as reports that TINKs are not as cytotoxic as their blood counterparts (63), we hypothesize that this subset is likely not cytotoxic, that it lacks the ability to perform antibody-dependent cell-mediated cytotoxicity due to lack of CD16 expression and that its major anti-tumor activity is likely mediated by secreted factors such as cytokines, chemokines, and growth factors. Additionally, the number of CD56bright TINKs may be underestimated due

to usage of an anti-CD56 antibody that is conjugated to a moderately bright fluorochrome (BV510). Consequently, future studies also need to re-confirm our observations using anti-CD56 antibodies conjugated to brighter fluorochromes (e.g., PE or BV421).

Intradermal AdV.DC treatments did not appear to have a significant effect on NK cell phenotype and function. There are three possible reasons for this: (1) the time between DC injection and blood draw led to the failure to detect AdV.DC impact; (2) intradermal AdV.DC immunizations are not the optimal route of administration to induce systemic NK cell activation; and (3) the AdV.DC vaccine as formulated was incapable of proper recruitment and activation of NK cells. One month of systemic HDI administration following AdV.DC immunizations, which has been shown to boost adaptive immune responses and NK cell functionality (12–15), appears to promote replenishment of naïve NK cells in circulation in these patients as measured by increased representation of CD56bright CD16<sup>−</sup> NK cells. There appears to be a role for CD56dim CD16<sup>+</sup> and CD56dim CD16<sup>−</sup> NK cells in response to tumors. CD56dim CD16<sup>−</sup> NK cell prevalence negatively correlated with lytic ability, and appeared to inversely correlate with the presence of cytotoxic CD56dim CD16<sup>+</sup> NK cells. The developmental stage of this subset, as well the as biological function, is equivocal. The prevalence of this subset in the tumor may indicate aberrant cytotoxic capability. It may also indicate that patient samples contain mostly partially differentiated NK cells as CD56dim KIR<sup>−</sup> CD16<sup>−</sup> perforin<sup>+</sup> cells have been reported to be an intermediate stage NK cell differentiation (64). Functional analyses of these cells isolated ex vivo would confirm the phenotype.

In summary, we performed a detailed phenotypic and functional analysis of blood NK cells isolated from melanoma patients treated with AdV.DC ± HDI. We show that melanoma patient NK cells display elevated activation levels and that CD56dim CD16<sup>−</sup> NK cells are a unique non-cytolytic subset in melanoma patients that may positively impact patients' clinical outcome.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the University of Pittsburgh Institutional Review Board (IRB) with written informed consent from all subjects (HCC protocols 04-001, 09-021 and 96-099). All subjects gave written informed consent in accordance with the

#### REFERENCES


Declaration of Helsinki. The protocol was approved by the University of Pittsburgh IRB.

#### AUTHOR CONTRIBUTIONS

LV and LB were accountable for the conception and design of research and writing the manuscript. LV performed most, while CC, PS, and JL performed some of in vitro experiments and data analysis. YL and FD performed all biostatistical analyses. JK was the clinical trial leader. CS processed and coordinated tumor biopsies. AW and AH processed samples for NanoString testing. AM-H and SW analyzed NanoString data.

#### FUNDING

This work was supported by research funding from the University of Pittsburgh Cancer Institute and National Institute of Health P50 CA121973-04 Skin SPORE (P.I. JK) Project 2 to LB and Developmental Research Program to LV. This project used the UPMC Hillman Cancer Center (HCC) Immunologic Monitoring and Cellular Products Laboratory (LB) and the HCC Flow Cytometry Facility (A. D. Donnenberg, Director) that are supported in part by National Institutes of Health Award P30CA047904.

#### ACKNOWLEDGMENTS

We appreciate the efforts of the UPMC Hillman Cancer Center Clinical Research Services team, particularly Elizabeth A. Rush for melanoma program specimen processing and assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00014/full#supplementary-material

patients with advanced melanoma. New Engl J Med. (2011) 364:2119–27. doi: 10.1056/NEJMoa1012863


**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 Vujanovic, Chuckran, Lin, Ding, Sander, Santos, Lohr, Mashadi-Hossein, Warren, White, Huang, Kirkwood and Butterfield. 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.

# Differential Cytokine Utilization and Tissue Tropism Results in Distinct Repopulation Kinetics of Naïve vs. Memory T Cells in Mice

Hye Kyung Kim<sup>1</sup> \*, Hyunsoo Chung<sup>1</sup> , Juntae Kwon<sup>2</sup> , Ehydel Castro<sup>1</sup> , Christopher Johns <sup>1</sup> , Nga V. Hawk <sup>1</sup> , SuJin Hwang<sup>2</sup> , Jung-Hyun Park 2† and Ronald E. Gress 1†

<sup>1</sup> Experimental Transplantation and Immunology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States, <sup>2</sup> Experimental Immunology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States

#### Edited by:

Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States

#### Reviewed by:

Sid P. Kerkar, Boehringer Ingelheim, United States Fernando Concha-Benavente, University of Pittsburgh, United States

> \*Correspondence: Hye Kyung Kim kihye@mail.nih.gov

†These authors share senior authorship

#### Specialty section:

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

Received: 15 October 2018 Accepted: 12 February 2019 Published: 04 March 2019

#### Citation:

Kim HK, Chung H, Kwon J, Castro E, Johns C, Hawk NV, Hwang S, Park J-H and Gress RE (2019) Differential Cytokine Utilization and Tissue Tropism Results in Distinct Repopulation Kinetics of Naïve vs. Memory T Cells in Mice. Front. Immunol. 10:355. doi: 10.3389/fimmu.2019.00355 Naïve and memory T cells co-exist in the peripheral T cell pool, but the cellular mechanisms that maintain the balance and homeostasis of these two populations remain mostly unclear. To address this question, here, we assessed homeostatic proliferation and repopulation kinetics of adoptively transferred naïve and memory T cells in lymphopenic host mice. We identified distinct kinetics of proliferation and tissue-distribution between naïve and memory donor T cells, which resulted in the occupancy of the peripheral T cell pool by mostly naïve-origin T cells in short term (<1 week), but, in a dramatic reversal, by mostly memory-origin T cells in long term (>4 weeks). To explain this finding, we assessed utilization of the homeostatic cytokines IL-7 and IL-15 by naïve and memory T cells. We found different efficiencies of IL-7 signaling between naïve and memory T cells, where memory T cells expressed larger amounts of IL-7Rα but were significantly less potent in activation of STAT5 that is downstream of IL-7 signaling. Nonetheless, memory T cells were superior in long-term repopulation of the peripheral T cell pool, presumably, because they preferentially migrated into non-lymphoid tissues upon adoptive transfer and additionally utilized tissue IL-15 for rapid expansion. Consequently, co-utilization of IL-7 and IL-15 provides memory T cells a long-term survival advantage. We consider this mechanism important, as it permits the memory T cell population to be maintained in face of constant influx of naïve T cells to the peripheral T cell pool and under competing conditions for survival cytokines.

Keywords: cytokines, apoptosis, migration, lymphopenia, proliferation

# INTRODUCTION

The peripheral T cell pool comprises a mixed population of recent thymic emigrants (RTE), naïve T cells and memory T cells, which all depend on IL-7 for survival and homeostasis (1–4). While the RTE and naïve T cell pool is constantly replenished by newly generated T cells from the thymus (5), most memory T cells are thought to have limited renewal capacities (6). Consequently, memory T cells face steep competition with newly arriving RTEs and pre-existing naïve T cells for IL-7 dependent signals to survive (7). IL-7 is a critical survival factor that upregulates anti-apoptotic Bcl-2 and Mcl-1 (8, 9), and it also promotes expression of trophic factors that are essential for T cell survival (10). The non-redundant requirement for IL-7 is illustrated in the severely compromised thymopoiesis of IL-7 deficient mice, and impaired survival of mature T cells in the absence IL-7 signaling (11, 12). Importantly, IL-7 is not produced by T cells so that T cells depend on exogenous IL-7 to survive. IL-7 is primarily produced by stromal cells and dendritic cells, and its expression is thought to be constitutive and developmentally set (13). Consequently, IL-7 availability constrains the size of the peripheral T cell pool (2, 14, 15). How the diversity and integrity of individual T cell subpopulations can be maintained in face of such competition is an intriguing question that has remained largely unresolved.

IL-7 signaling is transduced by the IL-7Rα and γc-chain complex (16). While expression of γc is considered to be constitutive (17), IL-7Rα expression is dynamically controlled during T cell development and differentiation (15, 18). Memory phenotype T cells express higher levels of IL-7Rα compared to naïve T cells (19), and it has been proposed that increased IL-7Rα expression would provide increased survival signals during effector to memory transition (20). Increased IL-7Rα expression is also utilized as a marker to identify memory precursor populations during an immune response (21). On the other hand, IL-7 signaling downregulates expression of its own receptor so that decreased IL-7Rα expression does not necessarily indicate less efficient IL-7 signaling (15). Thus, it remains unclear whether memory cells would utilize IL-7 more efficiently compared to naïve T cells. It is also not known how composition of the peripheral T cell pool is maintained when both memory and naïve T cells compete for the same resources to survive. Understanding these aspects of T cell homeostasis, however, has wide-ranging implications, and particularly in clinical settings of adoptive cell transfer (ACT) in cancer treatment or in T cell reconstitution after immune ablative procedures (22, 23). Studies in mice have shown that ACT into lymphopenic hosts strongly induces expansion of donor T cells Ex vivo expanded tumor infiltrating lymphocytes (TILs) into cancer patients was reported to better engraft in conjunction with a lympho-depleting regimen that creates lymphopenia (24). Moreover, depending on the differentiation status of donor T cells, such as naïve vs. memory or effector T cells, their anti-tumor activity, cytokine secretion and host grafting widely differed. The cellular and molecular basis of such distinct outcomes are still unresolved, but they remain of great interest to both clinicians and basic immunologists alike.

Here, we addressed these questions using mouse models of ACT, where distinct subsets of donor T cells were adoptively transferred into lymphopenic host mice and then monitored for their proliferation and expansion. Specifically, we examined competition of co-transferred naïve and memory T cells during IL-7-driven lymphophenia-induced homeostatic proliferation (25–27). Interestingly, short-term adoptive transfer (1 week) resulted in a preferential expansion and accumulation of naïveorigin T cells in the LN, so that they vastly outnumbered memory-origin T cells. Surprisingly, we found that such selective expansion of naïve T cells was limited to lymph nodes where IL-7 is abundant (13). In other organs, and specifically in non-lymphoid tissues, however, memory-origin donor T cells outnumbered naïve-origin donor T cells, indicating tissue-specific expansion of naïve vs. memory donor T cells. Mechanistically, we found that memory T cells were significantly less efficient to utilize and transduce signaling by IL-7, but that their ability to co-utilize IL-7 and IL-15 as homeostatic cytokines endows memory cells a competitive edge in their expansion over naive-origin T cells. Thus, memory T cells outcompete naïve T cells upon ACT into lymphopenic environments, and this process is controlled by their distinct utilization of homeostatic cytokines.

#### RESULTS

#### Lymphopenia-Induced Homeostatic Proliferation of Naïve and Memory T Cells

In this study, we defined T cells expressing large amounts of CD44 (CD44hi) as memory T cells (28), while T cells with low abundance of CD44 (CD44lo) are considered as naïve T cells. We previously demonstrated that naïve T cells contain a significant fraction of RTE, which are functionally distinct to truly mature naïve T cells (7). Consequently, a mixed population of RTE and naïve T cells cannot correctly represent the survival kinetic of naïve T cells. Thus, we used the Rag2-GFP transgene (Tg) to identify truly mature naïve T cells (29), and only considered Rag2-GFPneg CD44lo T cells as naïve phenotype cells (**Figure 1A**). To examine homeostatic expansion of naïve and memory T cells under competing conditions, next, we purified naïve and memory T cells, mixed them at 1:1 ratio, and then injected them into lymphopenic (Rag2−/−) host mice. Naïve- vs. memory-origin donor T cells were identified using CD45.1/2 congenic markers. After 5 days, we recovered donor cells from host lymph nodes (LNs) for further analysis. Here, we were observed preferential accumulation of naïve-origin T cells, which resulted in dramatically increased naïve/memory ratios (**Figure 1B**). Among the donor T cells, we further found a selective increase in CD8 T cell frequencies that was concomitant to a decrease in CD4 T cell frequencies (**Figure 1C**), because CD4 donor T cells failed to undergo effective proliferation (**Supplemental Figure 1A**). These findings agree with previous observations that CD8 T cells expand more vigorously than CD4 T cells under lymphopenic conditions (30–32). Collectively, these results indicate that naïve T cells are superior to memory T cells in repopulating the T cell pool.

#### Accelerated Proliferation of Memory T Cells Under Lymphopenic Conditions

To gain mechanistic insights into the distinct repopulation efficiencies, we examined proliferation of naïve- vs. memoryorigin CD8 T cells. To this end, we purified naïve and memory T cells and labeled them with Cell Trace Violet (CTV) before their adoptive transfer. Dilution of an intracellular dye such as CTV can serve as a faithful marker of proliferation, and thus accurately reports the proliferative history of a given cell population (33). Surprisingly, and contrary to our expectation, we found that memory CD8 T cells proliferated substantially faster than naïve T cells (**Figure 2A**), which resulted in increased naïve/memory CD8 T cell ratio after adoptive transfer (**Figure 2B**). Thus, while memory T cells undergo more

vigorous proliferation than naive T cells, paradoxically, memory donor T cells did not outnumber donor naïve T cells after homeostatic proliferation.

T cell exhaustion is a homeostatic mechanism that trims the size of the activated memory T cell pool (34). PD-1 is a marker for T cell exhaustion (35), and we wished to determine if the rapid and excessive proliferation could induce T cell exhaustion in adoptively transferred memory T cells. Surface analysis for PD-1 expression, however, did not show noticeable difference between memory and naïve CD8 donor T cells (**Figure 2C**). We also did not see increased caspase-3 activity in memory T cells (data not shown). Collectively, these results indicate that exhaustion or increased cell death are unlikely causes for inefficient expansion of memory T cells in adoptive transfer experiments.

#### Diminished IL-7 Signaling in Memory T Cells

Lymphopenia-induced homeostatic proliferation depends on IL-7 signaling (36). Thus, we wished to know if memory T cells would be less efficient in IL-7 signaling, which could result in their impaired expansion and accumulation upon adoptive transfer. To this end, we examined IL-7-induced STAT5 phosphorylation in naïve and memory CD8 T cells. Compared to naïve T cells, memory T cells were substantially blunted in their IL-7 response, as demonstrated in significantly reduced

expression on naïve and memory T cell origin donor T cells in LN (left) and spleen (right) of host mice after 5 days of adoptive transfer. Histograms are representative of 2 independent experiments. \*\*\*P < 0.001.

amounts of phosphorylated STAT5 (pSTAT5) relative to that of naïve CD8 T cells (**Figure 3A**). Detection of intracellular pSTAT5 was highly specific, because IL-7 signaling in STAT5-deficient T cells did not show any pSTAT5 activity in the same assay (**Supplemental Figure 1B**). To exclude a difference in signaling kinetics, we monitored pSTAT5 contents at early time points (10, 20, and 30 min) and also after prolonged IL-7 stimulation (2 and 4 h), and still found both CD4 and CD8 memory T cells being significantly blunted in their IL-7 response compared to naïve T cells (**Figure 3B**, **Supplemental Figure 1C**). Such reduced IL-7 responsiveness further translated into diminished downstream effector molecule activation, so that Akt and mTOR phosphorylation were significantly decreased in IL-7 signaled memory CD8 T cells compared to naïve CD8 T cells (**Figure 3C**). Akt and mTOR are serine kinases that upregulate T cell metabolism and provide anti-apoptotic signals (37). Consequently, we considered the possibility that suboptimal IL-7 signaling in memory T cells could result in increased cell death, which would lead to a preferential loss of memory T cells in mixed donor T cell adoptive transfer experiments.

# Increased Pro-survival Factor Expression in Memory T Cells

To examine if memory T cells would be more prone to apoptosis, we assessed expression of survival molecules in naïve and memory CD8 T cells. Bcl-2 is a major anti-apoptotic molecule downstream of IL-7 (8). We expected that memory T cells would express significantly smaller amounts of Bcl-2 than naïve T cells, because we found memory CD8 T cells to show decreased IL-7 responsiveness. Strikingly, and contrary to our expectation, qRT-PCR analysis revealed that memory T cells expressed significantly larger amounts of Bcl-2 mRNA transcripts than naïve T cells (**Figure 4A**), which further correlated with increased Bcl-2 protein expression (**Figure 4B**). In agreement, assessing the intracellular contents of active caspase-3, which is a measure of apoptosis (38), revealed that memory T cells were less apoptotic than naïve T cells (**Figure 4C**). Increased cell survival, however, would contradict our finding that memory T cells are less effective in IL-7 signaling than naïve T cells. To solve this conundrum, we examined expression of other prosurvival molecules, and we noted that Bcl-x<sup>L</sup> mRNA expression was highly upregulated in memory T cells compared to naïve T cells (**Figure 4A**). Bcl-x<sup>L</sup> is a potent anti-apoptotic protein that is induced by IL-15 signaling (39–42). IL-15 utilizes the IL-2Rβ/γc cytokine receptor complex for signaling (16, 43), and memory T cells, but not naïve T cells, express large amounts of IL-2Rβ (44). Therefore, these results suggest that the ability to co-utilize IL-15 together with IL-7 could provide survival and proliferative advantage to memory T cells over naïve T cells.

#### IL-15 Signaling Promotes Proliferation of Memory T Cells

Based on these observations, we wished to know if IL-15 indeed provides additional proliferative cues to memory T cells. If

stimulated with IL-7 (1 ng/ml) for 30 min and assessed for phosphor-STAT5 content in naïve and memory CD8 T cells. Graph shows summary of 5 independent experiments. (B) Kinetics of IL-7 (1 ng/ml)-induced STAT5 phosphorylation in naïve and memory T cells at 10, 20, 30, 60, 120, 240, and 360 min of stimulation. Graph shows summary of two independent experiments. (C) IL-7-induced phosphorylation of Akt and m-TOR in naïve and memory T cells. Graphs show summary of 6 independent experiments. \*P < 0.05; \*\*P < 0.01.

this would be the case, it would explain how memory CD8 T cells can outpace naïve T cells in proliferation, even as they are less efficient in IL-7 signaling. To this end, we adoptively transferred a 1:1 mixed population of naïve and memory donor T cells into Rag2−/−Il15−/<sup>−</sup> mice and assessed their proliferation. Interestingly, unlike in IL-15-sufficient Rag2−/<sup>−</sup> hosts, memory T cells in IL-15-deficient Rag2−/<sup>−</sup> (Il15−/−Rag2−/−) host mice did not proliferate more vigorously than naïve T cells, and we did not observe differences between naïve and memory T cell proliferation (**Figure 5A**). These results indicate that IL-15 is the driver of accelerated proliferation of memory-origin T cells during lymphopenia-induced homeostatic proliferation.

Consequently, when assessing the accumulation of naïvevs. memory-origin donor T cells in lymph nodes (LN) of Rag2−/−Il15−/<sup>−</sup> host mice, we did not notice any significant changes compared to IL-15-sufficient Rag2−/−Il15+/<sup>+</sup> hosts

(**Figure 5B**, left). Naïve-origin T cells still outnumbered memoryorigin donor T cell after 1 week of adoptive transfer, regardless of the presence or absence of host IL-15. On the other hand, we found a dramatic change in the naïve- vs. memory-origin T cell ratio in the spleen, where naïve-origin T cells were significantly more abundant than memory-origin T cells (**Figure 5B**, right). Thus, the lack of host IL-15 significantly impaired the expansion of memory T cells in the spleen, but not in the LN. These results suggest that IL-15 would contribute to the expansion of memory T cells in LNs. In contrast, IL-15 is abundantly expressed in the spleen (**Figure 5B**) (45), and can act as a major contributor to memory T cell proliferation. In agreement, the naïve- vs. memory-origin ratio was reversed in the spleen, so that memory-origin T cells outnumbered naïve-origin T cells. These results reveal a previously unappreciated aspect of lymphopeniainduced homeostatic proliferation that is associated with tissuespecificity and differential usages of homeostatic cytokines.

#### Distinct Tissue Migration of Adoptively Transferred Naïve and Memory T Cells

While IL-15's contribution would explain the preferential accumulation of memory-origin T cells in the spleen, it remained unclear to us why naïve-origin T cells would outnumber memory-origin T cells in the LN upon homeostatic proliferation. As a potential explanation, we considered that adoptively transferred memory-origin donor T cells would be inefficient in seeding the LN. In fact, memory T cells display an activated phenotype that comprises downregulation of lymphoid tissue homing and retention molecules, such CD62L and CD103 (46, 47). Thus, relative to naïve T cells, memory T cells would be impaired or delayed in entering lymph nodes after adoptive transfer. Accordingly, we hypothesized that, depending on whether the donor T cells would be of naïve or memory T cell origin, T cells would migrate and occupy survival niches in different organs. To test this idea, we performed short-term transfer experiments where we injected a 1:1 mixture of naïve and donor T cells into lymphopenic host mice. We harvested donor T cells after 3 days, instead of the usual 5 days, of injection to monitor migration in the absence of proliferation. In addition to LN (**Figure 6A**), we harvested T cells from other organs, such as spleen, lung and liver, and examined the naïve/memory-origin donor T cell ratio in these tissues (**Figure 6B**). As expected, we found significant and preferential accumulation of naïveorigin T cells in the LN (**Figure 6A**). Other organs, however, were preferentially seeded with memory-origin T cells, indicating distinct tissue migration between naïve and memory T cells (**Figure 6B**). To demonstrate that the selective accumulation of naïve-origin T cells in LN was mediated by LN-specific adhesion molecules, next, we asked if memory T cells would also accumulate in LN if they would express tissue homing molecules, such as CD62L. Notably, among CD44hi memory T cells, the central memory T cell population expresses large amounts of CD62L and differs from effector memory T cells that are absent for CD62L (**Supplemental Figure 2A**). Thus, we

expected that central memory donor T cells would accumulate in LN as is the case for naïve donor T cells. This was precisely the case, as we found that effector memory donor T cells that lack CD62L were substantially outnumbered by naïve donor T cells in the LN, but that central memory donor T cells which express CD62L were found in similar ratios to naïve donor T cells in the LN (**Supplemental Figure 2B**). Importantly, both effector and central memory donor T cells proliferated more vigorously than naïve donor T cells (**Supplemental Figure 2C**), effectively excluding delayed proliferation as a basis of impaired accumulation of effector memory T cells. Collectively, these results suggested that memory T cells survive and accumulate as efficient as naïve T cells, but that their initial migration and accumulation differ among tissues in the host.

### Co-utilization of IL-15 and IL-7 Promotes Long Term Survival Advantage to Memory T Cells

Because of such differences in homeostatic expansion among tissues, we considered that homeostatic expansion outside of the LN would result in accumulation of memory-origin T cells. This was indeed the case. To obtain a more comprehensive picture of donor T cell proliferation, first, we monitored T cell expansion beyond the short-term (5 days) adoptive transfer, and assessed donor T cells numbers at 1, 2, 4, and 6 weeks after injection. Strikingly, with increasing time of adoptive transfer, there was a dramatic increase in total donor T cell numbers that we could recover from LN and spleen of host mice (**Figure 7A**). However, it is important to point out that naïve- and memory-origin donor T cells accumulated in the host at unequal ratios. With increasing number of donor T cells, we found a preferential expansion of memory-origin T cells in both LN and spleen (**Figure 7B**). Initially, there was a preferential accumulation of naïve-origin donor T cells in the LN **(Figure 7B**, 1st week time point**)**. However, such skewed expansion was limited to the first week of adoptive transfer, and it was not found after 2 weeks or thereafter, and not in any other organ **(Figure 7B)**. Collectively, these results document a lagged response of memory-origin donor T cells in peripheral expansion, that is presumably driven by IL-15 in non-lymphoid tissues and outside of the LN.

To directly examine the role of IL-15 in this process, next, we performed the same adoptive transfer experiments into IL-15-deficient Rag2−/−Il15−/<sup>−</sup> mice (**Figure 7C**). Strikingly, in the absence of host IL-15, the expansion and accumulation of memory-origin donor T cells were significantly delayed, and the preferential expansion of naïve-origin T cells continued for another week or more (**Figure 7C**). As shown in **Figure 7C**, adoptively transferred memory-origin T cells were still outnumbered by naïve-origin T cells up to 2 weeks and in both LN and spleen in Il15−/−Rag2−/<sup>−</sup> host mice (**Figure 7C**). Eventually, memory-origin T cells caught up with naïve-origin T cells so that after 4 and 6 weeks of adoptive transfer, the naïve vs. memory-origin T cell ratio was reduced and we found them in similar frequencies (**Figure 7C**). Because we did not observe an effective expansion of memory-origin T cells after 6 weeks of transfer, these data further confirm the IL-15 requirement for effective memory T cell repopulation. Collectively, these results indicate that co-utilization of IL-7 and IL-15 significantly affects the initial expansion of memory T cells, and that it further provides long-term survival advantage to memory T cells during homeostatic proliferation.

#### DISCUSSION

Maintaining the memory T cell pool is a critical aspect in T cell immunology as it provides the reservoir for rapid and vigorous immune responses to re-challenging antigenic insults. Importantly, T cell memory is formed in the presence of antigens, but memory cells need to be maintained after clearance of pathogens. Therefore, memory T cells are thought to survive in the absence of TCR-mediated antigen stimulation and rather rely on homeostatic cytokines for their survival (48, 49). IL-7 is a key homeostatic cytokine for memory T cell survival, but its expression is scarce and limited to few tissues (13). Because RTEs and naïve T cells also require IL-7 signaling, competition for IL-7 has been proposed to be a mechanism to control the size of the peripheral T cell pool (7, 15). While the strict IL-7 dependency of each T cell subsets is well established, it is less well known how the competition among individual T cell populations would maintain the subset composition of the peripheral T cell pool.

shows result from each 6, 7, 10, and 8 experiments for 1 week, 2 weeks, 4 weeks, and 6 weeks, respectively. (B) Naïve to memory ratio among donor T cells recovered from spleen and LN of Rag2−/<sup>−</sup> host mice. Graph shows result from each 8, 9, 12, and 12 experiments for 1 week, 2 weeks, 4 weeks, and 6 weeks, respectively. (C) Naïve to memory ratio among donor T cells recovered from spleen and LN of Rag2−/−Il15−/<sup>−</sup> host mice. Graph shows result from each 6, 8, 6 and 5 experiments for 1 week, 2 weeks, 4 weeks, and 6 weeks, respectively. \*P < 0.05; \*\*P < 0.01; \*\*\*P < 0.001.

As a potential solution, our current study reports that memory-origin donor T cells outpace naïve-origin donor T cells during homeostatic proliferation. Specifically, we identified IL-15 as the driver for fast proliferation of memory T cells under lymphopenic conditions. IL-7 drives the expansion of both naïve and memory T cells, but IL-15 has an additive effect on IL-7-mediated proliferation of memory T cells. Consequently, memory T cells have a proliferative advantage over naïve T cells once entering the peripheral T cell pool. The selective effect of IL-15 on memory T cells was imposed by the distinct expression of IL-2Rβ, which is necessary for binding and signaling of IL-15 on target cells (50). Memory cells express high levels of IL-2Rβ, but naïve T cells only express very low levels of IL-2Rβ and are thus inefficient to bind and signal IL-15 (51).

Despite the contribution of IL-15 to accelerate their proliferation, it was curious that memory donor T cells were less efficient than naïve donor T cells to repopulate the lymphopenic environment of Rag2-deficient mice. Initially, we considered the possibility that excessive proliferation would be detrimental for memory T cell survival such as by inducing exhaustion that would result in increased cell death. Analysis for surface PD-1 expression and intracellular caspase-3 activity, however, suggested that this was not the case. Therefore, we faced a conundrum that, despite increased proliferation, diminished numbers of memory-origin donor T cells were recovered from host mice compared to naïve origin T cells. Previously, we showed that short term administration of recombinant IL-7 proteins resulted in tissue-redistribution of naïve and RTE cells, so that RTE preferentially accumulated in lymphoid tissues (3). Analogous to distinct trafficking of RTE and naïve T cells, here, we considered the possibility that distinct tissue migration and tropism of naïve vs. memory-origin donor T cells could provide the molecular basis for impaired expansion of memory T cells, and our current data are in support of this idea.

Memory T cells are more agile and migratory than naïve T cells, which agrees with their prime mission to survey tissues for pathogenic antigens (21). Accordingly, naïve and memory T cells express different sets of chemokine receptors and cell adhesion molecules (52). Naïve T cells express large amounts of the chemokine receptor CCR7 and the cell adhesion molecule CD62L which facilitate their migration and entrance into secondary lymphoid tissues. Memory T cells, on the other hand, express CCR9 and CXCR3, which promote trafficking to peripheral tissues (53, 54). Moreover, memory T cells, but not naïve T cells, preferentially home to the bone marrow, where they undergo expansion (55), and homeostatic proliferation (56). As a corollary, adoptively transferred memory and naïve T cells could disperse to distinct tissues and establish residency and undergo homeostasis. In agreement with this idea, we found that adoptively transferred memory-origin donor T cells preferentially migrated into non-lymphoid tissues, such as liver and lung, but then gradually re-appeared in secondary lymphoid organs. After 2 weeks of adoptive transfer, memory-origin donor T cells then outnumbered naïve T cells also in the LN. Thus, the maintenance of the memory T cell pool is driven by a slow kinetic of expansion that is regulated by two homeostatic cytokines and will eventually outcompete and outnumber naïve T cells. This observation raises two important issues; First, short-term (<1 week) analysis of adoptive transfer, which is usually the method of choice in assessing lymphopenia-induced homeostatic proliferation, provides an inaccurate picture of the in vivo events of T cell repopulation. As such, discovering the delayed kinetic of memory T cell expansion was enlightening, because it revealed that inefficient recovery of memory donor T cells at early time-points was not due to their failure to

expand. Instead, it was the inefficient recruitment of memoryorigin donor T cells into LNs which made it appear as if naïve T cells would be superior in their repopulation of the lymphopenic environment. Secondly, despite their better responsiveness to IL-7, naïve T cells are intrinsically less efficient in repopulating a lymphopenic environment compared to memory T cells. Naïveorigin donor T cells are less proliferative, and their expansion is mostly limited to LN tissues. Consequently, the homeostasis of the naïve T cell pool depends on thymic output and the continuous influx of newly generated naïve T cells. Collectively, these findings showed that the T cell subset origin of donor T cells determines the engraftment efficacy, repopulation kinetics, and tissue-distribution of adoptively transferred T cells.

The clinical implications of this study are manifold, and they provide new insights into designing ACT in cancer immunotherapies as well as for understanding CD8 T cellmediated GVHD upon allogeneic stem cell transplantations. Specifically, it has been a long-standing question which T cell subset would be the most effective in adoptive immunotherapies (57). There is a consensus emerging that attributes lessdifferentiated, naïve phenotype T cells being the subset with the greatest curative and anti-tumoral potential (22, 58, 59). The exact mechanism underlying this observation remains to be unraveled. However, multiple pathways have been proposed, such that undifferentiated T cells would possess greater proliferative potential, retain the ability to produce IL-2, and display greater anti-tumor efficacy (60, 61). In addition to these, it is also proposed that the lymphoid homing molecule L-selectin (CD62L), which is usually associated with a naïve phenotype promotes anti-tumor effects in ACT. Such propensity is illustrated in the superior effector function of naive T cells and central memory T cells both of which are marked by the expression of CD62L (62). On the other hand, there are also conflicting data about the role of CD62L in ACT (63), where the failure to express CD62L did not impair the function of donor T cells and did not alter the outcome of T cell adoptive immunotherapy in mice (63). Because CD62L expression is usually associated with a more undifferentiated phenotype, these results suggest that it is rather the cell intrinsic property than CD62L expression itself that confers superior function to the CD62L<sup>+</sup> subset in ACT. Along these lines, it would be important to ensure that the repopulated T cell pool in ACT or after ablative immune would retain a naïve phenotype to maximize its function, and our data indicate this could be achieved by minimizing the incorporation of memory-origin donor T cells. Further, these data now provide a mechanistic understanding for our previous observation in humans that CD4<sup>+</sup> naïve cells never reconstitute to baseline levels without thymic recovery (64)

These data also further reinforce the importance of the thymus in translational settings. Reconstituting the peripheral immune system after severe immune depletion such as chemotherapy, irradiation or other immune ablative events would benefit from increased thymus function to supply newly generated naïve T cells into the pool (65, 66). Without continuous thymic output, memory T cells would eventually outcompete and outnumber naïve T cells, resulting in diminished TCR diversity and compromising homeostasis of the peripheral T cell pool. Therefore, further investments to identify mechanisms that can rejuvenate the thymus or boost thymic output of naïve T cells are critical to replenish an immunocompetent and diverse peripheral T cell pool. Lastly, our current observations are in agreement with the seminal study by Surh et al. where they observed memory CD8 T cells to utilize either or both IL-7 and IL-15 for survival and homeostatic proliferation (67). However, our multipronged approaches of investigations on lymphopenia-induced proliferation under conditions of competition significantly expands the scope of our understanding, and now provide the molecular basis of distinct repopulation kinetics of naïve and memory T cells.

# MATERIALS AND METHODS

#### Mice

C57BL/6 (CD45.2) and C57BL/6 CD45.1 congenic mice were purchased from the Charles River Laboratories. Rag2−/<sup>−</sup> mice were purchased from The Jackson Laboratory. Rag2−/−Il15−/<sup>−</sup> mice were generated in house by breeding Rag2-deficient mice with IL-15-deficient mice. Rag2-GFP-Tg mice were previously described and obtained from the Jackson Laboratory (7). Mice with T cell-specific deletion of STAT5a/b were previously reported and maintained in house (68). Animal experiments were approved by the NCI Animal Care and Use Committee, and all mice were cared for in accordance with NIH guidelines.

# Flow Cytometry

Cells were harvested from the thymus, spleen, and lymph nodes. Data were acquired using an LSRII flow cytometer (BD Biosciences) and analyzed using FlowJo. Live cells were gated using forward scatter exclusion of dead cells stained with propidium iodide. Naïve and memory T cell subpopulations were electronically sorted using a FACSAria II (BD Biosciences) cell sorter, based on their GFP and CD44 expression levels. In brief, single cell suspensions were stained for TCRβ, CD4, CD8, and CD44 expression and resuspended in sorting buffer (0.5% BSA in Ca2+/Mg2+-free PBS) at 20 × 10<sup>6</sup> cells/ml and filtered through 0.45µm nylon meshes before passing through the cell sorter. Collected cells were washed once in PBS before further processing for tail vein injection or RNA isolation. The following antibodies were used for staining: TCRβ (H57-597), IL-7Rα (A7R34), CD44 (IM7), CD62L (MEL-14), CD4 (GK1.5), CD8α (53-6-7), and isotype control antibodies (eBioscience or BioLegend). Antibodies for pAkt (M89-61) and phosphormTOR (O21-404) were purchased from BD Biosciences, and used in staining kits from eBioscience following the manufacturer's instructions.

#### Lymphocyte Isolation

Single cell suspensions were prepared from lymph node, spleen, liver and lung. Mononuclear cells (MNC) from liver and lung were prepared using lymphocyte isolation protocols as previously described with minor modifications (69). In brief, liver tissues were pressed through a 70µm cell strainer (BD Biosciences) and resuspended in PBS. Cell suspensions were centrifuged at 100 g for 3 min, and supernatants were collected, spun down, and washed again with cold PBS. Liver samples underwent enrichment for lymphocytes by centrifugation in a two-step Percoll gradient (GE Life Sciences). Lymphocytes at the interphase were harvested, washed, and resuspended in cell culture media before further analysis. All liver MNCs were identified by expression of CD45.

#### In vitro IL-7 Stimulation

T cells were stimulated with IL-7 as previously described (7). In brief, single cell suspensions were adjusted to 5 × 10<sup>6</sup> cells/ml and stimulated with recombinant IL-7 (PeproTech) at 37 C for the indicated time. pSTAT5 contents were assessed after 30 min upon fixing and permeating cells with paraformaldehyde and acetone/methanol, followed by staining with anti-pSTAT5 specific monoclonal antibodies (clone 47, BD Bioscience).

#### Cell Trace Violet (CTV) Labeling and Adoptive Transfer

Donor T cells were electronically sorted from lymphocytes isolated from LN, which were pooled out of inguinal, axillary, cervical, and mesenteric area. Before injection, donor cells were loaded with CTV (Invitrogen) as previously described (7). 10 × 10<sup>6</sup> cells were tail-vein injected into Rag2−/<sup>−</sup> or Rag2−/−Il15−/<sup>−</sup> double deficient mice. Donor cells were recovered at indicated times, from spleen or lymph nodes for analysis.

#### REFERENCES


#### Statistical Analysis

Statistical tests were performed with Prism (GraphPad). Statistical significance was determined with Student's t-test. <sup>∗</sup>P < 0.05 was considered significant. ∗∗P < 0.01; ∗∗∗P < 0.001. ANOVA test was used to compare more than 3 groups of normally distributed data. Error bars indicate standard error of the mean (SEM).

#### AUTHOR CONTRIBUTIONS

HK designed, performed, analyzed experiments, and wrote the manuscript. HC, JK, EC, CJ, NH, and SH performed and analyzed experiments. J-HP and RG conceptualized the study, directed the experiments, analyzed data, and wrote the manuscript.

#### ACKNOWLEDGMENTS

This study was supported by the Intramural Research Program of the US National Institutes of Health, the National Cancer Institute, and the Center for Cancer Research.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00355/full#supplementary-material


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

Copyright © 2019 Kim, Chung, Kwon, Castro, Johns, Hawk, Hwang, Park and Gress. 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.

# Cross-Talk Between Antigen Presenting Cells and T Cells Impacts Intestinal Homeostasis, Bacterial Infections, and Tumorigenesis

Stephen J. Gaudino and Pawan Kumar\*

*Department of Molecular Genetics and Microbiology, Stony Brook University, Stony Brook, NY, United States*

dendritic cells, macrophages, and B cells. APCs interact with T cells to link innate and adaptive immune responses. By displaying bacterial and tumorigenic antigens on their surface via major histocompatibility complexes, APCs can directly influence the differentiation of T cells. Likewise, T cell activation, differentiation, and effector functions are modulated by APCs utilizing multiple mechanisms. The objective of this review is to describe how APCs interact with and influence the activation of T cells to maintain innate immunity during exposure to microbial infection and malignant cells. How bacteria and cancer cells take advantage of some of these interactions for their own benefit will also be discussed. While this review will cover a broad range of topics, a general focus will be held around pathogens, cancers, and interactions that typically occur within the gastrointestinal tract.

Innate immunity is maintained in part by antigen presenting cells (APCs) including

Keywords: APC, SFB, CRC, SCFA, TLR, PAMP

### INTRODUCTION TO INNATE IMMUNITY, APC ACTIVATION, AND T CELL FUNCTION

The immune system is divided into innate and adaptive responses. Adaptive immunity is regulated by B cells and T cells. Maturation of T cells occurs in the thymus and maturation of B cells occurs in the bone marrow. During antigen-dependent activation, B cells can develop into memory cells, which are activated upon subsequent exposure to the contacted antigen, or plasma cells, which secrete antibodies specialized to target that antigen (1). Similarly, T cells can develop into memory cells or effector cells. The two major types of effector T cells produced by the adaptive immune system are helper T cells (Th) and cytotoxic T cells (TC). T<sup>h</sup> cells are distinguished by their expression of CD4, subset-specific expression of transcription factors (T-bet, GATA3, and RORγt), and the release of cytokines that influence the activation and differentiation of other immune cells. Three major subsets of T<sup>h</sup> cells exist (Th1, Th2, and Th17), each of which is specialized for protecting against certain infections. Th1 cells primarily secrete interferon-γ (IFN-γ) which is associated with protection against intracellular microbes (predominantly viruses) and the onset of anti- or pro-tumorigenic effects, Th2 cells fight parasitic infections by secreting specific interleukin (IL) proteins including IL-4, IL-5, and IL-13, and Th17 cells fight microbial pathogens by secreting cytokines such as IL-17A, IL-17F, and IL-22 (2–6). T<sup>C</sup> cells are distinguished by their

#### Edited by:

*Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States*

#### Reviewed by:

*Dan Anthony Mitchell, University of Warwick, United Kingdom Derek Pociask, Tulane University School of Medicine, United States*

> \*Correspondence: *Pawan Kumar pawan.kumar@stonybrook.edu*

#### Specialty section:

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

Received: *30 August 2018* Accepted: *12 February 2019* Published: *06 March 2019*

#### Citation:

*Gaudino SJ and Kumar P (2019) Cross-Talk Between Antigen Presenting Cells and T Cells Impacts Intestinal Homeostasis, Bacterial Infections, and Tumorigenesis. Front. Immunol. 10:360. doi: 10.3389/fimmu.2019.00360* expression of CD8 and their ability to directly contact and kill transformed and infected cells (1, 7). T cells and B cells act together to establish an immunological memory against individual pathogens or cancer cells. Adaptive immunity can take a few days to fully develop, but once activated allows for the mounting of a rapid immune response upon subsequent exposures to the specific pathogen or cancer cell. Innate immunity, unlike adaptive immunity, is traditionally not based on immunological memory. However, certain innate cells, particularly natural killer (NK) cells, aid in the development of immunological memory against viruses. For instance, Ly49H<sup>+</sup> NK cells exposed to mouse cytomegalovirus (MCMV) expand to mount a primary immune response (8). This process is promoted by IL-18 signaling (9). Naïve mice that received an adoptive transfer of memory NK cells were able to mount a strong secondary response upon infection with MCMV (8). Signaling by IL-18, however, is not required for recall responses by memory NK cells (9).

Classical antigen presenting cells (APCs) are dendritic cells (DCs) and B cells (10). To mount an immune response, APCs must first recognize and bind their target. To do so, APCs express antigen-specific surface receptors including pattern recognition receptors (PRRs). PRRs detect pathogen-associated molecular patterns (PAMPs), which are produced by microbes, and damage-associated molecular patterns (DAMPs), which are produced by damaged or mutated host cells (11). Depending on the receptor, expression of PRRs can be constitutive or inducible (12, 13). One major family of PRR is the Toll-like receptors (TLRs). TLRs are typically expressed on the cell surface or within endosomes and are type I transmembrane proteins whose extracellular domains express leucine-rich repeats that are used to recognize and bind to specific PAMPs (14–16). Once the extracellular domain binds its target, the TLR activates a cytosolic signaling cascade which is initiated by an adaptor protein that interacts with the intracellular domain of the TLR. Depending on the TLR, the two adaptor sets that can be activated are TIRAP-MyD88 and TRAM-TRIF (14, 16–18). Another group of PRR is the nucleotide binding oligomerization domain (NOD) like receptors (NLRs). NLRs are present in the cytoplasm and, like TLRs, initiate signaling cascades upon binding to microbial PAMPs (14, 16). After binding to their appropriate PAMP or DAMP, APCs internalize their target by initiating phagocytosis, pinocytosis, or clathrin-mediated endocytosis. The pathway by which molecules are endocytosed determines how they will be degraded and then displayed by major histocompatibility complex (MHC) for T cell recognition (19, 20).

Two types of MHCs display antigens: class I MHCs and class II MHCs. While MHC I receptors are produced by all nucleated cells and display endogenous antigens to activate CD8<sup>+</sup> TC, only APCs produce MHC II receptors to display exogenous antigens and activate CD4<sup>+</sup> T<sup>H</sup> cells. Some APCs including DCs can also present exogenous antigens to the MHC I receptor to activate CD8<sup>+</sup> T cells during a process called cross-presentation (20–23). Presentation of antigens by either MHC I or MHC II receptors also depends on antigen composition (particulate vs. soluble), method of endocytosis, and degradation by lysosomal proteases (20). T<sup>C</sup> and T<sup>H</sup> cells utilize membrane-bound T cell receptors (TCRs) to bind MHC receptors (1). TCRs consist of two polypeptide chains (alpha and beta) that are linked together by disulfide bonds. T cells also produce co-receptors on their surface which further aid stabilizing interactions with MHCs of APCs. These include CD4 and CD8 (11, 20, 24). Costimulatory interactions between APCs and T cells can also occur, respectively, between B7 and CD28, ICAM-1 and LFA-1, and CD40 and CD40L (25, 26).

#### IMMUNITY TO BACTERIA

In this section, innate immunity will be discussed with respect to commensal and pathogenic bacteria. First, immunity will be reviewed within the context of commensals and their influence on altering the activity of Th17 and T regulatory (Treg) cells. Then, APC interactions with T cells, mainly regarding the expression of various TLRs, to fight pathogenic infections will be discussed.

#### IMMUNITY AND BACTERIAL COMMENSALS

The gut microbiota is composed of all the archaea, fungi, protozoans, viruses, and bacteria that inhabit the gastrointestinal tract. Specifically, the colon of a healthy 29 to 30-year-old male of average weight and height has been estimated to contain ∼ 3.8 × 10<sup>13</sup> bacteria (27). Commensal bacteria have been shown to modulate immune cell responses in the intestine. Reciprocally, immune cells interact with intestinal epithelial subsets to regulate colonization by commensal bacteria. Altered microbiota composition and aberrant immune responses to commensal bacteria, however, have been thought to play a role in the development of metabolic disorders (obesity and type II diabetes), autoimmune disorders (multiple sclerosis (MS) and type I diabetes), and inflammatory bowel disease (IBD) (28–31).

APCs, specifically DCs and macrophages, utilize PRRs to maintain intestinal homeostasis (discussed in "Immunity and Bacterial Pathogens" section). Macrophages plays an important role in maintaining tolerance against the commensal microbiota and food antigens. Although macrophages display robust bactericidal activity, they are not a key source of major inflammatory cytokines such as TNF-α, IL-1β, IL-6, and IL-23 (32, 33). However, macrophages constitutively produce and respond to IL-10. Macrophage-derived IL-10 is critical for Foxp3<sup>+</sup> Treg cell development, maintenance, and expansion (34– 36). DCs also maintain tolerogenic responses by interacting with adaptive immune cells. DCs presents luminal antigens in secondary lymphoid organs of the intestine. These include the Peyer's patches (PPs) and mesenteric lymph nodes (MLNs). DCs regulate the homing of lymphocytes to the intestine by inducing expression of gut homing receptors (CCR9 and α4β7) (37, 38). DCs also are the major source of IL-23 which, in combination with other cytokines, influences the differentiation of Th17 cells and promotes the generation of IL-22, a tissue protective cytokine. Defects in NOD2 and autophagyrelated genes (ATG16L and IRGM) in DCs of IBD patients Gaudino and Kumar APC and T Cell Cross-Talk

reveal compromised antigen presentation, cytokine release, and increased inflammation (39–41).

IBD is marked by chronic gastrointestinal inflammation due to the onset of disorders such as ulcerative colitis (UC) and Crohn's disease (CD). Increased expression of Th1-derived IFNγ and Th17 associated IL-17A and IL-22 are evident in the inflamed tissue of CD patient (42, 43). Genome-wide association studies show that genes (IL23R and STAT3) encoding Th17 differentiation pathways are associated with increased IBD risk. This suggests a therapeutic significance of targeting IL-23 or IL-17A in IBD (44, 45). Indeed, neutralizing IL-23 has been shown to be effective in reducing intestinal inflammation. Ustekinumab, an antibody directed against the IL-12p40 subunit, is used to block both IL-12 and IL-23 and is approved by the FDA for treatment of moderate to severe CD, whereas antibody targeting IL-23p19 subunit (IL-23) currently is in clinical trial for both CD and UC (46, 47)**.** Interestingly, a clinical trial in which an anti-IL-17A monoclonal antibody was used to treat CD was prematurely ended since treatment with these antibodies was associated with detrimental effects (48). Subsequently we and others have shown that IL-17A is critical for preserving the epithelial barrier and regulating gut microbiota colonization (30, 49, 50). Additionally, it has been shown that IL-23 is not required for IL-17A generation from γδ T cells during chemicalinduced colitis (49). This may explain why anti-IL-23 and IL-17A neutralization have counterintuitive outcomes. It remains unclear what regulates IL-17A generation in γδ T cells. It is possible that DC-derived IL-6 or other factors regulate IL-17A responses in the gut since CD103<sup>+</sup> CD11b<sup>+</sup> DCs have been shown to regulate Th17 differentiation in an IL-16 dependent manner (51).

The gut microbiota and microbial metabolites have been shown to regulate IL-17A responses in the gut. The intestinal microbiota, particularly segmented filamentous bacteria (SFB), has been shown to induce Th17 responses (52–54). Induction of Th17 cell responses by SFB is well-characterized. Monocytederived macrophages regulate SFB induction of antigen-specific Th17 cells (55, 56). Furthermore, colonization by SFB is associated with increased expression of IL-21 and isoforms of serum amyloid A (SAA). SAA indirectly promotes differentiation of Th17 cells by acting on DCs (53, 57). Along with SFB, a mixture of twenty bacterial isolates including Clostridium and Bifidobacterium species from an ulcerative colitis patient has been shown to induce Th17 activity (57). Additionally, Escherichia coli Schaedler and Morganella Morganii have been shown to regulate Th1 and Th17 cell differentiation via monocyte-derived DCs (58). Moreover, CD172α <sup>+</sup> lamina propria DCs promote microbial antigen-specific Th17 cell differentiation in responses to TLR5 activation (59). The microbiota, including SFB, induces Th17 responses; however, it is poorly understood how immune cells regulate functions of the gut microbiota such as colonization by SFB. We and others have shown that IL-17A and IL-22 regulate the gut microbiota, including SFB colonization (30, 60, 61). Furthermore, we show that intestinal regulation of the gut microbiota by IL-17A modulates systemic autoimmunity suggesting a yin-yang relationship between the gut microbiota and Th17 cell responses (30).

The differentiation of naïve T cells into pathogenic (α/β CD4<sup>+</sup> T cells that express high levels of IL-23R, coproduce IL-17A and IFN-γ/GM-CSF and induce autoimmunity) or non-pathogenic (α/β CD4<sup>+</sup> T cells that produce IL-17A and IL-17F but do not induce autoimmunity) Th17 cells is influenced by DC-derived cytokines. Naïve T cells exposed to TGF-β1 and IL-6 differentiate into non-pathogenic Th17 cells, but those exposed to TGF-β1, IL-6, and IL-23 or TGF-β3 and IL-6 develop into pathogenic Th17 cells (62). Signaling by IL-23 increases expression of T-bet and production of TGF-β3 by developing Th17 cells. Likewise, IL-23 signaling has been associated with increased expression of RORγt and production of GM-CSF, an essential cytokine for the progression of autoimmunity, by Th17 cells (63). Production of dietary-derived fatty acid metabolites has also been shown to alter the differentiation of T cells (64). For instance, stimulation by long chain fatty acids triggers naïve T cell differentiation into Th1 and Th17 cells via the upregulation of p38-MAPK. This, in turn, promotes the onset of autoimmunity (64).

While SFB have mainly been associated with Th17 cell differentiation, Bacteroides fragilis or Clostridia species have been shown to regulate the induction and activity of Treg cells (65, 66). Polysaccharide A derived from B. fragilis activates DCs in a TLR2-dependent manner to induce Treg cell differentiation and IL-10 generation (66, 67). A mixture of seventeen Clostridia species that induce Treg cell differentiation and function were isolated from a human fecal sample (65). When germ-free mice were inoculated with the mixture, an increase in Treg cell abundance and induction were observed. These changes may be due to an increased production of microbiota-dependent fatty acid metabolites, particularly SCFAs. This study shows that SCFAs stimulate secretion of TGF-β by epithelial cells to promote induction of Treg cells (65). Kashiwagi et al show that TGF-β derived from DCs via TLR2-Smad3 pathways is important for the generation of Treg cells in the lamina propria of mice that were inoculated with Clostridium butyricum (68). Subsequently, the importance of SCFAs particularly butyrate in regulating Treg differentiation has been shown by many studies (69, 70). Butyrate and propionate have been shown to directly modulate Treg generation by promoting histone H3 acetylation of the Foxp3 locus and protein (69, 70). Additionally, butyrate signaling in macrophages and DCs via GPR109a, a receptor for butyrate and niacin, has been shown to promote Treg cell development (71). Mice deficient in GPR109a have fewer IL-10 producing CD4 T cells (71). Colonic Treg cells express TCRs, including CT7, that most likely aid in the recognition of specific antigens derived from the commensal microbiota (72). These TCRs are unique to colonic Treg cells since they are not expressed by Treg cells outside the colon (72).

APCs also modulate commensal microbiota-dependent Th2 cell responses. Mice treated with propionate display enhanced production of macrophage and DC precursors in their bone marrow. However, these DCs are impaired in eliciting effector functions of Th2 cells in a house dust mite extract-dependent allergic inflammation model (73).

Along with Th17 and Treg cells, innate lymphoid cells (ILCs) maintain immunity by interacting with APCs to influence commensal bacteria and T cell effector functions. ILCs are Gaudino and Kumar APC and T Cell Cross-Talk

separated into three groups (ILC1, ILC2, and ILC3) based partially on the cytokines they secrete. Similar to Th17 cells, ILC3 cells secrete IL-17A and IL-22 (**Figure 1**) (74). IL-22 secreted from ILC3 can act on epithelial cells to induce expression of antimicrobial peptides. IL-23 derived from CD103<sup>+</sup> CD11b<sup>+</sup> DCs has been shown to regulate innate IL-22 responses following administration of bacterial flagellin (75). ILC3s also directly interact with T cells. MHC II receptors are expressed by CCR6<sup>+</sup> RORγt <sup>+</sup> ILCs which allows for direct binding and presentation of antigens to CD4<sup>+</sup> T cells (76). Upon interacting with T cells, intestinal ILCs maintain homeostasis by limiting immune responses against commensal bacteria (76). Mononuclear phagocyte-derived TNF-like ligand 1 (TL1A) has been shown to regulates ILC3-dependent regulation of IL-22 production and mucosal host defense during acute colitis (77). TL1A also regulates expression of costimulatory molecule OX40L in MHC II<sup>+</sup> ILC3s. This is required for antigen-specific T cell responses in a chronic colitis model (77).

# IMMUNITY AND BACTERIAL PATHOGENS

While commensal bacteria directly or indirectly (via APCs) influence the differentiation of T cells, APCs utilize TLRs to recognize specific microbial markers alter the function of T cells to fight bacterial pathogens. Cell surface TLRs aid in the phagocytosis of microbial pathogens by DCs and macrophages. Binding of surface TLRs to specific microbial PAMPs or whole bacteria such as E. coli or Staphylococcus aureus induces receptor phagocytosis (78). Activation of different APC TLRs results in the production of cytokines that promote the differentiation of naïve T cells into Th1 cells. This is evidenced by the activation of two TLRs, TLR2 and TLR4, that are expressed in immune cells and intestinal epithelial cells (79). TLR2 binds to Grampositive and Gram-negative bacterial components including lipoteichoic acid, porins, and peptidoglycan, but TLR4 binds only to lipopolysaccharide, a component specific to the Gram-negative outer membrane (79–81). When activated, TLR4 promotes the production of IL-12 p70, the active form of IL-12 which is composed of the p35 and p40 subunits. IL-12 p70 aids in the polarization of naïve T cells into Th1 cells (81). On the other hand, activation of TLR2 results in increased production of IL-12 p40 homodimer which acts as a receptor antagonist of IL-12 (81).

A third TLR expressed by immune cells and the intestinal epithelium is TLR5, a receptor that specifically binds to flagellin monomers of Gram-positive and Gram-negative bacteria (79). Flagella are long whip-like structures utilized by bacteria for motility, adhesion, and secretion of virulence factors. Numerous intestinal pathogens including Listeria monocytogenes, Salmonella typhimurium, and Campylobacter jejuni produce flagella for successful host colonization and invasion (82–84). TLR5 signaling in DCs results in activation of the IL-22/23 axis (85). IL-23 then stimulates the secretion of IL-17A by memory T cells and promotes naïve CD4<sup>+</sup> T cells to differentiate into Th17 cells (**Figure 1**) (86–88). Activation of TLR5 expressed by epithelial cells and DCs plays an important role in clearing pathogenic bacteria such as adherent invasive E. coli. Epithelial TLR5 signaling is important in limiting bacterial adherence in the intestine (85).

TLRs can also interact with and effect the signaling of other innate immune receptors to influence T cell function. This is displayed by interactions that occur between TLRs and NOD2. NOD2 is part of the NLR family, is encoded by the Card15 gene, and recognizes Gram-positive and Gramnegative peptidoglycan-derived peptides including muramyl dipeptide (MDP) (89, 90). Activation of NOD2 is associated with altered TLR2 signaling. As stated previously, TLR2 is activated in response to peptidoglycan. Both NOD2 and TLR2 activation result in NF-κB signaling. When NOD2 signaling is stimulated by MDP, TLR2-induced NF-κB signaling and production of IL-12 are inhibited. However, in Card15−/<sup>−</sup> APCs, NOD2 signaling does not occur but TLR2 signaling can still occur; when treated with peptidoglycan, these mice display higher production of IL-12 by macrophages due to increased NF-κB signaling. This enhances Th1 responses (production of IL-12, IFN-γ, and IL-18) (91). Besides being expressed by APCs, TLR9 and NOD2 are unique because they are both expressed by Paneth cells, a subset of antimicrobial peptide-producing epithelial cells within the crypts of the small intestine. While many antimicrobial peptides are constitutively expressed, signaling by bacterial antigens including flagellin, peptidoglycan, and lipopolysaccharide can further stimulate their production (92).

Enteric pathogens utilize virulence factors to avoid detection by and alter the response of the innate immune system. Two such pathogens are Salmonella enterica serovar Typhi and Salmonella enterica serovar Typhimurium. While S. Typhi and S. Typhimurium are closely related enteric pathogens, they promote different disease states in humans. S. Typhi causes typhoid fever by crossing the intestine and then spreading to systemic organs; on the other hand, S. Typhimurium infection is restricted to the intestine and typically causes enteritis (93). S. Typhi infection is specific to humans and does not occur in mice (94). This may be due to the ability of mice, but not humans, to produce TLR11 which is activated by flagellin94. Unlike S. Typhi, S. Typhimurium can infect multiple hosts including mice. Therefore, S. Typhimurium is used to study typhoid-like dissemination in susceptible strains of mice (94, 95). Adaptive B and T cell responses are required to provide protection in mouse model of Salmonella infection. In infected mice, S. Typhimurium invades ileal epithelial cells including microfold (M) cells. M cells are specialized epithelial cells that aid in the transcytosis of luminal antigens and microbes into PPs, a region of secondary lymphoid tissue that contains DCs, T cells, and B cells (96). Upon sensing the bacteria, epithelial goblet cells upregulate mucus production, specifically mucin 2, to restrict contact of S. Typhimurium with the epithelial layer (97). To circumvent this, S. Typhimurium expresses flagella and a type III secretion system (T3SS) that is encoded by Salmonella Pathogenicity Island 1 (SPI-1) (84, 93). Effector proteins are injected through the SPI-1 T3SS into epithelial cells to alter their cytoskeletal shape and tight junction integrity. While injection of effectors into M cells promotes optimal bacterial

invasion, SPI-1 mutants of S. Typhimurium can still invade M cells but to a lesser extent (**Figure 1**) (93). Once inside the PP, S. Typhimurium can infect DCs to be transported in a CCR7-dependent manner to MLNs (98). S. Typhimurium then spreads to systemic tissues including the liver and spleen and replicates within phagocytes such as macrophages (93). To avoid degradation within macrophages, S. Typhimurium expresses a second T3SS encoded by Salmonella Pathogenicity Island 2 (99). DCs infected by S. Typhimurium display inhibited antigen presentation which prevents stimulation of naïve T cells (100). S. Typhimurium can also directly infect T cells to inhibit their proliferation and secretion of cytokines including IL-2 and IFN-γ (100).

Another well-studied enteric pathogen, L. monocytogenes, alters APC function to influence T cell function. Similar to S. Typhimurium, L. monocytogenes replicates within macrophages. Infected macrophages release TNF-α and IL-12 which promote the secretion of IFN-γ from NK cells (101). Secreted IFN-γ then activates macrophages to produce reactive oxygen and nitrogen intermediate to prevent the escape of L. monocytogenes from phagosomes and aid in bacterial degradation (102). To promote its growth, L. monocytogenes induces production of type I interferons (IFNs). Escape of L. monocytogenes from macrophage phagosomes (which occurs due to the production of listeriolysin O, a pore-forming toxin) results in the upregulation of IFN-α and IFN-β by the infected macrophage (103, 104). Increased production of type I IFNs has been associated with increased apoptosis of T cells and greater production of IL-10 to favor bacterial proliferation (104).

Memory T cells also aid with combating infection by L. monocytogenes and comprise a major of component of intestinal T cells including tissue effector memory (TEM), central memory (TCM), and resident memory cells (TRM). While TEM and TCM cells both reside in the blood and spleen, TCM can reside within lymphoid tissues and TEM can reside in nonlymphoid tissue (105). TRM cells predominantly reside within the lamina propria and intraepithelial lymphocyte regions of the intestine and provide immune regulation throughout life (106). Upon oral infection with L. monocytogenes, CD103+ DCs acquire and process bacterial antigens. These DCs express CCR7 and travel from the LP via the lymphatics to the MLN (107, 108). In the MLN, DCs present L. monocytogenes antigens to naïve CD8 T cells. Activated T cells proliferate and differentiate into early effector cells that further differentiate into either shortlived effector cells or memory precursor effector cells (108). Memory precursor effector cells can differentiate into TEM or TCM and migrate to the intestine (108). DCs release retinoic acid which binds to the retinoic acid receptor of T cells. Upon binding to RA, T cells exhibit enhanced expression integrin α4β7 and CCR9, both of which direct migration to the LP of the small intestine (105, 109). Once at their site of infection, TEM cells express granzyme B. Expression of granzyme B, however, is downregulated in TCM cells (105).

Enteropathogenic E. coli (EPEC) and Enterohaemorrhagic E. coli (EHEC) are two clinically significant human pathogens of the intestine that are responsible for deaths caused by severe diarrhea. Citrobacter rodentium (C. rodentium) is a mouse pathogen that shares several clinical pathological mechanisms of EPEC and EHEC and, therefore, serves as a useful model to understand the innate and adaptive immune responses in the gut following infection as well as pathogenesis of IBD (110, 111). The attaching and effacing (A/E) lesion formed by EPEC, EHEC, and C. rodentium distinguishes them from other intestinal pathogens and commensal E. coli. A/E lesions form when the pathogenic bacterium binds to the intestinal epithelium, remodels the brush border, and injects effector proteins into the host cell via a T3SS. These effectors then influence the activity of host actin nucleation factors, N-WASP and Arp2/3, to promote actin polymerization. This results in the formation of an actin pedestal that raises the bacterium above neighboring epithelial cells to further assist its pathogenesis (112). Both innate and adaptive immune responses are required to control C. rodentium infection since mice deficient in Rag1 succumb to infection (113). C. rodentium infection leads to microbiota dysbiosis and the development of colitis (114, 115). Interestingly, microbiota dysbiosis is a key factor that influences the susceptibility of C. rodentium infection and immune responses (115, 116). Myd88-dependent TLR2 and TLR4 signaling is stimulated in epithelial and myeloid cells to recognize bacterial PAMPs. DCs and macrophages secrete proinflammatory cytokines including IL-12, IL-6, IL-23, and TNF-α in response to the activation of PRRs. Th17-, Th22-, and ILC3-derived IL-22 plays an important role in regulating C. rodentium infection (117, 118). IL-23 is required for provide protection against C. rodentium infection in a IL-22 dependent manner117. IL-22-dependent induction of antimicrobial peptide Reg3γ has been suggested to regulate C. rodentium infection. However, a recent study shows that Reg3γ <sup>−</sup>/<sup>−</sup> mice are equally susceptible to infection. It is possible that another Reg3 family member of antimicrobial peptide compensated for Reg3γ. It remains unclear how IL-22 regulates C. rodentium infection. Since IL-22Ra1 is expressed on absorptive cells (enterocytes), secretory cells (goblet, Paneth), and stem cells of intestine, future work should be directed toward understanding the effects of IL-22 on multiple intestinal cell lineages (119).

Yersinia species are Gram-negative bacterium, among them Y. pestis, Y. pseudotuberculosis, and Y. enterocolitica, are pathogenic to human. Y. pseudotuberculosis and Y. enterocolitica cause yersiniosis which leads to gastroenteritis and mesenteric lymphadenitis and may be fatal if disseminated into liver and spleen. Similar to Salmonella and other pathogens, Yersinia use T3SS to inject toxin (Yersinia outer protein, Yops) to cell cytoplasm. These Yops (YopE, YopJ, YopH, YopM, YopO, and YopT) disrupt intracellular signaling in macrophages which results into inhibition of cytokines (IL-1β, IL-18, required for innate immune cells recruitment) secretion and phagocytosis. Several studies show that YopE contain a dominant CD8 T cells epitope which is required to confer protection (120–124). Furthermore, CD8 T cells response are also required to provide protection from subsequent infection. CXCR1+ macrophages and/or DCdependent antigen presentation and local inflammatory conditions are critical for the development of heterogenous population (CD103+ or CD103–) of CD8 TRM cells in the intestine (125).

#### IMMUNITY IN RESPONSE TO CANCER

Colorectal cancer (CRC) consistently accounts for many of the cancer-related deaths in men and women. It is the third most common cancer in men and women and the fourth leading cause of cancer-related deaths. Although there are wide variations in global incidence and mortality rates related to CRC, rapidly transitioning countries (indicated by a medium to high human development index) tend to display increased CRC incidence and mortality (122). Over the past few years, the incidence of CRC has been declining in individuals ≥50 years old but is increasing in young adults (126). The onset of colorectal cancer has been associated with various factors including genetics, intestinal microbiota, and immune activity (127–129). Particularly, the release of cytokines by APCs has been shown to influence the development of cancer cells (87, 130, 131).

CRC cells express tumor-associated antigens (TAAs) and tumor-specific antigens (TSAs). While TSAs are unique to tumors, TAAs can be present on normal cells and tumor cells. Common TAAs and TSAs include carcinoembryonic antigens (CEA), Wilm's tumor gene 1 (WT1), Muc1, Her2, and p53 (132– 134). DCs can capture bodies of killed tumor cells, process TAAs, and present these antigens via MHC I or MHC II molecules, respectively, to TCRs of CD4<sup>+</sup> T cells or CD8<sup>+</sup> T cells (135). APC activation in response to cancer can also occur via the binding of APC-expressed PRRs to tumor-derived DAMPs. DAMPs may be released by tumor cells in response to anti-tumor therapy or stress pathways. With regard to CRC, three of the major DAMPs that stimulate anti-tumorigenic responses of DCs include High Mobility Group Box 1 (HMGB1), extracellular ATP, and calreticulin (CRT). Although typically restricted to the nucleus and cytoplasm, HMGB1 is secreted by necrotic tumor cells. Released HMGB1 can bind to DC-expressed TLR4 to enhance antigen presentation (136, 137). HMGB1 can also signal via DC-expressed RAGE membrane protein to promote NF-κB signaling and DC maturation (138). ATP released from dying tumor cells can bind to the P2XY receptor of DCs. This binding promotes the recruitment of DCs to the tumor stroma. CRT is typically expressed within the endoplasmic reticulum (ER) of tumorigenic cells. Chemotherapeutic agents induce the release of reactive oxygen species and induce ER stress. This promotes the transport of CRT from the ER to the cell surface where CRT serves as a signal for DC-mediated engulfment, degradation, and antigen presentation to cytotoxic CD8<sup>+</sup> T cells (139). While DCs are vital for antitumor immune responses, cancer cells utilize various mechanisms to evade immune detection. Cancer cells can downregulate TAAs, modulate antigen processing or presentation pathways, release cytokines that promote Treg function, secrete immunosuppressive factors, and express ligands that block immune checkpoints (140). Additionally, immunosuppressive cells within the tumor microenvironment such as tumor-associated macrophages (TAMs), cancer-associated fibroblasts, and Treg cells have been shown to inhibit antitumor immunity (140).

To promote immunity against tumors, DCs and activated T cells secrete multiple cytokines. Two important anti-tumorigenic cytokines are IL-2 and IL-15. These cytokines can induce similar biological effects, and the receptors for these cytokines share similar structural features including a common γ chain and IL-2/IL-15Rβ chain (141, 142). Binding of IL-2 or IL-15 to their receptor results in the activation of Janus kinases (JAK). JAK1 is activated when IL-2 or IL-15 bind to the β chain of the receptor, but JAK3 is activated when IL-2 or IL-15 bind to the or γ chain of the receptor (141). JAKs bind to their corresponding ligand to promote receptor dimerization. Dimerized JAKs phosphorylate each other, and phosphorylated JAKs phosphorylate a conserved tyrosine residue of signal transducers and activators of transcription (STAT) molecules which then enter the nucleus to regulate transcription (141, 143). While IL-2 is secreted mainly by stimulated CD4+ T cells, it is also produced by effector CD8+ T cells, DCs, and NK cells (141, 144–148). Signaling by IL-2 can influence the functions of CD4+ T cells, CD8+ T cells, and NK cells (141, 149– 152). IL-15 is expressed on the surface of a range of cell types including monocytes, macrophages, and DCs and influences the activation of DCs, proliferation of CD8+ T cells, and development of NK cells (142, 153–155). IL-2 and IL-15 may play a role in preventing colorectal cancer and have been studied as potential immunotherapy agents (141, 148, 156). However, treatment with these cytokines may have to be supplemented. For instance, survival of melanoma patients treated with IL-2 and a melanoma vaccine displayed increased survival compared to patients who were only treated with IL-2 (157). Studies have also modified the structure of IL-2 to increase its efficiency as an immunotherapeutic agent (158). IL-2 "superkines" have increased affinity for IL-2Rβ and display increased activation of NK and CD8+ T cells. Mice that were injected with melanoma, colon carcinoma, or lung carcinoma cells and then treated with modified IL-2 displayed decreased tumorigenesis compared to mice treated with unmodified IL-2 (158). The onset of colorectal cancer has commonly been studied by the administration of azoxymethane (AOM) combined with dextran sulfate sodium (DSS) to induce colitis-associated carcinoma in mice. Compared to wild-type mice treated with AOM/DSS, Il15−/<sup>−</sup> mice treated with AOM/DSS display increased tumorigenesis and decreased survival159. CD11c-Il15 mice display reconstituted production of IL-15 in APCs. Unlike Il15−/<sup>−</sup> mice which display decreased levels of NK and CD8+ T cells, CD11c-Il15 mice display similar levels of NK and CD8+ T cells as wild-type mice. CD11c-Il15 mice also display reduced AOM/DSS-induced tumorigenesis (159). While IL-2 and IL-15 have been shown to reduce the onset of cancer, other studies have shown that they may have little therapeutic effects or may even help promote tumorigenesis (141, 148). Thus, further studies need to be conducted to elucidate the relationship between these cytokines and the onset of cancer.

The presence of different T cell classes in the tumor microenvironment has been associated with changes in severity of prognosis. High levels of Th17 cells and expression of Th17 genes (including IL17A and Rorc) are associated with a poor prognosis for CRC, but patients with high expression of Th1 cytotoxic genes (including Ccl5, Stat1, and Il27) had improved occurrence of disease-free survival (160, 161). Along with being secreted by T cells, IL-27 is secreted by DCs and macrophages in response to bacterial and parasitic pathogenesis to inhibit Th1, Th2, and Th17 cell development and inflammatory responses (130, 162). Enterocyte-specific knockdown of IL-17RA, a common receptor for IL-17A, IL-17F, and IL-17C resulted in reduced tumor formation, thereby indicating a direct role for IL-17A or other IL-17 family member cytokines in tumorigenesis (163). Upregulation of IL-17A in colon tumors is dependent on upregulated IL-23 signaling by CD11b<sup>+</sup> cells (164). This increase in IL-23 by DCs, and subsequent increase in IL-17A, may be due to: (1) the microbiota and (2) an impaired epithelial barrier. Compared to the microbiota of healthy individuals, CRC patients possess an altered microbiota distinguishable by the presence of bacteria, such as Fusobacterium nucleatum, that promote the proliferation of CRC cells (165). Increased IL-23 may be caused by increased TLR/MyD88 signaling induced by the intestinal microbiota since knockout mice in Myd88−/−and triple knockout mice in Tlr2,4,9−/<sup>−</sup> display reduced tumor growth. Similarly, mice treated with antibiotics to induce a depleted commensal microbiota displayed reduced tumor size and expression of IL-23 and IL-17A (164). CRC cells have also been shown to be "leaky" since they express defective barrier proteins including mucin 2 and junctional adhesion molecules-A and -B (164). This potentially allows for the increased passage of microbes and microbial products into cancer cells to stimulate TLR signaling and, in turn, IL-23 expression by DCs which then activates Th17 cells to release IL-17A (164). Expression of IL-17 has also been linked to tumorigenic processes including the upregulation of proangiogenic factors and neovascularization (**Figure 2**) (166, 167).

IL-6 is another pro-inflammatory cytokine. IL-6 binding to Janus kinase promotes STAT3 activation (168). STAT3 then enters the nucleus to increase the transcription of anti-apoptotic genes including Bcl2 and Mcl1 and metastatic genes including Mmp1 and Mmp2 (169–171). Elevated levels of IL-6 have been associated with various cancers including lung, liver, pancreatic, and colorectal (172). Likewise, IL-6 signaling and the subsequent activation of STAT3 have been associated with the translocation of hMSH3, a DNA mismatch repair protein, from the nucleus into the cytosol (173). This translocation allows for the accumulation of tetranucleotide frameshift mutations to promote CRC (173). Some immune cells that produce IL-6 are DCs, macrophages, and Th17 cells (87, 131, 174). IL-23 has been shown to induce production of IL-6 by Th17 cells (87, 175). In turn, IL-6 aids in the differentiation of Th17 cells by promoting expression of IL-21 which induces production of IL-17A via activation of the transcription factors STAT3 and RORγt (176). IL-6 activation also promotes tumorigenesis by altering gene expression for the induction of cell proliferation, progression of the epithelial to mesenchymal transition, and resistance to anti-cancer drugs such as erlotinib (168). Signaling by IL-11, a member of the IL-6 family of cytokines, also promotes the development of CRC by activating STAT3 (177). IL-11

has been identified as a more dominant activator of STAT3 signaling and inducer of CRC than IL-6. Likewise, inhibition of IL-11 signaling has been linked with reduced tumor growth and burden. However, while hematopoietic cells produce IL-11, the secretion of IL-11 by non-hematopoietic cells is more so responsible for the progression of CRC (177). Along with IL-6 and IL-11, elevated levels of IL-22 have been associated with colon cancer. IL-22 has been shown to activate STAT3 signaling and further enhance the development of colon cancer (178, 179). Knockout mice of IL-22 binding protein, a soluble receptor for IL-22, display IL-22-dependent tumor growth (180). ILC3s have been shown to be a major source of IL-22 in mice for promoting tumorigenesis; however, human studies indicate that CCR6<sup>+</sup> Th17 cells are major producers of IL-22 (179, 181). Currently, IL-6 and STAT3 inhibitors are being investigated as potential therapeutic agents to inhibit tumor growth (182).

secretion of IL-17A by Th17 cells. IL-17A, in turn, is associated with angiogenesis.

In addition to STAT3, TNF-α is a critical cytokine for its role in regulating signaling pathways and inflammatory responses, especially in relation to pro- and anti-tumorigenic activity. Like other cytokines, TNF-α is produced by a range of cell types including epithelial cells, macrophages, neutrophils, NK cells, and T cells (183–186). Elevated levels of TNF-α have been associated with increased levels of infiltrating immature myeloid cells which develop into pro-tumorigenic cells such as tumor associated macrophages and neutrophils (TAMs and TANs) once they are exposed to the tumor microenvironment (187, 188). In some cell types, stimulation with TNF-α has been shown to encourage monocyte recruitment to tumors by promoting the expression of monocyte chemoattractant protein-1 (CCL2) (189, 190). Various cell types produce CCL2 including smooth muscle and mesenchymal stromal cells, and other monocytes and macrophages (191–193). TAMs produce cytokines such as IL-10 that inhibit T cell function. IL-10 has been shown to increase the expression of Mgat5, a glycosyltransferase that increases branching of glycoproteins on the surface of CD8<sup>+</sup> T cells and has been associated with the progression of cancer (194, 195). Increased glycoprotein branching hinders direct T cell interactions with APCs and decreases TCR signaling to promote poor sensitivity to antigens (194). While increased expression of IL-10 produced by TAMs has been correlated with the progression of certain cancers including non-small cell lung carcinoma, the role of IL-10 in promoting CRC is not fully understood as it displays both protective and inflammatory roles (196–198).

Programmed death 1 (PD-1) is expressed on the surface of T cells and interacts with PD-L1, a ligand mainly expressed by macrophages, activated DCs and cancer cells (199). Likewise, cytotoxic T lymphocyte-associated antigen 4 (CTLA4) is expressed on naïve or memory T cells and interacts with CD80/CD86 on the surface of DCs. Although this review will not cover interactions between PD-1 and PD-L1 or CTLA4 and CD80/86 in depth, interactions between PD-1 or CTLA4 and its ligand act to keep T cell activity in check by preventing overactive immune responses (199). Immunotherapy utilizing anti-PD-1 or anti-CTLA4 monoclonal antibodies is effective in fighting the progression of various cancers, and the intestinal microbiota may influence whether certain individuals respond more positively to therapy (199, 200). For instance, increased levels of Akkermansia muciniphila, a commensal bacterium, were observed in patients who favorably reacted to treatment (200).

# CONCLUSION

While much has been done to determine how APCs interact with T cells, several aspects should be expanded. First, how bacteria influence APC interactions with immune cells to combat microbial infections should be further investigated. The interplay between microbiota-derived signaling and the onset of cancers also requires better understanding. Although not widely covered in this review, potential immunotherapies regarding altering APC or microbiota function should be studied. Likewise, a better understanding of how tumors develop strategies to take advantage of immune cells to promote tumorigenesis and dampen T cell responses, especially regarding the PD-1 system, is required. Lastly, more investigation is needed to determine how factors derived from the microbiota, including SCFAs, and intestinal epithelial cells shape APC and immune cell interactions.

#### REFERENCES


# AUTHOR CONTRIBUTIONS

SG wrote the manuscript and designed the figures. PK edited the text.

#### FUNDING

This work was supported by the Crohn's and Colitis Foundation (476637), National Multiple Sclerosis Society (PP-1709- 29192), and The Research Foundation of SUNY to PK, and the National Science Foundation Graduate Research Fellowship to SG.

#### ACKNOWLEDGMENTS

We would like to acknowledge Michael Beaupre and Xun Lin for their input.


antibody, for moderate to severe Crohn's disease: unexpected results of a randomised, double-blind placebo-controlled trial. Gut. (2012) 61:1693–700. doi: 10.1136/gutjnl-2011-301668


gut commensal molecule via both innate and adaptive mechanisms. Cell Host Microbe. (2014) 15:413–23. doi: 10.1016/j.chom.2014.03.006


**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 © 2019 Gaudino 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.

# Adipose Tissue in Persons With HIV Is Enriched for CD4<sup>+</sup> T Effector Memory and T Effector Memory RA<sup>+</sup> Cells, Which Show Higher CD69 Expression and CD57, CX3CR1, GPR56 Co-expression With Increasing Glucose Intolerance

Celestine N. Wanjalla1,2 \* † , Wyatt J. McDonnell 1,2,3,4 \* † , Louise Barnett <sup>5</sup> , Joshua D. Simmons <sup>5</sup> , Briana D. Furch1,5, Morgan C. Lima1,5, Beverly O. Woodward1,5 , Run Fan<sup>6</sup> , Ye Fei <sup>6</sup> , Paxton G. Baker <sup>7</sup> , Ramesh Ram<sup>8</sup> , Mark A. Pilkinton1,2 , Mona Mashayekhi 9,10, Nancy J. Brown<sup>9</sup> , Simon A. Mallal 1,2,5,7,8, Spyros A. Kalams 1,2,5 and John R. Koethe1,2 \*

*<sup>1</sup> Division of Infectious Diseases, Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>2</sup> Center for Translational Immunology and Infectious Disease, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>3</sup> Vanderbilt Vaccine Center, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>4</sup> Department of Pathology, Microbiology, and Immunology, Vanderbilt University, Nashville, TN, United States, <sup>5</sup> Tennessee Center for AIDS Research, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>6</sup> Department of Biostatistics, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>7</sup> VANTAGE, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>8</sup> Institute for Immunology and Infectious Diseases, Murdoch University, Perth, WA, Australia, <sup>9</sup> Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, United States, <sup>10</sup> Division of Diabetes, Endocrinology and Metabolism, Vanderbilt University, Nashville, TN, United States*

Chronic T cell activation and accelerated immune senescence are hallmarks of HIV infection, which may contribute to the increased risk of cardiometabolic diseases in people living with HIV (PLWH). T lymphocytes play a central role in modulating adipose tissue inflammation and, by extension, adipocyte energy storage and release. Here, we assessed the CD4<sup>+</sup> and CD8<sup>+</sup> T cell profiles in the subcutaneous adipose tissue (SAT) and blood of non-diabetic (*n* = 9; fasting blood glucose [FBG] < 100 mg/dL), pre-diabetic (*n* = 8; FBG = 100–125 mg/dL) and diabetic (*n* = 9; FBG ≥ 126 mg/dL) PLWH, in addition to non- and pre-diabetic, HIV-negative controls (*n* = 8). SAT was collected by liposuction and T cells were extracted by collagenase digestion. The proportion of naïve (TNai) CD45RO−CCR7+, effector memory (TEM) CD45RO+CCR7−, central memory (TCM) CD45RO+CCR7+, and effector memory revertant RA+(TEMRA) CD45RO−CCR7<sup>−</sup> CD4<sup>+</sup> and CD8<sup>+</sup> T cells were measured by flow cytometry. CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA were significantly enriched in SAT of PLWH compared to blood. The proportions of SAT CD4<sup>+</sup> and CD8<sup>+</sup> memory subsets were similar across metabolic status categories in the PLWH, but CD4<sup>+</sup> T cell expression of the CD69 early-activation and tissue residence marker, particularly on TEM cells, increased with progressive glucose intolerance. Use of t-distributed Stochastic Neighbor

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Dorothy Ellen Lewis, University of Texas, United States Clovis Steve Palmer, Burnet Institute, Australia*

#### \*Correspondence:

*Celestine N. Wanjalla celestine.wanjalla@vumc.org Wyatt J. McDonnell wyatt.j.mcdonnell@vanderbilt.edu John R. Koethe john.r.koethe@vumc.org*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *28 October 2018* Accepted: *15 February 2019* Published: *19 March 2019*

#### Citation:

*Wanjalla CN, McDonnell WJ, Barnett L, Simmons JD, Furch BD, Lima MC, Woodward BO, Fan R, Fei Y, Baker PG, Ram R, Pilkinton MA, Mashayekhi M, Brown NJ, Mallal SA, Kalams SA and Koethe JR (2019) Adipose Tissue in Persons With HIV Is Enriched for CD4*<sup>+</sup> *T Effector Memory and T Effector Memory RA*<sup>+</sup> *Cells, Which Show Higher CD69 Expression and CD57, CX3CR1, GPR56 Co-expression With Increasing Glucose Intolerance. Front. Immunol. 10:408. doi: 10.3389/fimmu.2019.00408* Embedding (t-SNE) identified a separate group of predominantly CD69lo TEM and TEMRA cells co-expressing CD57, CX3CR1, and GPR56, which were significantly greater in diabetics compared to non-diabetics. Expression of the CX3CR1 and GPR56 markers indicate these TEM and TEMRA cells may have anti-viral specificity. Compared to HIVnegative controls, SAT from PLWH had an increased CD8:CD4 ratio, but the distribution of CD4<sup>+</sup> and CD8<sup>+</sup> memory subsets was similar irrespective of HIV status. Finally, whole adipose tissue from PLWH had significantly higher expression of TLR2, TLR8, and multiple chemokines potentially relevant to immune cell homing compared to HIV-negative controls with similar glucose tolerance.

Keywords: HIV—human immunodeficiency virus, adipose tissue, diabetes mellitus, TEMRA, T effector memory cells, GPR56, CX3CR1, memory T cells

#### INTRODUCTION

People living with human immunodeficiency virus (HIV) are at an increased risk of developing insulin resistance and overt diabetes mellitus, but the factors contributing to the high prevalence of metabolic disease in the HIV population are not fully understood (1–3). Since HIV was identified as the cause of acquired immune deficiency syndrome (AIDS) in the early 1980s, the metabolic consequences of altered adipose tissue function in people living with HIV (PLWH) have been a major research focus (4). Early studies identified accelerated lipolysis and hepatic lipogenesis as central energy metabolism abnormalities in untreated HIV infection (5–7), while earlygeneration nucleoside reverse transcriptase inhibitors (NRTIs) caused adipocyte mitochondrial damage and adipose tissue fibrosis (8–12), and treatment with early protease inhibitors was accompanied by accumulation of visceral adipose tissue (VAT), hyperlipidemia, and insulin resistance (13–16). More recently, several studies describe profound changes in adipose tissue T cell populations during chronic HIV and simian immunodeficiency virus (SIV; a non-human primate virus similar to HIV) infections, which may influence adipose tissue metabolic function. These include changes in T cell surface marker phenotypes, cytokine production, antigen receptor repertoire, and capacity for latent infection with HIV or SIV provirus (17–22). Notably, several studies found that HIV and SIV were accompanied by a substantial increase in the proportion of adipose CD8<sup>+</sup> T cells relative to CD4<sup>+</sup> T cells, which is strikingly similar to the enrichment in CD8<sup>+</sup> T cells also described in obesity (23–25).

Adipose tissue is a complex and vascularized cellular amalgam, comprised of multipotent adipocyte progenitors, mature adipocytes, fibroblasts, and immune cells of the adaptive and innate lineages. A diverse group of immune cells collectively identify and eliminate the range of viruses and other pathogens which can infiltrate adipose tissue, and these processes impact local levels of pro-inflammatory and other cytokines with subsequent effects on adipocyte regulation and energy storage and release. T lymphocytes play several beneficial and detrimental roles within this environment. Studies of obese humans and animals demonstrate an increase in adipose tissue CD8<sup>+</sup> T cells and CD4<sup>+</sup> TH1 cells, a decrease in T regulatory cells, and an increase in M1-phenotype (CD68+, tumor necrosis factor-α [TNF-α], interleukin [IL]-6, IL-12, and IL-23-producing) pro-inflammatory macrophages (23, 24, 26, 27) compared to non-obese controls. In animal models of obesity, the infiltration of CD8<sup>+</sup> T cells into adipose tissue precedes the recruitment of macrophages (23); and the resulting increase in local IL-6, TNF-α and other inflammatory mediators act on adipocyte surface receptors and other mechanisms to inhibit insulin signaling via reduced insulin receptor substrate-1 (IRS-1), phosphoinositide 3-kinase p85α, and glucose transporter type 4 (GLUT4) expression (17, 18, 23, 28–32).

The increase in the CD8<sup>+</sup> to CD4<sup>+</sup> T cell ratio observed in HIV infection is like that seen in diet-induced obesity, though the mechanisms underlying the accumulation of CD8<sup>+</sup> T cells in the adipose tissue of PLWH are not well-defined. Studies in macaques suggest that the high proportion of CD8<sup>+</sup> T cells is not due to depletion of CD4<sup>+</sup> T cells (18). Although lack of Ki-67 expression in adipose tissue T cells has been interpreted as lack of evidence for in situ proliferation, greater CD8<sup>+</sup> TCR clonality in subcutaneous adipose tissue (SAT) implies antigen specificity might drive the increase rather than stochastic recruitment of circulating CD8<sup>+</sup> T cells. This is further supported by the finding that CD8<sup>+</sup> and CD4<sup>+</sup> T cells in adipose tissue predominantly display a memory phenotype with increased levels of CD69 expression compared to those in blood (17, 18).

While prior studies have shown enrichment of CD8<sup>+</sup> over CD4<sup>+</sup> T cells in adipose tissue after HIV infection, there is a paucity of data on whether a particular subset of cells underlies this change, and whether adipose tissue T cell profiles differ according to insulin sensitivity in PLWH (as might be expected given prior findings in obesity-related insulin resistance). In this study, we hypothesized that the enrichment of CD8<sup>+</sup> T cells in the adipose tissue of PLWH could be attributed to an overrepresentation of one or a few memory cell subtypes, and that greater CD8<sup>+</sup> and CD4<sup>+</sup> T cell activation would characterize the adipose tissue of diabetic PLWH. We evaluated SAT CD4<sup>+</sup> and CD8<sup>+</sup> T cell subsets (including naïve cells, activated cells, and central memory [TCM], effector memory [TEM], and effector

memory revertant RA<sup>+</sup> [TEMRA] cells) in PLWH vs. HIVnegative controls, and among diabetic vs. non-diabetic PLWH.

#### MATERIALS AND METHODS

#### Study Participants

We enrolled 26 PLWH on long-term antiretroviral therapy (ART) with sustained virologic suppression from the Vanderbilt Comprehensive Care Clinic between August 2017 and June 2018. Hemoglobin A1c (HbA1c) and fasting blood glucose (FBG) were used to classify participants as non-diabetic (n = 9; HbA1c < 5.7% and FBG < 100 mg/dL), pre-diabetic (n = 8; HbA1c 5.7–6.5% and/or FBG 100–125 mg/dL), and diabetic (n = 9; HbA1c ≥ 6.5% and/or FBG ≥ 126 mg/dL, and on antidiabetes medications). A group of 8 HIV-negative, non- and pre-diabetic controls were enrolled from the community. The PLWH were on ART for at least 18 months, had HIV-1 RNA <50 copies/ml for the prior 12 months, CD4<sup>+</sup> count >350 cells/µl, and had no known inflammatory or rheumatologic conditions. We excluded persons with self-reported heavy alcohol use (defined as >11 drinks/week), any cocaine/amphetamine use, and those receiving corticosteroids or growth hormone.

All visits occurred in the Vanderbilt Comprehensive Care Clinic research suite or the Vanderbilt Clinical Research Center between 8 and 11 am. Participants fasted for a minimum of 8 h prior to blood collection for laboratory measurements and peripheral blood mononuclear cell (PBMC) separation (PLWH only). Blood glucose, HbA1c, high-sensitivity C-reactive protein (hsCRP), low-density lipoprotein (LDL), triglycerides, and highdensity lipoprotein (HDL) were measured in the fasting blood samples at the Vanderbilt Clinical Chemistry Laboratory.

#### Adipose Tissue Biopsy and T Cell Extraction

SAT biopsies were collected ∼3 cm to the right of the umbilicus after anesthetizing the skin with lidocaine and infiltrating 40 ml of sterile saline and lidocaine into the subcutaneous adipose tissue as tumescent fluid. We collected ∼5 grams of adipose tissue using a 2.1 mm blunt, side-ported liposuction catheter (Tulip CellFriendlyTM GEMS system Miller Harvester, Tulip Medical Products) designed for the extraction of viable adipocytes and stromal vascular cells during cosmetic adipose tissue transfer procedures (33). With this approach, adipose tissue is recovered in droplets generally <3 mm in diameter, limiting the need to mechanically mince the sample, and the tissue is placed in 40– 50 cc of cold saline and mixed to rinse. Any visible clots are removed before the sample is transferred to a 70µm mesh filter for repeated saline rinses with constant stirring. The adipose tissue is then placed in a gentleMACSTM Dissociator (Miltenyi Biotec) followed by incubation with collagenase (Roche Catalog #11088866001). Mononuclear cells were separated by Ficoll-Paque Plus density gradient and cryopreserved in fetal bovine serum (FBS) with 10% DMSO.

#### Flow Cytometry Analysis

PBMC and SAT mononuclear cell aliquots were processed and stained as previously published (21). In brief, matched cryopreserved PBMCs and SAT cells were quickly thawed at 37◦C and suspended in R10 media (RPMI with 10% FBS). These were then washed once in phosphate buffered saline (PBS) and stained with multiple fluorescent tagged antibodies: CD3- BV786 (Clone SK7; BD Biosciences #563800), CD4-PcPCy5.5 (Clone RPA-T4; BD Biosciences #560650), CD8-A700 (Clone PRA-T8; BD Biosciences #557945), CD57-FITC (Lot 4182924; BD Pharmingen # 555619), CX3CR1-PE (Clone 2A9-1; BD Biosciences #565796), CD45RO-PECF594 (Clone UCHL1; BD Biosciences #562299), CD14-V500 (Clone M5E2, BD Biosciences #561391), CD19-V500 (Clone HIB19; BD Biosciences #561121), LIVE/DEAD Fixable Aqua (ThermoFisher; L34957), CD69-APC (Clone FN50; BD Biosciences #560711), CCR7-V450 (Clone 150503; BD Biosciences #562555), GPR56-PECy7 (Clone CG4; BioLegend #358205), and HLA-DR APC Cy7 (Clone G46.6; BD Biosciences #561358). CCR7 and CX3CR1 antibody stains were performed at 37◦C; the remainder was stained at room temperature. Cells were analyzed on a 4-laser FACSAria III (all samples from PLWH) or 5-laser LSRII (HIV-negative SAT samples) (BD Biosciences, San Jose, CA). Bead calibration was used to standardize runs done on different days. Flow cytometry data were analyzed using FlowJo software (version 10.4.1) and Cytobank (version 6.3.1) (34, 35).

The T cell memory populations in our study are defined as naive T cells (TNai, CD45RO−CCR7+), T effector memory (TEM, CD45RO+CCR7−), T central memory (TCM, CD45RO+CCR7+), and T effector memory revertant/re-expressing CD45-RA (TEMRA, CD45RO−CCR7−). Representative gating strategies in **Supplementary Figure 1** and **Figure 1** show the phenotypic markers used to define the memory subsets (CD45RO and CCR7). Total memory cells are a combined group consisting of TEM, TCM, and TEMRA.

#### Adipose Tissue Gene Expression

Adipose tissue was rinsed, placed in cryovials and snap-frozen in liquid nitrogen immediately after collection for subsequent mRNA expression assays. mRNA was extracted after mechanical lysis with the Qiagen RNeasy Lipid Tissue Kit. We used the Nanostring nCounter Plex<sup>2</sup> human inflammation panel to quantify mRNA expression of over 250 genes spanning a broad range of relevant immune pathways including interleukin signaling, Ras, T-cell markers, and Toll-like receptor signaling. Adipose tissue mRNA expression levels were normalized using 14 synthetic spike-ins (6 positive controls and 8 negative controls) and 6 cellular housekeeping genes included in the assay (GAPDH, GUSB, HPRT1, PGK1, TUBB, and CLTC). We first calculated the coefficient of variation (CV) for the control genes. The CV of the positive controls is proportional to the technical variability introduced by the nCounter platform. The CV for the housekeeping controls is proportional to the confounding biological variation due to sample input. The mean endogenous CV shows the global noise of experimentally observed genes. After evaluating different normalization approaches based on CV values, we developed a normalization strategy including the following steps. First the background count levels were calculated using the mean of negative controls, then subtracted from each sample. The normalization factor for sample/RNA content was

calculated using the geometric mean of a set of pre-specified annotated housekeeping genes. The algorithm normalizes for sample or RNA content, i.e., "pipetting" fluctuations, using the geometric mean of pre-specified annotated housekeeping genes. The count data were then divided by the normalization factor to generate counts normalized to the geometric mean of housekeeping genes. None of the housekeeping genes differed significantly in their distribution between study groups.

#### Statistical Analyses

For comparisons, PLWH were grouped according to metabolic status as non-diabetic (n = 9), pre-diabetic (n = 8), and diabetic (n = 9). The 8 HIV-negative controls were matched to 9 PLWH with similar HbA1c and body mass index (BMI) values to yield a similar comparison group (one control was matched to two PLWH). Percentages of CD4<sup>+</sup> and CD8<sup>+</sup> subsets were compared between paired blood and adipose tissue samples using Wilcoxon signed-rank tests for paired data. Differences between the PLWH metabolic groups, and between HIVnegative controls and the PLWH, were calculated using Mann-Whitney and Kruskal Wallace tests. When significant betweengroup comparisons were noted among the PLWH, univariable and multivariable linear regression were used to assess the relationship of cellular populations with progressive glucose intolerance. Statistical analyses and graphs were performed using SPSS (IBM, Armonk, NY), R (www.r-project.org), and GraphPad Prism 7 (GraphPad Software, La Jolla, CA),

Adipose tissue genes were grouped according to immune system pathways specified in the NanoString kit, and normalized mRNA levels were compared between the PLWH and HIVnegative participants. DEseq2 was used to detect differential expression between two groups based on the normalized count data, taking into account technical and biological variability (36, 37). Differences in gene expression were calculated as fold-changes, and p-values were adjusted for multiple comparisons using the Benjamini-Hochberg procedure. Analyses were performed using R. Volcano plots displaying the NanoString data were generated using XL-STAT.

# RESULTS

### Clinical and Demographic Characteristics of PLWH and HIV-Negative Subjects

The non-diabetic (n = 9), pre-diabetic (n = 8), and diabetic (n = 9) groups of PLWH are compared in **Table 1**. Age, race and sex were similar across the groups, as were CD4<sup>+</sup> nadir, duration of ART, and the proportions receiving integrase inhibitor-based regimens (p > 0.05 for all comparisons). BMI and waist circumference increased with progressive glucose intolerance (p = 0.04 for both). The HIV-negative controls (n = 8) and comparator PLWH (9 out of 26 total) were similar in age, BMI, and HbA1c values (p > 0.05 for all), though the controls were more likely to TABLE 1 | Cohort demographic and clinical characteristics according to glucose tolerance.


*ART, antiretrovirals; AZT, azidothymidine; BMI, body mass index; FG, fasting glucose; HDL, high-density lipoproteins; IQR, interquartile range; LDL, low-density lipoproteins; TG, triglycerides; yr, year.*

*The bolded values are p-values* <*0.05, indicating statistical significance.*

be female and white (**Table 2**). Bold values in **Tables 1**, **2** indicate p < 0.05.

#### Adipose Tissue From PLWH Is Enriched in Memory CD8<sup>+</sup> and CD4<sup>+</sup> T Cells

We found no difference in the percentage of total CD8<sup>+</sup> and CD4<sup>+</sup> T cells (gated on CD3<sup>+</sup> T cells) between SAT and peripheral blood from PLWH **Figure 1A**. We also observed, as previously described by other groups, a higher percentage of total CD4<sup>+</sup> memory T cells (**Figure 1B**) and CD8<sup>+</sup> memory T cells (**Figure 1C**) in SAT compared to peripheral blood (p < 0.0001) (17).

### SAT From PLWH Is Enriched in CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA Cells Compared to Blood

Despite the growing body of literature on adipose tissue immune cells, the distribution of memory T cell subsets in adipose tissue from PLWH has not been characterized, and the role of memory TABLE 2 | Comparison of HIV-negative participants and comparator non- and pre-diabetic PLWH.


*The bolded values are p-values* <*0.05, indicating statistical significance.*

T cell subsets within tissue compartments in general is not wellunderstood. A prior study assessed the distribution of CD4<sup>+</sup> and CD8<sup>+</sup> TNai, TEM, TCM, and TEMRA subsets in lung, spleen, colon, ileum, jejunum, and lymph nodes, but not adipose tissue, Wanjalla et al. Adipose Tissue T Cell Subsets

from healthy donors, and found that CD4<sup>+</sup> TNai, TEM, and TCM proportions in lung and pulmonary and mesenteric lymph nodes were similar to the proportion in blood, while jejunum, ileum, and colon were roughly 3- to 4-fold enriched in TEM (38). In all tissues, CD4<sup>+</sup> TEMRA were relatively sparse (<10%). In contrast, ∼30% of CD8<sup>+</sup> T cells in blood, spleen, and lung had a TEMRA phenotype, while the CD8<sup>+</sup> T cells in jejunum, ileum, and colon were overwhelmingly TEM (38). A subsequent study in the adipose tissue of mice found approximately equal proportions of CD4<sup>+</sup> TEM and CD4<sup>+</sup> TRM cells (a "resident memory" phenotype defined by the authors as CD69 expression on TEM cells) in adipose tissue samples, and fewer TCM (39). However, CD8<sup>+</sup> memory cells were predominantly TRM in SAT, with similar, lower proportions of TEM and TCM. The same study looked at mesenteric adipose tissue in macaques and again found high levels of CD8<sup>+</sup> TRM, while CD4<sup>+</sup> cells were primarily TNai and TCM, with low levels (<10%) of TEM and TRM, potentially highlighting important differences between species.

We first assessed CD4<sup>+</sup> and CD8<sup>+</sup> memory subsets in SAT and blood from all 26 PLWH. Multiparameter gates were used to quantify memory CD4<sup>+</sup> and CD8<sup>+</sup> T cells within each sample as shown in **Supplementary Figures 1, 9**. Each sample had >100,000 ungated cells (**Supplementary Figures 10A–C**). SAT was significantly enriched in CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cells (**Figures 2A,B**) compared to blood but had fewer TNai and TCM cells. We then compared the average TNai, TCM, TEM, and TEMRA fractions within SAT and blood among the 26 PLWH (**Supplementary Figure 2**). We found that CD4<sup>+</sup> T cells in SAT can be ranked by frequency as TEM > TCM > TNai > TEMRA, whereas peripheral blood contains CD4<sup>+</sup> TCM > TNai > TEM > TEMRA. In contrast, SAT CD8<sup>+</sup> T cells are primarily TEMRA followed by TEM.

### The Relative Distribution of SAT Memory T Cell Subsets Does Not Differ With Metabolic Status in PLWH

The study of adipose tissue T cells in mice by Han et al. also found that recall responses of adipose tissue memory T cells were enhanced after antigen re-challenge, with downregulation of several metabolic pathways in whole adipose tissue (including lipid biosynthesis and cholesterol and long-chain fatty-acyl-CoA metabolic processes) and a detectable reduction in serum levels of adiponectin and cholesterol (39). These findings suggest a possible mechanism by which the accumulation, and subsequent stimulation, of memory cells in adipose may disrupt metabolic homeostasis.

In light of prior animal studies indicating a potential contribution of effector memory T cells to inflammation and impaired metabolic function in adipose tissue, we sought to characterize the distribution of SAT memory T cell subsets in PLWH to determine if there were clear differences between non-diabetics, pre-diabetics, and diabetics in the relative proportions of TNai, TCM, TEM, and TEMRA cells. We utilized tdistributed Stochastic Neighbor Embedding (t-SNE) to visualize groups of adipose tissue and blood CD4<sup>+</sup> and CD8<sup>+</sup> T cell populations based on 12-color flow cytometry staining. **Supplementary Figure 3** shows the distribution of TNai, TCM, TEM, and TEMRA cells in CD4<sup>+</sup> and CD8<sup>+</sup> T cells in the adipose tissue and blood from four representative non-diabetic and diabetic PLWH. The plots showed fewer TNai cells and enriched CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cell fractions in the SAT compared to the matched peripheral blood.

The t-SNE findings were congruous with our results from 2-dimensional flow cytometry gating. For CD4<sup>+</sup> T cells, the proportion of TNai and TCM cells were significantly lower compared to blood in all three groups, while the proportion of TEM and TEMRA cells was significantly higher (**Figure 3A**). However, none of the adipose memory subsets in the prediabetics or diabetics were significantly different from the nondiabetics in pairwise comparisons. There was less consistency in the relative proportions of CD8<sup>+</sup> memory cells in the blood and adipose tissue. While CD8<sup>+</sup> TNai and TCM cells were significantly lower in the SAT in all three groups (**Figure 3B**), SAT CD8<sup>+</sup> TEM cells were only significantly higher compared to blood in the diabetics, and SAT TEMRA cells were only significantly higher in the non- and pre-diabetic groups. In summary, these findings suggest that while adipose tissue is enriched in CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cells compared to blood, the relative distribution of naïve and memory T cells within SAT does not markedly vary by metabolic status in PLWH.

Next, we assessed SAT memory T cells according to BMI category and age (**Supplementary Figures 4**, **5**). We observed no significant differences in SAT CD4<sup>+</sup> and CD8<sup>+</sup> memory T cell populations according to BMI status. When we compared naïve and memory T cell subsets by age, we found the proportion of CD4<sup>+</sup> TNai cells in SAT from PLWH <35 years (34% of total CD4<sup>+</sup> cells) was significantly higher than in older subjects 35– 55 years (12%, p < 0.05) and those >55 years (16%, p < 0.01). The relative reduction in CD4<sup>+</sup> TNai cells appeared principally due to an increase in the proportions of TEM cells at higher ages; TEM cells constituted 29% of total CD4<sup>+</sup> cells in participants <35 years, but rose to 50% in those 35–55 years (p = 0.06) and 47% in those >55 years (p < 0.01). These differences were less pronounced for CD8<sup>+</sup> T cells; the TEM proportion in PLWH >55 years was significantly higher than those <35 years (38% vs. 22%, p < 0.05), but the difference was not significant for those ages 35– 55. Our findings were similar to a prior study of health donors, which found the proportion of total CD4<sup>+</sup> memory (CD45RO+) T cells in lung and mesenteric lymph nodes, spleen, ileum and colon also rose with increasing age (38).

### CD4<sup>+</sup> T Cell CD69 Expression Increases With Progressive Glucose Intolerance in PLWH

CD69 is an inducible, early-activation indicator which also serves as a putative tissue-resident marker on memory T cells in human, as well as in animal, mucosal and lymphoid tissues (38, 40, 41), but is largely absent on memory T cells in blood (38). CD69 has been used as a marker of adipose tissue resident memory cells in animals (39), including in SIV-infected macaques (18), as well as in prior studies of PLWH (17, 19). At present, there are few

FIGURE 2 | Subcutaneous adipose tissue has a higher percentage of TEM (CD45RO<sup>+</sup> CCR7−) and TEMRA (CD45RO<sup>−</sup> CCR7−) cells compared to blood in PLWH. (A) The bar graphs show frequencies of CD4<sup>+</sup> <sup>T</sup>Nai, TEM, TCM, and TEMRA cells in subcutaneous adipose tissue (SAT) and blood (PBMC) from all twenty-six subjects. (B) Frequencies of CD8<sup>+</sup> <sup>T</sup>Nai, TEM, TCM, and TEMRA cells in SAT and PBMC. The numbers in the bar graphs on the left indicate mean and diagonal lines indicate matched pairs of SAT and PBMCs. Wilcoxon matched pair signed test was used to calculate statistics \*\*\*\**p* < 0.0001; \*\*\**p* < 0.001.

FIGURE 3 | Analysis of CD4<sup>+</sup> and CD8<sup>+</sup> memory subsets by metabolic status in PLWH. (A) The bar graphs on the top row show frequencies of CD4<sup>+</sup> <sup>T</sup>Nai, TEM, TCM, and TEMRA cells in subcutaneous adipose tissue (SAT) and blood (PBMC) from all twenty six PLWH. Participants are grouped based on glucose tolerance: non-diabetic (Non-DM): hemoglobin A1c < 5.7% and fasting glucose < 100 mg/dL (*n* = 9); Pre-DM: hemoglobin A1c 5.7–6.4% or fasting glucose 100–125 mg/dL (*<sup>n</sup>* <sup>=</sup> 8); DM: hemoglobin A1c <sup>&</sup>gt; 6.5% and/or fasting glucose <sup>&</sup>gt; 126 mg/dL, and on anti-diabetes medication (*<sup>n</sup>* <sup>=</sup> 9). (B) Frequencies of CD8<sup>+</sup> <sup>T</sup>Nai, TEM, TCM, and TEMRA cells in SAT and PBMC. The box and whiskers plot indicate mean ± SD. Wilcoxon matched pair signed test was used to calculate statistics between matched PBMC and SAT T cells, Mann-Whitney test was used to analyze differences in SAT and PBMC T cell subsets between metabolic groups; \*\**p* < 0.01, \**p* < 0.05.

studies on CD69 expression on adipose tissue T cells in PLWH, none of which has looked at their link with metabolic status.

We measured CD69 expression on memory T cell subsets in SAT and blood (gating strategy shown in **Supplementary Figure 6**). As reported in prior studies of PLWH, CD69 expression was present on CD4<sup>+</sup> T cell subsets from SAT but almost absent in peripheral blood (**Figure 4**) (17, 19), whereas CD69 expression on CD8<sup>+</sup> T cells from SAT

TNai, TCM, TEM, and TEMRA cells expressing the CD69 activation and putative tissue-residence marker in subcutaneous adipose tissue (SAT) and blood (PBMC). The box and whiskers plot indicate mean ± SD. Wilcoxon matched-pair rank test was used to calculate differences between PBMC and SAT. Mann-Whitney test used to calculate differences between groups; blue lines and red \* depict differences between groups \*\**p* < 0.01, \**p* < 0.05.

and blood was similar (**Supplementary Figure 11A**)**.** While we observed the relative proportions of SAT CD4<sup>+</sup> and CD8<sup>+</sup> TCM, TEM, and TEMRA cells to be similar regardless of metabolic status in PLWH (**Figure 3**), expression of CD69 on CD4<sup>+</sup> T cells rose with progressive glucose intolerance in a step-wise progression from non-diabetic, to pre-diabetic, to diabetic. Compared to non-diabetics, diabetic PLWH had significantly higher CD69 expression on total CD4<sup>+</sup> T cells, TCM, and TEM (p < 0.01 for all three), as well as TNai and TEMRA (p < 0.05 for both, **Figure 4**). Similarly, the pre-diabetics also had higher CD69 expression on SAT total CD4<sup>+</sup> T cells and TNai, TEM, and TEMRA (p < 0.05 for all) compared to non-diabetics, but not on TCM (p < 0.07). In contrast, we did not observe any significant differences in CD69 expression on CD8<sup>+</sup> T cells according to metabolic status.

Given the stepwise progression of CD69 expression on CD4<sup>+</sup> T cell subsets with rising glucose intolerance, and the significant pairwise comparisons, a linear regression model was used to assess CD69 expression according to metabolic status. Mean CD4<sup>+</sup> T cell CD69 expression for non-diabetics, prediabetics, and diabetics is shown in **Table 3**. Frequencies of CD69<sup>+</sup> cells were natural-log transformed to improve normality (Shapiro-Wilk p < 0.01 for all untransformed CD4<sup>+</sup> and CD8<sup>+</sup>

T cell subsets). Expression of CD69 on total CD4<sup>+</sup> T cells rose with progressive glucose intolerance (**Table 3**, p = 0.004), which was robust to adjustment for BMI (p = 0.03 for metabolic status) and to age (p = 0.01 for metabolic status) in separate models (data not shown). Among CD4<sup>+</sup> T cell subsets, progression from non-diabetic to diabetic groups was accompanied by increased CD69 expression on TCM (p = 0.02), TEM (p = 0.04), and TEMRA (p = 0.04) cells.

### CD57 Expression Is Higher on SAT CD4<sup>+</sup> and CD8<sup>+</sup> T Cells, but Does Not Vary With Metabolic Status in PLWH

We previously reported a higher proportion of latedifferentiated, CD57<sup>+</sup> CD8<sup>+</sup> T cells in the SAT of non-diabetic PLWH compared to blood (37 vs. 23%, p < 0.01) (21). CD57 is a terminally-sulfated glycan carbohydrate epitope found on T cells and natural killer (NK) cells which serves as a marker of late differentiation, though there is limited consensus as to whether CD57 is a marker of an inability to proliferate in response to antigen stimulation, signifies reduced replicative



\**Univariable linear regression. The association of total CD4*<sup>+</sup> *T cell CD69 expression with progressive glucose intolerance was robust to adjustment for BMI (p* = *0.03) and to age (p* = *0.01) in separate models.*

*The bolded values are p-values* <*0.05, indicating statistical significance.*

capacity, or represents an increased susceptibility to activationinduced apoptosis (42–44). Prior studies have shown that CD57 expression on CD4<sup>+</sup> and CD8<sup>+</sup> T cells is higher in the blood of PLWH compared to that of HIV-negative controls (44–46). CD8<sup>+</sup> T cells expressing CD57 produce more interferon-γ and TNF-α after TCR stimulation than CD57<sup>−</sup> T cells, and CD57<sup>+</sup> CD8<sup>+</sup> T cells have a distinct gene expression profile characterized by greater cytotoxic effector potential (e.g., production of perforin, granzymes, and granulysin) (47, 48). Additionally, a higher percentage of CD57<sup>+</sup> expression on T cells has been implicated in other inflammatory diseases, such as rheumatoid arthritis (49) and beryllium-induced disease (50).

Given the potential pro-inflammatory effects of CD57<sup>+</sup> T cells in adipose tissue, we compared CD4<sup>+</sup> and CD8<sup>+</sup> T cell expression of CD57 in SAT vs. blood, and according to metabolic status (**Supplementary Figure 11**). In all three metabolic groups, expression of CD57 on SAT CD8<sup>+</sup> T cells was higher compared to blood, confirming our prior study findings (21). Among the memory subsets, CD57 expression was highest on CD4<sup>+</sup> TEMRA and CD8<sup>+</sup> TEM and TEMRA cells. However, we did not observe an increase in CD57 expression on either CD4<sup>+</sup> or CD8<sup>+</sup> memory cell subsets with progressive glucose intolerance, with the exception of higher CD57 expression on CD4<sup>+</sup> TCM in diabetic individuals compared to those without diabetes.

### CD4+CD69lo Cells Co-expressing CD57, CX3CR1, and GPR56 Are Associated With Increasing Glucose Intolerance

Adipose tissue serves as a reservoir for both latently HIV-infected CD4<sup>+</sup> T cells and free HIV RNA virus, and in the Genotype-Tissue Expression (GTEx) project adipose tissue contained one of the highest levels of CMV transcripts (17, 18, 21, 51). These findings suggest adipose tissue may also serve as a site for antiviral immune activity. Recent studies of CD4<sup>+</sup> TEMRA cells have identified major subsets based on G protein-coupled receptor GPR56 expression, with virus-specific cells more frequently GPR56<sup>+</sup> and more clonally expanded compared to GPR56<sup>−</sup> cells (52). Increased expression of GPR56 and killer-like receptors (KLR) has been linked to higher cytokine expression by memory CD4<sup>+</sup> T cells, including T cells obtained from liver tissue (53). CD4<sup>+</sup> and CD8<sup>+</sup> TEMRA cytotoxic T cells expressing GPR56 have also been associated with increased co-expression of CX3CR1 (52, 54, 55). CX3CR1 receptor is expressed on terminally differentiated T cells, gamma-delta T cells and NK cells, and has also been identified as a marker of anti-CMV T cells (55–58).

To assess the presence of GPR56<sup>+</sup> CX3CR1<sup>+</sup> T cells, and the parent memory cell population(s), we used t-SNE and viSNE to identify surface marker clusters that differed between nondiabetic, pre-diabetic, and diabetic individuals. We identified a group of cells that expressed CD57, lacked CD69, and also co-expressed GPR56 and CX3CR1. A representative plot of t-SNE maps generated from CD4<sup>+</sup> gated T cells showed that the CD57<sup>+</sup> CX3CR1<sup>+</sup> GPR56<sup>+</sup> co-expression was mainly on TEM and TEMRA cells (**Figure 5A**). Concatenated viSNE plots of non-diabetic, pre-diabetic and diabetic PLWH demonstrated two distinct clusters of cells: CD57−/<sup>+</sup> CD69<sup>+</sup> CX3CR1−/<sup>+</sup> GPR56−/<sup>+</sup> and CD57<sup>+</sup> CD69lo CX3CR1<sup>+</sup> GPR56<sup>+</sup> (**Figure 5B**). A significantly larger proportion of total CD4<sup>+</sup> and CD4<sup>+</sup> TEMRA cells in SAT from diabetics were CD57<sup>+</sup> CD69lo CX3CR1<sup>+</sup> GPR56<sup>+</sup> compared to SAT from non-diabetics (p = 0.051), and approached significance for CD4<sup>+</sup> TEM cells (p = 0.07) (**Figure 5C**). Of note, we observed a similar population of CD4+ TEMRA and TEM cells co-expressing CD57<sup>+</sup> CX3CR1<sup>+</sup> GPR56<sup>+</sup> in matched PBMC samples which, again, increased with glucose intolerance and were significantly different between diabetic and non-diabetic PLWH (total CD4<sup>+</sup> [p < 0.01], CD4<sup>+</sup> TEM [p = 0.05] and CD4<sup>+</sup> TEMRA [p < 0.05]) (**Supplementary Figure 6**). Similar plots of CD8<sup>+</sup> T cells were generated for adipose tissue (**Supplementary Figure 7**) and PBMC (data not shown) from PLWH. As with the CD4<sup>+</sup> T cells, CD57<sup>+</sup> CX3CR1<sup>+</sup> GPR56<sup>+</sup> co-expressing CD8<sup>+</sup> T cells were predominantly TEM and TEMRA, though there were no significant differences between non-diabetics and diabetics.

Lastly, we assessed CD57<sup>+</sup> CX3CR1<sup>+</sup> GPR56<sup>+</sup> co-expression on SAT CD4<sup>+</sup> and CD8<sup>+</sup> T cells from our HIV-negative controls (**Supplementary Figure 8**). We identified two distinct clusters of CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cells co-expressing CD57, CX3CR1, and GPR56 though the expression was more diffuse compared to SAT cells from PLWH, particularly for GPR56.

Taken together, our results demonstrate a population of CD57+CX3CR1+GPR56<sup>+</sup> co-expressing predominantly CD4<sup>+</sup> TEMRA cells in both blood and SAT of PLWH which appear to be associated with diabetes. The expression of the GPR56 and CX3CR1 markers may indicate these cells are virus-specific TEMRA cells, a population that could contribute to inflammation in adipose tissue.

# SAT CD4<sup>+</sup> and CD8<sup>+</sup> T Cell Memory Subsets Compared by HIV Status

Couturier et al. identified a major shift in the CD4:CD8 T cell ratio in PLWH compared to HIV-negative controls (17), a finding that has been replicated in subsequent HIV and SIV

studies (18, 19, 21). Specifically, adipose tissue stromal vascular fraction (SVF) CD3<sup>+</sup> T cells from individuals without HIV were predominantly memory CD4<sup>+</sup> CD45RO<sup>+</sup> T cells rather than memory CD8<sup>+</sup> T cells. In contrast, this distribution was reversed in PLWH, with more memory CD8<sup>+</sup> T cells in the adipose tissue, which represented an ∼50% enrichment in memory CD8<sup>+</sup> T cells over the peripheral blood, and could not be attributed to differences in peripheral blood T cell subsets

Phenotypic analysis of SAT memory T cell subsets in PLWH could provide insight on possible mechanisms contributing to the profound shift in the CD4:CD8 ratio that accompanies HIV infection. Therefore, we compared the proportion of CD4<sup>+</sup> and CD8<sup>+</sup> naive, TCM, TEM, and TEMRA cells between 8 non- and pre-diabetic HIV-negative persons and 9 PLWH; these subjects were selected from the 26 PLWH in our cohort based on similar age, HbA1c and BMI values (**Table 2**). As in prior studies, we found that SAT from PLWH is enriched in CD8<sup>+</sup> T over CD4<sup>+</sup> T cells (51 vs. 47%, respectively) compared to HIV-negative persons (21 vs. 66%; **Figures 6A,D)**. Although PLWH have a much higher proportion of CD8<sup>+</sup> T cells in the SAT, we found no significant difference in the overall proportions of SAT CD4<sup>+</sup> and CD8<sup>+</sup> memory cells between PLWH and HIV-negative persons (**Figures 6C,F**), and the distribution of CD4<sup>+</sup> and CD8<sup>+</sup> memory T cell subsets was remarkably similar (**Figures 6B,E**). However, three notable differences were present: compared to PLWH, the HIV-negative persons had a significantly higher percentage of CD4<sup>+</sup> TEM (58 vs. 39%, p = 0.02), a lower percentage of CD4<sup>+</sup> TCM (15 vs. 29%, p = 0.07) in their SAT (**Figure 6B**), and a significantly higher percentage of CD8<sup>+</sup> TCM compared to PLWH (6.4 vs. 3.4%, p = 0.03) (**Figure 6E**).

These findings suggest the profound change in the SAT CD4:CD8 T cell ratio observed in PLWH is not driven by the enrichment or depletion of a single memory T cell phenotype. Rather, it is phenotype agnostic and involves shifts in disparate naïve and memory T cell phenotypes from both the CD4<sup>+</sup> and CD8<sup>+</sup> lineages. This raises the possibility that a chemotactic signal from SAT in PLWH is recruiting CD8<sup>+</sup> T cells more robustly in the context of infection with HIV. However, the similar subset distributions suggest this signal, if present, may represent the amplification of a normal physiologic pathway as opposed to an "HIV-specific" process.

#### Comparison of SAT Gene Expression in PLWH and HIV-Negative People With Similar Glucose Tolerance

Given the increased proportion of CD8<sup>+</sup> T cells in SAT from PLWH, we performed a sub-study to investigate potentially relevant adipose tissue immune signaling pathways upregulated in the context of HIV infection. Extracted SAT mRNA from

6 PLWH and 7 HIV-negative individuals was assayed using a NanoString human inflammation panel to quantitate expression of over 250 genes representing a broad range of immune pathways (**Figure 7A**). The overall fold change in genes expressed from SAT of PLWH over HIV-negative is shown in the volcano plot (**Figure 7B**). In general, we found increased expression of chemokine receptors (CXCR2, CXCR1, CXCR4) and ligands (CCL5, CXCL5) in those with HIV. Additional genes including ALOX12 and IL12A were also elevated. We further analyzed inflammatory gene pathways defined by NanoString (**Figure 7C**). CXCR2, CXCR1, CXCR4, TLR2, and TLR8 gene expression were significantly higher in PLWH compared to HIV-negative (**Table 4**) whereas CD4 and CXCL9 were higher in HIVnegative individuals. Future studies are needed to identify the cells expressing these genes, and whether receptor and ligand expression might account for differences in the cells that traffic to SAT in PLWH.

# DISCUSSION

Several recent studies have uncovered a profound change in the balance of adipose tissue CD8<sup>+</sup> and CD4<sup>+</sup> T cells in PLWH, and similar studies in SIV-infected macaques indicate that the relative enrichment of CD8<sup>+</sup> T cells stems from viral infection rather than from ART treatment or CD4<sup>+</sup> cell depletion. However, the characteristics and consequences of this CD8<sup>+</sup> cell enrichment are generally unknown. Here, we demonstrate that the adipose tissue of PLWH with viral suppression on long-term ART is a reservoir of CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cells; two cell types with high pro-inflammatory potential when stimulated. Furthermore, we show that expression of CD69, a putative marker of TCR-linked activation and tissue residency, increased on CD4<sup>+</sup> TEM and TEMRA cells in a stepwise manner from non-diabetic, to pre-diabetic, to diabetic individuals. We also identify a phenotypically unique population of CD4<sup>+</sup> T cells co-expressing CX3CR1, GPR56, and CD57 that is specifically enriched in the SAT of diabetic PLWH. While the significance of these cells is currently unclear, CX3CR1 and GPR56 have previously been identified on cells with antiviral functions (55– 59). Finally, our results demonstrate that the enrichment of CD8<sup>+</sup> T cells over CD4<sup>+</sup> T cells in PLWH as compared to HIVnegative individuals is relatively non-specific in the sense that the change is not attributable to a profound increase or reduction in a single memory cell subset.

Our findings contribute to the growing body of literature on both adipose tissue T cells in general, as well as the immune

phenotype of adipose tissue in the context of metabolic disease among PLWH. Han et al. recently demonstrated that white adipose tissue in mice is a major reservoir for memory T cells with potent proliferative, effector, and protective potential (39). The adipose tissue T cells predominantly expressed CD44, a marker of antigen experience, and were CD62L negative, which is consistent with TEM and TRM populations in mice. Furthermore, approximately half the CD44+CD62L<sup>−</sup> CD8<sup>+</sup> and CD4<sup>+</sup> cells expressed CD69. The study also assessed recall responses of adipose tissue T cells in mice previously challenged with antigen, demonstrating an influx of monocytes and neutrophils, as well as highly reactive memory T cells. Specifically, these memory T cells indicated rapid and enhanced effector potential with upregulation of several genes involved in antimicrobial defenses. Furthermore, antigen re-challenge led to downregulation of several metabolic pathways (including lipid biosynthesis and cholesterol and long-chain fatty-acyl-CoA metabolic processes) as well as to a detectable reduction in serum levels of adiponectin and cholesterol, further highlighting the potential role memory T cells play in metabolic disease.

These animal studies suggest the accumulation of TEM and TEMRA cells in the adipose tissue of PLWH may represent a potent source of inflammation in the setting of antigen stimulation. Viral pathogens, including HIV and CMV, are found in adipose tissue and could serve as a chronic stimulus for TEM and TEMRA cells, with downstream effects on metabolic function (17, 21, 39, 51). Our observation of increasing TEM and TEMRA CD69 expression with declining glucose tolerance in PLWH may indicate the enrichment of a resident memory phenotype in diabetic individuals. Our study design precludes an assessment of whether the presence of increased CD69<sup>+</sup> TEM and TEMRA cells preceded or followed the development of glucose intolerance, though future longitudinal studies in PLWH with early indications of metabolic disease could address this question. Furthermore, future studies to identify the receptor specificity of the CX3CR1, GPR56, and CD57-expressing CD4<sup>+</sup> cells, and experiments to co-culture these cells with adipocytes, may help characterize the role of these immune cells in adipose tissue, identify their cytokine expression patterns, and explore the potential effects of these cells on adipocyte function.

An early study of SAT and VAT from PLWH by Couturier et al. identified major differences in CD4<sup>+</sup> and CD8<sup>+</sup> T cell populations compared to HIV-negative controls (17), which were subsequently reported in other HIV and SIV studies (18, 19, 21). In the HIV-negative controls, adipose tissue SVF CD3<sup>+</sup> T cells in SAT were predominantly memory CD4<sup>+</sup> CD45RO<sup>+</sup> T cells (61%) with fewer memory CD8<sup>+</sup> T cells (15%), while this distribution was reversed in PLWH, with more memory CD8<sup>+</sup> T cells (46%) compared to memory CD4<sup>+</sup> T cells (35%). This represented an ∼50% enrichment in memory CD8<sup>+</sup> T cells



over the blood in the subjects with HIV and was not reflective of differences in peripheral blood T cell subsets between the two groups.

Notably, the Couturier et al. study found significant differences in the rates of CD69 expression on memory CD4<sup>+</sup> and CD8<sup>+</sup> T cells in adipose tissue vs. blood: <5% of these cells expressed CD69 in the blood, compared to 60–67% in adipose from PLWH and 61–72% in the adipose from HIV negative persons. A similar study comparing activation markers on adipose tissue T cells in HIV-negative lean, overweight and obese individuals using fine needle aspiration found ∼5–10% CD69 expression on SAT CD4<sup>+</sup> T cells and ∼25% expression on CD8<sup>+</sup> T cells (60). We observed the highest CD69 expression on TEM cells in diabetic PLWH (mean 22%). This was significantly higher than CD69 expression on TEM cells from non-diabetics (mean 15%) and pre-diabetics (mean 18%), and 20-fold higher compared to TEM cells in blood. The reason for the lower CD69 expression in our cohort compared to Couturier et al. is unclear and may reflect differences in CD69 expression in adipose tissue samples collected after death or by surgical resection as opposed to liposuction aspirates processed within 30–60 min of collection in our study, differences in tissue processing to extract T cells, or could be explained by residual peripheral blood in samples obtained via liposuction.

Prior studies have also demonstrated a significantly higher proportion of SAT Treg cells (defined as CD25<sup>+</sup> FOXP3<sup>+</sup> CD4<sup>+</sup> T cells; a cell type thought to exert an anti-inflammatory effect and reported to be depleted in obese adipose tissue) in PLWH compared to HIV-negative persons, with no major differences in TH1 and TH17 pro-inflammatory subsets (19, 24). The proportion of TH1 CD4<sup>+</sup> T cells (expressing intermediate or high levels of T-bet) did not differ according to HIV status, while TH2 CD4<sup>+</sup> T cells (expressing GATA-3) were barely detected in the adipose tissue from both PLWH and HIV-negative persons. While CD4<sup>+</sup> T cell expression of HLA-DR (24%) in SAT was higher compared to blood in the HIV-negative, similar levels of SAT HLA-DR expression were observed in the PLWH. Furthermore, SAT CD4<sup>+</sup> T cell expression of PD-1 expression was much higher compared to PBMCs (45 vs. 3%), but again there were no significant differences in SAT according to HIV status (19).

Two studies of SIV-infected cynomolgus macaques confirm the adipose tissue CD8<sup>+</sup> T cell enrichment observed in PLWH is a viral phenomenon, rather than related to ART treatment (18, 61). In both studies, SIV infection was associated with a higher percentage of CD8<sup>+</sup> T cells in both the SAT and VAT compared to non-infected animals (18, 61). One study also demonstrated the inverted CD8:CD4 ratio was not driven by a reduction in the total number of CD4<sup>+</sup> T cells in infected animals; rather, SIVinfected animals had significantly higher density of CD8<sup>+</sup> T cells in VAT and a somewhat higher density in SAT (18). For both non-infected healthy and SIV-infected monkeys, the majority of the adipose tissue CD4<sup>+</sup> and CD8<sup>+</sup> T cells were memory T cells (>94% CD95+), with a large fraction of activated cells marked by expression of CD69<sup>+</sup> (62–84%) and CD25<sup>+</sup> (3–13%).

A central question that remains unanswered is whether the increase in the proportion of adipose tissue CD8<sup>+</sup> T cells in PLWH and macaques with SIV results from in situ proliferation vs. increased infiltration from the circulation. The high expression of CD69 on SAT memory cells suggests a "tissue resident" phenotype, but does not answer the question of whether these cells are expanded clones or prior migrants. On the one hand, a study by Damouche et al. found no significant differences in the proportion of CD4<sup>+</sup> or CD8<sup>+</sup> T cells expressing Ki-67, indicative of cycling and recently divided cells, between PLWH vs. HIV-negative controls (19). The authors suggested the low percentage of Ki-67<sup>+</sup> cells in SAT (<2%) from PLWH and HIVnegative subjects reflected minimal T cell proliferation within the tissue. Similarly, a study in macaques found no differences in proportions of Ki-67-expressing CD4<sup>+</sup> or CD8<sup>+</sup> T cells in animals with and without SIV, suggesting the higher density of CD8<sup>+</sup> T cells in adipose tissue does not result from recent proliferation (18). However, proliferation of memory T cells within adipose tissue is clearly demonstrated in other studies. Han et al. injected mice with pseudotuberculosis and evaluated memory T cells 4 weeks later. They showed proliferation of TEM and TRM by Ki67 expression, bromodeoxyuridine labeling, and cell cycle stage analysis of memory T cells (39).

The use of T cell receptor (TCR) sequencing of adipose tissue T cell subsets in future studies may provide insight into the clonal lineage of adipose tissue T cells. Previously, our group demonstrated that CD8<sup>+</sup> TCR clonality is higher in SAT compared to blood in PLWH using bulk TCRβ CDR3 deep sequencing, where bias-controlled V and J gene primers are used to amplify rearranged V(D)J segments (21, 62, 63). In that study, the 10 most prevalent TCRβ clones comprised a significantly larger percentage of total clones in SAT (25%) compared to paired blood (16%), and the Shannon's Entropy index, a measure of overall repertoire diversity, was lower in adipose tissue compared to blood (4.39 vs. 4.46, respectively). Additionally, V-J gene pairing and gene usage differed between blood and adipose tissue, albeit not statistically significant, potentially due to the small sample size. While these findings are intriguing, the lower proportion of CD4<sup>+</sup> and CD8<sup>+</sup> TNai cells in SAT demonstrated in the current analysis may have had a role in the higher SAT clonality scores reported in the prior study. In addition to clonal lineage, TCR sequencing may also inform our understanding of potential antigen targets for adipose tissue T cells in PLWH. A recent study in mice demonstrated that diet-induced obesity is characterized by increased adipose tissue CD8<sup>+</sup> T cell density, and the TCR repertoire of these CD8<sup>+</sup> T cells is more clonal and positively charged (in respect to amino acids). This work was also the first to demonstrate that isolevuglandins (a group of negatively charged reactive gamma-ketoaldehydes generated by free radical oxidation) presented on adipose tissue macrophages from obese fat can independently activate T cells, potentially highlighting a mechanism contributing to inflammation (25). Examining TCR charge and polarity, in addition to clonality, and investigating antigen presenting cells in the adipose tissue may be an important approach for further characterizing the adipose tissue immune milieu in PLWH vs. HIV-negative persons.

Profiling of adipose tissue mRNA expression in SIV-infected macaques showed significantly higher levels of IL-2, IL-7, and CCL19 in acute SIV compared to uninfected controls, which may contribute to the homing and survival of T cells (61). In our gene expression sub-study on whole adipose tissue of PLWH and HIVnegative controls with similar glucose tolerance, we found >2 fold expression of IL12A, CXCR1, CXCL5, ALOX12, and C9 in the PLWH compared to HIV-negative. Analysis of subgroups of inflammatory gene pathways revealed significantly higher levels of TLR2, TLR8, CXCR4, CCR7, CCL5, CXCR1, and CXCR2 and lower levels of CXCL9 and CD4 in the PLWH.

Previous studies have linked inducible and increased expression of TLR2 in PBMC and SAT of obese individuals with type 2 diabetes (64–67). TLR8 is an endosomal receptor that binds HIV-1 single-stranded RNA and is similar to TLR2 in its MyD88-dependent activation of NF-kB. Previous studies of TLR8 expression indicate that it is most often absent or present at low levels in adipose tissue (64), but that a positive correlation exists between plasma CRP levels and adipose tissue TLR8 expression. In our study, we are unable to discriminate whether the differences in TLR expression are due to alterations on T cells, macrophages, adipocytes or pre-adipocytes. However, we likely have identified a contribution of TLR8 to SAT inflammation that is more pronounced in PLWH compared to HIV-negative individuals with similar glucose tolerance and BMI.

Regarding chemokine ligands and receptors, M1 macrophages express higher levels of CCR7, CCL9, and CCL5 and lower levels of CXCR4 in comparison to M2 macrophages (68). In the context of HIV infection, CXCR1 and CXCR2 are of interest because HIV-1 matrix protein p17 has been shown to mimic IL-8 and binds CXCR2 with high affinity, stimulating pro-angiogenic ERK downstream (69). Angiogenesis has been linked to obesity via neovascularization-driven migration of adipocytes (70). Further studies of chemokine receptors on CD4<sup>+</sup> and CD8<sup>+</sup> T cells may provide insight into whether HIV infection results in signals that preferentially recruit CD8<sup>+</sup> T cells over CD4<sup>+</sup> T cells independent of obesity.

# LIMITATIONS

Our cohort study was cross-sectional and enrolled a total of 34 participants. We classified participants as non-diabetic and pre-diabetic based on fasting blood glucose and HbA1c values as recommended for clinical practice by the American Diabetes Association (diabetics were classified based on medication usage). In the future, the measurement of fasting insulin and use of the homeostatic model assessment (HOMA) may provide more nuanced stratification of participants. While a strength of our study was the collection of SAT via liposuction and rapid processing to extract T cells, samples may have contained residual peripheral blood, which would bias our comparisons of SAT and blood toward the null hypothesis. We were limited in the number of markers we could examine via flow cytometry and did not examine additional markers of tissue resident cells, as well as T regulatory cells and TH1/TH2 subsets. Our gene expression analysis of SAT from PLWH and HIV-negative persons was on whole adipose tissue as opposed to sorted SVF cells. Therefore, differences in genes cannot be directly linked to specific cell types.

# CONCLUSIONS

Although PLWH can survive decades on effective ART, this success is offset by the rising burden of metabolic diseases affecting the HIV population (1–3). Here, we demonstrate that the adipose tissue of PLWH is a reservoir of CD4<sup>+</sup> and CD8<sup>+</sup> TEM and TEMRA cells; two cell types with high pro-inflammatory potential when stimulated. Furthermore, we show that expression of CD69, a putative marker of TCR-linked activation and tissue residence, on CD4<sup>+</sup> T cells increases in a stepwise manner from non-diabetic, to pre-diabetic, to diabetic individuals, which may reflect an interaction between cells of the adaptive immune system and adipocytes in the context of metabolic disease. We also identify a population of CD57+CX3CR1+GPR56<sup>+</sup> coexpressing CD4<sup>+</sup> T cells that is specifically enriched in the SAT of diabetics. This could represent a group of virus-specific cells that contribute to inflammation in adipose tissue and potentially pre-dispose PLWH to metabolic disease, though further studies to assess the effects of these cells on adipocytes are needed. Finally, our results demonstrate that the enrichment of CD8<sup>+</sup> T cells over CD4<sup>+</sup> T cells in PLWH as compared to HIVnegative individuals is relatively non-specific in the sense that the change is not attributable to a profound increase or reduction in a single memory cell subset. Further studies are needed to understand the clonal lineage, TCR characteristics, antigen targets, chemokine receptors and the functional phenotype of CD57+CX3CR1+GPR56<sup>+</sup> T cells in the adipose tissue of PLWH, and to identify potential therapeutic targets.

# DATA AVAILABILITY

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

#### ETHICS STATEMENT

The study was carried out in accordance with the recommendations of the human experimentation ethical standards of the Vanderbilt Institutional Review Board. All subjects gave written informed consent, in accordance with the Helsinki Declaration of 1975, as revised in 2000. The protocol was approved by the Vanderbilt Institutional Review Board.

#### AUTHOR CONTRIBUTIONS

CW, WM, and JK: conceptualization and methodology; WM and RR: software; CW, WM, MP, and JK: validation; WM, CW, RF, and YF: formal analysis; CW, WM, LB, JS, BW, BF, ML, MM, PB, and RR: investigation; JK, SK, and SM: resources; WM and CW: data curation; CW, WM, and JK: writing—original draft; CW, WM, and RR: visualization; JK, SM, and SK: supervision; CW, WM, JK, RF, and YF: statistics; JK, NB, MP, and SK: project administration; JK and SM: funding acquisition; All authors: writing—review and editing.

#### REFERENCES


#### ACKNOWLEDGMENTS

This study was supported by NIH grants R01 DK112262 (JK and CW), R56 DK108352 (WM and JK), the Vanderbilt Clinical and Translational Science award from NCRR/NIH grant UL1 RR024975, the Vanderbilt Infection Pathogenesis and Epidemiology Research Training Program (VIPER) grant T32 AI007474, and the Tennessee Center for AIDS Research grant P30 AI110527. This study was also supported by American Heart Association grant 17SFRN33520059. The funding authorities had no role in study design; data collection, analysis, or interpretation; decision to publish; or preparation of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00408/full#supplementary-material

reverse transcriptase inhibitor therapy. AIDS. (2004) 18:815–7. doi: 10.1097/00002030-200403260-00015


in HIV-infected persons. J Acquir Immune Defic Syndr. (2018) 77:e14–21. doi: 10.1097/QAI.0000000000001573


CD8(+) T cells by persistent viruses and vaccines. Cell Rep. (2018) 23:768–82. doi: 10.1016/j.celrep.2018.03.074


response in human adipose tissue in obesity and type 2 diabetes. Am J Physiol Endocrinol Metab. (2007) 292:E740–7. doi: 10.1152/ajpendo.00302. 2006


**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 Wanjalla, McDonnell, Barnett, Simmons, Furch, Lima, Woodward, Fan, Fei, Baker, Ram, Pilkinton, Mashayekhi, Brown, Mallal, Kalams and Koethe. 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.

# BTLA/HVEM Signaling: Milestones in Research and Role in Chronic Hepatitis B Virus Infection

Xueping Yu1,2†, Yijuan Zheng1†, Richeng Mao<sup>2</sup> , Zhijun Su<sup>1</sup> \* and Jiming Zhang<sup>2</sup> \*

*<sup>1</sup> Department of Infectious Diseases, First Hospital of Quanzhou, Fujian Medical University, Quanzhou, China, <sup>2</sup> Department of Infectious Diseases, Huashan Hospital, Fudan University, Shanghai, China*

B- and T-lymphocyte attenuator (BTLA) is an immune-regulatory receptor, similar to CTLA-4 and PD-1, and is mainly expressed on B-, T-, and all mature lymphocyte cells. Herpes virus entry mediator (HVEM)-BTLA plays a critical role in immune tolerance and immune responses which are areas of intense research. However, the mechanisms of the BTLA and the BTLA/HVEM signaling pathway in human diseases remain unclear. This review describes the research milestones of BTLA and HVEM in chronological order and their role in chronic HBV infection.

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Ranjit Chauhan, Memorial University of Newfoundland, Canada Adam J. Gehring, University Health Network (UHN), Canada*

#### \*Correspondence:

*Jiming Zhang jmzhang@fudan.edu.cn Zhijun Su su2366@sina.com*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *23 October 2018* Accepted: *08 March 2019* Published: *29 March 2019*

#### Citation:

*Yu X, Zheng Y, Mao R, Su Z and Zhang J (2019) BTLA/HVEM Signaling: Milestones in Research and Role in Chronic Hepatitis B Virus Infection. Front. Immunol. 10:617. doi: 10.3389/fimmu.2019.00617* Keywords: B and T lymphocyte attenuator, herpes virus entry mediator, hepatitis B virus, milestones, T lymphocyte

#### INTRODUCTION

Lymphocyte activation is either triggered by the binding of an antigen to its T-cell receptor (TCR) or B-cell receptor (BCR), or by a co-stimulatory or co-inhibitory molecule. T-cells require a co-stimulatory or co-inhibitory molecule for activation, and the quality of T-cell activation is determined by multiple co-signaling molecules. These co-signaling molecules exert both positive stimulatory and negative regulatory functions, and act in a coordinated fashion to maintain homeostasis in the body (1). Co-signaling molecules can be classified into two major families based on their structure. The first is the CD28 immunoglobulin (Ig) superfamily (IgSF), which includes CD28, cytotoxic T-lymphocyte antigen-4 (CTLA-4), inducible costimulatory molecule (ICOS), programmed death-1 (PD-1), and B and T-lymphocyte attenuator (BTLA); and the second is the tumor necrosis factor receptor (TNFR) superfamily (TNFRSF) (2), which includes CD27, CD30, 4-1BB, herpesvirus entry mediator (HVEM), CD40, and OX40 (**Table 1**). Similar to PD-1 and CTLA-4, BTLA inhibits T-cell reactions and cytokine production. Studies on hepatitis B virus (HBV) infection revealed that BTLA is highly expressed in virus-specific T-cells, which have a potent inhibitory effect on events such as T-cell proliferation and cytokine secretion. In this review, we discuss the biological characteristics of BTLA and its ligand and explore their role in chronic HBV infection.

#### CHRONOLOGICAL MILESTONES IN BTLA RESEARCH

BTLA, also known as CD272, was first discovered by genetic screening in 2003 for its ability to inhibit Th1 cell expression (3). It is the third new member of the CD28 family discovered after PD1 and CTLA-4. In 2005, HVEM was identified as the specific ligand of BTLA (4). HVEM belongs to the TNFR family and not to the Ig family, thus shattering the perspective that receptors exclusively bind with ligands belonging to the same family. In the same year, the herpes simplex virus type 1 glycoprotein D (HSV1 gD) was found to bind to HVEM from the crystal structure of the BTLA-HVEM complex (5). In the subsequent year, the structure, distribution, biological



characteristics, and other aspects of BTLA and HVEM were summarized in a review by Murphy et al. (6), who also found a third motif, Grb-2, in the cytoplasmic domain of BTLA that recruits the PI3K-subunit p85, thus, leading to stimulation of the PI3K signaling pathway and subsequent T-cell activation (7). In 2008, the receptors HSV1 gD, LIGHT (also known as tumor necrosis factor superfamily 14) and CD160, were found to constitute the CD160/BTLA/LIGHT/HVEM signaling regulatory network and to share the same ligand HVEM as BTLA (8). The interactions within the CD160/BTLA/LIGHT/HVEM signaling regulatory network were summarized in the 2009 review by Cai and Freeman (9). Specific BTLA antibody clones such as 6F7 and 6H6 targeting the BTLA-HVEM pathway were summarized in the review by Crawford and Wherry (10). Between 2006 and 2010, the roles and mechanisms of BTLA and its ligands in human diseases [organ transplantation (11), intestinal inflammation (12), rheumatoid arthritis (13), and cancer (14)] and animal models [experimental cerebral malaria (15), mouse pancreatic transplantation (16)] were reported. In 2010, Murphy et al. extensively reviewed the biological characteristics and functional mechanisms of BTLA and its ligands and discussed newer findings (17). A new anticancer therapy based on the blockade of the BTLA signaling pathway was proposed next, which signaled the beginning of a new chapter in cancer intervention (18).

Since HVEM could interact with many co-signaling molecules, Kronenberg et al. proposed that the CD160/BTLA/LIGHT/HVEM signaling regulatory network plays a bidirectional regulatory role in various inflammatory, autoimmune, and infection immune reactions (19). Decreased BTLA levels could induce hyper-activation of T-lymphocytes in HIV patients thereby promoting disease progression (20). In 2012, the HVEM-BTLA signaling pathway was found to be upregulated in the hepatic tissue of HBV-related acute-onchronic liver failure (HBV-ACLF) patients, promoting disease progression (21). In addition, BTLA was also found to promote the development and progression of sepsis through inhibition of the innate immune response (22). In 2013, sirolimus was identified to promote the inhibitory effects of BTLA thereby enabling immune tolerance in kidney allograft (23). BTLA was reported to be a crucial molecular marker in "immunoparalysis" associated with sepsis (24), and it was shown to play a positive regulatory role in viral diseases; for example, mouse hepatitis virus-3 (MHV3) could induce BTLA signaling and cause acute liver failure through phagocyte activation and secretion of the inflammatory molecules TNF-α and FGL2 (25). This study advances our understanding of the conditions that determine the negative or positive regulatory functions of BTLA in humans. In 2015, the BTLA-HVEM signaling pathway was reported to help intestinal parasites (especially Strongyloides stercoralis) maintain an infection (26). In the following year, vaccines blocking the BTLA/CD160 signaling pathway were shown to activate the response of aged CD8<sup>+</sup> T cells to the influenza virus (27). Moreover, dendritic cells (DCs) were demonstrated to induce extrathymic T-cell tolerance in peripheral Treg cells through the BTLA-HVEM signaling pathway (28). In 2017, Shen et al. found that the CD8<sup>+</sup> BTLA<sup>+</sup> T-cells isolated lymphocytes from the liver tissue of chronic hepatitis B patients had a negative regulatory effect on Treg cells that helped HBV to avoid immune clearance (29). In 2018, BTLA was elucidated as a marker of a less cytotoxic T-cell subset in diffuse large B-cell lymphoma (30). Our summary listing the research milestones on the BTLA signaling pathway in chronological order aims to distill the information from previous findings and provide more explicit research directions for future studies (**Figure 1**).

# BIOLOGICAL CHARACTERISTICS OF BTLA

#### Structure and Distribution of BTLA

The human BTLA gene, located on chromosome 3 at 3q13, comprises 5 exons with a total length of 870 bp and 3 mRNA splice variants, which encode functional proteins that can be transcribed. Of note, a single nucleotide polymorphism (SNP) of BTLA, rs76844316, was reported to protect against chronic hepatitis B infection (31). BTLA is a type I transmembrane glycoprotein comprising 289 amino acids. Its protein structure is similar to those of CTLA-4 and PD-1 and includes an extracellular domain, transmembrane domain, and cytoplasmic domain (32). The cytoplasmic domain contains three conserved signals: a growth factor receptor-bound protein-2 (Grb-2) recognition motif, an immunoreceptor tyrosine-based inhibitory motif (ITIM), and an immunoreceptor tyrosine-based switch motif (ITSM) (6). ITIM is present in many inhibitory receptors, binding and activating the tyrosine phosphatases SHP-1 and SHP-2, which dephosphorylate tyrosine and inhibit protein tyrosine kinase (PTK)-dependent cell activation (33). The Grb-2 recognition motif recognizes the Grb-2 protein, recruits the PI3K protein subunit p85, and stimulates the PI3K signaling pathway, promoting cell proliferation and survival (7). Thus, the BTLA molecule exerts bidirectional regulatory effects: immunosuppressive effects like those on CTLA-4 and PD-1 proteins, and positive stimulatory effects like those on CD28 and ICOS proteins.

BTLA is widely expressed in the spleen, thymus, and lymph nodes and has relatively low expression or is even undetectable in the liver, kidney, heart, brain, and other organs. BTLA is constitutively expressed in the CD4/CD8 single-positive Tcells in the mouse thymus (34). Additionally, BTLA is highly expressed in the B-lymphocytes, splenic macrophages, and bone marrow-derived dendritic cells (35, 36).

#### BTLA Ligands

HVEM, a BTLA-specific receptor, was discovered in 2005 (4). HVEM is expressed in peripheral T- and B-cells, highly expressed in resting T cells, immature B cells, and memory B cells, but downregulated in activated T and B cells. Additionally, it is widely expressed in monocytes, dendritic cells, Treg cells, neutrophils, and NK cells (37, 38). HVEM binds with many co-signaling molecules, both co-stimulatory and co-inhibitory. The roles that both types of signaling molecules play in signaling pathways also differ and are known as the "molecular switch" models of activation and inhibition. Binding of HVEM to LIGHT or LIGHT-α exerts a positive stimulatory effect, stimulating lymphocyte proliferation, activation, and inducing inflammatory reactions; thus, providing a second stimulatory signal for T cell activation (4, 39). Binding of HVEM to BTLA and CD160 exerts an adverse regulatory effect, inhibiting Tand B-lymphocyte activation and proliferation and binding of HVEM to HSV-gD, which can promote HSV infection in target cells (4). Taken together, HVEM provides either an inhibitory or activating signal and bi-directionally regulates host immune function.

# FUNCTION OF BTLA IN IMMUNE CELLS

#### Function of BTLA in T-lymphocytes

Resting T-cells express high levels of BTLA and HVEM, and T cell activation increases or decreases BTLA and HVEM expression, respectively. The inhibition of T-cells by BTLA is stronger than the positive stimulatory effect of HVEM on T-cells and prevents the excessive activation of T-cells (40). Importantly, HVEM and BTLA in naive T-cells form a cis-heterodimeric complex, blocking the external CD160 and other co-signaling molecules from binding to HVEM and stimulating the NF-κB signaling pathway, thereby maintaining T-cell tolerance (41). Other studies have demonstrated that BTLA gene knockdown mice (Btla−/−) resist immune tolerance induced by high doses of oral or intravenous ovalbumin (OVA) and show increased infiltration by the inflammatory cells in multiple organs, which induces an autoimmune hepatitis-like disease (42, 43). Additionally, Liu et al. could not induce immune tolerance in Btla−/<sup>−</sup> mice injected with large doses of OVA, indicating that BTLA plays a vital role in inducing and maintaining T cell immune tolerance (44).

In addition to inhibiting antigen-specific TCR signalingmediated T-cell proliferation, activation (CD25, CD38), and cytokine (IL-2, IL-4, and IL-10) production (34), the BTLA molecule also crosslinks HVEM on Treg cells to facilitate their immunosuppressant effects (40). Further, BTLA inhibits IgG production by inhibiting secretion of IL-21 by follicular helper T cells (Tfh) and plays an essential role in the immunomodulation in body fluids (45). γδT cells play an important role in pathogen clearance and the anticancer process. Interestingly, BTLA inhibits γδT cell proliferation, and secretion of IL-17, TNFα, and other cytokines leading to decreased pathogen clearance and anticancer activity (46, 47).

In some instances, binding of BTLA and HVEM mediates immunosuppressive activity and transduces positive signals that promote the survival of effector T cells (33). Tarun et al. used the vaccinia virus to infect mice and found that the BTLA-HVEM co-signaling system significantly promotes the survival of antiviral effector CD8<sup>+</sup> T-cells and production of memory cells (48). Competitive stimulation with BTLA antibodies (3C10) can induce IL-10-dependent Treg cell production and helps prolong allogeneic heart transplantation in mice (49). Additionally, BTLA can increase the number and activity of γδT cells and reduce the symptoms of skin inflammation; Btla−/<sup>−</sup> mice have a reduced number of γδT cells and are susceptible to dermatitis. Moreover, BTLA/HVEM crosslinking was observed to suppress T-cell activation thereby preventing allograft rejection (21, 50). Variations of the therapeutic strategy that targets the BTLA-HVEM immune checkpoint pathway using specific antagonist antihuman antibodies has been published in numerous patents, and drugs capable of targeting BTLA-associated signaling pathways such as the HVEM-BTLA-CD160 pathway are currently in preclinical trials (10, 18, 51, 52).

#### Function of BTLA in B-lymphocytes

BTLA research has been focused on T-cells, and there are few studies on its function in B-cells. Previous studies have revealed that BTLA is an inhibitory receptor in the BCR signaling pathway. BTLA attenuates the BCR signaling strength by recruiting and phosphorylating the protein tyrosine kinase Syk and downregulating B-cell linker protein, phospholipase E2, and NF-κB (53). Ware et al. suggested that HVEM-BTLA signaling can inhibit CPG-mediated B-cell proliferation and cytokine secretion, and increase stimulatory molecules on their surface; however, this does not affect IL-8 and MIP-1β secretion, indicating that BTLA can partially, but not completely, inhibit B cell function (54). However, studies have also shown that BTLA expression in B cells is decreased in elderly patients, leading to reduced responsiveness to the trivalent influenza vaccine, and an inability to produce useful IgG antibodies and mount effective vaccination responses (55). Thus, BTLA can play bidirectional regulatory roles in specific cases.

#### Function of BTLA in Dendritic Cells

Latest research demonstrates that HVEM-BTLA signaling plays an important role in maintaining the stability of the internal environment for DCs. Lymphotoxin beta receptor (LT-βR) signaling can induce DC proliferation, whereas HVEM-BTLA signaling inhibits their proliferation, indicating that HVEM-BTLA signaling can regulate LT-βR signaling by feedback and maintain the stability of the internal environment for DCs (56). Interestingly, an adenoviral infection can cause immature DCs to express high levels of CCR7 and exhibit relatively strong migration ability. However, their immune tolerance is relatively poor, and the overexpression of BTLA promotes the maintenance of immune tolerance in these DCs (57). Additionally, BTLA<sup>+</sup> DCs in the thymus increase the expression of CD5 in peripheral T-cells through the BTLA-HVEM signaling pathway and promote the differentiation of these CD5+Tcells into Treg cells; thus, producing extrathymic T-cell tolerance (28).

#### Function of BTLA in Natural Killer T-cells

Like B- and T-lymphocytes, BTLA is expressed in the natural killer T (NKT) cells. Nakajima et al. established that BTLA−/<sup>−</sup> NKT mice secrete more cytokines (IFN-γ and IL-4) after αgalactosylceramide stimulation and Con A injection compared to wild-type mice, and develop Con A-induced hepatitis more easily(58). However, these phenomena were not observed in BTLA−/−NKT−/<sup>−</sup> mice. When BTLA−/−NKT and NKT cells were purified in vitro and injected into the NKT−/<sup>−</sup> mice, mice receiving BTLA−/−NKT cells were more susceptible to Con A-induced hepatitis, indicating that BTLA inhibits hepatitis induced by NKT cells. Similarly, Fu et al. also found in a Con Ainduced acute hepatitis model that NKT cells inhibit the release of cytokines (IFN-γ, IL-2, and IL-4) and liver tissue damage through upregulation of the HVEM-BTLA signaling pathway (59). Additionally, in mouse models of breast cancer, type I NKT cells express high levels of BTLA, and blocking the BTLA signaling pathway may promote infiltration of tumors by NKT cells and inhibit tumor growth (60).

#### FUNCTION OF BTLA IN CHRONIC HBV INFECTION

HBV infection severely endangers the health of humans. Globally, there are 240 million patients with HBV infection and every year 0.65 million patients die of HBV-associated endstage liver diseases, whose leading mortality causes include liver cirrhosis (LC), liver failure (LF), primary hepatic cell carcinoma (HCC). China is an endemic zone for HBV infection, and currently, there are 93 million individuals with chronic HBV infection, with ∼20 million chronic hepatitis B (CHB) patients. Thus, HBV infections put a heavy economic burden on the country and its citizens. However, the pathogenic mechanism of chronic HBV infection is not completely understood. Research has demonstrated that the HBVM-BTLA signaling pathway plays an important role in cancer (14, 61), intestinal inflammation (12), autoimmune diseases (4, 13), viral infection (62), transplant rejection (11, 63), and in continuous chronic HBV infection. In this section, we have provided a current summary of the literature review of BTLA's functions in chronic HBV infection.

#### Function of BTLA in CHB

The response of HBV-specific T cells (CTL) in CHB patients is extremely weak and can be undetectable. Additionally, the inability to clear HBV leads to continuous infection. Multiple reports suggest that this could be related to an increased expression of T-cell co-inhibitory molecules (e.g., PD-1). However, recent studies demonstrated no significant difference in the peripheral blood expression of BTLA in CD4<sup>+</sup> and CD8+Tcells in CHB patients and healthy individuals, (64, 65) and the expression levels were similar in the 4 subtypes of CD4+T and CD8+T cells (TEM-RA, Tnaïve, Tcm, and Tem) (64). Although these results suggest that BTLA does not contribute to chronic HBV infection (or CHB immune tolerance); however, there is a difference in BTLA expression levels in the CTL subtypes in the peripheral blood and liver tissue of CHB patients. In the peripheral blood, BTLA is primarily expressed in the Tcm subtype of T-lymphocytes, whereas BTLA in the liver tissue is primarily expressed in the Tem subtype. This difference may be due to an upregulation of BTLA expression during homing of the peripheral CD8+T-cells to the liver that prevents the excessive transition of CD8+T cells from the CM stage to the EM stage helping HBV evade immune clearance. Thus, it is believed that CD8+BTLA+T cells can negatively regulate Treg cells (29). Critically, during the four different phases of HBV infection [immunotolerant phase, immune clearance phase, nonreactive or minimally (non-) replicative phase, and reactivation phase], the immune reactions produced by HBV are different. Additionally, Zhou et al. established that the frequency of rs76844316 in the G allele of the BTLA gene was decreased in patients with severe CHB, which leads to increased sensitivity to HBV and association with severe disease (31). Therefore, the total BTLA expression level in CHB patients should not be viewed in isolation and BTLA expression in patients at different phases of HBV, across multiple severities of CHB should be analyzed. However, to our knowledge, no such data have been reported.

# Function of BTLA in HBV-LC and HCC

CHB is a progressive disease, and the Chinese "Guidelines for the Prevention of Chronic Hepatitis B (2015 edition)" indicates that every year, ∼2–10% of the CHB patients develop LC and 3–6% of the LC patients further progress into HCC (66). Liao et al. suggested that BTLA expression levels are significantly upregulated during the progression of CHB from HBV-LC to HCC, but the expression levels of other co-signaling molecules (CD28, ICOS, LIGHT) do not change significantly indicating that BTLA plays an important role in the progression of CHB (67). BTLA is expressed at high levels in the peripheral blood of patients with HBV-associated HCC and directly correlates with CD4+CD25+Treg cells. These findings indicate that BTLA may have a synergistic effect with CD4+CD25+Treg cells, inhibit Tcell activity and proliferation, and promote the immune evasion of tumors (68). Thus, blocking the BTLA signaling pathway inhibits T-cell function, "awakens" cancer recognition by the immune system, and clears tumor cells. Blockade of the HVEM-BTLA signaling pathway has been developed as a new anticancer method (18, 52) and has led to more anticancer drugs that target BTLA (**Figure 2**).

# Function of BTLA in HBV-ACLF

Liver failure is categorized as acute liver failure (ALF), subacute liver failure (SALF), acute-on-chronic liver failure (ACLF), and chronic liver failure (CLF) (69). Clinically, the most common form of HBV-associated liver failure is HBV-associated acute-onchronic liver failure (ACLF) in China, and it has a high mortality

FIGURE 2 | BTLA-associated signaling pathways (stimulatory/inhibitory) regulate the outcomes of HBV-ACLF, liver graft tolerance/rejection, liver cirrhosis and hepatocellular carcinoma. The "?" indicates that the role of the Grb-2 pathway in T-cell-related immune diseases requires further validation and has not been fully characterized yet. HBsAg, Hepatitis B antigen; TCR, T-cell receptor; MCH, major histocompatibility complex.

rate of 60–80% (70). ACLF was first coined in 1995 describing a condition arising from two simultaneous insults to the liver, one ongoing and one acute (71). The consensus statements of its definition, diagnosis and management were approved at the 2008 Annual Conference of the APASL in Korea (72), which was defined as an increasingly recognized syndrome characterized by an acute deterioration of liver function and organ/system failure (liver, kidney, brain, coagulation, circulation, and/or respiration) (70).

Since ACLF is a complex and dynamic disease, its diagnostic criteria consists of several components: the cause and timeframe of liver disease development and deterioration following an acute insult, whether patients have pre-existing chronic liver disease, the symptoms of liver failure to determine ACLF severity, and how to assess short- (28-day) and long-term (90-day) prognoses (73). HBV infection is one of the causes of acute liver injury, and BTLA/HVEM signaling contributed to HBV-ACLF pathogenesis (21). The timeframe of acute liver disease development after an acute insult has not been rigorously defined, but several studies have reported it to range from 2 to 8 weeks (73). The role of BTLA/HVEM signaling in determining this timeframe has not been characterized yet. BTLA/HVEM signaling pathway has been shown to prevent T-cell activation thereby promoting malignancy (**Figure 2**) (68). However, blocking this co-inhibitory pathway limited antitumor response against pre-existing tumor cells (74). The association of BTLA/HVEM signaling with the symptoms of liver failure in ACLF has not been elucidated. Although BTLA/HVEM signaling was shown to be implicated in poor HBV-ACLF outcome (21), further research is necessary to determine its potential as a biomarker for both short- and long-term clinical prognosis.

Currently, there are no effective therapeutic measures, and it is critical to search for early diagnostic markers or targets for pharmacological intervention. Upregulation of PD-1/PD-L1 (75) and BTLA/HVEM (21) pathways in the liver tissue of HBV-ACLF patients were reported. Additionally, BTLA is primarily expressed in fibrinogen-like protein-2 and CD68<sup>+</sup> phagocytes and is not expressed in the liver tissue of CHB patients or healthy individuals. Thus, BTLA has the potential to be a diagnostic

#### REFERENCES


marker for HBV-ACLF and provides a theoretical basis for HBV-ACLF immunotherapy.

#### CONCLUDING REMARKS

Since the discovery of BTLA in 2003, multiple studies have established that the HVEM-BTLA signaling pathway plays an essential immunomodulatory role in autoimmune disease, cancer, transplantation, infection, and other diseases. Recent studies on the HVEM-BTLA signaling pathways have unveiled the function and mechanism of BTLA, and targeted anticancer drugs for HVEM-BTLA are emerging. However, whether BTLA participates in inducing functional exhaustion of T-cells and its pathophysiological roles in associated diseases (e.g., HBV-ACLF) remain unknown. It is predicted that studies targeting BTLA will lead to a new revolution in unraveling the immune mechanisms, diagnosis, and treatment of chronic HBV infection.

#### AUTHOR CONTRIBUTIONS

XY and YZ searched, identified and reviewed the literature, and wrote the manuscript. RM made a table, gave critical comments, and revised the manuscript. JZ and ZS identified and reviewed the literature, wrote the manuscript, and revised the manuscript. All authors have made an intellectual contribution to the manuscript and approved the submission.

#### FUNDING

This study was supported by the National Natural Science Foundation of China (81400625, 81670528, and 81672009); Quanzhou high-level talent innovation project (2018C067R); Shanghai Pujiang Program (17PJD005); National Science and Technology Major Project of China (2017ZX10202202 and 2017ZX10202203-007); Natural Science Foundation of Fujian province (2016Y9065), Fujian Provincial Health commission Youth Research Project (2018-1-94, 2018-1-95), and Quanzhou science and technology project (2018Z069, 2018Z074).

interaction with herpesvirus entry mediator. Nat Immunol. (2005) 6:90–8. doi: 10.1038/ni1144


with high expression of checkpoints. Exp Hematol. (2018) 60:47–56.e1. doi: 10.1016/j.exphem.2018.01.003


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

# NF-κB Signaling and IL-4 Signaling Regulate SATB1 Expression via Alternative Promoter Usage During Th2 Differentiation

Satyajeet P. Khare1,2, Ankitha Shetty 1,3, Rahul Biradar <sup>1</sup> , Indumathi Patta<sup>1</sup> , Zhi Jane Chen3† , Ameya V. Sathe<sup>1</sup> , Puli Chandramouli Reddy <sup>1</sup> , Riitta Lahesmaa<sup>3</sup> and Sanjeev Galande<sup>1</sup> \*

*<sup>1</sup> Center of Excellence in Epigenetics, Indian Institute of Science Education and Research, Pune, India, <sup>2</sup> Symbiosis School of Biological Sciences, Pune, India, <sup>3</sup> Turku Center for Biotechnology, University of Turku and Abo Akademi University, Turku, Finland*

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

Reviewed by:

*Hu Zeng, Mayo Clinic, United States Geeta Upadhyay, Uniformed Services University of the Health Sciences, United States*

> \*Correspondence: *Sanjeev Galande sanjeev@iiserpune.ac.in*

#### †Present Address:

*Zhi Jane Chen, Faculty of Biochemistry and Molecular Medicine, University of Oulu, Oulu, Finland*

#### Specialty section:

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

Received: *20 November 2018* Accepted: *11 March 2019* Published: *02 April 2019*

#### Citation:

*Khare SP, Shetty A, Biradar R, Patta I, Chen ZJ, Sathe AV, Reddy PC, Lahesmaa R and Galande S (2019) NF-*κ*B Signaling and IL-4 Signaling Regulate SATB1 Expression via Alternative Promoter Usage During Th2 Differentiation. Front. Immunol. 10:667. doi: 10.3389/fimmu.2019.00667* SATB1 is a genome organizer protein that is expressed in a lineage specific manner in CD4<sup>+</sup> T-cells. SATB1 plays a crucial role in expression of multiple genes throughout the thymic development and peripheral differentiation of T cells. Although SATB1 function has been subjected to intense investigation, regulation of *SATB1* gene expression remains poorly understood. Analysis of RNA-seq data revealed multiple transcription start sites at the upstream regulatory region of *SATB1*. We further demonstrated that *SATB1* gene is expressed via alternative promoters during T-helper (Th) cell differentiation. The proximal promoter "P1" is used more by the naïve and activated CD4<sup>+</sup> T-cells whereas the middle "P2" and the distal "P3" promoters are used at a significantly higher level by polarized T-helper cells. Cytokine and TCR signaling play crucial roles toward *SATB1* alternative promoter usage. Under Th2 polarization conditions, transcription factor STAT6, which operates downstream of the cytokine signaling binds to the P2 and P3 promoters. Genetic perturbation by knockout and chemical inhibition of STAT6 activation resulted in the loss of P2 and P3 promoter activity. Moreover, chemical inhibition of activation of NF-κB, a transcription factor that operates downstream of the TCR signaling, also resulted in reduced P2 and P3 promoter usage. Furthermore, usage of the P1 promoter correlated with lower SATB1 protein expression whereas P2 and P3 promoter usage correlated with higher SATB1 protein expression. Thus, the promoter switch might play a crucial role in fine-tuning of SATB1 protein expression in a cell type specific manner.

Keywords: SATB1, alternative promoter, TCR signaling, cytokine signaling, STAT6

# INTRODUCTION

T-cells constitute the cell mediated, adaptive, immune system component in jawed vertebrates. Tcells develop in thymus and differentiate in periphery. The common lymphoid progenitor (CLP) cells develop in thymus into CD4 and CD8 single positive T-cells in a series of steps orchestrated by transcription factor network. The CD4<sup>+</sup> T-cells which enter peripheral blood further undergo differentiation into one of many functionally distinct T-helper cell subtypes depending on the antigen stimulus and cytokine environment [reviewed in (1, 2)]. Naïve CD4<sup>+</sup> T-cells respond to pro-inflammatory cytokines IL-12 and IFN-γ and differentiate along Th1 lineage aided by the master regulatory transcription factor T-bet. In contrast, IL-4 directs the differentiation of naïve cells toward the Th2 lineage via the transcription factor GATA-3. Similarly, IL6/TGFβ skew the differentiation of naïve cells toward Th17 lineage via the master regulatory transcription factor RORγT [also reviewed in (3)].

Special AT-rich sequence Binding protein 1 (SATB1) (4), plays a crucial role in the development of T-cells in the thymus as well as their differentiation in the periphery [reviewed in (5)]. In the periphery, SATB1 is expressed by T-helper cells where it activates genes in a locus-specific manner (6, 7). The importance of SATB1 in T-cells is underscored by the observation that Satb1- KO animals exhibit arrested thymic development (8). Recently, SATB1 was shown to be essential in specifying T-lymphocyte subsets by directing lineage-specific transcription programs (9).

In Th2 cells, SATB1 regulates expression of GATA3 in a Wnt/β-catenin signaling-dependent manner. Upon Wnt signaling, β-catenin translocates to the nucleus and binds to SATB1 to de-repress a cascade of genes crucial in differentiation (10). SATB1 also regulates downstream production of IL-5 cytokine by direct binding to the IL-5 promoter (7, 10). In contrast, during regulatory T (Treg) cell differentiation downregulation of SATB1 is essential (11). Treg cells are essential for immune tolerance. Treg cells respond to and secrete the cytokine TGF-β, express the master regulator transcription factor FOXP-3. FOXP-3 represses SATB1 transcriptionally by regulating its expression and post-transcriptionally by upregulating microRNAs that target 3' UTR of the SATB1 transcripts (11, 12). Interestingly, SATB1 is expressed at the Treg precursor stage of development and plays a crucial role in the lineage specification of Treg cells in the thymus (13).

Despite the importance of SATB1 in T-cell development and function, the mechanism that regulates its expression in Thelper cells remains poorly understood. In thymocytes, SATB1 gene is dynamically expressed throughout all the stages. The T-cell receptor (TCR) signaling has been shown to play an important role in SATB1 gene expression during early thymocyte development (14). Specifically, the transcription factor GATA-3 was found to directly regulate SATB1 expression in developing thymocytes by binding to the upstream regulatory region (14). Analysis of publicly available T-cell transcriptome data resulted in identification of a large regulatory region at the SATB1 gene locus. This large regulatory region codes for multiple SATB1 mRNA isoforms that differ in the transcription start sites corresponding to promoters. These isoforms that result from alternative promoter (AP) usage, differ in the sequence of the 5' UTR and splicing of the first exon that harbors them. Alternative promoters play crucial role in gene regulation in the determination of cell fate and function. APs allow diversification of transcriptional regulation enabling expression in various cell lineages and developmental stages. Use of APs results in mRNA isoforms that differ in the sequence of 5' UTRs that are crucial for post-transcriptional regulation [reviewed in (15)]. With this background, we studied the role of alternative promoters in SATB1 expression during T-helper cell differentiation.

Here, we show a complex mechanism of SATB1 regulation during peripheral T-helper differentiation. We found that SATB1 gene expression is regulated via alternative promoters (proximal P1, middle P2, and distal P3) during peripheral differentiation of CD4<sup>+</sup> T-cells. The helper T-cells rely on P2 and P3 promoter usage whereas activated T-cells and Treg cells preferentially use the P1 promoter, suggesting the importance of pro-inflammatory cytokines in promoter switching. Experiments performed using a Jurkat cell line based system suggested a crucial role of TCR signaling in P2 and P3 promoter usage. We identified STAT family of transcription factors that operate downstream of cytokine signaling and NF-κB that operates downstream of the TCR signaling as regulators of SATB1 P2 and P3 promoter usage. Finally, we find differential correlation between SATB1 isoforms that result from alternative promoter usage and SATB1 protein expression suggesting possible role of alternative promoters in regulation of protein expression.

# MATERIALS AND METHODS

# RNA-Seq Analysis

Publicly available human CD4<sup>+</sup> T-cell polyA RNA-Seq datasets [E-MTAB-2319 (16), GSE35871 (17), and GSE71645 (18)] were analyzed to identify SATB1 transcripts in various CD4<sup>+</sup> primary T-cells and cell-lines. In brief, reads were aligned to reference human genome assembly [hg38, Gencode (19)] using HiSAT2. Transcripts were assembled and merged using Stringtie (20). Merged transcriptome assembly was visualized on IGV Genome Browser (21). CpG island track was downloaded from UCSC genome browser for the hg38 genome assembly and was also uploaded onto the genome browser (22). SATB1 expression was analyzed in Th2 cells and induced Treg (iTreg) cells using featureCounts (23) and DESeq2 (24). Exon expression was analyzed by generating an exon-count matrix. The GlmQLFit test in EdgeR was applied for differential expression analysis (25). Normalized exon-counts were converted to FPKM for expression plot. Statistical significance of the number of overlapping differentially expressed genes between Jurkat cells and primary T-cells was tested using two-tailed hypergeometric test (26). Junction reads between SATB1 alternative first exons and second exon were plotted using bam files on the IGV genome browser.

# In vitro Differentiation of Naïve CD4<sup>+</sup> T-Cells and Inhibitor Treatment

Human naïve CD4<sup>+</sup> T-cells were isolated from total peripheral blood mononuclear cells (PBMCs) by magnetic bead based negative selection (130-094-131, Miltenyi Biotec). In brief, ∼100 mL human peripheral blood samples were collected from healthy volunteers. Blood samples were diluted in PBS and subjected to Ficoll Paque (17-1440-03, GE Healthcare) density gradient centrifugation for separation of PBMCs. Purity of naïve CD4<sup>+</sup> T-cells was confirmed and their in vitro differentiation was confirmed by flow cytometry analysis (16). In brief, naïve CD4<sup>+</sup> T-cells were plated at 1 million/ml density in complete RPMI medium. For non-specific activation, naïve cells were incubated in the presence of 2µg/ml anti-CD3 (130-093-387, Miltenyi Biotec), 2µg/ml anti-CD28 (130-093-375, Miltenyi Biotec). For Th2 differentiation, naive cells were incubated in presence of 2µg/ml anti-CD3, 2µg/ml anti-CD28, 10 ng/ml IL-4 (130- 093-915, Miltenyi Biotec), 10µg/ml anti-IFNγ (130-095-743, Miltenyi Biotec), and 10µg/ml anti-IL-12 (130-095-755, Miltenyi Biotec). Cells were harvested after 72–96 h of incubation. Th2 differentiation was confirmed by staining for intracellular IL-4 using anti-IL4 PE antibody (12-7049-42, eBiosciences). Naïve CD4+, Th0, and Th2 cells were subjected to RNA and protein isolation followed by qRT-PCR and Western blot analysis.

For studying the effect of NF-κB transcription factor on SATB1 alternative promoter expression, naïve CD4<sup>+</sup> cells were subjected to non-specific activation or differentiation for 48–72 h followed by treatment with 6µM NF-κB inhibitor (CID2858522, Tocris) for another 24 h. The cells were then harvested and subjected to either qRT-PCR analysis for SATB1 alternative promoter usage or Western blot.

For studies in mice, spleens were dissected from wild type (WT) and Stat4 or Stat6 knockout (KO) animals (Jackson Laboratories, Bar Harbor, ME, USA). Cells were extracted from tissues and were cleared using cell strainer (352340, Corning). Cells were then subjected to naïve CD4<sup>+</sup> T-cell isolation by negative selection using magnetic beads (CD4+CD62L<sup>+</sup> T cell Isolation Kit II, Miltenyi Biotec, 130-093-227). Similar to human samples, a fraction of naïve CD4<sup>+</sup> T-cells was activated in presence of plate bound anti-Cd28 500 ng/ml (BD 553295), anti-CD3 1µg/ml (BD 553238), IL-2 (R&D Systems 419 402-ML), or subjected to in vitro Th2 differentiation for 4 days in additional presence of Il-4 10 ng/ml (R&D Systems 404-ML), anti-Ifnγ 10µg/ml (BD 557530), and anti-Il-12 10µg/ml (BD 554475). A fraction of cells from WT and knockout animals were subjected to Gata3 (BD 560074), Ifn-γ (BD 554411), and Satb1 (BD 562378) staining to confirm polarization. The other fraction of cells was used for RNA isolation followed by qRT-PCR analysis.

#### RNA Sequencing and qRT-PCR Analysis

Total RNA was extracted using RNeasy Mini Kit (Qiagen, #74106). Isolated RNA were either subjected to PolyA enrichment using MicroPoly(A) Puris Kit (Thermo #AM1919) followed by sequencing (Accession number PRJNA503398) or cDNA synthesis using random hexamer primers for differential gene expression analysis. Alternative first exon specific forward primers (Exon 1a, 1b, and 1c corresponding to putative promoters P1, P2, and P3) were designed and used with Exon 2 specific reverse primer. Gene expression (SATB1, GATA3, IL2RA, 18s rRNA) analysis was also carried out using specific primers. Primer sequences used in qRT-PCR analysis are summarized in **Supplementary Table 1**, **Supplementary Information**.

#### ChIP-Seq Analysis

Publicly available mouse dataset for Stat6 ChIP-seq in Th2 and Stat4 ChIP-seq in Th1 [GSE22105 (27)], and H3K4me3 ChIP-Seq in naïve, Th2 and Th1 cells [GSE14254 (28)] were analyzed for binding on the mouse Satb1 locus. In brief, ChIP-seq reads were aligned to mouse Gencode mm10 genome assembly (19) using Bowtie2 (version 4.8.4) (29) and BWA (version 0.7.12) (30) for Stat4 or Stat6 ChIP-Seq and H3K4me3 ChIP-Seq datasets, respectively. For Stat4 and Stat6 ChIP-Seq, aligned files were subjected to peak calling using MACS (31). The bed files and the bigwig files were visualized on the IGV genome browser.

#### Jurkat Cell Culture and Treatments

Jurkat E6.1 cells were obtained from ATCC (TIB-152). Cells were cultured in RPMI 1640 complete medium (10% FBS) as per the ATCC guidelines. Cells were seeded at 0.2 million/ml density 1 day prior to the activation. Jurkat cells were activated with 0.1µg/ml Phorbol 12-myristate 13-acetate (PMA) (P1585, Sigma), 1µM Ionomycin (I0634, Sigma), and/or 50 ng/ml IL-4 (130095373, Miltenyi Biotec) as indicated. The cells were harvested 48 h after treatment. To dissect out the role of transcription factors in regulation of SATB1 promoters, Jurkat cells were treated with specific inhibitors 24 h before they were harvested. Following specific inhibitors were used−5 and 20µM JAK3 inhibitor (420122, Calbiochem), and 6µM NF-κB inhibitor (CID2858522, Tocris). After harvesting, Jurkat cells were subjected either to qRT-PCR analysis for SATB1 alternative promoter usage or for expression of various other genes (IL2RA, GATA3) or were subjected to western blot analysis.

### Western Blot Analysis

Cell lysates were prepared in RIPA buffer (150 mM NaCl, 1% IGEPAL, 0.5% Na-deoxycholate, 0.1% SDS, 50 mM Tris pH8) with protease and phosphatase inhibitors. Protein amounts were estimated by BCA method (Thermo Scientific # 23227). Equal amounts of protein were electrophoresed on a 10% SDS-polyacrylamide gel and transferred onto a PVDF membrane. Western blot analysis was performed using the following primary antibodies: rabbit anti-pSTAT6 (#9361, Cell Signaling Technology), rabbit total anti-STAT6 (#9362, Cell Signaling Technology), rabbit anti-SATB1 (#3650, Cell Signaling Technology), anti-GATA-3 (Abcam, ab106625), mouse anti-β-Actin (ac004, Abclonal), mouse anti-γ-Tubulin (#T6557, Sigma), and appropriate anti-mouse and anti-rabbit secondary antibodies. Densitometry analysis was performed using ImageJ (v 1.5.1) (32).

#### Statistical Analysis

Gene expression values were normalized to the Control group as indicated. Student's t-test was applied for twogroup comparisons. For multiple group comparisons, one-way ANOVA was performed with post-hoc Bonferroni correction using commercial software (Prism 5.0a, GraphPad).

# RESULTS

#### SATB1 Uses Alternative Promoters During Th2 Differentiation

We performed analysis on publicly available RNA-Seq datasets for primary CD4<sup>+</sup> T-cells and cell lines [E-MTAB-2319 (16), GSE35871 (17) and GSE71645 (18)] and found presence of three mRNA isoforms expressed at the SATB1 locus. These isoforms differ in their transcription start site (TSS) and therefore 5' untranslated region (5'UTR). Interestingly, these isoforms do not differ in the coding DNA sequence (CDS) and thus seem to code for identical SATB1 protein sequence. The three transcription start sites encompass a ∼20 Kb regulatory region. We marked the regions around the TSSs as putative promoters (P1 proximal, P2 middle, and P3 distal) (**Figure 1A**). We

mRNA isoforms of *SATB1* with alternative first exons (E1a, E1b, and E1c) that correspond to the usage of three alternative promoters (P1, P2, and P3, respectively). (B) Schematic of *in vitro* differentiation of naïve CD4<sup>+</sup> T-cells into Th2 cells. Naive CD4+ cells were treated with anti-CD3 and anti-CD28 along with IL4 for 96 h to induce Th2 differentiation. (C) Immunoblot assay performed as mentioned in methods using the antibody against GATA3, phospho-STAT6 (pSTAT6), total STAT6 and SATB1 for naïve CD4<sup>+</sup> and Th2 cells. Increase in expression of GATA3, pSTAT6, SATB1 confirms the differentiation of naive CD4<sup>+</sup> cells into Th2 cells. (D,E) qRT-PCR analysis for total *SATB1* expression and *SATB1* alternative promoter usage in naive and differentiated Th2 cells. A significant increase in expression of SATB1 is observed which corresponds to increased usage of *SATB1* P2 and P3 promoter. Error bars indicate SEM. (*N* = 6; \* <0.05, \*\* <0.01); *P*-values were calculated using Student's *t*-test.

checked the expression of these three isoforms in human naïve CD4<sup>+</sup> T-cells and those subjected to in vitro differentiation into Th2 phenotype (**Figure 1B**) by quantitative real-time PCR (qRT-PCR) analysis. Naïve cell isolation and Th2 differentiation was confirmed by flow cytometry analysis for CD4 and IL-4 expression, western blot analysis for expression of Gata3 and activated Stat6 (pStat6) (**Supplementary Figures 1A–D**, **Figure 1C**). Expectedly, total SATB1 protein and gene expression also increased during Th2 differentiation (**Figures 1C,D**). The proximal promoter P1 transcripts showed decrease in expression whereas the P2 and P3 promoter transcripts showed significant increase in expression upon Th2 differentiation (**Figure 1E**).

#### Stat6 and Stat4 Regulate Satb1 P2 Promoter Usage

We performed Chromatin Immunoprecipitation sequencing (ChIP-Seq) analysis on public datasets [GSE22105 (27)] to study the histone modifications typically associated with transcription activation [H3K4me3 (33)] upon Th2 differentiation. We found that as compared to naïve CD4<sup>+</sup> T-cells, the Th2 cells showed increase in H3K4me3 marks on P2 and P3 promoter (**Figure 2A**). We then analyzed the ChIP-Seq data for the master regulator transcription factors of Th2 differentiation, such as Gata3 (34) [GSE20898 (35)] and Stat6 (36) [GSE14254 (28)] for their involvement in Satb1 promoter regulation. We found that Stat6 exhibited differential occupancy on mouse Satb1 promoters. Interestingly, Stat6 occupied P2 and P3 promoter regions but not the P1 promoter (**Figure 2B**). To study the importance of Stat6 binding in the regulation of alternative promoter usage, we used Stat6 knockout (KO) mice. Naïve CD4<sup>+</sup> T-cells isolated from spleens of wild-type and Stat6-KO mice were subjected to Th2 differentiation conditions. Satb1 isoform expression analysis by qRT-PCR suggested that P2 promoter usage was significantly affected in Stat6-KO mice (**Figure 2C**). Stat6-KO also resulted in significant decrease in Satb1 protein levels (**Figure 2D**) suggesting a significant contribution of P2 and P3 promoters toward protein expression.

Increase in Satb1 expression has also been observed in other T-helper subtypes (**Supplementary Figure 3**). Activation of cytokine signaling and Stat family of transcription factors is a property shared by all T-helper cells [reviewed in (37)]. For example, Stat4 transcription factor plays a crucial role in differentiation of naïve CD4<sup>+</sup> T-cells into Th1 phenotype (38). We analyzed H3K4me3 marks and Stat4 occupancy at the Satb1 locus in Th1 cells. Similar to Th2, Th1 cells also showed an enrichment of both H3K4me3 marks and Stat4 occupancy at the P2 promoter. When naïve cells isolated from Stat4-KO mice were subjected to Th1 differentiation conditions, they failed to show an increase in P2 promoter usage, unlike the wild-type animals (**Supplementary Figure 4**).

Next, we analyzed publicly available transcriptome data [GSE71645 (18)] for the expression of SATB1 alternative first exons in CD4<sup>+</sup> T-cells subjected to activation by TCR signaling (Th0) and Th2 polarization conditions. We observed that P2 and P3 promoter usage was higher in Th2 cells as compared to Th0 cells underlining the importance of cytokine signaling in P2 and P3 promoter expression (**Figure 3A**). Total SATB1 expression was also higher in Th2 cells (**Figure 3B**).

To further validate the importance of Stat6 in P2 promoter usage, we established an in vitro system that mimics Th2 differentiation. We cultured Jurkat cells under conditions that mimic polarization signals (**Figure 3C**) with an aim to establish a system that permits easier manipulation at the genetic level, which will be required for our future studies. Upon treatment of Jurkat cells with PMA and Ionomycin [that mimic activation of TCR signaling (39)] in the presence of IL-4 cytokine, phospho-Stat6 (pStat6) levels were elevated along with increase in SATB1 P2 promoter usage (**Figures 3E,F**). We performed RNAseq analysis of Jurkat cells treated with activating (P+I) and polarizing (P+I+IL4) conditions and found significant overlap of upregulated genes between Jurkat cells activated in presence of IL-4 and Th2 cells (**Figure 3D**). Treatment of Jurkat cells subjected to polarizing conditions in the presence of JAK3 inhibitor resulted in decrease in P2 promoter usage in a dosedependent manner (**Figures 3G,H**), suggesting causal role of JAK/STAT signaling in SATB1 P2 and P3 promoter usage. Thus, the results obtained using the chemical inhibitors in the in vitro system corroborate those obtained from the in vivo experiments with genetic perturbation (knockout animals) and consolidate the role of JAK/STAT signaling in the P2 promoter usage.

### NF-κB Signaling Regulates SATB1 P2 Promoter Usage

We found that STAT-family of transcription factors downstream of the cytokine signaling positively regulate the SATB1 P2 promoter expression. However, when Jurkat cells were treated with IL-4 cytokine (in absence of PMA and Ionomycin), no increase in P2 and P3 promoter expression was observed (data not shown). The switch in promoter usage was observed only upon simultaneous activation of cytokine and TCR signaling. This observation does not rule out the role of TCR signaling in SATB1 alternative promoter usage. TCR signaling is mediated by NFAT transcription factors downstream of the CD3 signaling (40) and the AP-1 (41) and NF-κB (42) transcription factors downstream of the CD28 signaling. Synergistic activation of gene expression by cooperativity of STAT6 and NF-κB transcription factors has been reported (43–45). Therefore we checked if NF-κB and STAT6 co-regulate SATB1 P2 promoter expression by using a chemical inhibitor of NF-κB activation.

We subjected naïve human CD4<sup>+</sup> T-cells isolated from peripheral blood to Th2 differentiation. The differentiating Th2 cells that were treated with inhibitor that specifically affects NFκB activation downstream of PKCθ resulted in decrease in the P2 promoter usage. This decrease also coincided with decrease in SATB1 protein expression. No similar decrease was observed in the P2 promoter usage in Th0 cells (**Figures 4A,B**). We tested this observation in Jurkat cells subjected to activating (P+I) and polarizing conditions (P+I+IL4). IL2RA and GATA3 were used as positive controls for NF-κB inhibition in activated (46) and polarized Jurkat cells (47), respectively (**Figures 4C,D**). We also monitored if NF-κB inhibition affects phosphorylation of STAT6 in polarized Jurkat cells and found no significant difference in pSTAT6 levels upon NF-κB inhibition (**Figure 4E**). Similar to primary cells, SATB1 P2 promoter usage was significantly affected when polarized Jurkat cells were treated with the NFκB inhibitor. Unlike primary cells, usage of the P3 promoter was also affected though not to the same extent as the P2 promoter. Similar to Th0 cells, this inhibition was specific to polarized Jurkat cells since those activated with PMA and Ionomycin did not show any effect on usage of any of the SATB1 promoters. These results suggested a synergistic activation of SATB1 P2 promoter usage by IL4 and NF-κB signaling (**Figures 4F,G**).

# SATB1 P2 Promoter Usage Is Specific to T-Helper Cells

CD4<sup>+</sup> T-cells can differentiate into Th cell subtypes or regulatory T-cells. While SATB1 expression has been observed in T-helper cells, SATB1 expression is suppressed in regulatory T-(Treg) cells by both transcriptional and post-transcriptional mechanisms (11). Unlike T-helper cells, Treg cell differentiation is triggered by the anti-inflammatory cytokine TGF-β (48) which leads to the activation of SMAD transcription factors and those downstream of the TCR-signaling (49). We then studied the usage of SATB1

mice, respectively. Error bars represent SEM (*N* = 4); *P*-values were calculated using one-way ANOVA (\* <0.05). *Stat6* KO adversely affects *Satb1* alternative promoter usage. Unlike the wild type animals, no significant increase is observed in *Satb1* P2 and P3 promoter usage in cells from *Stat6* KO animals subjected to Th2 differentiation conditions. (D) Flow cytometry analysis for Satb1 protein expression in wild-type and *Stat6* KO, respectively under Th2 differentiating conditions. Satb1 protein expression is not enhanced when naive T-cells from *Stat6* KO animals are subjected to Th2 differentiation conditions.

alternative promoters in Treg cells using a publicly available RNA-seq dataset [E-MTAB-2319 (16)]. As expected, Treg cells showed lower usage of P2 promoter as compared to Th2 cells. Surprisingly, P1 promoter usage was significantly higher in Treg cells (**Figures 5A,B**), suggesting its association with activation of TCR signaling in the absence of the pro-inflammatory cytokine signaling. To test this hypothesis further, we studied SATB1 promoter usage in the Jurkat cell based system. We found that the SATB1 P1 promoter usage was higher in Jurkat cells activated with P+I and that the P1 promoter usage correlated with lower SATB1 protein expression levels (**Figures 5C–F**). Further analysis of Ribo-Seq and RNA-Seq data from Jurkat cells

FIGURE 3 | STAT6 regulates P2 promoter usage *in vitro*. RNA-Seq analysis of available data (GSE71645) for (A) expression of *SATB1* alternative first exons and (B) total *SATB1* expression. Expression analysis confirms higher usage of *SATB1* P2 and P3 promoters and higher total *SATB1* expression in human Th2 cells as compared to Th0 cells. Error bars represent min-max values of FPKM and normalized counts, respectively (*N* = 3); *P*-values calculated using EdgeR and DESeq2 respectively. (C) Schematic for the treatment of Jurkat cells under activating (PMA+Ionomycin; P+I) and polarizing conditions (+IL4). (D) Venn diagram showing an overlap of differentially expressed genes obtained by RNA-seq analysis between Th0 vs. Th2 cells and P+I vs. P+I+IL4 treated Jurkat cells. Significant overlap is observed between genes expressed higher in Th2 and P+I+IL4 treated Jurkat cells suggesting that Jurkat cell line-based model mimics Th2 differentiation condition. Significance of overlap was calculated using two-tailed hypergeometric test. (E) Immunoblot analysis confirms STAT6 activation (pSTAT6) only in Jurkat cells subjected to polarizing conditions. γ-Tubulin was used as loading control. (F) qRT-PCR analysis for *SATB1* P2 promoter usage in Jurkat cells activated with TCR and cytokine signaling. A significant increase in expression is observed in *SATB1* P2 promoter usage in Jurkat cells under polarization conditions (*N* = 5). (G) Immunoblot analysis for pSTAT6 levels and (H) qRT-PCR analysis for *SATB1* P2 promoter usage in Jurkat cells activated with TCR and cytokine signaling in presence of JAK3 inhibitor. A decrease in pSTAT6 levels and *SATB1* P2 promoter usage is observed upon JAK3 inhibition. Error bar represents SEM (*N* = 3); *P*-values calculated using Student's *t*-test (\* <0.05, \*\* <0.01).

revealed higher ribosome occupancy on the alternative first exons resulting from usage of the P2 and P3 promoters as compared to the P1 promoter. These results suggest differential translatability of SATB1 isoforms (**Supplementary Figure 5**).

# DISCUSSION

In summary, the results presented above demonstrate that SATB1 gene expression is orchestrated by an intricate regulatory network

FIGURE 4 | NF-κB regulates *SATB1* P2 promoter expression. (A) Quantitative RT-PCR analysis of *SATB1* alternative promoter usage in iNF-κB treated Th0 and Th2 cells. (B) Western blot for SATB1 expression in control and iNF-κB treated Th2 cells and densitometry analysis of expression (*N* = 4). Quantitative RT-PCR analysis of NF-κB target genes (C) *IL2Ra* and (D) *GATA3* confirms NF-κB inhibition in Jurkat cells subjected to activating and polarizing conditions. (E) Immunoblot analysis of activated STAT6 (pSTAT6) expression. Quantitative RT-PCR analysis of (F) *SATB1* alternative promoter usage and (G) total *SATB1* expression upon NF-κB inhibition in Jurkat cells. A significant decrease in *SATB1* P2 and P3 promoter usage was observed upon inhibition of NF-κB in polarized but not in activated Jurkat cells (*N* = 5). Error bar represents SEM and *P*-value calculated using one-way ANOVA (ns, not significant) (\* <0.05).

protein expression. Error bar represents SEM (*N* = 3 for Figures 3C,E,F).

of NF-κB signaling and cytokine signaling. During T-helper cell differentiation, the SATB1 gene is expressed via alternative promoter usage. The P1 promoter is preferentially used by the naïve CD4<sup>+</sup> T-cells and Th0 cells whereas the P2 and P3 promoters are preferentially used by the Th2 cells. STAT6 transcription factor that is downstream of cytokine signaling binds to the SATB1 P2 promoter and positively regulates its expression in Th2 cells. The NF-κB transcription factor which is downstream of the TCR signaling also regulates the P2 and P3 promoter usage. Finally, we observed that the P1 promoter was used more in Treg cells and Th0 cells which weakly correlated with protein expression. Whether this correlation results from differential translation of resulting isoforms or differential protein degradation in different cell types needs to be further studied. The results are summarized and represented schematically in **Figure 6**.

Various genes have been shown to be regulated via multiple promoters during immune cell activation (50–52). Substantial number of these genes switch promoters without change in the coding DNA sequence suggesting that the promoter switch may enable their expression under different transcription factor repertoires in different physiological contexts. The master regulator transcription factor of Th2 differentiation, GATA3 (53), and the key pro-inflammatory cytokine, IL4 (54), have also been shown to differentially use alternative promoters in naïve and differentiated Th2 cells. Similar to SATB1, GATA3 alternative

FIGURE 6 | Graphical representation of *SATB1* expression via alternative promoters in activated (Th0) and polarized (Th2) CD4<sup>+</sup> T-cells. Activation of non-specific TCR signaling in Th0 cells leads to *SATB1* P1 promoter usage. However, in the polarized Th2 cells, activation of cytokine signaling along with TCR activation leads to use of P2 and P3 promoters. Transcription factor STAT6, which acts downstream of the cytokine signaling and NF-κB, which acts downstream of the TCR signaling, positively regulate P2 and P3 promoter usage in polarizing conditions. The switch in promoter usage also correlates with change in SATB1 protein expression. The promoter switch may therefore enable regulation of SATB1 expression in a cell-type specific manner.

promoters are also under the dual regulation of IL-4 and TCR signaling. However, while GATA3 is specifically expressed in Th2 cells, the SATB1 locus responds to cytokine signaling during both Th1 and Th2 differentiation.

T-helper cell differentiation is triggered by antigens and cytokine cues leading to activation of STAT family of transcription factors by phosphorylation. STAT4 and STAT6 play a critical role in maintaining chromatin configuration and transcription of genes that drive Th1 and Th2 differentiation, respectively. STAT4 and STAT6 bind to specific DNA sequences in these two cell types but also regulate quite a few common target genes (27), SATB1 being one of them. STATs also shape the enhancer landscapes of T-cells. Interestingly, we did not observe expected motifs at the STAT4 and STAT6 occupancy sites in the SATB1 P2 promoter region. The nearest direct STAT binding site was observed upstream of the SATB1 regulatory region (data not shown). STAT6 also participates in the longrange intra-chromosomal interactions in Th2 cells to regulate expression of various gene loci (55). Therefore, the possibility that the upstream STAT binding site operates as an enhancer element to regulate P2 (and P3) promoter expression via promoter-enhancer looping cannot be ruled out.

We observed an increase in the usage of two out of the three SATB1 alternative promoters (P2 and P3) during T-helper cell differentiation. The usage of P2 promoter increases more than that of the P3 promoter. The P3 promoter region also harbors the start site of the divergent transcript from the SATB1-AS1 (SATB1 Anti-Sense RNA 1) gene. Bidirectional transcription is often associated with genes related to transcriptional regulation and development (56) and their promoters often coincide with large CpG islands (57). In agreement with this notion the SATB1 P3 promoter coincides with a CpG island. However, since the usage of both P2 and P3 promoters positively correlates with SATB1 protein expression, the need for the use of two different promoters remains unclear. The use of two different promoters might presumably enable rapid increase in SATB1 expression with an additive effect on transcription. Else, SATB1 isoforms may affect expression and translation of other isoforms. These possibilities need to be further studied. The Jurkat cell line based system developed here can be useful for the study of promoter functions by knock-outs or similar experiments.

The Jurkat cell-line based system was also crucial in suggesting the involvement of NF-κB in SATB1 P2 promoter usage. Jurkat cells treated with IL-4 cytokine alone did not result in phosphorylation of STAT6 or increase in P2 promoter usage (data not shown). However, IL-4 treatment in combination with activation using PMA and ionomycin (P+I) resulted in STAT6 activation and increase in P2 promoter activity. This suggested that either pSTAT6 alone is sufficient for increase in P2 promoter usage or the transcription factors downstream of the TCR signaling co-operate with STAT6 to drive the P2 promoter usage. Considering the role of NF-κB in regulation of cytokine signaling (58), we blocked NF-κB activation using chemical inhibitors and found that without affecting the global levels of STAT6 phosphorylation, the treatment resulted in reduced P2 promoter usage and also affected P3 promoter significantly in Jurkat cells. Synergistic activation of gene expression by NF-κB and STAT6 has been observed in past (43–45). Such synergism could be responsible for driving SATB1 alternative promoter expression. Whether STAT6 and NF-κB interact directly at the P2 promoter remains to be studied.

SATB1 protein is expressed in T-helper cells in a lineage specific manner (9). Even though it is expressed in all T-helper subtypes, its role is better studied in Th2 differentiation (7). Unlike the helper T-cells, SATB1 is repressed in regulatory T-cells (11) at both transcriptional and post-transcriptional level suggesting distinct effects of pro-inflammatory cytokines and anti-inflammatory cytokines on the SATB1 P2 promoter expression. Interestingly, even though the expression of SATB1 is suppressed in Treg cells, the P1 promoter is active. Transcription via the P1 promoter is therefore weakly correlated with SATB1 protein expression. We extended this observation to the primary CD4<sup>+</sup> T-cells and Jurkat cells and observed that activation of these cells by TCR-signaling alone resulted in higher P1 promoter usage and lower SATB1 protein expression as compared to the polarizing conditions. It is therefore possible that the P1 promoter is less translatable and is used when basal level of SATB1 expression is needed.

Transcription factor/s responsible for P1 promoter usage in activated T-cells have not yet been identified. Since NF-κB inhibition had no specific effect on P1 promoter expression, AP-1 and NF-AT transcription factors both also downstream of TCR signaling, can be studied. Also, the role of NF-κB in regulation of P2 promoter usage in T-helper cells may not be direct. Further experiments need to be performed to check the direct binding of NF-κB at the P2 promoter. A co-operative binding of STAT factors with NF-κB has been previously reported (59) and cannot be ruled out in case of P2 promoter usage.

SATB1 alternative promoter usage leads to change in the sequence of the 5' untranslated region. The 5' UTR has been shown to play a role in regulation of various genes by controlling the rate of translation via secondary structure formation, presence of upstream open reading frames, and via miRNA binding [reviewed in (60, 61)]. We observed weak correlation between P1 promoter usage and SATB1 protein expression in Treg cells and activated Jurkat cells. It would be of interest to test whether in 5′ UTR sequence differentially affects the translation efficiency of SATB1 transcripts originating from alternative promoter usage. Further differential degradation of SATB1 protein in activated vs. polarized cells also cannot be ruled out as a cause of this discrepancy.

In summary, we have elucidated a complex mode of regulation of SATB1 gene expression via alternative promoters during peripheral differentiation of T-cells. SATB1 regulation represents a unique example with conserved mode of regulation in multiple T-helper subtypes with possible effect on protein levels. We anticipate that understanding of the unique functions of these promoters in gene regulation and the physiological consequences of their expression/repression will be a prime focus of the future studies.

# DATA AVAILABILITY

The datasets generated for this study can be found in NCBI SRA, PRJNA503398.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the Institutional Human Ethics Committee, IISER-Pune with written informed consent from all subjects. All subjects gave written informed consent. The protocol was approved by the Institutional Human Ethics Committee, IISER-Pune.

Mice used in this study were maintained in the Central Animal Laboratory at Turku University. This study was carried out in accordance with appropriate guidelines for the care and use of laboratory animals and were approved by the "Finnish Animal Ethics Committee." The protocol was approved by the "Finnish Animal Ethics Committee."

# AUTHOR CONTRIBUTIONS

SK designed experiments, performed ChIP-Seq and RNA-Seq analysis, performed Jurkat cell-line experiments, T-helper cell differentiation, qRT-PCR analysis, and wrote the manuscript. AS and IP performed FACS staining, Stat4/Stat6 KO experiments, performed T-helper cell differentiation, and qRT-PCR analysis. RB performed Jurkat cell-line experiments. AVS assisted in Jurkat cell experiments and T-helper cell differentiation. PR assisted in RNA-Seq analysis. ZC provided expertise in Stat4/Stat6 KO mouse experiments, RL participated in experimental design, discussed and interpreted results and supported the research. SG designed the experiments, was involved in regular discussions and data interpretations, supported the research and wrote the manuscript.

#### FUNDING

Work in SG lab is supported by the Center of Excellence in Epigenetics program of the Department of Biotechnology, grant numbers BT/01/COE/09/07, BT/MED/30/SP11288/2015, and intramural funds from IISER Pune. RL has been supported by the Academy of Finland, AoF, Center of Excellence in Molecular Systems Immunology and Physiology Research (2012- 2017) grant 250114; by the AoF grants 292335, 294337, 292482, 298732, 298998, 315585, and the Sigrid Jusélius Foundation

#### REFERENCES


(SJF). SK, AVS, and PR were supported by Post-Doctoral fellowships from IISER-Pune, DST-SERB (N-PDF), and DBT (RA), respectively. ZC was supported by the Academy of Finland, AoF grant 258313.

#### ACKNOWLEDGMENTS

The authors thank Sarita Heinonen and Marjo Hakkarainen for excellent assistance and Imran Mohammad for technical help during the knock-out animal experiments. The authors also thank Mr. Dattatray Bhat and the KEM diabetes research unit, KEM hospital, Pune for help in collection of blood samples.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00667/full#supplementary-material


element in the IL-2 receptor alpha gene. Science. (1988) 241:1652–5. doi: 10.1126/science.2843985


**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 Khare, Shetty, Biradar, Patta, Chen, Sathe, Reddy, Lahesmaa and Galande. 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.

# Cytokines: Key Determinants of Resistance or Disease Progression in Visceral Leishmaniasis: Opportunities for Novel Diagnostics and Immunotherapy

#### Edited by:

*Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States*

#### Reviewed by:

*Anil Dangi, Duke University Medical Center, United States Jitendra Kumar Tripathi, University of North Dakota, United States*

#### \*Correspondence:

*Suresh K. Kalangi suresh.kalangi@ indrashiluniversity.edu.in orcid.org/0000-0002-7328-9322 Suresh V. Kuchipudi skuchipudi@psu.edu*

*†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: *22 September 2018* Accepted: *12 March 2019* Published: *05 April 2019*

#### Citation:

*Dayakar A, Chandrasekaran S, Kuchipudi SV and Kalangi SK (2019) Cytokines: Key Determinants of Resistance or Disease Progression in Visceral Leishmaniasis: Opportunities for Novel Diagnostics and Immunotherapy. Front. Immunol. 10:670. doi: 10.3389/fimmu.2019.00670* Alti Dayakar <sup>1</sup> , Sambamurthy Chandrasekaran<sup>2</sup> , Suresh V. Kuchipudi <sup>3</sup> \* † and Suresh K. Kalangi <sup>4</sup> \* †

*1 Independent Researcher, Vizianagaram, India, <sup>2</sup> Bio5 Institute, University of Arizona, Tucson, AZ, United States, <sup>3</sup> Animal Diagnostic Laboratory, Department of Veterinary and Biomedical Sciences, The Pennsylvania State University, University Park, PA, United States, <sup>4</sup> Department of Biosciences, School of Sciences, Indrashil University, Mehsana, India*

Leishmaniasis is a parasitic disease of humans, highly prevalent in parts of the tropics, subtropics, and southern Europe. The disease mainly occurs in three different clinical forms namely cutaneous, mucocutaneous, and visceral leishmaniasis (VL). The VL affects several internal organs and is the deadliest form of the disease. Epidemiology and clinical manifestations of VL are variable based on the vector, parasite (e.g., species, strains, and antigen diversity), host (e.g., genetic background, nutrition, diversity in antigen presentation and immunity) and the environment (e.g., temperature, humidity, and hygiene). Chemotherapy of VL is limited to a few drugs which is expensive and associated with profound toxicity, and could become ineffective due to the parasites developing resistance. Till date, there are no licensed vaccines for humans against leishmaniasis. Recently, immunotherapy has become an attractive strategy as it is cost-effective, causes limited side-effects and do not suffer from the downside of pathogens developing resistance. Among various immunotherapeutic approaches, cytokines (produced by helper T-lymphocytes) based immunotherapy has received great attention especially for drug refractive cases of human VL. Therefore, a comprehensive knowledge on the molecular interactions of immune cells or components and on cytokines interplay in the host defense or pathogenesis is important to determine appropriate immunotherapies for leishmaniasis. Here, we summarized the current understanding of a wide-spectrum of cytokines and their interaction with immune cells that determine the clinical outcome of leishmaniasis. We have also highlighted opportunities for the development of novel diagnostics and intervention therapies for VL.

Keywords: Leishmania, cytokines, T-cells, visceral leishmaniasis, diagnosis, immunotherapy

# INTRODUCTION

Leishmaniasis is a neglected tropical disease (NTD) caused by an obligatory intracellular protozoan parasite that belongs to the genus Leishmania. It is a vector-borne infection transmitted by female sandflies and the disease is highly prevalent in poor and malnourished populations of the world living in tropical and subtropical countries. The life cycle of Leishmania is simple and the parasite propagates in two different morphological forms. The promastigote stage of the parasite exists in the insect body fluids and enters the mammalian host when sandfly takes a blood meal. Promastigotes transform into amastigotes inside the mononuclear phagocytes of hosts and establish the infection by evading host defense system (1). The infected individuals could develop self-healing cutaneous ulcers to life-threatening visceral disease (2). World-wide, 0.7–1.0 million new cases of leishmaniasis and 20,000 to 30,000 deaths are reported each year (3). Visceral leishmaniasis (VL) or kala-azar is the deadliest clincial form of leishmaniasis, typically caused by L. donovani and L. infantum in the Old World and L. chagasi in the New World. Occasionally, L. tropica and L. amazonensis have also been found to cause VL in the Middle East and South America, respectively (4). The anthroponotic transmission of VL is prevalent in the Indian subcontinent (5). The annual report of global VL indicates that there are 50,000 to 90,000 new cases each year with high incidence in the Indian subcontinent and East Africa (3). VL is an opportunistic infection and has been identified as a coinfection in HIV patients (6). HIV infection amplifies the risk of developing active VL and the severity by 100–2,320 times (7). Fever, weight-loss, anemia, pancytopenia, hyperpigmentaion of skin and hepatosplenomegaly are some of the manifestations of VL and the mortality rate is over 95% (3). Children under the age of 1 year and adults above 50 years of age are highly susceptible to VL (8, 9). The susceptible host genetic background (10), nutritional status especially malnutrition (11) and immune suppression (12) ameliorates the clinical outcome of the disease. The current VL treatment relies mostly on chemical drugs like pentavalent antimonials (SbV), amphotercin B, miltefosine, and paromomycin etc. But their misuse, life-threatening toxicity, and development of resistance by the parasites (13) highlight the need for drug-sparing alternative therapeutic strategy to combat the clinical disease. Recently, immunotherapy has emerged as a promising option to control various diseases including VL. This review presents an in-depth critical analysis of immune responses to leishmaniasis and highlights prospective cytokine candidates that could be used for the diagnosis and therapy of VL.

#### Leishmania Infection and Innate Immune Cells

Leishmania infection in humans is usually subclinical and parasites may persist for life-time of the host through several escape mechanisms (14). For example, Leishmania blocks the maturation of complement system and C5–C<sup>9</sup> membrane attacking complex formation, reduces the expression of B7 and CD40 that are required for T-cell anti-parasitic activity, promote overexpression of the iron transporters, modifies the toll-like receptor (TLR)-2/TLR-4 signaling and inhibits Janus tyrosine kinase/signal transducer and activator of transcription (JAK/STAT) pathway in macrophages (M8s) thereby turnoff the cytokine cascade, and alters the expression profile of cytokines and chemokines etc. It is clear that Leishmania parasites manipulate several key aspects of host defense for their survival. Consequently, targeting immune components is a reliable method to combat the disease. In addition, host innate immune signatures that are specific to Leishmania infection could help early prediction of the disease outcome. These include aspects of innate immune response, such as frontline defense led by the natural killer (NK) cells, mononuclear and polymorphonuclear phagocytes (15). In general, Leishmania parasite resists their uptake by phagocytic dendritic cells (DCs) and M8s (16, 17) by inhibiting reactive oxygen species (ROS) production that delays phagolysosome formation (18) and blocks lysosomal proteolytic degradation (19). The complement protein C3b, a potent immune opsonin accelerates phagocytosis of Leishmania (17) by interacting with the parasite surface glycoprotein gp63 (20). M8s and DCs that engulfed Leishmania activate their TLR-9 signaling and produce interleukin (IL)- 12, which stimulates NK cells to produce interferon (IFN)-γ, a key cytokine that is responsible for skweing Th1 response (16, 21) and stimulate the M8s to produce ROS and nitric oxide (NO) for oxidative killing of intracellular amastigotes thereby protects the host (22–25). To establish an early infection, L. major inhibits the NK cell proliferation and IFN-γ production (26) and L. donovani evades inducible nitric oxide synthase (iNOS)-dependent killing of intracellular amastigotes in M8s via downregulation of iNOS mRNA expression (27, 28) and induction of arginase expression (29) as the arginine is a common substrate for both iNOS and arginase enzymes. Thus, M8s play a complex role in Leishmania pathogenesis and are associated with both survival and death of the parasites (30). Further, the induction of FasL-mediated apoptosis in Leishmania infected M8s is a part of host defense mechanisms in innate immunity (31). Similarly, other immune cells also play a key part in the early host defense against leishmaniasis. For example, a drastic reduction in IL-8 and eotaxin secretion from neutrophils and eosinophils, respectively (32), and an elevated number of IL-4 <sup>+</sup> neutrophils and IL-10<sup>+</sup> eosinophils and reduced number of IFN-γ <sup>+</sup> and IL-12<sup>+</sup> eosinophils are observed in active VL patients (33).

#### Origin of Th1-Th2 Dichotomy in Leishmaniasis

While the host innate immune response against leishmaniasis is important, it is now clear that the T-cell mediated immunity and the cytokines produced from various immune cells play a crucial role in determining the disease outcome (shown in **Figure 1**). However, the cytokines function in autocrine (locally) and paracrine (at a distance from the site of synthesis) fashion to regulate the immune response (34). A longitudinal study on Leishmania pathogenesis and disease recovery highlighted the role of helper T (Th)-cell responses (35). Therefore, immune cells and their cytokines have been recognized as potential targets for immunotherapy to modulate the activity of factors that

are crucial in the immune system for healing. In this context, the phenomenon called "Th1-Th2 dichotomy" became popular based on the role of the cytokines produced by these cells in disease progression and/or host protection. Mosmann et al. reported for the first time that the cloned murine Th-cells are in two functional subsets namely Th1 and Th2 based on the production of IFN-γ and IL-4, respectively (36). Thereafter, several studies demonstrated the key role of major cytokines [e.g., IL-10, Transforming growth factor (TGF)-β, IL-4, IL-6, IL-12, and IFN-γ] that implicated the role of Th1/Th2 balance in disease progression or host protection. In general, Th1 type response mediates host resistance and Th2 type response associates with disease progression (37). In resistant mouse strains, the abundance of Th1 type cytokines; IFN-γ, IL-2, and lymphotoxin spontaneously cleared the L. major infection, whereas, in susceptible mouse strains, infection led to the fatal disease by the action of Th2 type cytokines; IL-4, IL-5 and IL-10 (38). IL-4 and IL-10 associated with the visceralization of cutaneous L. major infection (39, 40). However, the Th1-Th2 dichotomy is more complex than previously recognized, which is more evident in certain cases of leishmaniasis (41), such as L. donovani infection where the susceptibility of mouse strains is variable (42). Unlike in cutaneous leishmaniasis (CL), T-cells with Th2 phenotype are difficult to demonstrate in the mouse model of VL (37, 40–43). Similarly, the association between Th1 response and disease resistance to VL is complex in humans (44, 45). Occasionally, individuals respond to the exposure of Leishmania antigens via T-cells even they have no prior exposure to the parasite; this is possible due to the cross-activity by other microorganisms (46).

# Immune Response During VL

Immune response in VL patients is characterized by abundant anti-leishmanial antibody titers and low or absence of Leishmania-specific T-cell proliferation and IL-2 and IFN-γ production. Recovery from VL is mostly dependent on the induction of T-cell immunity; preferably Th1 response, which is primed by IL-12<sup>+</sup> DCs and M8s (47, 48). IFN-γ produced from IL-12 primed T-cells induce NO-mediated killing of the parasites (49, 50). In contrast, VL progression in humans is associated with abundant production of Th2 type cytokines IL-10, TGF-β, and IL-4 or presence of IL-10<sup>+</sup> regulatory T cells (Tregs), which diminish the anti-parasitic activity of M1-type M8s and Th1 response (51–53). However, the presence of abundant IL-10 is crucial rather than a lack of IFN-γ in the VL clinical disease progression (54). IL-10 partially inhibits IFN-γ production but strongly resists IFN-γ mediated activation of M8s while killing the intracellular parasites (55–57). Likewise, the lack of IFN-γ may result in relatively higher levels of IL-10 in human leishmaniasis resulting in M8 deactivation (58) and parasite proliferation (59). Murine model of VL demonstrates higher disease susceptibility due to the presence of high IL-10 levels during initial phase of infection (60). The splenic infection of L. donovani causes a constitutive expression of chemokine ligand 2 (CCL2) or monocyte chemoattractant protein 1 (MCP-1), which triggers IL-4 secretion from Th2 cells that activates

the M8s in alternative manner. These M2-type M8s express arginase in abundant quantity and help in the biosynthesis of polyamines, which favor the survival and growth of the parasite (61). During chronic VL, the high expression of programmeded death protien-1 (PD-1) or Cytotoxic T-lymphocyte Antigen 4 (CTLA-4) causes unresponsiveness in CD4<sup>+</sup> T-cell, which produce TGF-β in abundant levels and helps in persistence of infection (62). Taken together, there is mounting evidence suggesting that Th1-Th2 imbalance and T-cell unresponsiveness are critical issues in VL pathogenesis.

#### Objective of the Review

Role of a whole host cytokines in the resistance and disease progression during VL is increasingly being uncovered. Till date targeting either Th1 or Th2 cytokines produced promising results for leishmaniasis cure. Th1/Th2 balance is not the only determinant of the outcome of leishmaniasis as previously thought because a range of other cytokines have recently been implicated in both disease progression and host protection (**Figure 2**). Hence, there is a need for in-depth analysis of the role of cytokines and Leishmania pathogenesis to get a comprehensive view of the complex interplay of Leishmania parasite and their hosts. This review aims to summarize and critically analyze the state-of-the-art knowledge relating to cytokines and VL pathogenesis. Special emphasis has been made for the identification of potential cytokine targets that could be used for the development of novel diagnostic assays and immunotherapies for the detection and treatment of VL.

# CYTOKINE RESPONSE IN LEISHMANIASIS

It is well-known that cytokines play a role in pathogenesis and hosts resistance of VL. Cytokines that play crucial role in Leishmania pathogenesis or host defense are tabulated in **Tables 1**, **2**, respectively. However, there are several cytokines that are play a dual role both in the disease progression and host resistance are summarized in **Table 3**. Cytokines targets for diagnosis and/or immunotherapy are shown in **Table 4**. Functions of individual cytokines as relates to pathogenesis and/or host resistance are discussed in detail in the sections below.

# IL-10 Is a Key Player in Disease Progression

IL-10 is an 18 kDa pleiotropic cytokine, primarily produced by alternatively activated M8s, DCs, and lymphocytes. As an immunoregulatory cytokine, IL-10 exerts multiple biological effects on different cell types (205). IL-10 is the product of Th2 subset, also known as cytokine synthesis inhibitory factor (CSIF) since it suppresses IFN-γ production from Th1 cells (112). IL-10 is known to inhibit production of cytokines like IL-1, IL-6, IL-12, and tumor necrosis factor (TNF)-α. In addition, Il-10 also inhibits M8 mediated activation of T-cell through the reduced expression of class II major histocompatibility complex (MHC) and co-stimulatory molecules on the surface of M8 and results in the inhibition of both innate and T-cell mediated immunity (188). The suppressive role of IL-10 in human VL results in the drastic fall in accumulation of monocyte derived macrophages, which is regulated by migration inhibitory factor (MIF). Further, IL-10 plays a substantial role in the pathogenesis of leishmaniasis by causing the downregulation of Th1 response, M8 activation (114) and antigen presentation by DCs (58). Furthermore, IL-10 inhibits the leishmanicidal functions of M8 (206) by diminishing the production of reactive nitrogen intermediates by M8, IFNγ by T and natural killer (NK) cells (115), and IL-12 mediated activation of M8 (48). High levels of IL-10 during the initial phases of infection due to decreased multifunctional CD4 T cells results in higher susceptibility to VL. Despite elevated levels of IFN-γ during the steady state of infection, parasite burden is not reduced due to higher levels of IL-10 (60). The unfavorable clinical outcome in localized CL was correlated with IL-10 but not with inadequate Th1 response (207). High levels of serum IL-10 is associated with symptomatic VL but absent in asymptomatic individuals. A key function of IL-10 is to protect the tissues from collateral damage due to excessive inflammation (116). However, in the face of parasitic infection an acute inflammatory response is necessary to control the parasite proliferation, hence, the anti-inflammatory role of IL-10 may TABLE 1 | Cytokines involve in the host protection.


help the disease progression (117). During active VL, CD8<sup>+</sup> Tcells could also play an important role in disease progression via abundant production of IL-10 (208). However, the role of IL-10 in VL appears to be species-specific as it was suggested that IL-10 may not be a regulatory cytokine in canine VL. In experimental CL, a group of Treg cells namely, CD4+CD25+Foxp3<sup>+</sup> and CD4+CD25−Foxp3<sup>−</sup> are possible source of IL-10 (209). In contrast, IL-10 in human VL is not produced from thymic Foxp3 Tregs; rather they are produced from IFN-γ co-producing CD4<sup>+</sup> T cells which are called type 1 regulatory (Tr1) cells (143). The role of Tregs was elucidated in modulating both Th1 (210, 211) and Th2 (210, 212) activity during murine L. major infection.

# TGF-β Functions Synergistically With IL-10 in Disease Progression

TGF-β is a 28 kDa homodimer and a potent anti-inflammatory cytokine produced by antigen-activated T-cells and mononuclear phagocytes (75). TGF-β has potent immunosuppressive effects in infectious and autoimmune diseases (118), which include inhibition of T-cell proliferation and M8 activation. TGFβ inhibits the functions of TNF-α and IFN-γ and controls the expression of inducible nitric oxide synthase (iNOS) and the development of Th1 and Th2 response. Unlike IL-10, the impact of TGF-β on parasite burden and IFN-γ dependent host resistance is marginal during L. donovani infection (119). Locally activated TGF-β favors the parasite growth by modulating innate and adaptive immune responses (120) and enhancing arginase expression (121, 122). In animal models, TGF-β secreted by Leishmania infected lymphocytes diverts the arginine pool from iNOS to arginase for the production of polyamines, which helps in the growth of the parasite (123). Both pro and antiinflammatory roles of TGF-β have been demonstrated (62, 124). L. chagasi infection induces TGF-β secretion by human M8s through activation of latent TGF-β itself. L. chagasi infection also induces the expression of TGF-β in spleen and liver tissues of both symptomatic and asymptomatic dogs (213). TGF-β exposure delays the killing of Leishmania parasite and TGF-β

#### TABLE 2 | Cytokines involve in the disease progression.


overexpression impairs the rate of cure in murine leishmaniasis models (125, 126). In human VL, the elevated levels of IL-10 and TGF-β postively correlate with the parasite load and with increased absolute numbers of FoxP3 Treg cells suggesting the role of Tregs in secretion of these cytokines. There is a significant correlation between the parasite load and circulating antigen specific TGF-β levels in VL patients suggesting its role in parasite multiplication and disease progression in humans (214).

### IL-4 Is Involved in the Pathogenesis of Leishmaniasis but Its Role in VL Is Conflicting

IL-4 is a 20 kDa Th2 subset cytokine that plays a critical role in the regulation of mast cell or eosinophil-mediated immune responses, B-cell mediated IgE class-switching and antibody production. It functions as a growth factor for mast cells and naive CD4<sup>+</sup> Th2 cells which produce anti-inflammatory cytokines IL-5, IL-10 and IL-13. Both IL-4 and IL-13 inhibit IFNγ-producing CD4<sup>+</sup> T-cells and suppress protective Th1 immune response (150) and trigger M8s to undergo alternative activation resulting in parasite survival and persistence of infection (151, 152). The similarities in IL-4 and IL-13 biological activities are due to a common receptor γ-chain that they both share, which is involved in the signal transduction via signal transducer and activator of transcription (STAT)-6 (215). Studies on murine model established the pathogenic role of IL-4 in leishmaniasis (39). IL-4 inhibits the oxidative burst by inducing low level production of reactive oxygen intermediates and NO in M8s (153, 154). L. major infected Langerhans cells show increased IL-4R expression and decreased IL-12 production in susceptible mice but not in resistant mice (155).

#### TABLE 3 | Cytokines with dual role in leishmaniasis.


IL-4 modulated antigen-uptake, endosomal processing, and humoral response are suggested to promote the disease development in Leishmania infection in humans (156). In murine L. donovani infection, IL-4 induces the host protective response (216) and vaccine mediated protection by IFN-γ secretion from CD8<sup>+</sup> T-cells (157). The peripheral blood mononuclear cells (PBMCs) harvested from cured VL patients produced IFN-γ or IL-4 in response to stimulation by L. donovani promastigote or amastigote crude antigen. In response to purified gp63 antigen, the proliferation capability of the same PBMCs was weak and produced IFN-γ or IL-4 (158). Cytokine analysis in VL unveils the induced expression of IL-10 and/or IL-4 mRNA in tissues and abundant levels of IL-4 in circulation of patients with progressive disease (217). Likewise, the conflicting role of IL-4 in VL is described, though it has a leading role in pathogenesis of VL as it is belonging to Th2 phenotype and anti-inflammatory cytokine subset. Recent studies on human splenic aspirates suggest that blockade of IFN-γ and TNF-α results in increased production of IL-4 which does not contribute to parasite replication and IL-10 production. The biological role of IL-4 in target organ of human VL still remains an outstanding question (218).

#### IL-13 Promotes Host Protection in VL and Its Role Is Leishmania Species-Specific

IL-13 is a 12-kDa cytokine that is expressed by Th2 and is important in host protection against Leishmania infection. For TABLE 4 | Cytokines used in diagnosis and chemo/immunotherapy of VL.


example, IL-13 knock-out mice infected with L. donovani show retarded hepatic granuloma formation and maturation, depleted IFN-γ secretion and enhanced production of IL-4 and IL-10 (159). IL-13 protects rats from L. major infection through the production of IL-1β and IL-12 (160), which is in contrast to the earlier studies that showed pathogenic role of IL-13 (16, 219). However, studies with BALB/c mice infected by L. mexicana and L. amazonensis have established the pathogenic impact of IL-4 and IL-13 (161). The susceptibility to L. (V.) panamensis infection is predominantly associated with IL-13 but not IL-4. The parasite species and the host genetic background may also influence the dual role of IL-13 and it may not be a potential target for immunotherapy (162).

#### Targeting Endogenous IL-6 May Offer Better Protection

IL-6 is a 26 kDa pleiotropic cytokine produced by a number of cell types, including monocytes, endothelial cells and Tlymphocytes (220). The main biological activities associated with IL-6 are the induction of acute-phase protein synthesis in hepatocytes, terminal differentiation of B-cells and activation of T-cells (221). It also induces the production of antiinflammatory proteins, such as IL-1 receptor antagonist (IL-1rα) and soluble TNF receptor (222). IL-6 plays a major role in the switching of monocytes from DC to M8s (130). IL-6 favors the development of Th2 response, which suppresses the activation and antimicrobial effect of M8s (75). Contradicting roles of IL-6 have been demonstrated in experimental CL and VL models (223–225). IL-6 has been shown to either promote (226, 227), suppress (228), or do not change (229) the intracellular host defense to leishmaniasis. Function of endogenous IL-6 as a host suppressive cytokine in case of L. donovani infection has outshined its potential pro-host defense effect.

During adoptive transfer of testing splenic DCs, IL-6 induces leishmanistatic effect but not host suppressive effect (226). IL-6 inhibits the IFN-γ mediated gene expression (131) and absence of IL-6 receptor signaling in L. donovani liver infection contributes to enhanced Th1-type response, accelerated tissue inflammation, and rapid parasite killing (132). IL-6 induces the secretion of IL-27 which in turn induces IL-10 production (133– 135) in the L. donovani infected mouse model (54). However, L. donovani-infected IL-6−/ <sup>−</sup> mice do not show effect on the IL-10 expression (226). This observation raises a potential possibility to target endogenous IL-6 as an anti-leishmanial therapeutic strategy (230). Expansion of CD25−FoxP3−IL-10+CD4<sup>+</sup> T-cells in vivo and therapeutic efficacy of adoptively transferred DCs against L. donovani infection are regulated by DC-derived IL-6 (226). IL-6 is produced by dogs with active leishmaniasis and is a key player in the pathogenesis of canine leishmaniasis (144, 231). The presence of TNF-α and IL-6 transcripts was found in both Leishmania antigen stimulated and unstimulated cells of asymptomatic infected and uninfected dogs (232). Further, increased anti-leishmanial antibody titers (hypergammaglobulinaemia) in canine VL are usually associated with high levels of IL-6 (224). Contrary to murine VL, the role of IL-6 in human VL is associated with disease severity and death, which is due to the inhibition of TNF-α in the early phase of infection and later by inhibiting the Th1 responses (190, 233).

# IL-27 Contributes to VL Pathogenesis and a Potential Target for Anti-VL Therapy

IL-27 is a member of the IL-6/IL-12 cytokine family and a heterodimer composed of EBI3 and p28. The main cellular source of IL-27 is CD14<sup>+</sup> spleen cells (140) in particular M8s and DCs. The anti-inflammatory properties of IL-27 have been demonstrated in various models of infectious diseases and autoimmunity (234). IL-27 mediates anti-inflammatory response by suppressing Th17 cells (138, 139) and inducing IL-10 secretion from activated CD4<sup>+</sup> T-cells via autocrine action of IL-21 (140). IL-27 plays a multifaceted role characterized by the induction of T-bet (141) in turn inhibition of parasite driven Th2 and Th17 development and Th1 polarization via IL-10 mediated feedback mechanism. IL-27 plays a critical role in the induction of IFN-γ and IL-10 from CD4<sup>+</sup> Tcells, and suppression of inappropriate Th17 development to achieve immune-balance during intracellular parasite infections. In L. major infection, early burst of IL-4 suppresses IL-27 mediated development of normal Th1 by inducing IL-6 and TGF-β and promote the development of Th17 cells (142). In contrast, IL-27 is not required for the normal development of Th1 response to L. donovani infection (140) but induces IL-10 production (143, 145).

Absence of IL-27 receptor signaling in L. donovani liver infection contributes to the accelerated Th1-type response, tissue inflammation, and rapid parasite killing with reduced parasite burdens in spleen and liver (71, 146). Blocking IL-27 evoke different responses in different mice models. For example, blocking IL-27 results in reduced parasite loads in BALB/c mice and augmented parasite burden is seen in C57BL/6. This dichotomy in the production of IL-27 could be due to the consequence of host immune modulation by the parasite to establish infection (235). IL-27 levels were elevated in human plasma with active VL and splenic mRNA levels of IL-27 and IL-21 were higher in pre-treated biopsies compared with posttreatment samples (140). During VL caused by L. infantum, the sequential pathway of TLR3 and TLR9 recruitment, production of type I IFN and activation of IRF1 in macrophages is induced by IL-27. The secretion of IL-27 increases the Th1 response but also dampens the production of IL-17 which directly impacts the reduced recruitment of neutrophils to target organs (236). Inhibition of IL-27 could be targeted for design of anti-VL treatment in the future.

# IL-5 Exerts Moderate Effects on VL Progression

IL-5 is a glycosylated homodimeric 45–60 kDa protein, functions as an anti-inflammatory cytokine and is produced by Th2 cells, mast cells, and eosinophils. IL-5 stimulates the B-cell growth and promotes the production of cytotoxic T-cells from thymocytes; however, the key function of IL-5 is in the activation, maturation, and survival of eosinophils. Eosinophils activated by IL-5 expel antibody bound parasites while releasing proteins associated with cytotoxic granules. In the case of CL and MCL, PBMCs induce secretion of both IL-4 and IL-5 at the site of the lesion (127, 128), which results in declined Th1 polarization. Several studies have reported that IL-4, IL-5, IL-10, and IL-13 provide favorable atmosphere for intracellular parasite growth and dissemination (129). Patients with chronic lesions produce abundant levels of IL-5 and IL-13, which further inhibits parasite killing by an additive effect of IL-13. IL-5 plays a minor role in the susceptibility to L. major infection in BALB/c mice (237).

#### IL-9 Increases Disease Susceptibility

IL-9 is a 14 kDa pleiotropic cytokine produced by Th-cells, primarily identified as a mouse T-cell growth factor (237) and mast cell differentiation factor. Erythroid precursors, Blymphocytes, eosinophils, bronchial epithelial cells and neuronal precursors are the secondary targets of IL-9 (238). It is a Th2 type cytokine produced via both IL-4 dependent (239) and IL-4 independent (240) pathways and involved in the physiological regulation of Th1-Th2 balance in vivo. Very little is known about the role of IL-9 in leishmaniasis. In L. major infection, IL-9 is transiently expressed in susceptible BALB/c as well as in resistant C57BL/6 and DBA mice during early days of infection, but 4-weeks onwards, its expression was only seen in susceptible mice but not in resistant mice (136). In vivo neutralization of IL-9 delays disease progression in BALB/c mice by inducing protective Th1 response suggesting, IL-9 promotes susceptibility to L. major infection. Further, IL-9 induces classical M8 activity and production of IFN-γ in L. major infected BALB/c mice, which serves as a model system to study the role of IL-9 in human diseases (137).

# IL-33 Is a Prognostic Cytokine for VL Pathogenesis

IL-33 is a member of IL-1 family, which includes IL-1 and IL-18. IL-33 is a crucial player in the defense against nematode infections and allergic reactions, since it causes Th2-type immune response via inducing the production of IL-5 and IL-13 by Tcells, mast cells, basophils, and eosinophils. In addition, IL-33 also induces non-Th2-type inflammation, suggesting its proinflammatory role like IL-1 and IL-18. Schmitz et al. first reported that IL-33 functions through ST2 (IL-1R4) orphan receptor present on different immune cell types (241). During L. major infection in BALB/c mice, ST2-expressing CD4<sup>+</sup> T-cells accumulated in local lesions (147). However, administration of polyclonal anti-ST2 antiserum depleted ST2-expressing cells as well as Th2 cells/cytokines and induced Th1 cytokine production, which in turn reduced the lesion development (148). Rostan et al. reported that the serum IL-33 levels were higher in VL patients besides the presence of IL33<sup>+</sup> cells in liver biopsy of a patient. Similar results were observed in BALB/c mice infected with L. donovani, additionally, ST2<sup>+</sup> cells were also observed in mouse liver. ST2 deficient BALB/c mice had shown strong Th1-type immune response via IFN-γ and IL-12 that controlled the hepatic parasite load and hepatomegaly. Recombinant IL-33 treatment of L. donovani infected BALB/c mice dramatically reduced the Th1 immunity and infiltration of polymorphonuclear neutrophils (PMNs) and monocytes in liver. In summary, IL-33 could be a very useful cytokine to determine the host susceptibility and disease prognosis of VL (149).

#### IFN-γ Is an Anti-leishmanial Cytokine

IFN-γ is a homodimeric glycoprotein consisting of two subunits each about 21 to 24 kDa. It is the most potent type II interferon that helps in M8 activation to the leishmanicidal state (63). The main cellular sources of IFN-γ production are activated CD4<sup>+</sup> and CD8<sup>+</sup> T-cells, and NK cells in response to IL-12 signaling (242). Of the several anti-leishmanial cytokines (23), IFN-γ is the most significant cytokine in host protection, which plays a prominent role in M8 priming (64) to produce leishmanicidal molecules (243). IFN-γ acts as monocyte-activating factor (65) and enhances release of oxygen radicals, secretion of proinflammatory cytokines (TNF-α, IL-l, and IL-6) (66), expression of MHC class-II, and antigen-presentation. In addition, IFNγ blocks the production of IL-10, which decreases all the above functions of monocytes (67). Several studies demonstrated that the leishmanicidal activity of M8s can be induced by a variety of cytokines, either alone or in combination. For instance, lipopolysaccharide (LPS) is required to induce the M8 leishmanicidal activity in vitro (244). In human VL, the response is predominantly Th2-type with absence of PBMCs derived IFNγ (47). But drug treatment induces a shift in the response so that individuals cured of VL often respond to Leishmania antigen by the production of both IFN-γ and IL-4 (158). In CL patients, however, the response is mainly dominated by IFN-γ and IL-4 is rarely detected (245), indicating that the immunological response to Leishmania in these individuals does not polarize as observed in inbred mouse strains. In vitro studies with T-cell clones (246) and in vivo studies using models of CL (40, 43) have demonstrated that IFN-γ can inhibit the expansion of CD4<sup>+</sup> Th2-cells, resulting in the preferential expression of Th1 cellmediated response. Reciprocal regulation is provided by the action of IL-10 on the IFN-γ producing capacity of Th1-type CD4<sup>+</sup> T-cells (16). A recent study showed that the variation in single nucleotide polymorphisms (SNPs) of IFN-γ gene at the position +874 (A/T) influences the susceptibility to VL such that individuals in southwest of Iran with an AT genotype are susceptible and those with a TT genotype are resistant to VL (247).

### IL-12 Is a Promising Candidate for VL Immunotherapy

IL-12 is a heterodimer consisting of two subunits (35 and 40 kDa) linked by a disulfide bond, mainly produced by activated M8s and DCs. It is a proinflammatory cytokine that plays a key role in bridging innate and adaptive immune responses (248). Protective immunity against leishmaniasis is typically associated with the production of IL-12 (16, 219). IL-12 drives Th1 response and induces IFN-γ production from both NK cells and T-cells (68), and mediates the leishmanicidal activity by inducing NOS2 expression and NO production (146). In addition, IL-12 also mediates T-cell proliferation and lymphokines production (71, 72). The presence of IL-12 reduces the ability of CD4<sup>+</sup> T-cells to produce IL-4 and increases the ability to produce IFN-γ. Thus, IL-12 and NK cells seem to play important role in determining the development of Th1 response (249). In vivo studies showed that IL-12 produced in infected mice (219, 250) is important to control Th2 expansion and to promote Th1 type response (69, 70). Neutralization of IL-12 leads to disease exacerbation in L. major and L. donovani infections (49, 69, 77). In contrast, the addition of IL-12 to lymphocyte cultures from VL patients restored IFN-γ production and increased cytotoxic activity of NK cells (48).

#### Endogenous TNF-α Offers Protection in VL

TNF-α is a 51 kDa homodimeric cytokine, mainly secreted by the activated M8s, T-cells, NK cells and mast cells. TNF-α is important in mediating both innate and adaptive inflammatory responses. The regulation of TNF-α production appears to be important because it has potential role in the formation and maintenance of granuloma (76). Antiparasitic activity of TNFα is mediated through activation of infected M8s for the destruction of intracellular amastigotes (73). TNF-α production is absent in susceptible mice but present in L. major infected resistant mice. Protective role of TNF-α in L. major infection is characterized by M8 activation, NO production and parasite clearance or suppression of visceralization (74). Protective Tcell response induced by TNF-α in L. major infected mice is due to the induced production of parasite-specific IgG1 and IgG2a. Acute infection with L. braziliensis resulted in the lack of production of parasite-specific IgG1 and IgG2a antibodies (251). The role of TNF-α in L. braziliensis infection is attributed to controlling the parasite numbers in the skin, lymph nodes and spleen and wound healing process (75). Brazilian patients with MCL had increased levels of TNF-α in both sera (252) and tissue lesions (253).

Treatment with TNF inhibitors, such as pentoxifylline in combination with anti-leishmanial pentavalent antimony, pentoxifylline promotes the re-epithelialization of mucosal tissues (254). However, infection of TNF−/ <sup>−</sup> mice with L. major shows some delay but no defect in antigen-dependent T-cell activation (74). IFN-γ independent anti-leishmanial mechanism mediated by endogenous TNF-α was described in IFN-γ knockout (GKO)-1 mouse infected with L. donovani (77, 199). The L. donovani infection provoked endogenous TNF-α level are enough to offer initial resistance to the parasite invasion and critical for the resolution of visceral infection. This is contrasting with the effect of exogenous TNF-α, which has no protective role in established infection and its continuous administration leads to impaired anti-leishmanial activity (255). The polymorphism and upregulation of TNF2 promoter transcription could be involved in enhancing clinical VL infection (256). TNF-α cannot be considered as a good marker of active disease in both human VL and canine VL due to its labile nature (224). The production of TNF-α follows biphasic kinetics due to its effect on target cells mediated by membrane-bound receptors (117). The high expression of IL-32 (especially γ-isoform) in CL and mucosal lesions is associated with endogenous TNF-α production but not with IL-10, suggesting the inflammatory role of IL-32 in host defense against Leishmania infection (257). Absence of IL-32 leads to high infection index but its overexpression opposed the parasite growth via NO cathelicidin and β-defensin 2 syntheses (258). In response to excessive inflammation, increase in the levels of TNF-α might promote the generation of IL-10 producing T-cells as a homeostatic response (78).

# IL-2 Promotes Anti-leishmanial T-Cell Response

IL-2 is a 15 kDa cytokine, produced by activated T-cells and was initially identified as a T-cell growth factor. IL-2 stimulates the proliferation and differentiation of B-cells, NK cells, monocyte/M8s, oligodendrocytes and lymphocyte activated killer (LAK) cells. IL-2 does not directly stimulate the intracellular antimicrobial activity of M8s (259) but exerts a range of immunoregulatory effects on T-cells and NK cells and induces the production of IFN-γ (79, 80). IL-2 may act as a susceptibility factor in leishmaniasis (250, 260) by inducing the production of IL-4 from CD4<sup>+</sup> T-cells (81). But in IL-4 deficiency, the inhibition of disease progression is attributed to IL-13 and IL-2. Interestingly, in CL, both IL-2 and IL-15 are attributed in host protection, while in MCL IL-2 is only protective but not IL-15 (200). Rapid production of IL-2 was observed after successful treatment or acquisition of resistance against L. donovani infection but not during acute phase (261, 262). The endogenous IL-2 could act as a defensive cytokine, only when the mice subsequently challenged after a prior infection with the parasite (82). In contrast, exogenous IL-2 exerts the antileishmanial action using L3T4<sup>+</sup> and Lyt-2<sup>+</sup> T-cells in acutely infected euthymic mice. IL-2 exerts leishmanicidal activity in splenocytes in vitro even in the absence of IFN-γ (83).

#### IL-15 Synergizes IL-2 and IL-12 in Host Defense and Has Scope in VL Therapy

IL-15 is a 14–15 kDa cytokine with four α-helix bundles and plays a central role in the innate and adaptive immune responses to infections (86). The main source of IL-15 is activated peripheral monocytes (263). Due to the common receptor βchain, immunological functions of IL-15 are similar to IL-2 (84), which includes the induction of T-cell proliferation (87), inhibition of T-cell apoptosis and preservation of memory Tcells (88), B-cell maturation and isotype switching (264). IL-15 also stimulates the proliferation and activation of NK cell (265) and induces the production of IFN-γ and TNF-α, synergistically with IL-12 (85). Nevertheless, IL-15 also shows distinct biological functions from IL-2 due to a different α-chain (266, 267). The possible pleiotropic role of IL-15 is reflected by its action on both Th1 and Th2 subtypes and the ability to induce the activity of IFN-γ and IL-12 (89) as well as IL-5 and IL-13 (90) in various experimental models. Endogenous IL-15 stimulates protective Th1 response by inducing the downregulation of IL-4<sup>+</sup> Th2 cells (86). IL-15 could be a potential therapeutic agent in acute VL since it upregulates IL-12 production and Th1 development (91). In contrast, other studies have demonstrated that endogenous IL-15 is not necessary for basal expression of IL-12 and M8 activation and is not able to influence the IL-12 activity and Th1 development in acute VL (268). IL-15 in combination with IFN-γ and/or IL-12 may increase the efficacy of conventional antimonial therapy for VL, because of low toxicity in vivo (201).

# IL-17 Role Is Contradictory in Leishmaniasis

IL-17 is a 35 kDa proinflammatory cytokine, primarily produced by activated T-cells (CD4<sup>+</sup> > CD8+) (269) and also by other subsets of T-cells including NKT cells and Th17 cells (96). The development of Th17 subset from naïve T-cells happens in the presence of IL-6 and TGF-β <sup>+</sup> Tregs (270). IL-17 stimulates different immune cells to produce inflammatory molecules including TNF-α, IL-1, and chemokines (163). At the site of inflammation, IL-17 affects the neutrophil function, reduces the apoptosis, and promotes the secretion of pro-inflammatory cytokines as well as tissue damaging molecules (164, 165). IL-17 induces the secretion of granulocyte macrophage-colony stimulating factor (GM-CSF) and G-CSF, which increase the production of neutrophils, monocytes, and chemoattractants for neutrophils (CXCL8, CXCL1, and CXCL6) as well as Th1 cells (CXCL10) (96, 97). IL-17 induces the production of IL-6, which mediates both proinflammatory and regulatory functions (96). IL-17 of Th17 subset and Th1 subset play a complementary role in the host protection from L. donovani infection. Contrastingly, the susceptibility for L. major infection is not only associated with uncontrolled Th2 immunity (271) but also with excessive IL-17 secretion, which mediates neutrophil recruitment (166). Other studies have also demonstrated that IL-17 dependent neutrophil recruitment is essential only during the late stages but not early stages of L. major infection (272). Mucosal disease caused by L. (V.) braziliensis is also associated with elevated levels of IL-17 response (167). Treating VL using curdlan, a bglucan immunomodulatory molecule induces Th1 cytokines with IL-12, IL-22 and IL-23 (273), while another immunomodulator Astrakurkurone is effective against experimental VL by inducing IL-17 along with IFN-γ (274). While these reports are suggestive of a protective role of IL-17 in VL, other reports suggested the involvement of IL-17 in exacerbating experimental VL in murine model (275) raising questions about its precise role in VL pathogenesis.

# IL-18 Protects From VL but Favors Other Forms of Leishmaniasis

IL-18 is a 22 kDa pleiotropic cytokine produced by activated M8s and Kupffer cells of liver (276). IL-18 promotes Th1 and NK cell development (168), induces IFN-γ production by lymphocytes and NK cells and synergizes with IL-12 (169, 170). IL-18R is expressed on Th1 cells but not on Th2 cells. IL-18 induced Th1 subset produces IFN-γ which indirectly regulates the expansion of Th2 cell (171). IL-18 promotes NK cell activity due to a constitutive expression of IL-18R on NK cells (277) and stimulates TNF-α secretion by human PBMCs (172). IL-18 also induces the activation of memory cells and in combination with IL-12 it prevents the reinfection of BALB/c mice with L. major (173). IL-18−/ <sup>−</sup> mice are highly susceptible to L. donovani infection when compared to the wild-type mice. However, endogenous IL-18 induces an initial IFN-γ independent antileishmanial effect in L. donovani infection (174). Paradoxically, IL-18 can also stimulate Th2 cytokines production, such as IL-4 from basophils (175) and CD4<sup>+</sup> T-cells (176) and IL-13 from mast cells. While IL-18 protects BALB/c mice from L. donovani infection, it increases the susceptibility of BALB/c mice to L. major infection. Notably, the resistance and susceptibility of BALB/c mice to L. mexicana infection are not mediated by IL-18 and are influenced by different genetic and immunoregulatory controls (278). IL-18−/ <sup>−</sup> BALB/c mice are highly resistant to L. mexicana infection due to increased IFN-γ production and antigen-specific IgG2a, reduced splenic IL-4, antigen-specific IgG1 and total IgE (177). IL-18 plays a critical role in the regulation of Th1 and Th2 balance in vivo, which frequently determines the outcome of many important infectious and autoimmune diseases (168).

#### IL-22 Offers Protection From VL

IL-22 is a 16.7 kDa cytokine, primarily produced by Th17 cells and to a lesser extent by Th1 and NK cells (96). Immunological functions of IL-22 are associated with the epithelium and mucosal surfaces (92), which include promoting inflammatory response and tissue repair (93, 279). IL-22 stimulates the production of pro-inflammatory molecules, such as S-100A proteins and CXCL5 (93). IL-17 and IL-22 act synergistically on epithelial cells to produce an antimicrobial peptide called β-defensin (95). IL-22 is also involved in protecting the liver (94) during chronic infections. During L. donovani infection, both IL-17 and IL-22 are produced by PBMCs and may exert complementary function along with Th1 cytokines (96, 97). The production of IL-22 requires IL-6 but not TGF-β (98).

#### Role of IL-1 Is Protective in VL but Contradictory in Other Forms of Leishmaniasis

IL-1 is synthesized as ∼35 kDa precursor, from which two functional agonistic proteins (IL-1-α and IL-1-β each 17 kDa M.W.) and IL-1Ra, receptor antagonist of IL-1R1, are produced. IL-1 is a potent proinflammatory "alarm cytokine" that synergizes the functions of TNF-α and is produced by M8s. IL-1 builds inflammation by inducing the expression of adhesion molecules on endothelial cells and leukocytes (280, 281). IL-1β, along with other proinflammatory cytokines, is released into the periphery during infection and coordinates immuneto-brain communication (180). IL-1 mediated inflammation is coordinated by adaptive T-cell response and controls the parasite dissemination (178, 179). IL-1 is responsible for regulating the delicate balance between inflammation and immunity which decides the fate of the disease progression in leishmaniasis. In L. major infection, the acute levels of IL-1α, IL-1β, and IL-1Ra are adequately downregulated unlike in L. amazonensis infection (282). Disease progression is inhibited with IL-1α treatment in L. major susceptible BALB/c mice during T-cell differentiation. IL-1β enhances activation of DCs and T-cell priming but do not affect the cytokine profile of DCs and pathogenic Th-cells (178). Contrastingly, Voronov et al. reported that BALB/c mice deficient in IL-1 family genes showed delayed disease progression with L. major infection due to apparent Th1 response even at late stages of the disease. IL-1α deficient mice were slightly more resistant to L. major infection than IL-1β KO mice (181). In L. amazonensis infection, IL-1β treatment induced DCs and cytokine production remains lower than that of L. major infection. IL-1 therapy in murine CL results in a wide range of outcomes depending on the course of treatment and parasite species involved. In this context, IL-1-based treatment may be effective for L. major but not L. amazonensis infection. The decreased production of IL-1 has been associated with L. donovani infection of murine peritoneal M8s in vitro and human circulatory monocyte population (182, 183). Similarly, human PBMCs failed to produce IL-1 in response to Leishmaniaantigen stimulation in vitro, during acute VL. However, following anti-leishmanial therapy, IL-1 and TNF-α levels are usually recovered, which correlates with clinical cure (184). Recombinant IL-1α induces mature granuloma formation in liver and IFNγ production from spleen cells but is not able to clear the parasite (185).

### IL-3 Is Likely a Host Protective Cytokine in VL

IL-3 is a 28 kDa glycoprotein derived from T-cells and supports the viability and differentiation of hematopoietic progenitor cells (283, 284) and monocytes (283). With the combination of GM-CSF, M-CSF, and IFN-γ, IL-3 shows an additive effect on human M8s in the induction of oxidative burst and TNF-α secretion to inhibit the replication and growth of Leishmania parasite. In contrast, IL-3 promotes the infection in murine model of CL highlighting the species-specific differences in the role of IL-3 in leishmaniasis (186). In combination with M-CSF, IL-3 induces superoxide ions production to kill the parasite and may involve in myelopoiesis during acute VL.

# IL-7 Shows Additive Effect With IFN-γ Against Leishmania

IL-7 is a 17 kDa glycoprotein derived from bone marrow stromal cells (285) and regulates a wide variety of functions including multiple effects on B-cells and proliferation of thymocytes (99– 101), NK cells (102) and mature T-cells (103). IL-7 induces the production of cytotoxic T-cells with alloreactive, antitumor, and antiviral activities (104). It was reported that IL-7 shows potential effects on monocytic lineages (105, 286). IL-7 enhances the synthesis and secretion of various inflammatory cytokines, such as IL-6, TNF-α, IL-1α, IL-1β, and M8 inflammatory protein (MIP) 113 by human circulatory monocytes. IL-7 is not as effective as IFN-γ but shows an additive effect with the combination of suboptimal concentrations of IFN-γ in killing the Leishmania amastigotes by inducing the production of TNF-α (105) and NO.

#### IL-8 Attracts Neutrophils to the Site of Infection

IL-8 is a non-glycosylated proinflammatory cytokine with a M.W. of 8 kDa, which is primarily produced from neutrophils and from other cell types including epithelial cells, keratinocytes, fibroblasts and endothelial cells. In mouse system, IL-8 has two functional homologs like MIP-2 (CXCL2/Groβ) and KC (CXCL1/Groα). During L. major infection in humans, IL-8 promotes the recruitment of neutrophils at lesion sites (106). In mice infected with L. major, IL-8 causes transient production of KC mRNA in the skin, which may associate with granulocyte recruitment (107) which is yet to be demonstrated in vivo (272). Notably, a reduced neutrophil count during active VL is associated with lower IL-8 levels in serum (32). It was identified that polymorphisms at IL-8 −251 position are associated with impaired IL-8 activity and the development of active VL in Iranian individuals (108) but such a correlation was not observed in Brazilian VL patients (287).

#### IL-23 May Offer Protection From VL in Association With IL-12p40

IL-23 is a pleiotropic cytokine produced by M8s and DCs which acts on receptors expressed by T-cells, NK cells and NKT cells (288). During L. donovani infection in BALB/c mice, the IL-23p19 mRNA expression in the liver tissue was comparable to that of wild-type. IL-12 independent protection in visceral infection (76, 109, 110) was mediated by IL-18 and probably by IL-23 also, since the p40 subunit of IL-12 shared by both IL-12p70 and IL-23 (174). IL-23p19 may pair with IL-12p40 to become biologically active (111), which are crucial for host protection. Therefore, IL-23 alone or in combination with other cytokines may be a possible option in immunotherapy of VL.

# INTERPLAY OF T-CELL SUBSETS VIA CYTOKINES IN LEISHMANIASIS

It is well established that the complex interplay of pathogens with their hosts is predominantly regulated by host-specific Th1/Th2 subset cytokines in the vicinity of several regulatory cytokines. In this context, previous studies have demonstrated that IL-10 produced by CD4+CD25<sup>+</sup> Tregs (211) is important for parasite persistence in mice (289). In human VL, the elevated level of IFN-γ mRNA in lymphoid organs (spleen and bone marrow) is accompanied by an abundant expression of IL-10 (55, 290, 291) where the predominant source of IL-10 is Foxp3−CD25−CD3<sup>+</sup> cells (143). However, in leishmaniasis, healing is predominantly associated with diminished expression of IL-10 mRNA (55, 290). The role of Th17 subset in human VL is unveiled by a longitudinal study in Sudan, which illustrated the protective role of Th17 subset that are employed by an induced production of IL-17 and IL-22 from L. donovani-specific T-cells (292). In fact, Th17 cells are pleiotropic in nature, responsible for either protection or pathogenesis and frequently associated with recruitment of neutrophils. Th17 cells under the influence of IL-27 producing CD4<sup>+</sup> T-cells diminish IL-17 and IL-22 secretion (293). Disease progression in pre or post-treated Indian VL patients is linked with serum IL-27 and splenic IL-27 transcript but not with splenic IL-17 transcript (140).

The pathogenic role of IL-27 in active VL is linked with suppression of Th17 cytokines production and expression of transcription factors. Consequently, IL-27 promotes the parasite dissemination by inducing antigen-specific IL-10<sup>+</sup> Tcell differentiation and expansion, and by inhibiting activation of effector Th17 lineage. Moreover, the host negotiates the Th17 response to control the pathogenic implications of VL that are driven by the parasite. Th9 cells are not the unique source of IL-9 production as Tregs and Th17 cells also produce IL-9 in lesser quantity. Initially, Th9 subset was thought of a splinter group of Th2 but now it is an independent lineage. Predictably, Th9 subset has a similar detrimental role like Th2 in the development of CL in the mouse model. Since, IL-4, IL-21, TGF-β, and IFN-α/β seem to activate Th9 cells to produce IL-4, IL-9, and IL10, which are involved in the pathogenesis of CL (136, 137). The "B-helper" follicular T-cell (Tfh) lineage is also implicated in leishmaniasis progression, which is the source for bulk production of IL-4 in the draining lymph nodes of susceptible mice infected by L. major (294). As a sequel of VL, PKDL patients' carries high plasma IL-10 levels (295). Immunologically, PKDL is characterized by hyper T-cell response and significant production of both Th1 and Th2 cytokines from PBMCs in response to crude L. donovani antigen (296). IL-10 levels in the skin and plasma could be used to predict the severity of PKDL pathogenesis and the chance of VL succession to PKDL.

# CYTOKINES IN VL DIAGNOSIS AND IMMUNOTHERAPY

As immunotherapy is mandatory for refractive cases of leishmaniasis, cytokines received great attention in the search for novel approaches for the diagnosis and immunotherapy of VL (summarized in **Table 4**). For the first time, Reed et al. used the lymphokines obtained from murine spleen cell culture supernatant and encapsulated in liposomes against VL challenge and reported a significant reduction in the liver parasite burden, highlighting the importance of lymphokines in leishmaniasis healing (297). In general, the absence of leishmanial-antigen stimulated lymphocyte proliferation and IFN-γ production are indicators for the clinical evaluation of a VL patient (45). However, these two parameters may also be used as coordinates in assessing the level of protection conferred by vaccine antigens (194). Later, several studies have tested the effect of direct administration of recombinant Th1 cytokines and monoclonal antibodies against Th2 cytokines alone or in combination against leishmaniasis. For example, prophylactic anti-IL-10R treatment induces the rapid control of experimental VL and antimonials activity in IL-10 knock-out or transgenic mice (51), and IFN-γ production from T-cells with an active VL (55). Further, it was reported that IL-10R inhibition in L. donovaniinfected mice controlled the parasite burden in liver, increased IFN-γ titers in serum, and iNOS production in macrophages altogether reduced the VL fatality (298). The therapeutic efficacy of anti-IL-10R and anti-GITR (glucocorticoid-induced TNF receptor-related protein) was tested against L. donovani challenge in C57BL/6 mice. Blocking IL-10 alone could reduce the parasite burden in spleen and liver but combination of these antibodies did not inhibit the parasite proliferation in spite of the significant increase in IFN-γ and TNF-α production (299). In another study, the blockade of IL-2 and IL-10 was effective in the reduction of parasite load in early and later phases of L. donovani infection in BALB/c mice (300). IL-10 neutralization in splenic aspirate cellsincreases IFN-γ and TNF-α production and reduces parasite burden in VL patients (187). In L. chagasi infected Brazilian population, the antigen-stimulated PBMCs derived IL-10 titers were higher in acute VL than after cure. However, Leishman skin test (LST) or Montenegro test positivity is not directly correlated with the IL-10 production in asymptomatic individuals (113). The balance between IL-10 and IL-12 determines the effectiveness of chemotherapy (17).

IL-4 induced in VL is usually associated with impaired treatment (191, 192). IFN-γ, IL-4, and IL-13 are upregulated in active VL, however, their levels are significantly declined after cure (193). The disease relapse in human VL patients is associated mostly with IL-10 rather than IL-13 and is influenced by IL-10<sup>+</sup> IFN-γ <sup>+</sup> antigen-specific T-cells (189). Blocking IL-4, IL-13, and TGF-β with receptor fusion antagonists substantially controlled the parasite replication but the clearance of visceral infection is marginal and had no synergistic effect with Sb<sup>V</sup> (119). Though IFN-γ, IL-6, IL-27, TNF-α, and IL-10 levels increased in Brazilian patients with active VL caused by L. infantum, the clinical manifestation are strongly correlated with IL-6, IL-27, TNF-α, and IL-10 (190). In general, TNF-α synergizes IFN-γ in the activation of M8s and clearance of parasite but it is found to elevate in serum despite the low TNF-α <sup>+</sup> monocytes in the circulation of active VL patients (301). To surpass the side-effects, rIFN-γ and muramyl tripeptide (MTP-PE) encapsulated in liposomes at varying doses of intravenous (i.v.) injections causes substantial reduction in the splenic parasite burden in murine VL (302). rIFN-γ was tested in combination with Sb<sup>V</sup> against VL patients from Brazil, Kenya, and India and the therapeutic efficacy was found to be 82.3, 75, and 87%, respectively (195–197).

IL-12 orchestrates acquired resistance in liver during intracellular L. donovani infection and parasite killing (174). IL-12 restored responses from PBMCs of VL patients much better than the treatment with anti-IL-10 alone or in combination with anti-IL-4 [53]. Hence, it is clear that successful VL therapy is associated with restoration of IFN-γ and IL-12 production (47). IL-12 was used as an effective adjuvant for a killed vaccine against L. major (198). The treatment of susceptible BALB/c mice with recombinant IL-12 mediates disease healing, which is associated with induced production of Th1 cytokines (81, 200) and suppression of IL-4 (75). The treatment for arthritis with anti-TNF-α results in increased susceptibility to VL (202, 203), suggesting that TNF-α could act as a primary effector component. Upon liposomal amphotericin B treatment, the plasma IL-15 levels were found to be increased in VL patients (86), suggesting the role of IL-15 as a marker in VL diagnosis and a target in the VL therapy. In opportunistic HIV co-infection, high levels of serum TNF-α and IFN-γ are the predictors for onset of acute VL infection (204). Recently, we demonstrated cytokine role and therapeutic potential of recombinant leptin (adipokine) in BALB/c mice with experimental VL caused by L. donovani. The serum leptin levels and splenic Th1 cytokine response were found to be reduced in active disease. Upon leptin administration, host protective responses including Graz-A<sup>+</sup> CD8<sup>+</sup> T-cells, IFN-γ, IL-12, and IL-2 production were found to be restored (303). Hence, low systemic leptin levels could be of prognostic and diagnostic value in the assessment of clinical VL.

#### LIMITATIONS AND FUTURE PROSPECTIVE

Although cytokine-based immunotherapy is a promising approach for VL cure, there are certain limitations associated with this strategy. The production of recombinant cytokines as large molecules used in therapeutics is very expensive and they must be administrated via injections, which is certainly painful to the patients. Administering high dose of cytokines could result in side effects characterized by malaise and influenza-like syndromes. As the cytokines have short half-life in plasma, multiple doses are need which further increases the sideeffects (304). A crucial aspect is that the cytokine therapy for leishmaniasis must be cost-effective over conventional treatment in order to be practical. It is possible that a combination therapy comprising a potent anti-leishmanial cytokine with the combination of an inhibitor (monoclonal antibody) targeting disease promoting cytokine or with current drug options could be a future prospective of leishmaniasis treatment. However, there is possibility that the different combinations of cytokines may produce a divergent immune response. Hence, it is important to further investigate the effect on immune response to develop a clinically relevant combination therapy. This is particularly important since several cytokines share common signaling cascades as outlined in this review, which affects the outcome of treatment. The gene manipulation strategy using advanced molecular biology tools may produce desired version of cytokines with small ligand-tags that have the potential to increase the halflife of cytokines from minutes to days in the blood by tethering with albumin protein, which further could reduce the number of required doses. After finding a successful combination of these against leishmaniasis, it would be optimal to design the chimeras of cytokines without losing native structural and functional properties. After administration, the chimeric cytokines should splice inside the body fluids and act independently. As a second option, cytokine and drug combination was also shown to be a reliable strategy against leishmaniasis. For example, combination of IFN-γ with antimony against experimental VL showed that antimony dosage required for leishmanicidal activity was reduced by 4- to 10-folds with IFN-γ combination (305). This is a string indication that administering a drugcytokine mix could address the drug toxicity and possible development of resistance. In another study, the pre-treatment for 20 days with IFN-γ before antimony therapy has cured the VL in 4 out of 9 Indian patients and rest of them had shown reduced parasitemia in spleen aspirates (197). As mentioned earlier, CTLA-4 and PD-1 causes T-cell unresponsiveness, so targeting these for leishmaniasis treatment may yield promising results. A study showed that the anti-CD40 and anti-CTLA-4 with the combination of Sb<sup>V</sup> against L. donovani infection in a mouse model increased IL-12 and IFN-γ production, T-cell activation and function, and synergistic with Sb<sup>V</sup> while increasing the parasite death (306). Similarly, administration of chimeric fusion protein OX-40L-Fc and anti-CTLA-4 improved granuloma maturation and CD4<sup>+</sup> T-cell proliferation to

augment the killing L. donovani parasite but had no effect on IL-10 and TGF-β production (307). Liposomal amphotercin B treatment with the combination of recombinant human granulocyte macrophage colony-stimulating factor (rHuGM-CSF) cured the VL clinical symptoms and splenomegaly in a patient suffering from HIV and VL (308). Another promising therapeutic option is administration of anti-leishmanial drugs and immunomodulators together. For example, the suboptimal doses of miltefosine with the combination of a single dose of TLR-ligand called Pam3Cys (tripalmytoil-Cysteine), an immunomodulator, significantly promoted the healing of L. donovani infection in mice by increasing the production of Th1/Th2 cytokines, reactive oxygen and nitrogen intermediates, and H2O<sup>2</sup> (309). Cytokine producing immune cell-based therapy either alone or in combination with drugs has recently emerged as a potential treatment for cancer and other infectious diseases. Glycosphingophospholipid (GSPL), a β-(1–4)-galactose terminal NKT-cell ligand of L. donovani antigen induces inflammatory signaling cascade to kill the intracellular parasite, induces effector T-cell response and controls the acute parasite load to an undetectable level in experimental VL (310). DCs could also be an attractive option as they are important antigen-presenting cells at the interface of innate and acquired immunity and can suppress early dissemination of the parasite to the lymphoid tissues mediated by IL-10 (311). Combination of bone marrowderived DCs pulsed with L. donovani antigen and antimony treatment completely cleared the infection from the spleen and liver (312) by inducing Th1 cytokines production (313). Cytokines are the key players in the determination of disease outcome during various immunotherapies. It is important to remember that measuring the levels of a pro- or antiinflammatory cytokine alone to predict the disease severity may not be reliable. Measuring the ratio of cytokines is a promising approach. For example, IFN-γ/IL-10 ratio is predictive of disease severity in VL (314).

# REFERENCES


# CONCLUSIVE REMARKS

As the cytokines are the key focus of various immunotherapies against leishmaniasis, it is essential to understand their role in detail with possible scope in developing novel diagnostics and targeted therapy for VL. There are key set of cytokines that are involved in the disease progression namely IL-10, TGFβ, and IL-4 and host protection namely IFN-γ, IL-12, TNFα, and IL-2 during VL. Notably, there are other cytokines that are also involved in the pathogenesis and host defense during VL. However, their role appears to be complex and is dependent on the Leishmania species and the type of clinical disease. For example, cytokines namely, IL-1, IL-13, IL-17, and IL-18 are involved in the host defense during VL but have an opposite effect by promoting the disease in CL. Nonetheless, cytokines involved in the host protection e.g., IL-15, IL-22, and IL-23 and pathogenesis e.g., IL-33, IL-27, IL-9, and IL-21 can be explored further as promising targets in diagnosis and immunotherapy of VL.

# AUTHOR CONTRIBUTIONS

Topic selection and content development was done by AD and SKK. First draft was prepared by AD and SKK. SVK have corrected and revised to final version. SC has been supportive in images and tables.

#### ACKNOWLEDGMENTS

Authors would like to thank Prof. Vadlakonda Laxmipathi Rtd. Professor Kakatiya University, India for his valuable guidance and critical suggestions during review writing. Authors are grateful to Dr. S. S. Mohanraj, Scientist B, CTSSS central silk board, Kota, Chhattisgarh, India for his kind help in polishing the Figures.


of infection with Leishmania chagasi. J Infect Dis. (1993) 167:411–7. doi: 10.1093/infdis/167.2.411


leishmaniasis due to Leishmania major. J Infect Dis. (1998) 177:1687–95. doi: 10.1086/515297


mononuclear cells from patients with active visceral leishmaniasis in Sicily. J Immunol. (1988) 140:2721–6.


of phase-specific immunotherapy. J Immunol. (2006) 177:4636–43. doi: 10.4049/jimmunol.177.7.4636


**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 Dayakar, Chandrasekaran, Kuchipudi and Kalangi. 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.

,

# Tumor-Associated Disialylated Glycosphingolipid Antigen-Revealing Antibodies Found in Melanoma Patients' Immunoglobulin Repertoire Suggest a Two-Direction Regulation Mechanism Between Immune B Cells and the Tumor

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Else Marit Inderberg, Oslo University Hospital, Norway Meenu Sharma, University of Texas MD Anderson Cancer Center, United States*

#### \*Correspondence:

*Beatrix Kotlan kotlanbea@gmail.com*

#### Specialty section:

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

> Received: *09 January 2019* Accepted: *11 March 2019* Published: *05 April 2019*

#### Citation:

*Kotlan B, Horvath S, Eles K, Plotar VK, Naszados G, Czirbesz K, Blank M, Farkas E, Toth L, Tovari J, Szekacs A, Shoenfeld Y, Godeny M, Kasler M and Liszkay G (2019) Tumor-Associated Disialylated Glycosphingolipid Antigen-Revealing Antibodies Found in Melanoma Patients' Immunoglobulin Repertoire Suggest a Two-Direction Regulation Mechanism Between Immune B Cells and the Tumor. Front. Immunol. 10:650. doi: 10.3389/fimmu.2019.00650* Beatrix Kotlan<sup>1</sup> \*, Szabolcs Horvath<sup>2</sup> , Klara Eles <sup>2</sup> , Vanda K. Plotar <sup>2</sup> , Gyorgy Naszados <sup>3</sup> Katalin Czirbesz <sup>4</sup> , Miri Blank <sup>5</sup> , Emil Farkas <sup>6</sup> , Laszlo Toth<sup>6</sup> , Jozsef Tovari <sup>7</sup> , Andras Szekacs <sup>8</sup> , Yehuda Shoenfeld<sup>5</sup> , Maria Godeny <sup>3</sup> , Miklos Kasler 9,10 and Gabriella Liszkay <sup>4</sup>

*<sup>1</sup> Molecular Immunology and Toxicology, National Institute of Oncology, Budapest, Hungary, <sup>2</sup> Center of Surgical and Molecular Pathology, National Institute of Oncology, Budapest, Hungary, <sup>3</sup> Center of Image Analysis and Radiological Diagnostics, National Institute of Oncology, Budapest, Hungary, <sup>4</sup> Department of Oncodermatology, National Institute of Oncology, Budapest, Hungary, <sup>5</sup> Zabludowitz Center for Autoimmune Diseases, Sheba Medical Center Affiliated to Sackler Faculty of Medicine, Tel-Aviv University, Tel Aviv, Israel, <sup>6</sup> Center of Oncosurgery, National Institute of Oncology, Budapest, Hungary, <sup>7</sup> Department of Experimental Pharmacology, National Institute of Oncology, Budapest, Hungary, <sup>8</sup> Agro-Environmental Research Institute, National Agricultural Research and Innovation Centre, Budapest, Hungary, <sup>9</sup> National Institute of Oncology, Budapest, Hungary, <sup>10</sup> Ministry of Human Capacities, Budapest, Hungary*

There is far less information available about the tumor infiltrating B (TIL-B) cells, than about the tumor infiltrating T cells. We focused on discovering the features and potential role of B lymphocytes in solid tumors. Our project aimed to develop innovative strategies to define cancer membrane structures. We chose two solid tumor types, with variable to considerable B cell infiltration. The strategy we set up with invasive breast carcinoma, showing medullary features, has been introduced and standardized in metastatic melanoma. After detecting B lymphocytes by immunohistochemistry, VH-JH, Vκ-Jκ immunoglobulin rearranged V region genes were amplified by RT-PCR, from TIL-B cDNA. Immunoglobulin variable-region genes of interest were cloned, sequenced, and subjected to a comparative DNA analysis. Single-chain variable (scFv) antibody construction was performed in selected cases to generate a scFv library and to test tumor binding capacity. DNA sequence analysis revealed an overrepresented VH3-1 cluster, represented both in the breast cancer and the melanoma TIL-B immunoglobulin repertoire. We observed that our previously defined anti GD3 ganglioside-binder antibody-variable region genes were present in melanoma as well. Our antibody fragments showed binding potential to disialylated glycosphingolipids (GD3 ganglioside) and their O acetylated forms on melanoma cancer cells. We conclude that our results have a considerable tumor immunological impact, as they reveal the power of TIL-B cells to recognize strong tumor-associated glycosphingolipid structures on melanomas

**204**

and other solid tumors. As tumor-derived gangliosides affect immune cell functions and reduce the B lymphocytes' antibody production, we suspect an important B lymphocyte and cancer cell crosstalk mechanism. We not only described the isolation and specificity testing of the tumor infiltrating B cells, but also showed the TIL-B cells' highly tumor-associated GD3 ganglioside-revealing potential in melanomas. The present data help to identify new cancer-associated biomarkers that may serve for novel cancer diagnostics. The two-direction regulation mechanism between immune B cells and the tumor could eventually be developed into an innovative cancer treatment strategy.

Keywords: B lymphocytes, glycosphingolipids (GSLs), immune regulation, immunoglobulin, metastatic melanomas, natural antibody, tumor-associated antigen

#### INTRODUCTION

Tumor immunological investigations have reached an immunotherapeutic breakthrough in cancer research and therapeutics (1, 2). However, new strategies are still needed to solve the cure of various cancer types. Earlier, cancer research was more focused on T cell-related immunological mechanisms (3). Later, it became evident that immune B cells are also essential components in the "anti-tumor battle" of the immune system. However, there is far less information available about B cells accumulated in the tumor [tumor infiltrating B (TIL-B) cells] (4, 5), than about T cells (tumor infiltrating T cells) (6, 7). This is probably because the very small amounts of TIL-B cells are hard to discover. More sophisticated detection and characterization methods are required to reveal the minute amounts of immune B cells in the cancerous tissue. New strategies, involving molecular immunological and biotechnological techniques, led to pioneering results on TIL-B cells in invasive breast carcinoma with medullary features (8, 9). Further studies in melanoma and some other solid tumors enlarged our knowledge on TIL-B cells (10–12). There are still unanswered questions: Why are TIL-B cells accumulated in the tumor? What are the main roles of TIL-B cells? What are the features of the TIL-B cell antibody repertoire? Do TIL-B cells have an essential impact on cancer cell recognition and elimination? The lack of interest and limited methodology have hindered the discovery within this important field until recently.

Tumor infiltrating B lymphocytes have been recognized as a new hallmark of breast cancer (13). Answering questions about TIL-B cells' functions remains a major challenge. To help address this issue, our previous study on breast carcinomas has been extended to malignant melanomas. We are convinced that systematic studies, various immunological and molecularbiological investigations and DNA sequence analysis have the power to reveal characteristics of cells hidden in minor quantities in the tumor microenvironment. Through the comparison of different cancers, we hope to gain insight into the way TIL-B cells perform specific tasks in the tumor microenvironment and to establish their relationship to cancer cells. We defined earlier essential TIL-B cell-targeted components in breast carcinomas, that is, the tumor-associated disialylated glycosphingolipids (GD3 gangliosides) (9). The difference in carbohydrate profiles between normal and cancerous tissues is a major issue. The characteristic carbohydrate expression associated with malignant transformation is caused by "aberrant glycosylation." Complex carbohydrate cell membrane components are the glycoproteins, proteoglycans, and glycosphingolipids (14, 15).

Due to the importance of disialylated GSLs in cancer progression, it seemed obvious to detect these cancerassociated membrane structures on melanoma cells and tissue sections. Parallelly, we examined in detail the B cell infiltration and antibody profile in melanomas. We hypothesized a causal link between the TIL-B produced antibody repertoire and the extensive expression of highly tumor-associated glycosphingolipid membrane structures on the cell surface of highly malignant solid tumors. Our project with innovative technological strategies not only opens up for a deeper understanding of these key elements in cancer progression and metastases, but will help to build various types of new diagnostic possibilities linked to these glycolipid-based cancer membrane structures.

#### MATERIALS AND METHODS

We set up a complex strategy with immunologic, molecular genetics, and biotechnological methods and processed cancerous tissue specimen in order to approach the questions on tumor infiltrating B (TIL-B) cells.

#### Cancerous Tissue

Cancerous tissue punch biopsy samples and peripheral blood of patients with melanoma were tested with ethical permissions (ETT TUKEB 16462-02/2010, 336/2014.9710-1/2015/EKU) received for our project. Previously stored minor tissue samples,

**Abbreviations:** AP, alkaline phosphatase; ATCC, american type cell collections; AEC, 3-amino-9-ethylcarbazole (substrate); BSA, bovine serum albumin; CSC, cancer stem cells; GD3 ganglioside type, disialylated glycosphingolipids; ELISA, enzyme labeled immunosorbent assay; FBS, fetal bovine sera; FITC, fluorescein isothiocyanate; FDA, food and drug administration; GSL, glycosphingolipid; H&E, haematoxilin & eosin; HJLCT, Harry J Loyd Charitable Trust; HSA, human serum albumin; HRP, hydrogen peroxide; IF, immunofluorescence; Ig, immunoglobulin; IHC, immunohistochemistry; Ministry of Human Capacities in Hungary, hungarian medical research council; NM, nodular melanoma; O'AcGD3, O' acetylated GD3; PFA, paraformaldehyde; PBMC, peripheral blood mononuclear cells; PBS, phosphate buffered saline; PNPP, p-Nitrophenyl Phosphate; PCR, polymerase chain reaction; scFv antibody, single-chain-variable antibody; SSM, superficial spreading melanoma; TIL-B, tumor infiltrating B cells.

from surgically removed melanomas and breast carcinomas, were investigated as well. Fresh melanoma samples were used to set up primary cultures and were fresh frozen for immunohistochemistry and molecular processes. Formalinfixed paraffin-embedded melanoma tissue sections were made additionally.

#### Cancer Cell Lines

Primary tumor cell cultures were developed from fresh melanoma tissue samples. The melanoma cell line (SK-Mel28, M24) was purchased from American Type Cell Collections (ATCC), A-2058 melanoma cell line was an earlier generous gift to the laboratory from professor Dr. Meenhard Herlyn and professor Dr. Lance A Liotta. Primary breast cancer cultures that were developed in the course of previous studies and MDA MB-231, ZR751 cell lines purchased from ATCC were investigated. Cells were maintained in steadystate culture conditions in RPMI-1640 culture media (Sigma, St. Louis, MO, USA) supplemented with 5% fetal bovine sera (FBS), penicillin/streptomycin) (1:100) (Sigma) and grown until confluent. To set up primary cell cultures, 0.8µg/ml gentamicin and amphotericin B (70µg/ml) (Sigma) were added to the media. Cells were grown in 25 or 75 cm<sup>3</sup> tissue culture flasks (Greiner) under culture conditions (37◦ , humidified thermostat with 5% CO2). Growth rates and viability were followed by inverted and normal light microscopy (Olympos and Nikon, both from Shinjuku, Tokyo, Japan).

#### Immunohistochemistry

Formalin-fixed and paraffin-embedded tissue sections were deparaffinized in Xilol and Ethanol. After this, hydrogen peroxidase blocking antigen retrieval was performed by heatexposure in a Microwave (Meditest MFX800-3). Slides were blocked with 3% bovine serum albumin (BSA) in PBS and then reacted (4◦C, overnight or 37◦C, 30 min) with the monoclonal antibodies of interest. Tumor infiltrating B cells were detected by B cell-specific monoclonal antibodies (CD20) (DAKO). Novel disialylated glycosphingolipid-specific antibodies (GD3) (Calbiochem and Axxora/Alexis, Abcam, London, UK,), HCBC3 anti GD3 antibody (provided by Dr. Mark C Glassy) and our selected disialylated GSL-specific antibody fragments were tested. IHC was performed with Supersensitive TM One Step Polymer IHC Detection Kit System (BioGenex), using mainly ImPact TM AEC substrate (Vector).

#### PCR Amplification of TIL-B Ig V Regions

Minor frozen tissue samples were homogenized under liquid nitrogen and stored in TRIZOL at −70◦C until RNA was

FIGURE 1 | Detection of tumor infiltrating B lymphocytes in malignant melanoma and invasive ductal breast carcinoma. Immunohistology with haematoxilin & eosin staining showed the same abundant immune cell infiltration in nodular (A) and superficial spreading melanoma (B) as in invasive breast carcinoma with medullary features (C) black arrows. B lymphocytes could be defined with CD20 monoclonal antibody, One Step Polymer HRP detection system and AEC substrate in melanoma (D,E) and breast carcinoma by immunohistochemistry (100x). Black arrows are pointing to immune cell infiltration (immunohistology) or CD20 positive B cells labeled in red (immunohistochemistry) (D,E). Investigations in breast carcinoma postulate that CD5 positive B cells are also present (F).

extracted according to the manufacturer's instructions (RNeasy Mini kit; Qiagen, Hilden, Germany and NucleoSpin RNA XS, Macherey-Nagel, Düren, Germany). cDNA was synthetized (Pharmacia Biotech kit) to amplify the human Ig VH-JH, Vκ-Jκ, and Vλ-Jλ encoding regions with specific primers designed previously (16). Polymerase chain reaction (PCR) was performed (35 cycles: 1 min, 94◦C; 1 min, 60◦C; and 1 min, 72◦C) in a PerkinElmer/Cetus thermocycler.

#### Cloning and Sequencing

Immunglobulin heavy and light chain variable region (VH-JH, Vκ-Jκ, and Vλ-Jλ) gene PCR products were purified, and then blunt end ligated into pUC18 (SmaI/BAP) plasmid vector (Pharmacia Biotech) before transformation into Escherichia coli TG1 bacteria. Gene-insert positive clones were selected by PCR screen technique (17). Sequencing of the plasmid dsDNA minipreps (QIAprep Spin kit; Qiagen) was performed by automatic sequencing (Dye Terminator Sequence Reaction Kit, DyeEx Spin kit (Qiagen; ABI PRISM Software, automatic sequencer of Perkin Elmer, and partly with commercially available sequencing service (Invitrogen, San Diego, CA).

#### Comparative DNA Sequence Analysis

Comparative DNA sequence analysis was performed first using BIOEDIT (18), Clustal X 1.8 (19), and TREEVIEW 1.5.2 (20). In later work phases, we had the Vector NTI 11 available to make all the sequence homology analyses. For comparative DNA sequence analysis, we used KABAT National Institute of Health (http://immuno.bme.nwu.edu), New Kabat Database Server: george at immuno.bme.nwu.edu, and IMGT, the international ImMunoGeneTics database <sup>R</sup> , (www.imgt.org), (http://imgt.cines.fr) and (http://imgt.cnusc.fr:8104). Databank search via National Center for Biotechnology Information Blast server to GenBank/European Molecular Biology Laboratory Net databases was conducted to find homologous sequences and the generated data was termed as Blastn result.

# Construction of scFvK for Phage Library Generation

Assembly reactions of rearranged Ig V region H and L chain genes were conducted by a three-step PCR amplification, using a linker peptide (Gly4Ser3) coding sequence. Purified and suitable restriction enzyme digested VH-JK fragments were ligated into a phagemid vector (21), according to the methods we described earlier (17). However, slight modifications in terms of the library generation and panning process against membrane preparations were made. According to our previous antibody repertoire analysis in breast carcinomas, Vκ light chains were represented with a broader variability than Vλ light chains. Therefore, as a first choice, we were more interested in the Vκ representatives in melanoma.

FIGURE 2 | Strategy to harness the tumor infiltrating B cells' tumor antigen revealing capacity for diagnostics and therapeutics. This schematic flow chart clearly shows, how cancerous tissues containing immune B cells can be processed and investigated in the course of various immunological and molecular biological techniques. As a result of our strategy antibody fragments with tumor antigen specificity could be obtained, representing binding to cancer cells. Small picture (in lower-right corner) shows melanoma cells reacted with anti GD3 ganglioside antibody fragment in immunohistochemistry, with FITC labeled antibody and DAPI nucleus staining.

# Soluble scFv Enzyme Labeled Immunosorbent Assay (ELISA)

Ninety-six-well Nunc MaxiSorp <sup>R</sup> flat-bottom plates were precoated (16 h, 4◦C) with 1–10 µg of native tumor cell membrane preparations. Plates were washed repeatedly and blocked with 200 µl of 2% BSA in PBS. Soluble fractions of the test antibody fragments and control antibodies were incubated in triplicates for 16 h at 4◦C. Detecting second antibody alkaline phosphate conjugated anti-c-myc (Sigma-Aldrich) and p-Nitrophenyl Phosphate (Sigma-Aldrich) substrate was used according to standard conditions. HCBC3 (anti GD3) and HCBD1 (anti GD2) monoclonal antibodies were Prof Dr. Mark C Glassy's generous gifts for testing.

#### Immunofluorescence—Flow Cytometry, FACS Analysis

Melanoma cells were cultured until reaching confluence, harvested by EDTA with 0.02% PBS and incubated at 37◦C for 30 min (or at 4◦C overnight) with anti-ganglioside monoclonal antibodies (CVL-MAB0014-1) (Axxora, Farmingdale, NY, USA), MA1-25302 (Pierce, Thermoscientific, Rockford, IL, USA), AB13779 (Abcam, London, UK). Cancerous cell suspensions were reacted with unique GD3 ganglioside-specific antibodies (Abcam, London, UK), Calbiochem), HCBC3 anti GD3 antibody or soluble fractions of our expressed disialylated GSL binder antibody fragments. First and second antibody reactions were followed by wash steps with 1% BSA PBS and PBS. Antimouse (Fab')2 phycoerythrin (DAKO) or anti-mouse (Fab')2 FITC (Sigma) was used as second label antibody. Melanoma patient-derived primary cell suspensions of melanoma cells (SK-Mel 28, A-2058) were investigated by flow cytometry (CyFlow SL-Green, FloMax, Partec, Munster, Germany) and in some cases by FACSAvia Sorter/Beckton Dickinson. Forward and side scatter dot plots and immunohistological curves were evaluated for antigen expression intensity and the percentage of positive cells with data analyzing software Flo Max (Partec).

# Immunofluorescence—Confocal Laser Microscopy

Minor melanoma tissue samples were snap-frozen with isopentane in liquid nitrogen. Fresh frozen cancerous tissue cryostat sections [6–8µm, freezing media (Bio-Optika, Milano, Slee Cryostat mnt (Auroscience)] of melanomas were fixed in 4% paraformaldehyde (PFA) PBS for 15 min. Three percent BSA PBS was used for blocking, before the monoclonal antibodies specific to tumor-associated disialylated glycosphingolipid antigens (Calbiochem, Abcam) were added for an overnight incubation at 4◦C. Indirect immunofluorescence with FITC-labeled second antibodies (anti mouse IgG FITC from DAKO), or biotinylated rabbit anti mouse IgG (Fab')2 (1:100) and Streptavidin FITC

(Vector Laboratories, Burlingame, CA, USA) with propidiumiodide (nuclear staining) was used. Chamber slides (Nunc Lab Tech) were used in certain cases to culture cancerous cells and IF label them in situ. Indirect immunofluorescence (FITC) labeling was detected in confocal laser microscopy (Nikon Eclipse E600, Nikon Model C1-Lu3, Tokyo, Japan) or conventional IF microscopy.

#### RESULTS

### Representative B Cell Antigenic Pattern Detected in Tissue Sections of Malignant Melanomas Enable the Investigations of TIL-B at a Molecular Level

Malignant melanoma tissue sections served as subjects for tracking TIL-B cells. As among solid tumors, invasive breast carcinomas with medullary features are of particular interest in respect to TIL-B cells, and we used those as positive controls. Immunohistology with Haematoxilin & Eosin showed a high level of tumor cell infiltration in the majority of the paraffin-embedded tissue sections in both solid tumors. IHC with CD20cy monoclonal antibody represented variable to considerable B cell infiltration in the investigated numerous tissue sections (**Figure 1**). Due to the substantial B cell infiltration found in invasive breast carcinoma with medullary features and metastatic melanomas, our complex strategy involving immunologic, molecular genetics, and biotechnological methods could be properly executed (**Figure 2**). By processing cancerous tissue specimen according to the methodological steps described in our Flowchart, we could successfully approach the questions on TIL-B cells. Tissue sections, cultured cancerous cells, and immune B cells found in the tumor microenvironment were all subjects of a next methodological pathway. Small pictures in the Flowchart show two essential milestones we've reached, demonstrating that the technology works. Fresh primary cell cultures of melanomas react well with specific anti-tumor antibody fragments of TIL-B origin.

### Heavy and Light Chain Immunoglobulin Variable Region Genes Could be Amplified and Cloned for Comparative DNA Sequence Analysis and scFv Antibody Fragment Construction

Amplified melanoma TIL-B heavy and light chain immunoglobulin variable region gene PCR products were

FIGURE 4 | Methodology flow chart on the selection of tumor binder scFv fragments by ELISA. The flow chart presents the new tumor infiltrating B cell antibody fragment phage display technology. It describes the biotechnological processes and our detection system. A master plate was prepared that enabled the detection of our scFvK phage display library in a soluble scFv and a phage displayed scFv form after rounds of panning reactions. Antibody fragments could be selected in the course of an ELISA using 95 well maxisorbe plates precoated with native cancer membrane preparations.

FIGURE 5 | Comparing the tumor membrane binding capacity of antibody fragments. This is an enzyme labeled Immunosorbent Assay performed with test antibodies and antibody fragments in Nunc MaxiSorp® flat-bottom 96 well plates precoated with native tumor cell membrane preparations. IDC breast primary tumor cell culture membrane coated plates were reacted with antibodies of interest (A): 1/MDA MB-231 panned sol scFv library (dark blue), 2/G2 sol scFvK (purple), 3/B2 sol scFvK (light green), 4/anti mucin antibody fragment (green), 5/HCBC3 anti GD3 antibody (orange), 8/medium control (yellow), 9/test/PBS background control (light blue). In the ELISA blocking assay two antibodies were used subsequently in one reaction (B): 1/G2 scFv (purple), 2/HCBC3 (anti GD3) and G2scFv (orange), 3/HCBD1 (anti GD2) and G2scFv (violet), 4/IgG and G2scFv (dark green), 5/medium control (yellow), 6/background control with PBS (light blue). Optical density (OD) was measured at λ: 405 nm when p*-*Nitrophenyl Phosphate substrate was added after using alkaline phosphatase conjugated antibody.

successfully cloned with an appropriate vector and bacterial transformation system. QiaGuick Qiagen plasmid preparations were qualified for a subsequent DNA sequence analysis. Ig variable region heavy and light chain DNA region inserted bacterial clone plasmid preparations were subjects for subsequent three-step PCR amplification, according to previously used techniques (19) for the scFvK antibody fragment construction (**Figure 3**). ScFVK antibody fragment phagemid library (9 × 10<sup>10</sup> member size), generated with vector system (pHEN1, pCANTAB) was suitable for further testing. We could improve our phagemid ELISA tumor binder antibody selection technique with the usage of native cancer membrane fractions, obtained from melanoma primary cell cultures (**Figure 4**). The technology is suitable to define cancer binder scFv antibody fragments in their soluble form or as a protein expressed on the phage. Soluble scFv ELISA proved to be a reliable test system to analyze the characteristic tumor cell binding potential of the selected scFv antibody fragments (**Figure 5**). The technology is appropriate to compare the binding efficiency of antibodies of different origin and check for eventual cross-reactivity. Anti GD3 ganglioside binding capacity of our selected antibody fragments could be strengthened and further characterized by blocking ELISA experiments.

A comparative DNA sequence analysis was performed and carefully evaluated by various DNA sequence analysis softwares and databases, as specified in the Materials and Methods section. We examined sequences from different melanoma clones but made comparisons with our previously defined disialylated GSL-specific antibody variable region coding DNA sequences, obtained from invasive breast carcinoma with medullary features. The present DNA analysis was compared to previously established essential comparative DNA analysis, indicated by the homology ribbon (**Figure 6**). Sequences belonging to the defined clusters were grouped into a tree structure by TreeView Analysis. We could identify members of the overrepresented VH3/1 cluster. The tendency for how the nearest VH3 sequences build groups is depicted in **Figures 7A,B**. Data show that among melanoma TIL-B antibody variable region heavy chain expressed genes, DNA sequences with extremely high homology to human anti-GD3 ganglioside antibody fragment could be found. A comparative DNA sequence analysis, on expressed tumor infiltrating B cell antibody fragment variable region genes, revealed highly homologous (98%) sequences to previously confirmed unique GD3 ganglioside binders. These data summarized first results of the comparative DNA sequence analysis, however, a more detailed DNA sequence analysis performed with Vector NTI advanced will be available shortly in an invited manuscript we are going to submit to Immunome Research.

### Evidence of Strong Disialylated Glycosphingolipid Expression on Primary Melanoma Cell Cultures and Established Cell Lines

Immunofluorescence FACS analysis with GD3-specific monoclonal antibodies could unambiguously define a strong expression of these unique tumor-associated antigens on primary melanoma cell cultures (**Figure 8A**). The great majority of the cell population showed significantly high immunofluorescence positivity, as compared to the background control values (**Figure 8B**). Mean fluorescence intensity was about 100 times higher than the value of the negative control. Results suggest the importance of these tumor-associated molecules for a strong cancer progression. These data strengthen our earlier findings, when fresh breast cancer cell lines were set up from invasive ductal carcinomas of the breast and used for testing tumor binder antibody fragments in the form of native membrane preparations. A considerable expression of disialylated GSLs could be shown with our selected soluble B2scFv and G2scFv

Difference to compiled germline is shown with green label, while alteration found to blastn results are depicted with a purple lane. Consensus sequences defined in this homology ribbon alignment serve as control in the melanoma antibody repertoire analysis.

antibody fragments on A2058 melanoma cell lines by indirect immunofluorescence FACS analysis (**Figures 8C,D**). Soluble preparations of these antibody fragments proved to be a good form for use of these molecules in immune assays like ELISA and immunofluorescence. Several other cell lines have been tested by us in the course of cell or cell membrane-based ELISA and IF-FACS (data will be shown elsewhere). Chamber slide cultures of SK-Mel 28, M24, and M2058 melanoma cell lines were set up. These cells showed strong disialylated GSL, that is GD3 ganglioside expression with our soluble scFv antibody fragments, in an immunofluorescence assay, visualized by confocal laser microscopy (**Figures 9A–C**). The present technique enabled the

careful detection of GD3 gangliosides on melanoma cryostat tissue sections. Our selected scFvK antibody fragment gave an even stronger reaction with the disialylated GSLs (**Figure 9D**), when compared to the commercially available antibody (**Figure 9E**). Immunofluorescence labeling was validated by controls (**Figure 9F**). This is an important step forward, taking into consideration the difficulties and special care needed to detect GD3 gangliosides in general. so Detecting and also defining these highly tumor-associated GSLs is of potential diagnostic value, as they are functionally so very relevant to cancer progression and metastases.

# Novel Strategy for Harnessing the Host Immune System to Build New Cancer Diagnostics and Therapeutics

Our original theory could now be proved in melanoma. The methodological pathway, with immunological, molecular genetics, and biotechnological techniques is suitable to harness the cancerous tissue together with its tumor microenvironment for the search for cancer antigens. The acquired results can be used as cornerstones to reveal highly cancer-associated biomarkers. One unique impact of the strategy is that it can identify glycolipid-based structures, members of the glycosphingolipid family that play an essential role in cancerogenesis by their abnormal glycosylation. TIL-B cell derived scFvK disialylated GSL binder antibody fragments have the capacity to strongly label cancerous melanoma cells, as seen in the Flow chart (**Figure 2)**. Our strategy and new data can now be further exploited with antibody engineering techniques to harness the tumor infiltrating B cell antibody repertoire to build novel diagnostics and immune-based therapeutics.

#### DISCUSSION

Our immunohistochemistry results support the presence of B cells in melanomas. The amount of these cells varies

greatly, but the methodological Flow chart we developed is sensitive enough for the investigations. We can conclude that our complex new strategy, involving cellular immunological, molecular genetics, and biotechnological techniques processing the cancerous tissue itself, is suitable to approach the questions on TIL-B cell characteristics. This technology, with DNA analysis and TIL-B scFv antibody fragment phage display, enables the revealing of human antibody fragments with specificity to unique glycosylated residues in disialylated glycosphingolipids, GD3 gangliosides. This is an important finding, taking into consideration the difficulties that hindered detection of GSLs and that cancer cells are well characterized by aberrant glycosylation of the surface membrane (15, 21). Our original hypothesis that TIL-B cells would eventually have functions in terms of specific tumor-recognizing capacity in solid tumors, could be proved now in malignant melanoma. The study not only strengthened the hypothesis but would serve as an additional source, besides peripheral blood and lymph nodes, to obtain tumor-binder antibodies of human origin. We emphasize the importance of antibody repertoire analysis at the DNA level, in order to better understand the precise nature of natural human antibodies. Our results suggest that scFv antibody fragment construction based on heavy and light chain immunoglobulin variable region genes, originated from B cells in melanoma, provide a new source for antibody fragments with binding capacity to glycolipid and glycoproteinbased tumor-associated antigens. Our results presented here, evidencing the strong disialylated glycosphingolipid expression on primary melanoma cell cultures and established cell lines, underline the importance of the defined human GD3 ganglioside specific antibody fragments. As the detection of these disialylated glycosphingolipids on malignant cells and tissues have been challenging, the present findings serve as a solution to that problem. The special functional characteristics of these key disialylated glycosphingolipid molecules provides promise of a further diagnostic and/or therapeutic usage (23, 24).

Ongoing investigations show how the tumor changes its environment to modify and impair functions of the host immune apparatus. In addition to some well documented mechanisms, there are essential results available to support the hypothesis

on shedding of certain tumor-associated GSLs (25). GSLs that shed from cancerous cells may serve to protect tumor cells from host immune destruction. Certain sialic acid-containing GSLs (gangliosides) have potent immune regulatory properties: inhibitory effects on antibody production in vivo, on the generation of antibody-synthesizing cells in vitro (26) and on lymphocyte proliferative responses to mitogens and antigens in vitro (27). Both bovine and human brain gangliosides, and gangliosides shed by antigen-stimulated lymphocytes (27) were shown to exhibit such immune regulatory properties.

As tumor-derived gangliosides affect immune cell functions and reduce the B lymphocytes' antibody production, we suspect an important B lymphocyte and cancer cell crosstalk mechanism. In order to reveal the cornerstone molecules of that regulation mechanism, the molecularly processed TIL-B cells and tumorassociated glycosphingolipid containing cancerous tissues are subject to a subsequent high-throughput gene expression data analysis.

Our finding, that TIL-B cells produce antibodies that are specific to tumor-associated disialylated glycosphingolipids, is of special interest, taking into consideration the importance of these molecules in tumor progression, invasion, metastases, and signal transduction (28, 29). We postulate that TIL-B cells, by producing unique anti-GD3 disialoganglioside-specific antibodies in the tumor-microenvironment serve a potential anti-tumor and an immunological regulation mechanism. Our novel tumor immunological protocol involving TIL-B antibody phage display and immunoglobulin profile analysis (30) to define the sialylated glycosphingolipid-binding capacity in a broader group of patients with metastatic melanoma is under evaluation. Antibodies with binding capacity to sialylated glycosphingolipid structures can neutralize shed gangliosides. This helps to restore the antitumor reaction that is diminished by shed tumorassociated molecules (31) and might influence the strength of the immune response in general.

We wonder, how far cancer cell editing is influenced by GSLs? It is well known that variation in sialylated glycosphingolipid composition, changes in membrane and/or cytoplasmic localization and the intensity of tumor antigen shedding into the microenvironment are important factors (32). Dramatic changes in glycolipid composition and metabolism were observed in spontaneous tumors, in addition to virally or chemically transformed cells (33, 34). Present data urge extensive immunological and biochemical analysis of sialylated glycosphingolipids, and further understanding of TIL-B cells' potential "crosstalk" in relation to cancer cells.

Interestingly, GD3 ganglioside directed monoclonal antibodies can bind to themselves and to each other. It is expressed within the VH region of anti-GD3 monoclonal antibodies, while the homophilic binding epitope must be bound to a surface. Homophilic binding may strongly contribute to the apparent avidity of mouse anti-GD3 for GD3 ganglioside. Shared or cross reactive idiotype may increase the avidity of antigen binding (35, 36). A subset of antibodies which carry an internal image of their own antigens would play an important role in the regulation of the humoral immune response through an idiotype network. The potential use of anti-idiotypic antibodies for cancer treatment continues to be a challenge for tumor immunotherapy. Failures and hopes obtained, with ganglioside mimicking anti-idiotypic antibodies and the existence of a natural response against gangliosides, suggest that these glycolipids could be idiotypically relevant antigens (37, 38). Our present results help to better understand this essential field. They provide relevant TIL-B Ig VH genes for further studies. Anti-idiotypic antibodies as cancer vaccines, with achievements and possible future improvements, have always been an expectant area of interest (39–41).

According to the above findings, TIL-B cells' unique GD3 ganglioside-specific antibody producing capacity offers a potential immune regulatory mechanism in the tumor-microenvironment. These results show novel interactions between the immune components of the tumor microenvironment and the cancer cells. This potential tumorimmunological regulation mechanism would complete existing theories, summarized on galectin-glycan interactions and B cells (42).

Antibodies to disialylated GSLs can neutralize shed gangliosides and thus help to restore the diminished T and B cell functions, caused by GSLs. However, as only a few antigens have been identified by TIL-B cells (43–46) our TIL-B disialylated ganglioside binder antibody fragments could be a real asset. Based on the results, we can conclude that TIL-B cells' immunoglobulin variable region gene usage with unique disialylated glycosphingolipid revealing capacity is a subject to be harnessed by further antibody engineering, in order to develop cancer targeting tools. Consequently, present data would lead to new early diagnostics and innovative immune-based cancer therapeutics.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Bioethic Codex, Principles and practice of medical and biological/clinical research (2016. január) (https:// ett.aeek.hu/bioetikai-kodex/ guidelines (that was written and approved by Board of Directors of Medical Research Council, Hungary), with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Medical Research Council, Hungary, coding number of the Ethical permission we received for our project: (ETT TUKEB 16462-02/2010, 336/2014.9710-1/2015/EKU).

# AUTHOR CONTRIBUTIONS

BK was responsible for the main idea, new methodology development, realization of preparative and experimental works, and wrote the manuscript. SH took minor cancerous tissue samples. KE took punch biopsies from cancerous tissues. VP carried out microscopic evaluation of histology and immunohistochemistry. GN summarized and updated patients' history data. KC, working in melanoma patient care, dealt with patients' formal consent, as well as minor cancerous tissue and blood sampling. MB, an experienced scientist in immunology, gave methodological suggestions in high-precision processes. EF, Head Surgeon in our Institute, arranged for access to the surgical materials. LT, Head Surgeon in our Institute, arranged necessary ethical permission issues. JT gave technical help in immunofluorescence confocal laser microscopy. AS helped with scientific technical issues, methodological obstacles, manuscript editing. YS, an experienced clinician scientist in immunological disorders, helped evaluating natural human antibody data. MG has strong experience in image analysis in radiological diagnostics and detects cancer metastases. As former director of the National Institute of Oncology, MK's suggestions and arrangements enabled to expand this tumor immunological project into a further melanoma patient follow-up study. GL, an experienced clinician, Head of the Oncoteam and Melanoma Patient Care Unit, arranged for innovative treatment strategies for melanoma patients and reported according to our Ethical Permission.

# FUNDING

BK received the following grants which supported the present work: OTKAT048933, Harry J. Lloyd Charitable Trust Melanoma Research Award, Fulbright: No120610, No1214104. An innovative collaborative research project grant by the National Scientific Research Fund of Hungary (OTKA K109865, 2013–2017) provided additional support.

# ACKNOWLEDGMENTS

Here we thank for the helpful suggestions of Professors Dr. Jean-Luc Teillaud (Paris, France), Dr. Mark C. Glassy (San Diego, CA, USA), Dr. Mepur H. Ravindranath (Los Angeles, CA, USA), and for the generous technical support of Mr. Arend Fock (Appel, Germany), as well as the voluntary technical assistance of Mrs. Sylvia Kotlan (Budapest, Hungary), that aided the efforts of the authors.

REFERENCES

#### 1. Fox BA, Schendel DJ, Butterfield LH, et al. Defining the critical hurdles in cancer immunotherapy. J. Translat. Med. (2011) 9:214–21. doi: 10.1186/1479-5876-9-214


46. Garaud S, Zayakin P, Buisseret L, Rulle U, Silina K, deWind A, et al. Antigen specificity and clinical significance of IgG and IgA autoantibodies produced in situ by tumor-infiltrating B cells in breast cancer. Front Immunol. (2018) 9:2660–71. doi: 10.3389/fimmu.2018. 002660

**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 Kotlan, Horvath, Eles, Plotar, Naszados, Czirbesz, Blank, Farkas, Toth, Tovari, Szekacs, Shoenfeld, Godeny, Kasler and Liszkay. 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.

# Emerging Role of Lymphocyte Antigen-6 Family of Genes in Cancer and Immune Cells

#### Geeta Upadhyay\*

*Department of Pathology, John P. Murtha Cancer Center, F. Edward Hebert School of Medicine, Uniformed Services University of the Health Sciences, Bethesda, MD, United States*

Stem Cell Antigen-1 (*Sca-1/Ly6A*) was the first identified member of the Lymphocyte antigen-6 (*Ly6*) gene family. Sca-1 serves as a marker of cancer stem cells and tissue resident stem cells in mice. The Sca-1 gene is located on mouse chromosome 15. While a direct homolog of Sca-1 in humans is missing, human chromosome 8—the syntenic region to mouse chromosome 15—harbors several genes containing the characteristic domain known as LU domain. The function of the LU domain in human *LY6* gene family is not yet defined. The *LY6* gene family proteins are present on human chromosome 6, 8, 11, and 19. The most interesting of these genes are located on chromosome 8q24.3, a frequently amplified locus in human cancer. Human *LY6* genes represent novel biomarkers for poor cancer prognosis and are required for cancer progression in addition to playing an important role in immune escape. Although the mechanism associated with these phenotype is not yet clear, it is timely to review the current literature in order to address the critical need for future advancements in this field. This review will summarize recent findings which describe the role of human LY6 genes—*LY6D*, *LY6E*, *LY6H*, *LY6K*, *PSCA*, *LYPD2*, *SLURP1*, *GML*, *GPIHBP1,* and *LYNX1*; and their orthologs in mice at chromosome 15.

Keywords: LY6D, LY6E, LY6H, LY6K, TGF-β, Immune, Oncology

#### INTRODUCTION

Sca-1 is among the first identified members of the murine Ly6 gene family (1, 2). The Ly6 gene family belongs to the superfamily of lymphocyte antigen-6 (Ly6)/urokinase-type plasminogen activator receptor (uPAR) proteins. This superfamily is characterized by the presence of LU domain. LU domain is a 60–80 amino acid domain, which is composed of 6–10 cysteines arranged in a specific spacing pattern that allows distinct disulfide bridges which create the three-fingered (3F) structural motif. Three-fingered structural motif are ancient proteins. The LU domain, a three-fingered motif in Ly6/uPAR family are believed to be the evolutionary ancestors of 3FTx toxins found in snake venom. The LU domain in human LY6/uPAR family is not toxic, exact function of LU domain is not yet defined. LU domain is also found in extracellular domains of cell-surface receptors with membrane spanning domain (activin type 2 receptor and bone morphogenetic type IA receptor), or in GPI-anchored proteins CD177 or in secreted globular proteins such CD59 antigen, SLURP1/2.

The LY6 gene family of proteins are located on human chromosomes 6, 8, 11, and 19 and the orthologs exist on syntenic areas of mouse chromosomes. The human chromosome 8 harbors genes namely PSCA, LY6K, SLURP1, LYPD2, LYNX1/SLURP2, LY6D, GML, LY6E, LY6L, LY6H, and

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Sandra A. Jablonski, Georgetown University, United States William Garrow Kerr, Upstate Medical University, United States Priyabrata Mukherjee, University of Oklahoma Health Sciences Center, United States*

> \*Correspondence: *Geeta Upadhyay geeta.upadhyay@usuhs.edu*

#### Specialty section:

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

Received: *31 August 2018* Accepted: *27 March 2019* Published: *24 April 2019*

#### Citation:

*Upadhyay G (2019) Emerging Role of Lymphocyte Antigen-6 Family of Genes in Cancer and Immune Cells. Front. Immunol. 10:819. doi: 10.3389/fimmu.2019.00819* GPIHBP1; while the syntenic mouse chromosome 15 contains genes Psca, Slurp1, Lypd2, Slurp2, Lynx, Ly6d, Ly6g6g, Ly6k, Gml, Gml2, Ly6m, Ly6e, Ly6i, Ly6a, Ly6c1, Ly6c2, Ly6a2, Ly6g, Ly6g2, Ly6f, Ly6l, Ly6h, and Gpihbp1. The human chromosome 19 harbors genes LYPD4, CD177, TEX101, LYPD3, PINLYP, PLAUR, LYPD5, and SPACA4; while the syntenic mouse chromosome 7 contains genes Lypd5, Plaur, Pinlyp, Lypd3, Tex101, Lypd10, Lypd11, Cd177, Lypd4, and Spaca4. The human chromosome 11 harbors genes ACRV1, PATE1, PATE2, PATE3, and PATE4, CD59; while the syntenic mouse chromosome 9 contains genes Pate4, Pate2, Pate13, Pate3, Pate1, Pate10, Pate7, Pate6, Pate5, Pate12, Pate11, Pate9, Pate8, Pate14, and Acrv1. The human chromosome 6 harbors genes LY6G6C, LY6G6D, LY6G6F, LY6G5C, and LY6G5B while the syntenic mouse chromosome 17 contains genes Ly6g6c, Ly6g5c, Ly6g5b, Ly6g6d, Ly6g6f, and Ly6g6e (2). Many of the mice Ly6 genes were lost in humans, perhaps due to their redundant function during evolution (3).

In this mini review, we will focus on the role of human LY6 gene family located on the chromosome 8 namely LY6D, LY6E, LY6H, LY6K, PSCA, LYPD2, SLURP1, GML, GPIHBP1, and LYNX1 and their orthologs in mice at chromosome 15. We chose to focus on this set of genes as they have shown to be increased in human cancer. It is to note that the most widely studied murine Ly6 gene on chromosome 15, Sca-1/Ly6A does not have a human ortholog. It is not clear which gene in human Ly6 gene family may be functionally similar to Sca-1 or the multiple genes may perform several important Sca-1 functions. Sca-1 protein is the most common cell surface marker that is used to enrich adult hematopoietic stem cells (HSCs). Sca-1 protein expression is variable depending on the stages of differentiation of HSCs. Sca-1 protein expression is reduced in HSCs that have differentiated to common myeloid progenitors and then re-expressed in subsets of myeloid progenitors. In a similar pattern, HSCs differentiating into progenitor population suppress Sca-1 expression and immature thymocytes have turned off the Sca-1 protein expression whereas mature single positive thymocytes and peripheral T- cells regain Sca-1 protein expression (4). Sca-1 protein expression has also been identified as an important regulator of tumor progression in mouse models of cancer (5–7). Although the role of LY6 genes on human chromosome 8 in immune cells is not yet established, their role in cancer progression is rapidly emerging. Therefore, we will discuss recent advances, current research gaps and critical need for future research regarding members of the LY6 gene family namely LY6D, LY6E, LY6H, LY6K, PSCA, LYPD2, SLURP1, GML, GPIHBP1, and LYNX1; and their orthologs in mice at chromosome 15.

#### Lessons Learned From the Phenotype of Knockout Mice

The knockout mice models for mouse Ly6 genes with human orthologs—namely Ly6E, Ly6K, Lynx1, Slurp1, and Gpihbp1 have been described (**Table 1**). The other family members have not been characterized in the knockout mice model. The phenotype and the characterization of the known knockout mice models are as described below, which offer important insight into the function of these genes.

**Ly6E:** The mouse Ly6E gene was deleted by gene targeting in mice to create homozygous knockout mice (8). Animals with a Ly6E−/<sup>−</sup> mutation showed embryonic lethality at E14.5. It was though that embryonic lethality was due to cardiac malformation. It was later resolved that Ly6E knockout induced lethality was due the critical role of Ly6E in the trophoblast stem cells (9). The trophoblast forms the outer layer of fetal part of the placenta. Ly6E is expressed specifically in the syncytiotrophoblast (SynT-I) cells (10). Ly6E was found to be a possible receptor for Syncytin A ligand (11). The syncytiotrophoblast layer of fetal placenta play important role in connecting with maternal placenta for the proper exchange of nutrients. In the absence of Ly6E in the fetal placenta, this fetal-maternal placental vascularization is not well-formed causing harm to the fetus. The role of Ly6E in human placental development has not yet been described. Thymic development was found to be normal in Ly6E−/<sup>−</sup> mice, suggesting that Ly6E may play a redundant role in thymic development and development of immune cells.

**Ly6K:** The mouse Ly6K gene knockout mice were generated using a targeting vector substituting exons 2, 3, and 4 (12). Adult Ly6K−/<sup>−</sup> mutant male mice were found to be infertile. Female Ly6K−/<sup>−</sup> mutant mice had normal fertility. Ly6K is not expressed in mature spermatozoa, however the spermatozoa


*The mouse Ly6 genes namely Ly6E, Ly6K, Slurp1, Lynx1, and Gpihbp1 have mouse orthologs and they have knockout mice described. The mouse Ly6 genes namely Pcsa, Lypd2, Ly6D, Gml, and Ly6H have human orthologs but the knockout mice have not been described. The mouse Ly6 genes Sca1, Ly6I/M, Ly6C1, Ly6C2, Ly6B, Ly6G, BC025446, 9030619P08Rik don't have human orthologs. Among these genes only Sca1 has been described in a knockout mice model.*

Frontiers in Immunology | www.frontiersin.org

from Ly6K−/<sup>−</sup> mutant male mice were unable to migrate into the oviduct (12). The reason for Ly6K associated migration defects in sperm of adult mice is not yet known. The role of Ly6K in human sperm related infertility is not yet described.

**Slurp1:** The mouse Slurp1 gene is a secreted member of the Ly6 gene family. The deficiency and mutations in human SLURP1 gene causes Mal de Meleda (MDM), a rare autosomal recessive genetic disease, characterized by inflammatory palmoplantar keratoderma (13, 14). Slurp1−/<sup>−</sup> mutant mice exhibit this rare palmoplantar keratoderma, and show metabolic phenotypes such as reduced adiposity, protection from obesity on a high-fat diet, low plasma lipid levels, and neuromuscular abnormalities such as hind-limb clasping (15).

**Gpihbp1:** Glycosylphosphatidylinositol-anchored highdensity lipoprotein-binding protein 1 (Gpihbp1) deficient mice on a regular chow diet display accumulation of chylomicrons in the plasma and high plasma triglyceride levels (16). Gpihbp1−/<sup>−</sup> mice show reciprocal metabolic perturbations in adipose tissue and liver due to defective lipolysis (17). These results suggest that human GPIHBP1 gene may play important role in the lipolytic processing of triglyceride-rich lipoproteins.

**Lynx1:** Ly6 neurotoxin1 (Lynx1) knockout mice display increased visual cortex plasticity in mice (18). The Lynx1−/<sup>−</sup> phenotype was partially derived from the inhibition of nicotinic acetylcholine receptor signaling by Lynx1 (18). These results suggest that human LYNX1 gene may play important role in nicotine response in the brain.

### Expression of Mouse Ly6 Genes in Immune Cells

The expression of many mouse Ly6 gene family proteins are lineage specific and their expression coincides with the differentiation stages of leukocyte cell populations as seen in the mouse model. The role of human LY6 genes in immune cells differentiation is still not clear. These properties have made them attractive targets to be used for subset identification of leukocytes in vitro and antibody-mediated depletion of specific immune cell populations in vivo. Recently, Nigrovic et al. summarized the mRNA expression of mice Ly6 genes in B-cells, T-cells, NK cells, monocytes, and dendritic cells based on data from the Immgen database (3). Ly6E RNA expression was expressed on all tested immune cell subtypes in the Immgen database. Ly6D RNA was expressed in B-cells, T-cells and dendritic cells. Lypd2, which was described in non-classical monocytes (19). Ly6F, Ly6H, Ly6K, Gml, Psca, Gpihbp1 RNA expression was not found to be associated with expression on immune cells in mice. Interestingly, the mice genes namely Sca1, Ly6B, Ly6C, Ly6G, Ly6I/Ly6M, and Ly6F RNA expression have found to be expressed on many immune cells types in mice. These genes do not have human orthologs. It is yet to be discovered which human genes may function similar to these mice genes. The mRNA expression for Ly6A/Sca1 was found to be present in Bcells, T-cells and dendritic cells (3). Ly6B RNA expression was found in NK cells, monocytes and neutrophils but absent from B-cells, T-cells and dendritic cells. Ly6C RNA expression was found to be present in all subsets except T cells. Ly6G RNA expression was only present in neutrophils and absent from all other cell types analyzed. Ly6I/Ly6M RNA expression was present only in monocytes and neutrophils and absent in all other cell types. It has been shown that RNA and protein expression for some of the human LY6 gene family members such as GML (20, 21), LY6K (22, 23), PSCA (24, 25), and LY6E (23, 26) have immune-modulatory properties in tumor microenvironment.



LY6E and LY6D are the only two genes on human chromosome 8 with a mouse ortholog on chromosome 15, for which some data is available to describe their expression and function in immune cells using mouse models, as discussed below. These genes have been shown to be important in the proliferation and differentiation of immune cells using mainly in vitro methods (3, 27). The mechanistic basis behind the differentiation specific expression remains to be fully understood.

**Ly6D:** The mRNA for mouse Ly6D gene was found to be present in B-cells, T-cells and dendritic cells and absent in NK cells, monocytes, and neutrophils (3). Recent in vitro studies indicate that Ly6D gene is expressed in common lymphoid progenitors which arise from hematopoietic stem cells and give rise to B-lineage lymphocytes (28). Mouse Ly6D gene expression is associated with B-cell specification (29). Ly6D and SiglecH gene expression positive cells from IL7R positive lymphoid progenitor cell populations were committed to become plasmacytoid dendritic cells, while double negative cells were uncommitted. Plasmacytoid dendritic cells are an immune subset devoted to the production of high amounts of type 1 interferon in response to viral infections (30).

**Ly6E:** The mouse Ly6E gene is expressed in peripheral Bcells, immature T-cells, activated T cells, thymus stromal cells, and macrophages (3). A functional role of Ly6E was reported in maintenance of self-renewal of erythroid progenitors (31).

#### Expression of Human LY6 Genes in Normal (Non-lymphoid) Tissues

**LY6D:** Human LY6D RNA and protein is expressed in normal esophageal and skin as published by the human protein atlas on www.proteinatlas.org (32–36).

**LY6E:** Human LY6E RNA is expressed in normal liver, fetal placenta, lung, and spleen (32–36).

**LY6H:** Human LY6H RNA is expressed in brain (32–36).

**LY6K:** Human LY6K RNA and protein is expressed in normal testis (32–36).

**PSCA:** Human PSCA RNA is expressed in prostate, urinary bladder, and esophagus (32–36).

**LYPD2:** Human LYPD2 RNA is expressed in esophagus and tonsil (32–36).

**SLURP1:** Human SLURP1 RNA is expressed in esophagus and skin (32–36).

**GML:** Human GML RNA is expressed in testis and adrenal gland (32–36).

**GPIHBP1:** Human GPIHBP1 RNA is expressed in adipose, lung, breast, brain, heart and soft (32–36).

#### Expression of Human LY6 Genes Is Associated With Cancer and Outcome of the Disease

The LY6D, LY6E, Ly6K, and Ly6H RNA is expressed at high levels compared to adjacent normal tissues in a multitude of tumors including ovarian, colorectal, gastric, breast, lung, bladder, brain and CNS, cervical, esophageal, head and neck, and pancreatic cancer. The increased expression of LY6D, LY6E, LY6K, and LY6H was associated with poor survival in ovarian, colorectal, gastric, breast, and lung cancer (37) (**Table 2**). The casual reason for the increased expression of LY6 family proteins and the poor survival outcome is not yet established. As discussed further in the review under cell signaling heading, LY6 proteins play important role in TGFβ signaling, AKT pathways and immune regulation. A cumulative effect of Ly6

downstream pathways may lead to increased aggressiveness of cancer cells and leading to poor survival outcome. More mechanistic studies will need to be performed to determine the precise pathways responsible for poor survival outcome in patients.

**LY6D**: Recent clinical outcome data added and published by Km plotter web tool show that increased RNA expression of LY6D is associated with poor prognosis in renal clear cell carcinoma and pancreatic ductal adenocarcinoma [**Figure 1**, (38)]. Recently, LY6D expression—in addition to OLFM4 and

S100A7—was found to be associated with distant metastasis of estrogen receptor positive breast cancer (39). LY6D has also been shown to be increased in aggressive forms of head and neck cancer (40).

**LY6E:** Recent data also show that increased expression of LY6E is associated with poor overall survival of renal papillary cell carcinoma and is a good prognostic marker for renal clear cell carcinoma [**Figures 2A,B**, (38)]. These new data indicated that increased expression of LY6E is associated with poor overall survival of pancreatic ductal adenocarcinoma [**Figure 2C**, (38)]. The use of genome wide data analysis has prompted several new reports showing increased expression of Ly6E in bladder cancer, gastric cancer (39, 40). The LY6E gene has been also associated with more aggressive stem like cells in hepatocellular carcinoma, pancreatic carcinoma, colon, and kidney (41–43).

**LY6H:** Recent data shows that increased expression of Ly6H is associated with poor overall survival of renal clear cell carcinoma and pancreatic ductal adenocarcinoma [**Figure 3**, (38)].

**LY6K:** Increased expression of LY6K has also been reported in metastatic ER positive breast cancer (41–43), esophageal squamous cancer (44), gingivobuccal cancers (45), bladder cancer (46), and lung cancer (47). Recent data also show that increased expression of LY6K is associated with poor overall survival of renal clear cell carcinoma, renal papillary cell carcinoma and uterine corpus endometrial carcinoma [**Figure 4**, (38)].

**PSCA:** Prostate stem cell antigen (PSCA) RNA is expressed at high levels compared to adjacent normal tissues in prostate and pancreatic tumors as well as in glioma (48, 49). PSCA is downregulated in esophageal squamous cell carcinoma where it can act as a tumor suppressor by facilitating the nuclear translocation of RB1CC1 (50). Bioinformatics studies have indicated that polymorphisms rs2294008 in the PSCA gene may be prognostic in nature, however this association remains inconclusive due to contradictory observation and a lack of in vitro or in vivo experimental evidence (51–53).

**LYPD2:** LYPD2 RNA is expressed at high levels compared to adjacent normal tissues in cervical and head and neck cancer are associated with a favorable prognosis as shown by the human protein atlas (32).

**SLURP1:** SLURP1 RNA expression is reduced in metastatic melanoma (13). The prognostic value of Slurp1 is unknown.

**GML:** GML RNA expression is increased in non-small cell lung carcinoma (NSCLC) expressing wild-type P53 or P53 negative tumors. This specific expression of GML in P53 negative tumors was found to be linked to cisplatin sensitivity in NSCLC (54). The prognostic value of GML is not clear.

**GPIHBP1:** GPIHBP1 RNA and protein is increased in renal cancer (32–35). The prognostic value of GPIHBP1 is unknown.

# Role of Human LY6 Genes in Other (Non-cancer) Diseases

**LY6D:** A single nucleotide polymorphism rs2572886 found in the human LY6D and Ly6PD2 genes is associated with cellular infection susceptibility to HIV-1 in lymphoblastoid B-cells and in primary T-cells and was also associated with accelerated disease progression in one of two cohorts of HIV-1–infected patients (55).

**LY6E:** LY6E expression is up-regulated during chronic HIV infection (26, 56). Using a HIV pathogenesis model, it was shown that LY6E can down-regulate monocyte responsiveness by modulating CD14 expression via an unknown mechanism (26). Recently, LY6E was shown to enhance viral infectivity IFN dependent fashion (57).

**GPIHBP1:** GPIHBP1 is an endothelial cell specific protein. It facilitates triglyceride (TG) lipolysis in the endothelial cells surface by binding to lipoprotein lipase (LPL), which releases nutrients to surrounding tissues (58). Four missense mutations in highly conserved residues in GPIHBP1 (C65Y, C65S, C68G, and Q115P) gene were identified in severe chylomicronemia leading to buildup of triglyceride causing health issues such as pancreatitis (59).

# Cell Signaling Pathways Associated With Human LY6 Genes

The LY6 proteins are glycosylphosphatidylinositol (GPI) anchored cell surface proteins, the regulation of LY6 RNA and protein is not yet understood very well.

#### TGF-β Signaling

Molecular analysis showed that LY6E and LY6K contribute to tumorigenic progression by increased TGF-β signaling, immune escape and increased INF-γ signaling (23).

#### PI3K Signaling

The PTEN and PI3K/Akt signaling pathways are involved in LY6E-mediated increase in HIF-1α transcription (60).

#### Nicotinic Acetylcholine Receptor Signaling

LYNX1, SLURP1, PSCA, and LY6H can modulate α7 nicotinic acetylcholine receptor (nAChR) signaling. The α7 nicotinic acetylcholine receptor (nAChR) protein is expressed in the central and peripheral nervous system (CNS), muscle, lung, and placenta. The nAChR signaling is associated with nicotine addiction, cancers of lung and liver, and preeclampsia (61, 62). In the brain, activation of α7nAChR in macrophages inhibits production of inflammatory cytokines (63). In colorectal cancer,

the activation of α7nAChR in tumor macrophages inhibits colorectal cancer metastasis through the JAK2/STAT3 signaling pathway (64). α7nAChR has been considered an important drug target for the inhibition of lung cancer (65). On a cautionary note, α7nAChR is such an important neurotransmitter receptor in the CNS and muscle, that targeting nicotinic signaling directly via α7nAChR may cause multiple unwanted side effects. Nicotinic signaling can be modulated by multiple members of the LY6 gene family including LYNX1/2, SLURP1/2, and PSCA (66). Nicotinic signaling can affect glutamatergic signaling in in the hippocampus which may impact learning and memory. One member of LY6 family, LY6H have shown to play important role in glutamatergic signaling in the brain (67). It is yet to be determined if LY6H or other members of LY6 gene family regulate cross talk of nicotinic signaling and glutamergic signaling in the brain.

#### CONCLUDING REMARKS

LY6 gene family members have the potential to be used as therapeutic targets. Several approaches including small molecules and antibody neutralization may be suitable in targeting LY6 family proteins. Ly6E protein expression was identified as a highly promising target for molecular antibody drug conjugates directed to solid tumors in animal models which show high RNA expression of LY6E (68). LY6D, LY6E, LY6H, and LY6K

#### REFERENCES


may be used in targeted cancer therapy, provided that future indepth mechanistic studies reveal the signaling networks of these proteins in physiology and disease [**Figure 5**, (37)]. GPIHBP1 mimicking small molecules will have the potential to treat defective lipolysis of triglyceride due to mutated GPIHBP1. In depth analysis of regulation of LY6 gene expression and identification of downstream targets of LY6 proteins are required to better understand LY6 biology. The α7 nicotinic acetylcholine receptor (nAChR) signaling is at the center of a number of diseases including schizophrenia, Alzheimer's disease, chronic pain and inflammatory diseases (69). Therefore, the LY6 gene family members such as LYNX1, SLURP1, PSCA, and LY6H which modulate α7 nicotinic acetylcholine receptor (nAChR) signaling may be considered for the tissue-specific modulation of the nAChR pathway.

#### AUTHOR CONTRIBUTIONS

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

# FUNDING

The work was supported by NCI R01CA227694 and USUHS startup funds to GU.


carcinoma. R Soc Open Sci. (2018) 5:181006. doi: 10.1098/rsos. 181006


**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 © 2019 Upadhyay. 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.

# ATP Triggers Human Th9 Cell Differentiation via Nitric Oxide-Mediated mTOR-HIF1α Pathway

Suyasha Roy and Amit Awasthi\*

*Immuno-Biology Lab, Translational Health Science and Technology Institute, Faridabad, India*

Interleukin 9 (IL-9)-producing helper T (Th9) cells have a crucial effector function in inducing allergic inflammation, autoimmunity, immunity to extracellular pathogens and anti-tumor immune responses. Although the cytokines that lead to the differentiation of human Th9 cells have been identified, other factors that support the differentiation of Th9 cells have not been identified yet. Here we show that the extracellular ATP (eATP) induces the differentiation of Th9 cells. We further show that eATP induces the production of nitric oxide (NO), which create a feed forward loop in the differentiation of human Th9 cells, as inhibition of purinergic receptor signaling suppressed the generation of human Th9 cells while exogenous NO could rescue generation of Th9 cells even upon inhibition of purinergic receptor signaling. Moreover, we show that ATP promotes mTOR and HIF1α dependent generation of Th9 cells. Our findings thus identify that ATP induced nitric oxide potentiate HIF1α-mediated metabolic pathway that leads to IL-9 induction in Th9 cells. Here we identified that the ATP-NO-mTOR-HIF1α axis is essential for the generation of human Th9 cells and modulation of this axis may lead to therapeutic intervention of Th9-associated disease conditions.

#### Edited by:

*Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States*

#### Reviewed by:

*Manu Rangachari, Laval University, Canada Paula M. Oliver, University of Pennsylvania, United States*

#### \*Correspondence:

*Amit Awasthi aawasthi@thsti.res.in*

#### Specialty section:

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

Received: *11 March 2019* Accepted: *02 May 2019* Published: *20 May 2019*

#### Citation:

*Roy S and Awasthi A (2019) ATP Triggers Human Th9 Cell Differentiation via Nitric Oxide-Mediated mTOR-HIF1*α *Pathway. Front. Immunol. 10:1120. doi: 10.3389/fimmu.2019.01120* Keywords: T helper cell (Th), inflammation, T helper 9 cell, cytokines, transcription activation

# INTRODUCTION

Interleukin (IL) 9, a pleotropic cytokine, was initially described as T cell growth factor, as it signals through common gamma chain on T cells (1, 2). In addition to T cells, the role of IL-9 was also demonstrated as mast cell growth factor (3). IL-9 functions through IL-9 receptor consisting of IL-9 receptor α chain and the γ chain (4). Upon binding to IL-9 receptor, IL-9 induces effector functions and activates mast cells, eosinophils during allergic inflammation (5, 6). The most precise role of IL-9 was found to be associated with human allergic inflammation, both IL-9 and IL-9R was found to be genetically linked to human asthma (7). Consistently, subsequent studies have shown that neutralization of IL-9 with anti-IL-9 antibody reduced the allergic inflammation in animal model of asthma, suggesting an important role of IL-9 in disease pathogenesis (7). Consistently, pulmonary overexpression of IL-9 was found to enhance immunopathology in allergic inflammation in asthma in mice (8).

Initial studies have identified IL-9 as Th2 cytokine, as in vivo neutralization of IL-4 substantially blocked the production of IL-9 during Leishmania infection (9). Most of the initial studies on IL-9 were conducted in Th2-biased Balb/c animal models, and therefore it was suggested that IL-9 enhance Th2-associated disease pathogenesis in Leishmania infection as well as allergic inflammation in asthma. Based on these studies, it was clearly established that IL-9 is primarily produced by T cells, its production is found to be increased with the expansion of Th2 cells. The clarity of IL-9 induction in T cells came up with the identification of a T cell population, which predominantly produce IL-9 without expressing lineage-specific cytokines of Th1, Th2 and Th17 cells (10, 11).

The identification of differentiation factors of Th9 cells led to reconcile the association of IL-9 with Th2 cells, as IL-4 is one of the Th2 cytokines required in combination with TGFβ1 to induce the developmental program for Th9 cells (10, 11). The developmental pathway of Th9 cells and iTregs is reciprocally regulated. While TGF-β1 induces the expression of Foxp3, IL-4 not only suppresses the TGF-β1-induced expression of Foxp3 but together with TGF-β1 induces IL-9-producing Th9 cells. Similar to murine Th9 cells, TGF-β1, and IL-4 were also found to induce the differentiation of human Th9 cells (10, 12).

Since IL-9 is primarily associated in allergic inflammation, the functions of Th9 cells was found to be associated in allergic diseases. In addition, Th9 cells are also crucial for the pathogenesis of IBD, EAE and anti-tumor immunity. In recent studies, using the mice model of cancer, the anti-tumor functions of Th9 cells were described (13, 14). It was shown that IL-21 and IL-1β enhance the anti-tumor functions of Th9 cells in IFN-γ dependent manner (15, 16). Subsequent studies have shown that Th9 cells could suppress the tumor growth in antigen-specific and antigen non-specific manner. In addition, it was recently demonstrated that GMCSF and TNF-α profoundly enhanced the anti-tumor functions of Th9 cells (17, 18). Moreover, the role of Th9 cells in eradicating advance tumor were also identified and suggested that IL-9, Eomes, and Traf6 of Th9 cells are essential in eradicating advance tumor (19).

Although the differentiating cytokines that lead to Th9 cell induction were identified, other environmental cues like nutrients, local oxygen levels, and metabolites are yet to be defined. Here we report that the extracellular ATP (eATP) promotes human Th9 cell differentiation, as blocking of purinergic receptor signaling suppressed the generation of Th9 cells. We further show that eATP induces the production of nitric oxide (NO) and creates a feed-forward loop to potentiate the human Th9 cells differentiation. However, ATP and NO independently activate mTOR-HIF-1α pathway, which in turn support Th9 cell differentiation. ATP triggers NO-mediated mTOR-HIF-1α signaling, which increases glycolysis and IL-9 production in human Th9 cells. Our findings identify for the first time that ATP-induced NO potentiates mTOR-HIF-1α-mediated metabolic signaling pathway that is required for IL-9 induction in Th9 cells, thus may lead to therapeutic intervention of Th9 associated disease conditions.

#### MATERIALS AND METHODS

#### Human T Cell Culture

Human Th9 cells differentiation is induced as described earlier (12, 20). All human experiments were performed in accordance to the approved guidelines of Human Ethics Committee of THSTI. Human blood samples were collected from healthy individuals after the written informed consent. Healthy individuals were enrolled in this study based on the inclusion/exclusion criteria prescribed by the Human Ethics Committee of THSTI. Naïve T cells (CD4+CD25−CD45RA+) were sorted by FACS Aria III (BD Bioscience) with ∼95% purity from PBMCs isolated from healthy donors using Ficoll-Paque Plus (GE Healthcare) gradient centrifugation. Sorted naïve CD4<sup>+</sup> T cells (0.1 × 10<sup>6</sup> )/well were stimulated with plate-bound antihCD3 (10µg/ml; UCHT1, Bio X cell) and soluble anti-hCD28 (2µg/ml; 28.2, Bio X cell) in round-bottom 96-well plate. The cells were cultured for 6 days in the presence of IL-12 (10 ng/ml) for Th1, IL-4 (10 ng/ml) for Th2, TGF-β1 (5.0 ng/ml) and IL-4 (10 ng/ml) for Th9, TGF-β1 (5.0 ng/ml), IL-1β (12.5 ng/ml), IL-6 (25ng/ml), IL-21 (25 ng/ml) and IL-23 (25ng/ml) for Th17, TGF-β1 (5ng/ml) and IL-2 (50 U/ml) for Treg cell differentiation (12, 21).

Hypoxia experiments were carried out in a hypoxia chamber (Coy Laboratory Products, Grass Lake, MI, USA) where cells were subjected to 1% oxygen at 37◦C in 5% CO2. For pharmacological chemical treatments, cells were incubated with 1 mM 2-DG (Sigma Aldrich, St. Louis, MO, USA); 50 nM rapamycin (Sigma Aldrich); 5µM acriflavine (Sigma Aldrich); 0.5 mM L-NIL (NO inhibitor) (Sigma Aldrich)**;** 30µM Suramin (Sigma Aldrich), 100µM NOC-18 (NO donor) (Sigma Aldrich), 50µM ATP (Sigma Aldrich) at the onset of culture.

#### Lentiviral Transduction

For transduction, the lentiviral plasmid was packed as described earlier using HEK293T packaging cell line with X-tremeGENE 9 Transfection Reagent (Roche, Germany). Naïve T cells (0.1 × 10<sup>6</sup> )/well were transduced with lentivirus upon activation with plate-bound anti-hCD3 (10µg/ml; UCHT1, Bio X cell) and soluble anti-hCD28 (2µg/ml; 28.2, Bio X cell). The virus-containing media was replaced with fresh medium containing TGF-β1 (5.0 ng/ml) and IL-4 (10 ng/ml) for Th9 differentiation and cultured for 6 days at 37◦C.

#### Nucleofection of Human T Cells

Nucleofection of naïve human CD4<sup>+</sup> T cells was performed as described earlier (20). Naïve T cells (1.5 × 10 6 ) were resuspended in 100 µl AmaxaTM 4D-NucleofectorTM Solution and nucleofected with PBS/pU6-HIF1 alpha RNAi plasmid 2 (Addgene, Plasmid #21104) on 4D-NucleofectorTM X unit (Lonza, Walkersville, MD, USA) using P3 Primary Cell 4D-NucleofectorTM X Kit (Lonza, Walkersville, MD, USA) as per manufacturer's protocol.

#### Flow Cytometry

For intracellular cytokine staining, cells were restimulated on day 7 post-culture with Phorbol 12-myristate 13-acetate (PMA; 50 ng/ml; Sigma-Aldrich), ionomycin (500 ng/ml; Sigma-Aldrich), and Golgi stop (BD Biosciences) for 5 h at 37◦C. Intracellular and surface staining was performed as described earlier. Data were acquired on a FACSVerse (BD Biosciences) and analyzed using FlowJo software (Tree Star, Ashland, OR, USA) (10, 12).

#### qPCR

mRNA was extracted from cells after in vitro differentiation using the RNeasy Mini Kit (Qiagen, Venlo, The Netherlands) and converted to cDNA using iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA). qPCR was performed with the SYBR Green Gene Expression Assay using the ABI Fast 7500 Dx qPCR system (Applied Biosystems) according to the manufacturer's protocol (10). The target gene expression was normalized to GAPDH, and the fold change was calculated as 2−11CT comparative threshold. Briefly qPCR results were analyzed with SDS2.1 software. The cycling threshold value of the endogenous control (gapdh/bactin) was subtracted from the cycling threshold value of the target gene to generate the change in the cycling threshold (1CT). The relative expression of each target gene is expressed as the "fold change." We used this previously used formula (POWER(2, -1CT)<sup>∗</sup> 10,000 to calculate the relative expression of gene (20).

The following primers sets were used: GAPDH (forward, 5′ - ACAGTTGCATGTAGACT-3′ ; reverse, 5′ -TTTTTGGTTGAG CACAGG-3′ ), Il9 (forward, 5′ -GACATCAACTTCCTCATC-3′ ; reverse, 5′ -GAGACAACTGGTCTTCTGG-3′ ), HIF1α (forward, 5 ′ -AAAATCTCSTCCSSGAAGCC-3′ ; reverse, 5′ -AATGTTCC AATTCCTACTGC-3′ ), Nos2 (forward, 5′ -AGCTCAACAACA AATTCAGG-3′ ; reverse, 5′ -ATCAATGTCATGAGCAAAGG-3 ′ ), mTOR (forward, 5′ -AGCAGAGAAAGGTTTTGATG-3′ ; reverse, 5′ -GATCTCCTCCATCTCTTCTC-3′ ), IRF4 (forward, 5 ′ -TGACTCTATGCTTTGGAGAG-3′ ; reverse, 5′ -GCTAAACT CCTAAGTACGTG-3′ ). All the primer sets were purchased from Sigma-Aldrich.

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

The concentration of human IL-9 was measured by ELISA carried out with paired antibodies according to manufacturer's instructions (BioLegend). The plates were analyzed on a SynergyTM HT Multi-Detection Microplate Reader, BioTek (Winooski, VT, USA) (10).

#### Nitrite Determination

The nitric oxide production was estimated in cell culture supernatants by measuring the nitrite concentration, a stable NO product, using Griess Reaction. 50µl of each experimental sample and 100 µl of Griess reagent (0.1% naphthylethylenediamine dihydrochloride and 1% sulphanilamide in 2% phosphoric acid) was added in duplicates in 96-well microtiter plates (22). After incubating for 10 min in dark, absorbance was read at 540 nm using ELISA Plate Reader (Bio-Rad Laboratories).

#### Lactate Assay

Accumulation of lactate, the end-product of glycolysis, were determined in cell culture supernatants harvested on day 7 using the Lactate Colorimetric Assay Kit II (Sigma-Aldrich, MAKO65, USA) according to the manufacturer's protocol.

# ATP Determination

The amount of ATP secreted in the cell culture supernatants harvested on day 7 post-culture was determined by ATP Determination Kit (Molecular Probes, invitrogen detection technologies; A22066, USA) according to manufacturer's instructions as described earlier (23).

### Statistical Analysis

All the statistical analysis was done using GraphPad Prism 7.0 software (La Jolla, CA, USA). Two-tailed Student's t-test were used for comparison of means between two groups and one-way ANOVA test followed by Tukey's multiple comparison's test was used for comparison of means between more than two groups. Multiple groups with two variables were compared using twoway ANOVA followed by Tukey's multiple comparison's test. All the data are presented as mean ± SEM. P < 0.05 were considered statistically significant for all the experiments.

# RESULTS

# ATP Enhances Human Th9 Cell Differentiation

T cells activation and differentiation relies on aerobic glycolysis, instead of oxidative phosphorylation, in order to meet their increased biosynthetic energy demands (24–26). As a result of glycolysis and metabolic activities, lactate accumulate upon T cell activation (24, 27, 28). To understand the metabolic requirements during human Th9 cells differentiation, we measured the lactate production in effector subsets of human Th cells. Interestingly, our data indicates that human Th9 cells produce substantially higher levels of lactate as compared to other Th subsets (**Figure 1A**). In fact, Th2, Th9, and Th17 cells, which produce IL-9, also produce higher amounts of lactate as compared to non-IL-9-producing Th0 and Th1 cells (**Figure 1A**). Since glycolysis leads to lactate production coupled with ATP generation (24, 29, 30), therefore we tested whether Th9 cells also generate higher levels of ATP to support the energy demands of Th9 cells. As compared to Th0, Th9 cells generate significantly higher amounts of ATP (**Figure 1B**), which led us to hypothesize whether ATP contributes to the differentiation of human Th9 cells. Although the role of eATP has been established in Th17 and Tregs cells differentiation and functions (23, 31, 32), the contribution of ATP in the differentiation of human Th9 cells is not identified yet. To understand the role of ATP in the differentiation of human Th9 cells, we supplemented Th9 culture condition with eATP. Our data indicate that the supplementation of eATP in the presence of TGF-β1 and IL-4 further enhanced IL-9 production in Th9 cells (**Figure 1C**). To further substantiate our claim, we blocked the P2X and P2Y P2-purinergic receptors using suramin to decipher whether blocking the binding of eATP affects IL-9 induction in human Th9 cells. Consistently, suramin inhibited Il9 expression and IL-9 production in human Th9 cells (**Figures 1C,D**). Since IRF4 is crucial for the induction and differentiation of Th9 cells (33), therefore we tested whether eATP enhances the expression

of IRF4 in human Th9 cells. Our data clearly indicate that eATP enhances the expression of IRF4 in human Th9 cells (**Figure 1C**), suggesting that eATP contribute to enhanced differentiation of human Th9 cells. Taken together these data collectively indicates that ATP is essential for enhancing IL-9 induction in Th9 cells.

#### ATP Is Essential for NO-Mediated Induction of IL-9 in Human Th9 Cells

The role of NO was tested in T cells differentiation, as NO was found to suppress IL-17 while it promotes IL-9 in Th17 cells (22, 34). Furthermore, the role of NO was also found to directly influence IL-9 induction in Th9 cells, as NOS2 deficiency was found to attenuate IL-9 induction and differentiation of Th9 cells (22, 24). It was shown that ATP contribute to the generation of NO production (35). Based on this observation, we wanted to test whether ATP promotes induction of IL-9 in human Th9 cells by potentiating NO production. To do this, we first tested whether eATP enhances the generation of NO in human Th9 cells. Differentiation of human Th9 cells in the presence of eATP has found to increase the expression of NOS2 and production of NO in Th9 cells (**Figure 2A**). Consistently, blocking of P2X and P2Y P2-purinergic receptors by suramin suppressed the expression of NOS2 and NO production in human Th9 cells, indicating a possibility that ATP-mediated NO might be essential for enhanced IL-9 production in human Th9 cells. Interestingly, we found that supplementation of NO in Th9 cell cultures enhances IL-9 production in human Th9 cells (**Figures 2B–E**). Contrary to NO supplementation, NO inhibition suppressed IL-9 induction in human Th9 cells (**Figures 2B-E**). We further tested whether ATP can rescue the Th9 cells differentiation in the absence of NO. Our data indicate that eATP was able to rescue IL-9 production in the absence of NO generation in human Th9 cells (**Figures 2F,G**). Taken together, it implies that ATP-NO creates a feed forward loop that is essential for the production of IL-9 and Th9 cells differentiation.

#### ATP-mTOR Pathway Induce Human Th9 Cells Differentiation

To understand the molecular mechanisms of contribution of eATP in enhancing human Th9 cells differentiation, we hypothesized to test whether ATP-activated mTOR pathway plays a role in the differentiation of human Th9 cells, as it was shown that ATP activates mTOR pathway (36, 37). To test our hypothesis, we differentiated human naïve CD4<sup>+</sup> T cells into Th9 cells in the presence of eATP. As shown earlier, while eATP increased IL-9 production, it also enhanced

IL-9. Data are representative of mean ± SEM from three independent experiments (*n* = 3). \**P* < 0.0332, \*\**P* < 0.0021, \*\*\**P* < 0.0002, \*\*\*\**P* < 0.0001; one-way ANOVA followed by Tukey's test (A–D, F).

mTOR and pS6 kinase expression, a downstream molecule in mTOR pathway, in human Th9 cells (**Figures 3A,B**). Next we tested whether mTOR is required for the differentiation of human Th9 cells. To do this, we suppressed mTOR using mTOR shRNA during human Th9 cells differentiation. Interestingly, mTOR inhibition significantly blocked human Th9 cells differentiation and IL-9 generation, as determined by IL-9 production and IRF4 expression (**Figures 3C,D**), suggesting that ATP-mediated activation of mTOR is essential in differentiation of human Th9 cells. Consistently, similar to mTOR shRNA inhibition, rapamycin, a pharmacological inhibitor of mTOR, also suppressed mTOR, IRF4 and IL-9 in human Th9 cells (**Figures 3E,F**). To further understand the role of ATP-NOmTOR axis in IL-9 production and Th9 cells differentiation, we tested whether ATP-mediated NO is essential for the induction of IL-9 in human Th9 cells through mTOR. We found that rapamycin suppressed the expression of NOS2 and production of NO in Th9 cells (**Figure 3G**). Since mTOR inhibition shown to suppress glycolysis, therefore we tested the expression of Glut1 and glycolytic genes in the presence of rapamycin. Our data suggests that rapamycin block the expression of Glut1 as well as other glycolytic genes (**Supplementary Figure 1**). Consistently, we have also found that, similar to rapamycin, mTOR shRNA also blocked the NOS2 expression in human Th9 cells (**Figure 3H**). To further understand the functional role of NOS2 and NO in the activation of mTOR-mediated Th9 cells differentiation, we tested whether inhibition or supplementation of NO with NOI or NOD, respectively block or enhance mTOR pathway in human Th9 cells differentiation. Our data indicate that inhibition of NO suppressed mTOR expression

FIGURE 3 | ATP-mTOR pathway induce human Th9 cells differentiation. (A) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence of ATP and Suramin followed by examination of mRNA expression of mTOR. (B) intracellular cytokine staining of the phosphorylation of S6 in the presence and absence of ATP in Th9 cells. (C) Sorted naïve T cells were activated with anti-CD3/CD28 for 24 h and differentiated under Th0 and Th9 skewing conditions, and transduced with control lentivirus carrying scramble shRNA or mTOR shRNA-expressing lentivirus followed by examination of mRNA expression of IL-9 and IRF4 and ELISA for IL-9 production. (D) Intracellular cytokine staining for IL-9 production. (E) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence and absence of RAPA (rapamycin) followed by examination of mRNA expression of IL-9, IRF4 and mTOR and ELISA for IL-9 production. (F) Intracellular cytokine staining for IL-9 production. (G) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence and absence of RAPA (rapamycin) followed by examination of mRNA expression of NOS2 and nitrite measurement in the culture supernatants. (H) Sorted naïve T cells were activated with anti-CD3/CD28 for 24 h and polarized under Th0 and Th9 skewing conditions, and transduced with control lentivirus carrying scramble shRNA or mTOR shRNA-expressing lentivirus followed by examination of mRNA expression of NOS2. (I) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence of NOD and NOI (NOD-NO donor; NOI-NO inhibitor) followed by examination of mRNA expression of mTOR. (J) Intracellular cytokine staining of the phosphorylation of S6 in the presence and absence of NOD in Th9 cells. (K) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days alone, in the presence of RAPA (Rapamycin) and RAPA+NOD followed by examination of mRNA expression of IL-9 and IRF4. (L) Intracellular cytokine production of IL-9 and estimation of IL-9 production in the culture supernatants by ELISA. (M) Examination of mRNA expression of NOS2 and mTOR. Data are representative of mean ± SEM from three independent experiments (*n* = 3). \**P* < 0.0332, \*\**P* < 0.0021, \*\*\**P* < 0.0002, \*\*\*\**P* < 0.0001; one-way ANOVA followed by Tukey's test (A,E,G,I,K,M), two-way ANOVA followed by Tukey's test (C,H).

while supplementation of NO enhanced pS6 Kinase activation in human Th9 cells (**Figures 3I,J**). As indicated that NO is essential for mTOR activation in Th9 cells, we next tested whether NO can rescue human Th9 cells differentiation in the absence of mTOR activation. Interestingly, supplementation of NO rescued Th9 differentiation as evident by IL-9 and IRF4 induction (**Figures 3K,L**) by rescuing NOS2 and mTOR expression in human Th9 cells (**Figure 3M**). These observations together indicate that ATP-NO-mTOR axis is required for human Th9 cells differentiation.

#### ATP-Induced HIF-1α Is Required for Human Th9 Cells Differentiation

It is demonstrated that mTOR signaling leads to HIF-1α activation, which play critical role in differentiation and interplay between effector and regulatory T cells (24, 38, 39). Our data indicated that eATP-NO-mTOR axis is essential in production of IL-9 in human Th9 cells. We tested whether ATP promotes IL-9 and differentiation of Th9 cells by inducing HIF-1α. Our data indicated that eATP not only enhanced IL-9 production in Th9 cells (**Figure 1**) but also increased the expression of HIF-1α at both mRNA and protein level in human Th9 cells (**Figure 4A**). Consistently, inhibition of eATP signaling by suramin inhibited the expression of HIF-1α in human Th9 cells (**Figure 4A**). These observations clearly indicated the role of HIF-1α in the differentiation of human Th9 cells.

FIGURE 4 | ATP-induced HIF-1α is required for human Th9 cells differentiation. (A) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence of ATP and Suramin followed by examination of mRNA expression of HIF1α and intracellular cytokine staining for HIF1α. (B) Sorted naïve T cells were nucleofected with naked scramble shRNA and HIF1α shRNA and were differentiated under Th0 and Th9 polarizing conditions for 6 days analyzed for mRNA expression of IL-9, IRF4, HIF1α and IL-9 production in the culture supernatants by ELISA (C) Intracellular cytokine staining of IL-9 in Th9 cells. (D) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence and absence of acriflavine (ACF) followed by examination of mRNA expression of HIF1α and intracellular cytokine staining for HIF1α. (E) mRNA expression of IL-9, IRF4 and IL-9 production in the culture supernatants estimated by ELISA. (F) Intracellular cytokine staining for IL-9 expression (G) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in normoxia (21% oxygen) and hypoxia (1% oxygen), respectively. Total RNA was extracted, reverse transcribed and real-time PCR was done for analyzing mRNA expression of IL-9 and IRF4 and IL-9 production in the culture supernatants estimated by ELISA. (H) Intracellular cytokine staining of HIF1α and (I) mRNA expression of HIF1α. (J) Sorted naïve T cells differentiated under Th0, Th1, Th2, Th9, and Th17 polarizing conditions for 6 days in normoxia (21% oxygen) and hypoxia (1% oxygen), respectively followed by measurement of lactate production in culture supernatants. (K) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the absence and presence of 2-DG followed by estimation of ATP production in culture supernatants. (L) mRNA expression of IL-9 and IRF4. (M) Estimation of IL-9 production in culture supernatants by ELISA and intracellular cytokine production of IL-9. Data are representative of mean ± SEM from three independent experiments (*n* = 3). \**P* < 0.0332, \*\**P* < 0.0021, \*\*\**P* < 0.0002, \*\*\*\**P* < 0.0001; one-way ANOVA followed by Tukey's test (A,D,E,K–M), two-way ANOVA followed by Tukey's test (B,G,I,J).

Next we assessed the role of HIF-1α in human Th9 cells differentiation by blocking HIF-1α using shRNA, and found that shRNA-mediated silencing of HIF-1α significantly suppressed IL-9 production as well as the differentiation of human Th9 cells indicated by the attenuated expression of IRF4 and IL-9 (**Figures 4B,C**). Consistently, we found that acriflavine, which suppresses the hetero-dimerization of HIF-1α and HIF-1β, also suppressed the expression of HIF-1α at mRNA and protein level (**Figure 4D**). HIF-1α is induced during glycolysis upon T cells activation and differentiation. HIF-1α in turn support glycolysis, as blocking of HIF-1α known to suppress glycolysis. Our data demonstrated that blocking of HIF-1α using acriflavine suppressed the expression of Glut1 together with glycolytic genes (**Supplementary Figure 2A**). Our data further suggests that acriflavine suppressed both IRF4 and IL-9 expression in Th9 cells (**Figures 4E,F**), implying the role of HIF-1α in human Th9 cell differentiation. To further support the role of HIF-1α in human Th9 cells differentiation, we found that human Th9 cell differentiation was enhanced under hypoxic condition, as indicated by the enhanced expression of IRF4, IL-9, and increased production of IL-9 together with HIF-1α expression (**Figures 4G–I**). Our data indicate that Th9 cells, as compared to other Th cells, are highly glycolytic in nature, as they produce an increased amount of lactate (**Figure 4J**). The lactate production was further enhanced in Th9 cells upon culturing them in hypoxic condition (**Figure 4J**). In addition, Th9 cells cultured in hypoxic condition were shown to have enhanced expression Glut1 and the genes that are associated with glycolysis (**Supplementary Figure 2B**). As lactate is generated due to higher consumption of glucose through glycolysis, we tested whether blocking glycolysis using 2-DG suppresses human Th9 cells differentiation. Sorted human naïve T cells were cultured in Th9 condition with or without 2-DG and expression of glycolytic genes were tested. Our data indicated that 2-DG

FIGURE 5 | NO and HIF1α synergistically promote human Th9 cell differentiation. (A) Sorted naïve T cells differentiated under Th0 and Th9 polarizing conditions for 6 days in normoxia (21% oxygen) and hypoxia (1% oxygen), respectively. Total RNA was extracted, reverse transcribed and real-time PCR was done for analyzing mRNA expression of NOS2. (B) Nitrite measurement in Th0 and Th9 cells under normoxia and hypoxia. (C) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence of acriflavine (ACF) followed by examination of mRNA expression of NOS2 and nitrite measurement in culture supernatants. (D) Sorted naïve T cells were nucleofected with naked scramble shRNA and HIF1α shRNA and were differentiated under Th0 and Th9 polarizing conditions for 6 days analyzed for mRNA expression of NOS2. (E) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the presence of NOD and NOI (NOD-NO donor; NOI-NO inhibitor) followed by examination of mRNA expression of HIF1α and intracellular cytokine staining of HIF1α in the presence and absence of NOD in Th9 cells. (F) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days alone, in the presence of ACF (Acriflavine) and NOD+ACF followed by examination of mRNA expression of IL-9 and IRF4 and IL-9 production in the culture supernatants estimated by ELISA. (G) Intracellular cytokine production of IL-9. (H) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days alone, in the presence of ATP and ATP + ACF followed by examination of mRNA expression of IL-9 by qPCR. Data are representative of mean ± SEM from three independent experiments (*n* = 3). \**P* < 0.0332, \*\**P* < 0.0021, \*\*\**P* < 0.0002, \*\*\*\**P* < 0.0001; one-way ANOVA followed by Tukey's test (C,E,F,H), two-way ANOVA followed by Tukey's test (A,B,D). (I) Schematic representation of ATP mediated NO and mTOR-HIF1α signaling induces human Th9 cell differentiation.

significantly suppressed the expression of glycolytic genes in Th9 cells (**Supplementary Figure 3**). Our data further indicated that blocking of glycolysis with 2-DG suppressed ATP generation and IL-9 production (**Figures 4K–M**). Taken together these data indicated that ATP-HIF-1α axis is essential for the differentiation of human Th9 cells.

#### NO and HIF-1α Synergistically Promote Human Th9 Cell Differentiation

Our data indicated that ATP-mediates NO production, which lead to further enhancement of human Th9 cells differentiation via mTOR and HIF-1α pathway. Balance between NO and free radicals, generated in ETC pathway, dictate the stability and functions of HIF-1α (40). It was found that NO stabilizes HIF-1α by making it resistant to PHD-mediated degradation (41). Our data suggest that NO is essential for the differentiation of Th9 cells, therefore we tested whether NO generated in human Th9 cells is linked to HIF-1α-mediated generation of human Th9 cells. To test our hypothesis, we tested and found Th9 cells cultured under hypoxic environment enhances IL-9 and NO production in Th9 cells (**Figures 5A,B**), suggesting a potential link of HIF-1α in generation of NO in Th9 cells. To directly test the role of HIF-1α in the production of NO in Th9 cells, we suppressed the HIF-1α function by acriflavine, which is known to suppress transcriptional activity of HIF-1α. Interestingly, inhibition of HIF-1α transcriptional functions significantly blocked the expression of NOS2 mRNA as well as NO production in Th9 cells (**Figure 5C**).

To further validate our observations, we inhibited HIF-1α using HIF-1α shRNA and determined NOS2 mRNA expression in human Th9 cells. Consistently, our data indicate that suppression of HIF-1α functions inhibited NOS2 expression in Th9 cells (**Figure 5D**). To further understand the association of HIF-1α and NO, we further tested whether NO modulate HIF-1α expression in human Th9 cells. To do this, we generated human Th9 cells either in the presence of NO inhibitor or NO donor to test its effect on HIF-1α expression. Our data indicated that NO donor enhanced while NO inhibitor suppressed HIF-1α expression, respectively and IL-9 induction in human Th9 cells (**Figure 5E**), indicating that HIF1α-NO work in feed-forward loop to promote differentiation of human Th9 cells. Similar to HIF-1α expression, NOD or NOI, respectively enhanced or suppressed the expression of Glut1 and glycolytic genes (**Supplementary Figure 4**). This led us to hypothesize whether HIF-1α-mediated NO production is critical for IL-9 induction and human Th9 cells differentiation, and if that is the case, whether exogenous NO can overcome the requirement of HIF-1α in human Th9 cells differentiation. To do this, we suppressed the functions of HIF-1α using acriflavine in the presence or absence of NOD. Interestingly, we found that the supplementation of exogenous NO could rescue the HIF-1α inhibition in human Th9 cells differentiation, indicating that HIF-1α-mediated generation of NO is critical for differentiation of Th9 cells (**Figures 5F,G**). Taken together the data suggests that HIF-1α is essential for NO-mediated human Th9 cells differentiation. To further understand whether eATP enhances IL-9 induction in Th9 cells is dependent on HIF-1α, we blocked HIF-1α functions using acriflavine (ACF) in the presence of eATP, and found that eATP-mediated enhancement of IL-9 in Th9 cells is suppressed in the presence of HIF-1α inhibitor, suggesting that HIF-1α is crucial for enhancing IL-9 by ATP (**Figure 5H**).

Based on our data, we proposed a schematic model of human Th9 cell differentiation, in which TGF-β and IL-4 initiate the differentiation of human Th9 cells further potentiated by ATP-mediated purinergic signaling. ATP further induces the NO production in human Th9 cells and activates interdependent mTOR-HIF-1α pathways and together contributes the promotion of human Th9 cell differentiation in a feed-forward loop (**Figure 5I**).

# DISCUSSION

Effector T cell subsets differentiate from naïve CD4<sup>+</sup> T cells in the presence of specialized cytokine milieu (42). As compared to other effector T cell subsets, IL-9-producing Th9 cells are found to be less well-characterized, though the cytokines and transcription factors that leads to the induction of human Th9 cells are identified (20, 43–45), the role of other environmental factors, such as metabolites in the differentiation of Th9 cells are yet to be identified. In this study we have shown the extracellular ATP (eATP) induces the differentiation of Th9 cells, as inhibition of purinergic receptor signaling suppressed the generation of human Th9 cells. We further demonstrated that ATP-mediated NO is essential for the differentiation of human Th9 cells, as exogenous NO could rescue the generation of human Th9 cells even in the absence of purinergic receptor signaling. Our findings further identify as to how ATP-nitric oxide potentiate mTOR-HIF-1αmediated pathway that leads to the induction of IL-9 in human Th9 cells.

Initially, IL-9 found to be produced by activated T cells and suggested to be a T cell growth factor (1, 46). Before the identification of Th9 cells, IL-9 thought to be a Th2 cytokine, and as a result the role of IL-9 was tested in Th2-associated disease (47) models (9). Although IL-9 found to be produced by subsets of T cells, such as Th2, Th17, iTregs as well as ILCs, Th9 cells found to exclusively produce IL-9 (20, 48). IL-9 found to be associated with human conditions, such as allergy and asthma, as both IL-9 and IL-9R found to have genetic association with human asthma. Consistently, over-expression of IL-9 within the lung was found to be associated with enhanced infiltration of eosinophils and lymphocytes.

Th9 cells were found to be most potent anti-tumor T cells. Although IL-9 was found to be associated with multiple diseases, the clarity of IL-9 functions in immune responses associated with diseases came only after the identification of Th9 cells (10, 11, 49).

Activation of naïve CD4<sup>+</sup> T cells in the presence of TGFβ1 and IL-4 induces the differentiation of IL-9-producing Th9 cells. Combination of TGF-β1 and IL-4 induces a distinct differentiation program as compared to TGF-β1 or IL-4 alone (10, 11). While TGF-β1 induces the generation of induced Tregs (iTregs) by inducing the expression of Foxp3, addition of IL-4 suppresses Foxp3 expression resulting the induction of IL-9 producing Th9 cells (10). Although the differentiation factors of Th9 cells was identified, other factors that enhances the differentiation of Th9 cells yet to be defined. Considering the important role of metabolic checkpoints, we identified as to how cellular ATP contributes to the differentiation of human Th9 cells. The role of ATP in regulation of T cells differentiation and functions has been identified. During the activation and differentiation of T cells, ATP is generated to fuel the T cell proliferation (29). An additional role of ATP as an extracellular signaling molecule has also been demonstrated, as it mediates cell to cell communication in an autocrine and paracrine manner. Under physiological conditions, cellular ATP is released into extracellular milieu, and therefore activates purinergic receptors signaling upon its binding to P2X and P2Y surface receptors. The role of ATP was found to be associated with Th17 cells differentiation, as ATP found to enhance the expression of TGFβ, IL-6, IL-23p19, and thus enhanced Th17 cell generation and exacerbated T-cell-mediated colitis in mouse model (23, 24). Similar to Th17 cells, we have shown here that eATP signals through purinergic receptors and enhances the human Th9 cells differentiation. However, it is not clear whether ATP controls, if at all, the generation of Foxp3<sup>+</sup> Tregs and Th9 cells reciprocally, as it was established in case of Foxp3<sup>+</sup> Tregs and Th17 cells (31, 50).

The role of NO was identified in differentiation and functions of Th cells, as NOS2-deficient mice were found to harbor enhanced frequency of Th17 cells with the reduction in Tregs cells in EAE, a mouse model of multiple sclerosis, indicating that NO is suppressed in Th17 cells differentiation (51). Subsequent study has identified that NO induces nitration of tyrosine residue in ROR-γt leading to the inhibition of RORγt-mediated induction of IL-17 in Th17 cells (52). In addition to Th17 cells, the role of NO was also identified in Th9 cells, as NO suppresses IL-17 and enhances IL-9 in Th17 cells. It was also shown that NO enhances differentiation of Th9 cells associated with lung pathologies. How NO generated during human Th9 cells differentiation is not clearly understood. Here we found that ATP-mediated NO production is essential to promote Th9 cells differentiation. We have shown that ATP-NO forms a feed-forward loop to promote IL-9 production in Th9 cells.

We further tested as to how ATP-NO modulate mTOR signaling in human Th9 cells. The role of mTOR pathways was found to be associated in Th9 cells differentiation, however, how ATP-NO axis regulate the mTOR pathway in human Th9 cells differentiation was not clearly understood. Our data indicated that eATP modulated mTOR pathway by enhancing the activation of S6 Kinase in human Th9 cells. Inhibition of mTOR suppressed both IL-9 and NOS2 induction in human Th9 cells. We further shown that ATP-mediated NO induces mTOR activation in human Th9 cells differentiation.

Since mTOR pathways leads to HIF-1α pathway, and the role of HIF-1α was suggested in the both Th9 and Th17 cells differentiation. However, it was not really known whether ATP-NO axis mediates HIF-1 alpha activation that leads to the differentiation of human Th9 cells. To connect this axis, our data indicates that ATP-induced the NO-mTOR-dependent induction of HIF-1α in human Th9 cells differentiation. Interestingly ATP and NO found to induce HIF-1α activation, which we found to further support the differentiation of human Th9 cells differentiation, as inhibition of HIF-1α leads to the inhibition of IL-9 induction in human Th9 cells.

Altogether, our work suggested that eATP enhances the differentiation of human Th9 cells and might function as a checkpoint of downstream mTOR-HIF-1α axis. Our data further emphasized the role of NO which is induced by eATP, which further potentiates glycolytic activity dependent on HIF-1α, and modulation of ATP-NO-HIF-1α axis may confer and contribute to the anti-tumor functions of Th9 cells.

#### ETHICS STATEMENT

All human experiments were performed in accordance to the approved guidelines of Human Ethics Committee of THSTI. Human blood samples were collected from healthy individuals after the written informed consent. Healthy individuals were enrolled in this study based on the inclusion/exclusion criteria prescribed by the Human Ethics Committee of THSTI.

#### AUTHOR CONTRIBUTIONS

SR designed, performed the experiments, and analyzed the data. AA wrote the paper, designed, and supervised the study. SR and AA contributed to writing and editing the paper.

#### ACKNOWLEDGMENTS

This work was supported by Wellcome Trust/DBT India alliance intermediate fellowship (IA/I/12/1/500524), Department of Biotechnology, Government of India and Core grant of Translational Health Science and Technology Institute. SR was supported by a Ph.D. fellowship from Council of Scientific and Industrial Research (CSIR).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01120/full#supplementary-material

Supplementary Figure 1 | Inhibition of mTOR signaling downregulates glycolysis resulting in diminished human Th9 cells differentiation. Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the absence and presence of Rapamycin (Rapa) followed by relative mRNA expression of glycolytic genes examined by qPCR. Data are representative of mean ± SEM from three independent experiments (*n* = 3). <sup>∗</sup>*P* < 0.0332, ∗∗*P* < 0.0021, ∗∗∗*P* < 0.0002, ∗∗∗∗*P* < 0.0001; one-way ANOVA followed by Tukey's test.

Supplementary Figure 2 | HIF1α is required for glycolytic activity in human Th9 cells. (A) Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the absence and presence of acriflavine (ACF) followed by relative mRNA expression of glycolytic genes examined by qPCR. (B) Sorted naïve T cells differentiated under Th0 and Th9 polarizing conditions for 6 days in normoxia (21% oxygen) and hypoxia (1% oxygen), respectively followed by mRNA

expression of glycolytic genes. Data are representative of mean ± SEM from three independent experiments (*n* = 3). <sup>∗</sup>*P* < 0.0332, ∗∗*P* < 0.0021, ∗∗∗*P* < 0.0002, ∗∗∗∗*P* < 0.0001; one-way ANOVA followed by Tukey's test (A), two-way ANOVA followed by Tukey's test (B).

Supplementary Figure 3 | Blocking glycolysis inhibits glycolytic genes in human Th9 cells. Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the absence and presence of 2-DG followed by examination of mRNA expression profile of glycolytic genes. Data are representative of mean ± SEM from three independent experiments (*n* = 3).

#### REFERENCES


<sup>∗</sup>*P* < 0.0332, ∗∗*P* < 0.0021, ∗∗∗*P* < 0.0002, ∗∗∗∗*P* < 0.0001; one-way ANOVA followed by Tukey's test.

Supplementary Figure 4 | Nitric oxide (NO) is crucial for enhanced glycolysis in human Th9 cells. Sorted naïve T cells were differentiated under Th0 and Th9 polarizing conditions for 6 days in the absence and presence of 2-DG followed by examination of mRNA expression profile of glycolytic genes. Data are representative of mean ± SEM from three independent experiments (*n* = 3). <sup>∗</sup>*P* < 0.0332, ∗∗*P* < 0.0021, ∗∗∗*P* < 0.0002, ∗∗∗∗*P* < 0.0001; one-way ANOVA followed by Tukey's test.


via IL-9 receptor signaling in intestinal epithelial cells. Nat Immunol. (2014) 15:676–86. doi: 10.1038/ni.2920


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

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

# CTL-Derived Exosomes Enhance the Activation of CTLs Stimulated by Low-Affinity Peptides

Shu-Wei Wu<sup>1</sup> , Lei Li <sup>1</sup> , Yan Wang<sup>2</sup> and Zhengguo Xiao<sup>1</sup> \*

*<sup>1</sup> Department of Animal and Avian Sciences, University of Maryland, College Park, MD, United States, <sup>2</sup> Department of Cell Biology and Molecular Genetics, University of Maryland, College Park, MD, United States*

Cytotoxic T cells (CTLs) bind to peptides presented by MHC I (pMHC) through T cell receptors of various affinities. Low-affinity CTLs are important for the control of intracellular pathogens and cancers; however, the mechanisms by which these lower affinity CTLs are activated and maintained are not well understood. We recently discovered that fully activated CTLs stimulated by strong-affinity peptides in the presence of IL-12 are able to secrete exosomes that, in turn, stimulate bystander CTLs without requiring the presence of antigen. We hypothesized that exosomes secreted by high-affinity CTLs could strengthen the activation of low-affinity CTLs. Naive OT-I CD8+ cells were stimulated with altered N4 peptides of different affinities in the presence or absence of Exo. The presence of Exo preferentially increased cell proliferation and enhanced the production of IFNγ in CTLs stimulated by low-affinity peptides. The expression of granzyme B (GZB) was augmented in all affinities, with higher GZB production in low-affinity stimulated CTLs than in high-affinity stimulated ones. Exosomes promoted the rapid activation of low-affinity CTLs, which remained responsive to exosomes for a prolonged duration. Unexpectedly, exosomes could be induced quickly (24 h) following CTL activation and at a higher quantity per cell than later (72 h). While exosome protein profiles vary significantly between early exosomes and their later-derived counterparts, both appear to have similar downstream functions. These results reveal a potential mechanism for fully activated CTLs in activating lower-affinity CTLs that may have important implications in boosting the function of low-affinity CTLs in immunotherapy for cancers and chronic viral infections.

#### Keywords: CTLs, IL-12, exosomes, activation, low-affinity, N4 peptides

# INTRODUCTION

Single positive CD4+ or CD8+ T cells mature in and migrate from the thymus following positive and negative selection to ensure this T cell pool remains self-restricted and non-autoaggressive (1, 2). Selection depends upon the affinity of the T cell receptor (TCR) for the peptide/MHC complex (pMHC) (2–7). Most CTLs (CD8+ T cells) are low-affinity (8, 9), but high-affinity CTLs are considered more essential to the immune response due to their more robust function and increased sensitivity to detection (10–12). The presence of CTLs with diverse affinities has been confirmed throughout the immune response (13, 14) via improved, more sensitive assays for detecting low-affinity CTLs (15–17). Of note, a similarly prominent existence of low-affinity

#### Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Karsten Sauer, Torque Therapeutics, United States Geeta Upadhyay, Uniformed Services University of the Health Sciences, United States*

> \*Correspondence: *Zhengguo Xiao xiao0028@umd.edu*

#### Specialty section:

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

Received: *16 December 2018* Accepted: *20 May 2019* Published: *04 June 2019*

#### Citation:

*Wu S-W, Li L, Wang Y and Xiao Z (2019) CTL-Derived Exosomes Enhance the Activation of CTLs Stimulated by Low-Affinity Peptides. Front. Immunol. 10:1274. doi: 10.3389/fimmu.2019.01274* polyclonal CD4+ T cell responses has also been reported (14, 18, 19). Low-affinity CTLs are important to fighting infection and malignant cells (12, 18, 20–22), particularly in the presence of multiple epitopes or where immune escape mutations occur (23). A greater breadth of recruited TCR affinities has been positively associated with improved host protection (12, 24). Low-affinity CTLs can become effectors despite the reduced magnitude of their immune activity compared to their highaffinity counterparts (25). Memory low-affinity CTLs are induced and maintained during infection (12, 26, 27) and can mount a robust recall response (28). How this low-affinity CTL response is initiated and maintained, however, is not wellunderstood (12, 27).

It has been elegantly demonstrated that affinity affects the kinetics of CTL expansion and contraction as well as egress from draining lymph nodes (12). Low affinity-primed CTLs expand to a lesser degree and contract earlier than high affinity-primed CTLs, and also exit lymphoid organs sooner and are released into circulation earlier in the adaptive immune response. These low affinity-primed CTLs may contribute to the early control of infection, whereas high affinity-primed CTLs are released later to take over the remainder of the CTL response (12). This was further supported by another recent report that presented evidence that low affinity-primed CTLs accumulate at efferent lymphatic vessels and are disseminated earlier than high affinityprimed ones, leading to rapid elimination of targets outside the lymph nodes (27). Low affinity-primed CTLs may be at least partially responsible for early control of microbial infections, serving as a critical part of the adaptive immune response together with their high-affinity counterparts (27).

TCR signaling may differ between low- and high-affinity CTLs. Reduced TCR affinity is generally associated with a reduced CTL response (12, 25, 27, 29–31). However, how the activation of CTLs is directly affected by TCR affinity remains controversial (5, 29, 30, 32–41). In a recent report, CTLs stimulated with peptides of different affinities nonetheless achieved a similar effector protein profile (42). The TCR signaling triggered by weak ligands may be different from that induced by strong ligands (43), as demonstrated by a unique pattern of ZAP-70 phosphorylation (44), representing an altered TCR activation pathway not explained by dose effects (44, 45). In addition, TCR affinity seems to affect transcription factor expression. Low affinity is associated with high eomes expression at high antigen doses (46), whereas strong TCR affinity reduces the ratio of Bcl6 to Blimp-1 and eomes to T-bet (46, 47). In addition, strong affinity induces higher expression, and low affinity induces reduced expression of both BATF and IRF4 (27). These reports suggest that low- and high-affinity CTLs may possess different TCR signaling pathways and may also be responding differently to other stimuli, such as exosomes.

Recently, we reported that antigen-stimulated CTLs secrete exosomes and that the presence of IL-12 changes their morphology and influences the enrichment of the proteins contained therein (48). More important, these IL-12 conditioned, CTL-derived exosomes can activate bystander naive CTLs without antigen stimulation (48). In this project, we examined the functions of these CTL-derived exosomes on CTLs stimulated with altered peptides of different affinities, using a simple OT-I cell in vitro stimulation model.

### MATERIALS AND METHODS

#### Purification of Naive OT-I CD8+ T Cells

OT-I mice were euthanized, and peripheral lymph nodes were collected. The harvested lymph nodes were homogenized in 15 mL glass grinders in Allos medium (49, 50). After washing with Allos medium several times and filtering through a 70µm nylon filter (VWR, Radnor, PA), cells were incubated together with FITC-labeled antibodies specific to B220, CD4, CD44, CD11c, and I-Ab for negative selection (Biolegend, San Diego, CA). The suspension was subsequently incubated with Anti-FITC conjugated magnetic MicroBeads (Miltenyi Biotech, Auburn CA) and passed through separation columns attached to a MACS magnet. Cells that did not bind to the column were collected, and their purity was confirmed (>95% CD8+ and <0.5% CD44hi cells).

#### Activation of Naive CTLs for Exosome Production

Flat-bottom Microtiter plates (Greiner bio-one, Frickenhausen, Germany) were coated with recombinant MHC I (DimerX H-2Kb: Ig fusion protein; BD Pharmingen, San Jose, CA) and the costimulatory molecule B7-1/Fc chimeric protein (R&D Systems, Minneapolis, MN) (49, 50). The coated plates were pulsed with N4 peptides. This MHC I/N4 plus B7-1 provided two signals (2SI): the first signal to the specific TCR expressed on the surface of OT-I CD8+ T cells, and the second signal (costimulation), thus designated as "2SI" stimulation. For 2SI stimulation, purified naive OT-I CD8+ T cells were placed at 3 × 10<sup>5</sup> cells in 1.5 mL Allos medium in each well of a 24-well plate with IL-2 at 2.5 U/mL. For three signal stimulation (3SI), naive OT-I CD8 T cells were stimulated with 2SI and supplemented with 2 U/mL of murine rIL-12 (R&D Systems, Minneapolis, MN), as previously described (48, 51). Supernatant from 2SI or 3SI stimulated CTLs was harvested three days after stimulation for exosome purification. Exosomes from 2SI were designated as "2SI-exo," whereas those from 3SI as "Exo" or "3SI-exo." D1 exosomes (D1E) were purified from the CTL supernatant after a one-day stimulation with 3SI.

#### Purification of Exosomes

Exosome-free medium was generated by centrifugation at 100,000 g overnight. Naive OT-I cells were seeded and incubated for 1 or 3 days, and then the cell supernatants were harvested for exosome purification. Briefly, cells were centrifuged at 300 g for 5 min remove cells and followed by 2,000 g for another 30 min to remove debris. The supernatant was collected and filtered through a 0.22µM filter (Corning, NY). Exosomes were precipitated by PEG6000 (Millipore, Darmstadt, Germany) overnight and pelleted by ultracentrifugation at 100,000 g twice for 70 min at 4◦C (Beckman Optima XPN-80, Beckman Coulter, Indianapolis, IN). The pellets were collected and washed with cold 1XPBS and followed by ultracentrifugation at 100,000 g twice for 70 min at 4◦C (Beckman Optima XPN-80, Beckman Coulter, Indianapolis, IN). Purified exosomes were examined for protein concentrations by Commassie plus Protein Assay Reagent (Thermo Scientific, Rockford, IL) and stored at −80◦C until use. Size distribution of exosomes was estimated by a Malvern Zetasizer Nano ZS90 (Malvern, UK) (48).

# Preparation of Cellular Fractions From 2SIor 3SI-Stimulated CTLs

Freeze-thaw lysis and sonication were used (52–54) 10 million 2SI- or 3SI-stimulated CTLs were harvested as a cell pellet three days after each stimulation. Each cell pellet was resuspended in 100 ul of cold 1xPBS, which was followed by three cycles of freeze-thaw on dry ice. To further disrupt cell structure, each sample was sonicated for 10 s on ice for six times with 30 s intervals between pulses at 20 kHz on a sonicator 350 (Plainview, NY). The treated sample was then resuspended into 5 mL (total volume), centrifugated at 2,000 g for 30 min at 4◦C. The pellet was resuspended in 100 µL 1xPBS and labeled "debris." Ten µL of "debris" was added to each well (96-well plate). The supernatant was filtered with a 0.22µm syringe filter (GVS, Sanford, ME), and flow through was collected and labeled as "soluble fraction." Protein concentrations were determined using BCA assay (48).

### Exosome Functions on Low- or High-Affinity, Peptide-Stimulated CTLs

Flat-bottom Microtiter plates (Greiner bio-one, Frickenhausen, Germany) were coated only with recombinant MHC I (DimerX H-2Kb: Ig fusion protein). The coated plates were pulsed with altered N4 peptides of different affinities, N4, A2, Y3, Q4, T4, and V4 (high to low affinity) (12). Purified naive OT-I CD8+ T cells were placed at 5 × 10<sup>4</sup> cells in 200 µL Allos medium in each well in a 96-well plate, with IL-2 at 2.5 U/mL, unless indicated otherwise. Purified exosomes were added into wells either at the beginning of or after stimulation at the indicated times, at a concentration of 33µg/mL as used in human T cells (55) and our previous report (48). Cells were harvested for assay at the indicated times after stimulation.

#### Transmission Electron Microscopy

Exosomes containing 0.03–0.3 ρg protein were suspended in 2% glutaraldehyde, applied to a Formvar-coated grid, and negatively stained with uranyl acetate. Electron microscopy was performed using a Zeiss EM10 transmission electron microscope at an accelerating voltage of 80 keV 26, as described previously (48).

#### Western Blot

An aliquot of 10 µg exosome proteins was separated by electrophoresis on a 10% SDS-polyacrylamide gel and analyzed by Western blot (48). First, the proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (BioRad, Hercules, CA) and blocked with 20 mM Tris-HCl, 150 mM NaCl, 0.05%, and pH 7.6 Tween-20 blocking solution (TBST) containing 1% bovine serum albumin (BSA). The membrane was incubated with the first antibody at room temperature (RT) for 1 h and then washed 3 times using TBST to remove excess antibody. The membrane was then incubated with horseradish peroxidase (HRP)-conjugated secondary antibody for 1 h at RT. Signals were detected with a SuperSignal West Pico Chemiluminescent Substrate (Thermoscientific, Rockford, IL) and a Gel Doc imaging system (Biorad, CA).

#### Cell Proliferation

Purified naive OT-I CD8+ cells were washed in Hank's Balanced Salt Solution (1 × HBSS) (Corning, Manassas, VA) and resuspended in 1 x HBSS at 1 × 10<sup>7</sup> cells/mL containing carboxyfluorescein diacetate succinimidyl diester (CFSE) for a final concentration of 0.5µM and incubated for 5 min at 37◦C before being transferred to cold Allos. Cells were then washed twice with Allos before plating (48).

#### Cell Staining and T Cell Activation Analysis

T cell activation markers were examined by flow cytometer (BD Biosciences, San Diego, CA) and analyzed using FlowJo software (FlowJo, Ashland, OR) (48); markers included CD25, IFNγ, and GZB. IFNγ expression was induced by incubating cells in RP-10 with 0.2µM OVA257−<sup>264</sup> peptide and 1 µL Brefeldin A (BioLegend, San Diego, CA) for 3.5 h at 37◦C. Cells were then fixed with 4% fixing buffer at a 1:1 ratio for 15 min at 4◦C, followed by permeabilization in saponin-containing Perm/Wash buffer (Biolegend, San Diego, CA) for another 15 min at 4◦C.

#### Killing Assay

The CellTiter-Glo <sup>R</sup> (CTG) killing assay is based on the number of viable cells left in the culture after cytotoxic T lymphocyte killing of the target cells (56). B16.OVA melanoma cells adhere to plastic surfaces and can efficiently present OVA257−<sup>264</sup> peptide; activated OT-I T cells recognize H-2K<sup>b</sup> /OVA257−<sup>264</sup> and initiate specific killing of these B16.OVA cells (57, 58). B16.OVA cells were seeded onto 96-well white plates at 30,000 cells/well in 100 uL Allos medium, and activated OT-I cells were added to each well as effectors to target cells (B16.OVA cells) at a ratio of 10:1. After overnight incubation, T cell suspensions (both OT-I cells and B16.OVA) were removed by washing three times with Allos medium. Luminescent signals (relative luminescent unit, RLU) from a 96-well plate was measured by the addition of 200 µL of 50% Cell Titer Glo (Promega, Madison, WI) followed by measurement of luminesce using a plate reader (Bio-Rad). The kill percentage of the B16.OVA cells by effector OT-I cells was calculated according to the following equation: Killed % = 100% × (RLU of untreated B16.OVA cells—RLU of B16.OVA cells cultured with OT-I cells)/RLU of untreated B16.OVA.

#### Proteomics

Exosomes containing 10 µg of proteins were incubated in 8 M urea at room temperature to disrupt their membranes. Tryptic digestion Exosomal proteins were reduced with DTT, alkylated with iodoacetamide, and digested with 0.5 µg Trypsin/LysC Mix (Promega, Madison, WI) at 35◦C, first at 4 M for 1 h, then further diluted with 0.8 M urea to activate the trypsin and incubated overnight. Tryptic digests were acidified with 2 µL TFA and desalted with C18 TopTip (Glygen Corp., Columbia, MD). Eluted peptides were vacuum-dried and dissolved in 35 µL solvent A (2.5% ACN, 0.1% formic acid in water). Peptide concentration was estimated using a Qubit 3.0 Fluorometer. LCMSMS analysis was carried out using a Dionex U3000 nanoHPLC system interfaced to a Thermo Scientific orbitrap Fusion Lumos mass spectrometer. Samples were analyzed in randomized order with a solvent blank between samples. For each sample, 1 µg of tryptic digest was injected into an AccalaimTM PepMapTM 100 trap column (5µm, 100 Å, 300µm × 5 mm), and desalted at 5 µL/min with 100% Solvent A for 5 min. The peptides were then separated with an Accalaim PepMapTM 100 nano column (3µm, 100 Å, 75µm × 250 mm) using a linear gradient of 2–52% solvent B (75% ACN, 0.1% formic acid) over 160 min. Precursor masses were detected in the Orbitrap at R = 120,000 (m/z 200). Fragment masses were detected by linear ion trap at unit mass resolution. Data dependent MSMS was carried out at a cycle time of 3 s. Dynamic exclusion was at 30 s. Protein identification and label-free quantification were carried out using Proteome Discoverer software (v. 2.2, Thermo Scientific), and mouse proteome was downloaded from Uniprot (uniprot.org) using both Sequest HT and Mascot search engines. M oxidation and NQ deamidation were set as variable modifications, and carbomidomethylation of C was set as the fixed modification. Precursor mass tolerance was 20 ppm, later filtered to 5 ppm in the report. Fragment mass tolerance was 0.6 Da.

### Enriched Pathway Analysis of Differentially Expressed Genes

Data were uploaded into the Ingenuity Pathways Analysis (IPA) software (Ingenuity Systems, http://www.ingenuity.com). The IPA database is maintained and edited by humans and contains genes, proteins, and RNA not only to find associations between expression data and canonical pathways but also build new networks. The significance of associations was computed using the right-tailed Fisher's exact test. All signaling pathways identified by IPA with a P-value of less than 0.05 have a statistically significant, nonrandom association.

#### Statistical Analysis

We used an unpaired, two-tailed Student's t-test in GraphPad (Prism 5.0 software; GraphPad Prism, La Jolla, CA, USA) for statistical analysis of significance.

# RESULTS

#### Exosomes Preferentially Enhance CTL Proliferation Stimulated by Low-Affinity Peptides

Peptide/TCR affinity is the main contributor to CTL responses, so we first tested the effects of CTL-derived exosomes on OT-I cells stimulated by peptides of different affinities. Lowaffinity peptides alone were able to drive the proliferation of naive CTLs, albeit at a slower rate (**Figure 1A**), consistent with previous reports (6, 13, 48). OT-I T cells treated with intermediate- and higher-affinity peptides (peptides N4, A2, Y3, and Q4) underwent similar rounds of cell cycles after a 48 h stimulation period. Low-affinity peptide stimulation (peptides T4 and V4) resulted in fewer cells entering the cell cycle; most remained undivided with peptide V4 (**Figure 1A**). The presence of exosomes (Exo) from the supernatant of fullyactivated CTLs (48) led to a 2- to 3-fold increase in final cell numbers following 2 days of stimulation with low-affinity peptides (T4 and V4) in comparison to corresponding peptideonly controls; this was not recapitulated with intermediate or high peptide affinities (**Figure 1B**). These final differences from low-affinity peptides were similarly reflected by CFSE dilution. The addition of Exo to low-affinity peptide T4 stimulation drove OT-I cells into further divisions than T4 alone, whereas Exo did not change significantly in the dividing of CTLs stimulated by intermediate or high affinity peptides Y3 and N4 (**Figure 1C**). Moreover, Exo pushed more cells into the cell cycle during low-affinity peptide (T4) treatment, and more of these dividing cells entered division 3 by reducing the number of cells in division 1 (**Figure 1D**). Thus, Exo preferentially enhanced the proliferation of CTLs stimulated with low-affinity peptides, with no such effect on intermediate- or high-affinity, peptidestimulated cells.

# Exosomes Preferentially Enhance the Expression of IFNγ in Low-Affinity CTLs

IFNγ is critical for early protection against infections (59– 62) as well as important for CTL function (63–65). IFNγ was barely detectable in any of the peptide-only treatments (**Figure 2B**), consistent with the fact that IFNγ production requires the presence of third signal cytokines such as IL-12 (50, 65). The presence of Exo slightly elevated IFNγ production in high affinity-stimulated CTLs (**Figures 2A,B**). Interestingly, IFNγ production increased when affinity decreased, with the highest production occurring in response to low-affinity peptides T4 and V4 (**Figures 2B,C**), suggesting a functional preference of Exo for low-affinity CTLs. Exosomes have been reported to mediate transcription signals during the immune response (66). In all treatments with Exo, the expression of T-bet (67) increased, with the most significant increase in the presence of low-affinity peptides; no effects were observed on eomes expression (**Figure 2D**). Exo can therefore preferentially enhance IFN-γ production in low affinitystimulated CTLs, and this is positively associated with Tbet expression. However, this preference was not the case in GZB regulation by exosomes. First, GZB production was higher in low affinity-stimulated CTLs than in high affinitystimulated CTLs (**Figures 3A,B**), consistent with a recent report using an in vivo system (27). Second, Exo increased GZB production at all affinities, with low-affinity-stimulated CTLs producing the highest total amounts of GZB (**Figures 3B,C**). In addition, killing ability was generally low for all affinities, and the addition of Exo enhanced the killing ability slightly, but significantly, only in low affinity-stimulated (T4) OT-I cells (**Supplementary Figure 1**). The expression of CD25 in low affinity-stimulated OT-I cells was higher than that in high-affinity stimulated OT-I cells, and the addition of Exo increased CD25 expression at all affinities (**Figures 3D,E**), without significant differences among them (**Figure 3F**). It thereby seems that exosomes preferentially enhance the activation of low affinitystimulated CTLs.

FIGURE 1 | Exosomes preferentially enhance CTL proliferation stimulated by low-affinity peptides. Purified naive OT-I cells were labeled with CFSE and stimulated with peptides in the presence or absence of Exo. Peptides ranged from high affinity (peptide N4) to low affinity (peptide V4). (A) Representative CFSE dilution in OT-I cells stimulated for 2 days by peptides with diverse affinities presented by plate-bound recombinant MHC I (Dimer X). (B) Fold change caused by CTL-derived exosomes (Exo) in the cell number of OT-I cells compared to peptide-only controls. Comparisons were based on Student's *t*-test between T4 and each of other peptides. (C) Representative histograms. Black lines: peptide only. Red lines: peptide plus Exo. (ND: undivided cells; D1: cells divided once; D2: second division; D3: third division). (D) Effects of Exo on divisions of OT-I cells stimulated by different affinities. Data in (A,C) are representatives of at least 5 experiments. Asterisks indicate statistical significance. \**P* < 0.05; \*\**P* < 0.01; \*\*\**P* < 0.001 by unpaired, two-tailed Student's *t*-test, which will be the same in the rest of this study.

without the Exo for 2 days. (A) Representative dot plots and gating for IFN-γ expression. (B) Representative data on the production of IFN-γ from a set of peptides with altered affinities. (C) Pooled data from multiple experiments on the production of IFN-γ. (D) Representative histograms of T-bet/eomes expression affected by Exo. Black lines: peptide only. Red lines: peptides plus Exo. Statistics were based on Student's *t*-test. Data in (A,B) are representatives of at least 5 experiments.

#### Exosomes From Partially Activated CTLs Fail to Activate Low Affinity-Stimulated CTLs

We found that exosomes derived from 2SI stimulation (antigen + costimulation) did not activate bystander CTLs (48). To test if these 2SI-exo had any effects on low-affinity stimulated CTLs, purified OT-I cells were stimulated with low-affinity peptide T4 in the presence or absence of 2SI-exo or 3SIexo for 2 days. In contrast to 3SI-conditioned exosomes (3SI-Exo, the same as Exo), 2SI-exo had no effect on low-affinitystimulated CTLs, demonstrated by unaltered expression patterns of IFNγ/GZB/CD25/T-bet, and instead appeared to inhibit their division (**Figure 4**). IL-2 is important for CTL activation (68, 69). To test if IL-2 played a role in exosome (3SI-exo) effects, purified OT-I cells were stimulated with low-affinity peptide T4 in the presence or absence of Exo and/or IL-2 neutralizing antibodies (70). The depletion of IL-2 completely abolished the effects of Exo as demonstrated by diminished expression of IFNγ/GZB/CD25 to a level even below T4 only controls (**Supplementary Figure 2**), suggesting that IL-2 is important to both peptide stimulation and exosome effectiveness. Of interest, despite the fact that IL-2 neutralization greatly reduced CTL activation at Y3 (intermediate affinity) and N4 (high affinity), the cell cycle progression was suppressed but not stopped (**Supplementary Figure 2**), suggesting IL-2 may not be definitely required for the cell cycle progression of CTLs.

The standard protocol for exosome purification is based on differential centrifugation and participation, which should remove most cell debris, but small contamination is still possible. To test the effects of potential contaminants in purified exosomes, cell debris and soluble fractions, another potential contaminant in exosomes, from activated CTLs, were prepared and added to high- or low-affinity stimulated CTLs, at a high ratio (debris from 20 cells to one stimulated cell) or protein concentration (5-fold of exosomes). When CTLs were stimulated with high-affinity N4, cell debris from both 2SI and 3SI inhibited GZB production, with minimal effects on proliferation (**Supplementary Figure 3**). In low-affinity stimulation, GZB production was also decreased by the presence of cell debris, but while 2SI-debris seemed to inhibit CTL proliferation, 3SI-debris did not (**Supplementary Figure 3**). Surprisingly, the

FIGURE 3 | Exosomes induce the strongest activation in low-affinity CTLs. Naive OT-I cells were labeled with CFSE and stimulated with peptide with or without Exo for 2 days. (A,D) Representative dot plots and gating for granzyme B (GZB) (A) and CD25 (D) expression. (B,E) Representative data on the production of GZB (B) or CD25 (E) from a set of peptides with altered affinities. (C,F) Pooled data from multiple experiments on the production of GZB (C) and the expression of CD25 (F). Statistics were based on Student's *t*-test. Data in (A,B,D,E) are representatives of at least 5 experiments.

Naive OT-I cells were labeled with CFSE and stimulated with peptide with or without two different types of exosomes for 2 days. 3SI-exo or 2SI-exo: exosomes secreted by 3SI-stimulated or 2SI-stimulated CTLs (48). Data are representatives of at least 4 experiments.

soluble fraction from 3SI-CTLs did enhance GZB production, but dramatically inhibited cell proliferation, in both high- and low-affinity stimulation conditions (**Supplementary Figure 3**). The effects of soluble fractions were dose-dependent and became undetectable at a concentration of 1/5 exosomes (**Supplementary Figure 4**). These data suggest that the effects of potential contaminants are different from the effects of exosomes, and the level of contamination is unlikely to be as high as was tested in this experiment. Thus, the functions of exosomes (**Figures 1**, **2**) are likely to be due to the exosomes, not potential contamination from exosome-producing cells.

#### CTL-Derived Exosomes Enhance Early Activation of Low-Affinity-Stimulated CTLs

To test if the activation of low affinity-stimulated CTLs was accelerated by CTL-derived exosomes, Exo was provided at the beginning of stimulation. IFNγ and GZB, which were undetectable at 12 h, were enhanced by Exo after 24 h (**Figure 5** and data not shown). Most of these IFNγ-producing cells were also GZB+ (data not shown). The positive control peptide T4+costimulation (B7-1)+IL-12 (3SI) resulted in the highest production of both molecules at 24 h, whereas T4+costimulation (B7-1) (2SI) only induced low levels of GZB and IFNγ, suggesting that third signal cytokines may contribute to the activation of low-affinity CTLs in a similar pattern to but at lower levels than high-affinity CTLs (50, 65). Interestingly, T-bet was detectable in T4-stimulated CTLs and enhanced by Exo to a level close to that of 3SI (**Figure 5**), suggesting that this IFNγ enhancement by Exo may be regulated through T-bet (71, 72). CD25 expression was also increased by Exo, although this enhancement was only detectable after 24 h (**Figure 5** and data not shown). Interestingly, CD25 expression peaked at 24 h (**Figure 5**), and started to decline at 48 h after stimulation (**Figure 4**). These data demonstrate that CTL-derived exosomes can enhance the early

activation of low affinity-stimulated CTLs in a pattern similar to 3SI.

# Low-Affinity CTLs Remain Responsive to Exosomes for a Prolonged Period of Time

The time points of available Exo for low-affinity-stimulated CTLs may vary in vivo. To examine how low-affinity primed CTLs respond to Exo, primed CTLs were exposed to Exo at different time points after priming. Exo were added for the final 6 or 12 h during a total incubation time of 72 h, and exosomes added at the last 12 h were able to enhance expression of IFNγ, GZB, and CD25 (**Figure 6**), although to lower levels of IFNγ compared to a full 48 h exposure (**Figure 2**). Interestingly, exposure to exosomes for just 6 h enhanced IFNγ, but decreased GZB, suggesting differential responses to exosomes by low-affinity CTLs primed at different times. Nevertheless, the swift responsiveness of primed CTLs suggests that CTL-derived exosomes may perform posttranscriptional regulatory functions, possibly via regulatory or signaling proteins.

# Activated CTLs Secrete Exosomes at Early Stages of Activation

We next explored how quickly high-affinity CTLs could produce exosomes after being activated. Naive OT-I cells were stimulated with 3SI (with high-affinity peptide N4) (48, 50) and the supernatant was harvested for exosome isolation at days 1 (24 h) and 3 (72 h) after activation. Exosomes were smaller at day 1 (D1E) than at day 3 (D3E) (**Figure 7A**). Most classical exosomal proteins were not detectable in D1Exo, such as Tsg101 and Alix, while flotillin was detected at variable levels among the different batches (**Figure 7B**). The yield of exosomes in D1E was about 40% that of D3E (**Figure 7A**), but time for exosome production was two-thirds shorter. The D3E were derived from many more cells (8- to 10-fold after massive division) than the D1E from undivided CTLs. Thus, early-activated effector CTLs produced more exosomes than late effectors on a per cell basis. Importantly, D1E induced similar levels of IFNγ and GZB as D3E in low affinity-stimulated CTLs (**Figure 7C**). It appears that exosome secretion can begin soon after activation, and despite their differences in protein profiles (**Figure 7B**), D1E and D3E were similarly effective in enhancing the function of low affinitystimulated CTLs.

#### Early and Late CTL-Derived Exosomes Share Effector Functions Based on Protein Profiles

We examined the protein profiles of exosomes derived from CTLs activated in the presence of IL-12 together with antigen/costimulation for either 24 h (D1E) or 72 h (D3E) (**Figure 7**). A total of 2097 proteins were identified from three biological replicates at each time point (**Supplementary Table 1**). Despite the fact that most proteins were detected in both exosome populations, there were substantial quantitative differences. Using 1.5-fold as the cutoff, 467 proteins were enhanced (or unique) in D1E (**Supplementary Table 2**), and 233 in D3E (**Supplementary Table 3**). We next explored how these skewed (differential) protein profiles were associated with downstream effector function using Ingenuity Pathway Analysis (IPA) (73). The two different profiles were predicted to relate to similar biological functions, such as enhancing the activation, movement of cells, inhibiting apoptosis, and cell death (**Table 1**). This result is consistent with functional data (**Figure 7**) and suggests that

FIGURE 6 | Low-affinity-primed CTLs are sensitive to CTL-derived exosomes. Naive CTLs were stimulated with low-affinity peptide T4/MHC I for a total of 72 h. Exosomes were added to stimulated cells at 60 or 66 h after stimulation, indicated as final 12 or 6h.

TABLE 1 | Predicted shared effector functions of differential exosomal proteins from CTLs stimulated for 1 and 3 days *in vitro*.


*Differential proteins either in D1E (*Supplementary Table 2*) or D3E (*Supplementary Table 3*) were analyzed using Ingenuity Pathway Analysis (IPA).*

the exosomes generated by early and late effectors may contain different proteins but induce similar functions.

#### DISCUSSION

Low-affinity CTLs are critical components of the immune response. We found that fully activated CTLs can secrete exosomes that preferentially enhance the activation of low-affinity CTLs, suggesting potentially interconnected communication between fully activated, high-affinity CTLs, and low-affinity CTLs through exosome secretion with several noticeable features. First, the original exosome-secreting CTLs required 3SI stimulation; 2SI stimulation-induced exosomes failed to enhance the activation of low-affinity CTLs. Second, this communication begins early, as a large quantity of functional exosomes were secreted shortly (24 h) after 3SI stimulation. Third, although we cannot exclude the possible function of shared common molecules in both early and late CTL-derived exosomes, pathway analysis suggests that the differential proteins between exosomes from these two stages share common downstream effector functions. Finally, lowaffinity CTLs can respond to exosomes for a prolonged period of time.

Although studied in a simplified in vitro system in this project, these data may recapitulate similar communication among CTLs of different affinities in animals. The output by the thymus of naive CTLs with varying affinities is a continuous process; thus, they may not be activated at the same time. Infection alters inflammation profiles and kinetics, and the necessary third signal cytokines may not be equally

available to CTLs in different tissues (74). Low-affinity CTLs may also be exposed to exosomes at different time points after priming. Our data suggest a potential communication through exosome secretion between fully activated, high-affinity CTLs and low-affinity CTLs, which needs to be further tested in animals.

Exosomes can be detected at relatively high concentration (0.1–20 × 10<sup>9</sup> particles/mL) in human plasma (75–77), with many potential sources such as epithelial and immune cells (78, 79). Mature mouse dendritic cells can produce exosomes at about 10µg/mL in the supernatant after 48 h culture (80). After stimulation with 3SI for 1 day, CTLs can produce exosomes at 0.9µg/mL in their supernatant from 3 × 10<sup>5</sup> cells (**Figure 7A**), equal to 30µg/mL based on 10<sup>7</sup> cells, suggests that activated CTLs may be an important source of exosome production. Exosomes have been mostly investigated as biomarkers for diagnosis and drug delivery (78, 79), but recently, some clinical trials have been carried out to test the function of in vitrogenerated exosomes in cancer patients, using relatively low doses based on the number of MHC II molecules detected in exosomes (81, 82). Despite differences among disease models (82), preclinical experiments in mice mostly used an exosome dose in the range of 0.1–50 µg/mouse, administered mostly subcutaneously; a higher dose may be required for a dermatology model (82). Nevertheless, it seems CTL-derived exosomes may be important under certain physiological conditions, which we will further examine in animals.

The mechanisms that underlie the preference of CTL-derived exosomes for low-affinity CTLs are not known. TCR signaling triggered by weak ligands appears different from that triggered by strong ones (43) and cannot be explained by dose effects (44, 45). One explanation may be the skewing of low-affinity CTL activation toward a profile similar to 3SI stimulation, but different from high-affinity stimulation. Mechanisms for primed low-affinity CTLs may also be time-dependent. Initial exposure to exosomes induced an activation status in primed low-affinity CTLs resembling 3SI stimulation, whereas late exposure led to enhanced IFNγ but reduced GZB expression. These data indicate, at least conceptually, that the mechanisms for exosome activation of low-affinity CTLs may not be a simple switch from A to B, but rather a status-based response potentially driven by different exosomal proteins. In addition, despite the similarity in size and expression of exosomal markers, exosomes are generally heterogenous (82); thus, the effects could be induced by different proteins in the same exosomes, or, different exosomes from the same population. Pathway manipulation, at least in vitro, may be useful in narrowing down the number of potential targets; this is ongoing in our lab.

Exosomes from fully activated CTLs can preferentially enhance the function of low-affinity CTLs to a certain level, but not to the level of fully activated CTLs stimulated by 3SI. This was demonstrated by activation markers, effector molecule expression, and killing ability assays. We speculate two possible functions for exosome-facilitated communication between highaffinity and low-affinity CTLs. Either this communication can transform a functional heterogenous CTL population against one pathogen or cancer to a more homogeneous population, or exosomes secreted by fully activated CTLs may not influence direct killing but rather support general immune regulatory function through critical factors like IFNγ, thus affecting multiple cell types, including CD4 T cells.

It is important to note that the presence of IL-2 is required for the function of exosomes, with implications for the initiation of the immune response and autoimmune disease. During infection or an anti-tumor immune response, strongly activated CTLs in the presence of IL-12 and IL-2 will secrete exosomes quickly to promote the function of a robust population of lowaffinity CTLs. When IL-2 is no longer available, such as when the infection is resolved, exosomes can no longer affect lowaffinity CTLs. If fully activated CTLs can be induced to secrete functional exosomes in humans in ways similar to mice, these functional exosomes could be injected together with IL-2 to boost low-affinity CTLs in immunotherapy for chronic infections and cancers. T cell-derived exosomes are found to be able to induce the production of inflammatory factors such as IL-6, IL-8, and MCP-1 (83). Because most autoantigens target low-affinity CTLs, the secretion of stimulatory exosomes from fully activated CTLs is a legitimate concern in the induction of autoimmunity. Based on the definitive IL-2 requirement for the function of exosomes on low affinity CTLs, we speculate that this concern is limited to the period of IL-2 production. Conversely, if CTLderived exosomes are indeed involved in autoimmune disease, neutralization of IL-2 could become an effective method to dampen exosome function.

Communication between high-affinity and low-affinity CTLs might occur through secretion of different exosomes. Cells may adopt different methods to generate and protein-load exosomes in response to diverse environments. Although both early and late stimulations can induce abundant exosome production in CTLs, the sizes of the vesicles and types of proteins contained therein differ noticeably, suggesting that the duration of stimulation directs both exosome formation and its content. The production of exosomes follows one of two pathways, ESCRT (Endosomal Sorting Complex Required for Transport)-independent or ESCRT-dependent (84) modalities. Tsg101 and Alix are major components of the ESCRT-dependent pathway (85). A lack of these proteins in early exosomes suggests that early-stimulated CTLs might utilize ESCRT-independent pathways to generate exosomes, whereas late exosome formation in CTLs follows a more measured, ESCRT-dependent route. It is also not clear if the late-isolated exosomes represent an accumulation of total exosomes from early and late stimulations. Further experiments will be necessary to unpack specific mechanisms of CTL exosome secretion at different time points.

While there are undoubtedly many factors involved in the initiation and maintenance of the low-affinity CTL immune response, our data indicate that exosomes secreted by fully-activated CTLs could preferentially enhance the activation of CTLs stimulated by low-affinity peptides, thus providing the first evidence that CTL-derived exosomes could contribute to a previously unappreciated communication between fully activated, high-affinity CTLs and low-affinity CTLs (**Figure 8**).

#### REFERENCES


#### ETHICS STATEMENT

Mice were maintained under specific pathogen-free conditions at the University of Maryland, and these studies have been reviewed and approved by the Institutional Animal Care and Use Committee.

#### AUTHOR CONTRIBUTIONS

ZX conceived the study and coordinated the study. ZX, S-WW, YW, and LL designed, performed, and analyzed the experiments and wrote the manuscript.

#### ACKNOWLEDGMENTS

We would like to thank Dr. Jameson (UMN) for the scientific discussion that initiated this project. We also thank Robert Joseph Bonenberger Jr., Tim Maugelfor, and Ken Class for technical assistance. We acknowledge support from the following sources: GRANT11885997 (USDA), GRANT12684386 (USDA), MAES Competitive Program (UMD) and 1R01GM131054 (NIH). Part of the data in this manuscript was from the thesis of S-WW in supervision of ZX: **Figures 1A,C**, **2A,B**, **3B,E**, **7A,B**.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01274/full#supplementary-material


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

Copyright © 2019 Wu, Li, Wang and Xiao. 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.

# Targeting Adenosine in Cancer Immunotherapy to Enhance T-Cell Function

Selena Vigano1†, Dimitrios Alatzoglou1†, Melita Irving<sup>1</sup> \* † , Christine Ménétrier-Caux <sup>2</sup> , Christophe Caux <sup>2</sup> , Pedro Romero<sup>3</sup> and George Coukos <sup>1</sup> \*

<sup>1</sup> Department of Oncology, Ludwig Institute for Cancer Research Lausanne, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, <sup>2</sup> Department of Immunology Virology and Inflammation, INSERM 1052, CNRS 5286, Léon Bérard Cancer Center, Cancer Research Center of Lyon, University of Lyon, University Claude Bernard Lyon 1, Lyon, France, <sup>3</sup> Department of Oncology, University of Lausanne, Lausanne, Switzerland

#### Edited by:

Anil Shanker, Meharry Medical College, United States

#### Reviewed by:

Per Thor Straten, Herlev Hospital, Denmark Mikhail M. DIkov, The Ohio State University, United States

#### \*Correspondence:

Melita Irving melita.irving@unil.ch George Coukos george.coukos@chuv.ch

†These authors have contributed equally to this work

#### Specialty section:

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

> Received: 12 January 2019 Accepted: 10 April 2019 Published: 06 June 2019

#### Citation:

Vigano S, Alatzoglou D, Irving M, Ménétrier-Caux C, Caux C, Romero P and Coukos G (2019) Targeting Adenosine in Cancer Immunotherapy to Enhance T-Cell Function. Front. Immunol. 10:925. doi: 10.3389/fimmu.2019.00925 T cells play a critical role in cancer control, but a range of potent immunosuppressive mechanisms can be upregulated in the tumor microenvironment (TME) to abrogate their activity. While various immunotherapies (IMTs) aiming at re-invigorating the T-cell-mediated anti-tumor response, such as immune checkpoint blockade (ICB), and the adoptive cell transfer (ACT) of natural or gene-engineered ex vivo expanded tumor-specific T cells, have led to unprecedented clinical responses, only a small proportion of cancer patients benefit from these treatments. Important research efforts are thus underway to identify biomarkers of response, as well as to develop personalized combinatorial approaches that can target other inhibitory mechanisms at play in the TME. In recent years, adenosinergic signaling has emerged as a powerful immuno-metabolic checkpoint in tumors. Like several other barriers in the TME, such as the PD-1/PDL-1 axis, CTLA-4, and indoleamine 2,3-dioxygenase (IDO-1), adenosine plays important physiologic roles, but has been co-opted by tumors to promote their growth and impair immunity. Several agents counteracting the adenosine axis have been developed, and pre-clinical studies have demonstrated important anti-tumor activity, alone and in combination with other IMTs including ICB and ACT. Here we review the regulation of adenosine levels and mechanisms by which it promotes tumor growth and broadly suppresses protective immunity, with extra focus on the attenuation of T cell function. Finally, we present an overview of promising pre-clinical and clinical approaches being explored for blocking the adenosine axis for enhanced control of solid tumors.

Keywords: adenosine, cAMP, CD73, CD39, cancer immunotherapy, T cells, tumor microenvironment

# INTRODUCTION

IMT has led to unprecedented clinical success for some advanced cancer patients and has been accepted as a new pillar of cancer therapy (1). Thus, the identification of biomarkers predicting response to IMT, as well as the development of combinatorial strategies for increasing its effectiveness in more patients, and against a broader range of tumor-types, have become important areas of research (2). The nucleoside adenosine, involved in the regulation of multiple diverse physiological processes either as an intracellular metabolite of nucleic acid synthesis and energy-charge regulation or as an intercellular messenger in neurological, cardiovascular and immunological systems, has recently emerged as a major immuno-metabolomic checkpoint in tumors (3). Conditions of stress, such as hypoxia, lead to the accumulation of extracellular adenosine, predominantly derived from enzymatic ATP catabolism, which can act directly on tumor cells expressing adenosine receptors to promote their growth, survival and dissemination. In addition, adenosine, which under physiological conditions serves as an immuno-regulatory molecule to protect normal tissues from uncontrolled inflammation, can impair antitumor immunity, both through the attenuation of protective immune cells including T cells, NK cells, and dendritic cells (DCs), and by enhancing the suppressive capacity of T regulatory cells (Tregs), and myeloid-derived suppressor cells (MDSCs), amongst others. Here we review the targeting of the adenosine pathway to promote immune function and tumor control, with focus on T-cell activity, important experimental findings and an overview of clinical testing.

#### REGULATION OF ADENOSINE LEVELS IN HEALTHY vs. MALIGNANT TISSUE

Extracellular adenosine, a nucleoside and derivative of ATP, is involved in the regulation of diverse physiological processes including vasodilation (4), kidney-exerted water reabsorption (5), pain perception (6), and fine-tuning of the sleep–wake cycle (7). Even though levels of extracellular adenosine within healthy tissues are negligible (8–11), upon injury this nucleoside sharply accumulates at the interstitium where it potently restricts immune responses (12) and directly promotes wound healing (13). Under homeostatic conditions in healthy tissues, the cytosolic concentration of ATP ranges from 1 to 10 mM (14), while its extracellular levels are negligible (15). This sharp gradient can be rapidly disrupted however upon breaches of the plasma membrane induced by necrosis, apoptosis or mechanical stress, as well as by regulated ATP efflux. The latter, induced by a variety of stimuli including hypoxia, ischemia and inflammation, has been shown to extensively occur via exocytosis, transmembrane transfer through ATPbinding cassette (ABC) transporters, as well as by diffusion through a variety of anion channels or non-selective plasma membrane pores formed by connexins, pannexin-1 or the ATP receptor P2X7R (16–18). For instance, stimulated T cells release ATP through pannexin-1 hemi-channels and via exocytosis (19, 20).

Once in the extracellular space, ATP undergoes rapid stepwise dephosphorylation by ecto-nucleotidases (21, 22) including the E-NTPDase CD39, which converts ATP or ADP to ADP or AMP, respectively, and the 5′ -nucleotidase CD73, which dephosphorylates AMP to adenosine (18, 23) (**Figure 1**). Additional enzymes whose ecto-activity contributes toward extracellular adenosine generation are other E-NTPDases, members of the ecto-phosphodiesterase/pyrophosphatase (E-NPP) family, nicotinamide adenine dinucleotide (NAD+) glycohydrolases, the prostatic acid phosphatase (PAP), and the alkaline phosphatase (ALP) (21) (**Figure 1**). Briefly, the co-enzyme NAD+, another key cellular component whose extracellular concentration significantly rises in injured tissue (24, 25), is converted to adenosine diphosphate ribose (ADPR) by the NAD<sup>+</sup> glycohydrolase CD38 (26), while ADPR as well as ATP are metabolized to AMP by the E-NPP CD203a (27). Moreover, PAP, which is predominantly, but nonexclusively, expressed in prostate tissue (28), is capable of converting extracellular AMP to adenosine (29), whereas ALP catalyzes the hydrolysis of ATP, ADP and AMP to adenosine (21). Finally, adenosine can also be produced intracellularly either by S-adenosylhomocysteine hydrolase (SAHH)-exerted hydrolysis of S-Adenosylhomocysteine (SAH), a metabolite of the transmethylation pathway, or due to soluble CD73-mediated catabolism of AMP, a nucleoside participating in multiple cellular processes and whose concentration rises within cells of low energy charge (30) (**Figure 1**). Intracellularly-generated adenosine can be secreted in a diffusion limited-manner through bidirectional equilibrative nucleoside transporters (ENTs) (31). However, although there is evidence suggesting that hypoxia can boost intracellular adenosine production (32, 33), the contribution of this pathway toward injurycaused interstitial adenosine buildup is considered minor due to concurrent hypoxia-induced downregulation of the aforementioned transporters (34, 35). Given its diverse effects, adenosine presence at the extracellular space is subject to tight spatiotemporal control (12, 13, 36). For instance, extracellular accumulation of adenosine is counteracted by its inward transfer through ENTs or concentrative, sodium gradient-dependent, symporters (31) as well as by the function of intra/extracellular adenosine deaminase (ADA) and of cytosolic adenosine kinase (ADK), which respectively convert adenosine to inosine or AMP (37) (**Figure 1**).

In contrast to homeostatic conditions, ATP levels are highly elevated in the TME as a result of necrosis, apoptosis, hypoxia, and persistent inflammation (17, 18), and intra-tumoral adenosine levels can reach micromolar concentrations (9, 10, 38). ATP catabolism in tumors is primarily mediated by CD39 and CD73 (39–41), and high expression of these ecto-nucleotidases is strongly associated with poor clinical outcome for patients suffering a variety of cancer-types (3, 42, 43). In particular, CD39 and/or CD73 (over)expression has been detected on the surface of tumor cells (39, 44–51), cancer-associated fibroblasts (CAFs) (52–54), mesenchymal stem cells and stromal cells (55–57), endothelial cells (ECs) (45, 46, 51), myeloid derived suppressor cells (MDSCs) (58–60), tumor associated macrophages (TAMs) (53, 61), Tregs (46, 62–64), Th<sup>17</sup> cells (65) and of antigen experienced/exhausted conventional CD4<sup>+</sup> and CD8<sup>+</sup> T cells (64, 66–68). In addition, CD39/CD73-bearing exosomes (69, 70), released by tumor cells (71), Tregs (72), and MDSCs (57, 73) further contribute to adenosine generation. Currently, hypoxia as well as incessant inflammation are considered to be the main drivers of intra-tumoral CD39 and CD73 overexpression. Namely, hypoxia-induced (74, 75) HIF1α (76–79) and Sp1 (80) activity promotes expression of these ecto-nucleotidases. Along the same lines, signaling pathways initiated by inflammationassociated molecules, such as IL-2 (81), IL-6 (66, 82), IL-1β (83), TNFα (83–85), type I IFNs (86, 87), IL-27 (66, 88), TGFβ (82, 89, 90) as well as by inducers of the Wnt (91, 92) or cAMP (83, 93–95) signaling pathways also boost CD39 (66, 81, 82, 88, 89, 95) and CD73 (81–87, 89–94) levels.

catabolism of AMP, and it can be exported by ENTs in a diffusion-limited manner. On the flip side, the combination of CD26-bound ADA activity and of adenosine cellular uptake, either through equilibrative ENTs or via concentrative CNTs, limits interstitial adenosine levels. Intracellularly, adenosine can be eliminated via its conversion to SAH by SAHH, to AMP by ADK, or to inosine by ADA. SAHH, S-adenosylhomocysteine hydrolase; SAH, S-Adenosylhomocysteine; ENTs, equilibrative nucleoside transporters; CNTs, concentrative nucleoside transporters; ADK, adenosine kinase; ADA, adenosine deaminase.

Although CD39 and CD73-mediated catabolism of extracellular ATP is considered to account for the bulk of intra-tumoral adenosine generation, expression levels of ectoenzymes participating in alternative adenosine production pathways also rise in the advent of cancer. For instance, CD38 is frequently upregulated within neoplastic tissues (26, 96, 97) and sporadic evidence suggests that CD203a levels also increase on TME components (98, 99). Along the same lines, the serum concentration of PAP increases during prostate cancer progression (100) while others suggest it gets upregulated on cancerous tissue as well (28). Finally, several studies have demonstrated elevated levels of ALP on cancer cells (101, 102) as well as a correlation of serum ALP levels and disease stage (103– 105). Critically, the relative contribution of these alternative adenosine-producing pathways toward intra-tumoral buildup of this nucleoside remains to be determined. Finally, along with aberrant production, defective uptake resulting from the downmodulation of equilibrative (106, 107) as well as concentrative (108–110) nucleoside transporters, also driven by hypoxia (34, 35, 111), further contributes to adenosine accumulation in the TME.

# ADENOSINE RECEPTOR SIGNALING

Four adenosine receptors (ARs), all coupled to G-proteins, have been identified; A1R, A2AR, A2BR, and A3R (112, 113). While A1, A2A, and A3 are described as high affinity adenosine receptors (EC<sup>50</sup> in the range of 0.1–0.7µM), A2BR is considered as low affinity because it is activated only in the presence of high concentrations of adenosine (EC<sup>50</sup> of 15–25µM), such as may be found in the TME or under other pathological conditions. Upon adenosine binding, these GPCRs induce the replacement of GDP bound by the heterotrimeric G proteins, a class of GTP hydrolases, with GTP thus resulting in the dissociation of the latter into Gα monomers and Gβγ dimers, now free to modulate downstream effectors before their GTP hydrolysis-induced reassociation (114).

Of the four classes of Gα proteins characterized to date, namely Gα<sup>s</sup> , Gα<sup>i</sup> , Gαq/11, Gα12/13, only Gα<sup>s</sup> and Gα<sup>i</sup> directly influence the activity of adenylyl cyclases (AC), enzymes that catalyze the cyclization of intracellular ATP into cyclic adenosine monophosphate (cAMP) (114). In terms of function, triggering of the Gαs-coupled A2AR and A2BR promotes AC activity (115). In contrast, stimulation of the Gαi-paired A1R and A3R inhibits cAMP generation (115). Although modulation of intracellular cAMP content constitutes a crucial aspect of extracellular adenosine-exerted regulation, stimulation of its receptors induces a variety of cAMP-independent biochemical effects, such as A1R/Gα<sup>i</sup> , A2BR/Gαq/11, A3R/Gαq/11-induced stimulation of phospholipase C (PLC) activity and A1R/Gα<sup>i</sup> , A2AR, A2BR/Gq/11, A3R-mediated ERK activation (115). Finally, elevation of extracellular adenosine levels induces receptor-independent boosting of AMP-activated protein kinase (AMPK) via intracellular transfer of this nucleoside followed by its conversion to AMP (116, 117).

#### ADENOSINE-INDUCED INTRACELLULAR cAMP ACCUMULATION IMPAIRS T CELL-MEDIATED ANTITUMOR RESPONSES

It is now understood that T cells play a major role in tumor control (118–120). As will be discussed however, elevated levels of adenosine in the TME can potently impair T-cell function by inducing accumulation of intracellular cAMP.

#### Levels of Adenosine Receptors on the T Cell Surface

Murine (121–127) and human (128–132) T cells express all four ARs, and levels of A2AR (122, 124–127, 129), A2BR (126, 127, 130), and A3R (127, 131) increase upon T cell activation. However, the biology of T cells is primarily affected by the predominantly expressed A2AR (122, 123, 128, 132). Of note, similarly to CD39 and CD73, A2AR, and A2BR are upregulated due to hypoxia-induced HIF1α (133) transcriptional activity. Moreover, mRNA levels of both A2AR and A2BR are upregulated in T cells specifically upon provision of anergic stimulus (134). Validating these findings, adoptively transferred tumor-specific T cells isolated from tumors contained twice the A2AR mRNA levels than counterpart T cells isolated from spleens of tumorbearing mice (135). Since triggering of the different ARs initiates diverse and even antagonistic signaling pathways, the net cellular effects of adenosine are determined by the relative surface expression of its receptors. It is clear, however, that treatment of human (136, 137) or murine (38, 126, 138, 139) T cells with adenosine or adenosine analogs induces A2AR- (38, 126, 137–139) as well as A2BR- (38, 136) mediated intracellular cAMP build-up.

# The Mechanics of cAMP-Mediated T Cell Suppression

The secondary messenger of adenosine cAMP, also a derivative of ATP, is involved in a diverse range of cellular functions including metabolism, transcription, and growth, while oscillations of its levels within distinct cell populations are paramount for the regulation of multiple bodily functions, such as endocrine, cardiovascular, neuronal, and immune processes (140). The intracellular concentration of cAMP is determined by the antagonistic activities of ACs, and of cAMP-specific phosphodiesterases (PDEs), proteins that hydrolyze cAMP to 5 ′ -AMP. Although cAMP can diffuse within the cytosol, the co-localization of the highly-targeted AC and PDE activities in particular subcellular regions results in the formation of distinct cAMP microdomains within which co-localized cAMP effectors are activated by in-situ generated cAMP before its swift degradation (141, 142). The formation of such microdomains is mediated by AKAPs, scaffold proteins shown to bind ACs, PDEs as well as effectors of the cAMP-signaling pathway (143, 144). Of the 10 currently identified AC isoforms, T cells express AC3, AC6, AC7 and AC9 (145, 146) with most cAMP production catalyzed by AC7 (146). As previously described, A2AR and A2BR are coupled to Gα<sup>s</sup> which stimulates the activity of ACs. Of the 11 PDE families characterized to date, isoforms belonging to the relatively strong-affinity (147) cAMP-binding families of PDE1 (145, 148), PDE3 (145, 149), PDE4 (145, 149), PDE7 (145, 149– 151), PDE8 (145, 151, 152), and PDE11 (145) have been observed within T cells, with most cAMP hydrolysis carried out by PDE3 and PDE4 isoforms (148, 149, 153). Of note, cAMP levels in T cells can also be augmented by additional factors in the TME including prostaglandin E<sup>2</sup> (PGE2) (154), norepinephrine (155), histamine (156), the neuropeptides VIP and PACAP (157, 158), and low pH (159). Additional phenomena contributing toward cAMP build-up within effector T cells include TCR triggering (160, 161) as well as direct cAMP transfer by tumor cells (162) or Tregs (163) via gap junctions.

Accumulation of cAMP within the T cell cytosol induces the activity of protein kinase A (PKA) and of exchange protein directly activated by cAMP (EPAC). PKA, the dominant effector of the cAMP signaling pathway (164) is an heterotetramer comprising two catalytic (C) subunits, maintained in an inactive state by tethering to two regulatory (R) subunits (165). Binding of cAMP to the R-subunits induces a conformational change resulting in the release of the C-subunits (166). As a result, liberated PKA C-subunits within T cells phosphorylate a wide variety of substrates affecting multiple signaling pathways (167). It is well established that sustained PKA activity disrupts signaling induced by triggering of the TCR, of the co-stimulatory receptor CD28 (168, 169) as well as by the IL-2 receptor (IL-2R) (170). Negative regulators of these signaling pathways, whose activity is bolstered by PKA, include Csk (171), SHP-1 (172), SHIP1 (173), HPK1 (174), and PP2A (175). Conversely, PLCγ1 (176, 177), Raf-1 (178, 179), JAK3 (170), RhoA (180, 181), VASP (182) as well as the transcription factors NFAT (183, 184) and NFkB (185, 186) constitute mediators or endpoint effectors of the aforementioned axes whose activity is dampened by PKA.

PKA activity also significantly affects cytoplasmic potassium concentration within T cells by inhibiting the activity of Kv1.3 (187) and KCa3.1 (188, 189), channels which are responsible for the bulk of potassium efflux by T cells (190). In a negativefeedback fashion, PKA induces reduction of the cytosolic cAMP concentration by directly phosphorylating AC6 in an inhibitory fashion (191) as well as isoforms of PDE3 (192), PDE4 (193, 194), PDE8A (195) in a stimulatory manner. At the transcriptional level, PKA augments the activity of CREB cAMP responsive element binding (CREB), cAMP responsive element modulator (CREM) and activating transcription factor-1 (ATF-1) (196), which induce or counteract the transcription of multiple inflammation-relevant genes such as IL-2 (197–199), IFNγ (200–202), IL-4 and IL-13 (203, 204), IL-17 (205–208), and FoxP3 (209, 210). Specifically, PKA promotes the transcriptional activity of CREB by phosphorylating it thus increasing its affinity for its co-activators CBP and p300 (211), and by promoting the nuclear localization of CRTC (212), another family of CREB co-activators. Finally, PKA directly phosphorylates and activates ATF-1 (213) as well as distinct CREM isoforms (214) in a way similar to CREB.

The guanine nucleotide exchange factor EPAC1 is another effector of cAMP in T cells (215, 216). cAMP binds to the cAMPresponsive N-terminal region of EPAC1 and induces an open conformation rendering its catalytic core accessible to its effectors (217, 218). The most heavily characterized EPAC1 effector in T cells is the anergy-associated GTPase Rap1 (219, 220) which in its GTP-bound form is targeted to the plasma membrane (221) where it inhibits TCR-induced MEK-ERK activation by sequestering Raf-1 (220, 222).

# Overview of the Inhibitory Effects of cAMP on T-Cell Biology

A variety of molecules, including cAMP analogs, direct AC activators (e.g., forskolin and cholera toxin) and PDE inhibitors have been used to elucidate the diverse effects of intracellular cAMP accumulation on T-cell biology. In the presence of such molecules (223–228) as well as by A2AR triggering (125, 126, 229) the capacity of previously unstimulated T cells, CD4<sup>+</sup> or unfractionated, to differentiate post-activation toward cells that produce Th1 (125, 126, 223–225, 229) or Th2 (226–229)-signature cytokines is drastically diminished. This occurs in a PKA-dependent fashion (230, 231) through multi-level disruption of TCR- or CD28-induced signaling (122, 232). Intriguingly, A2AR agonist-induced impairment of IFNγ production remains evident even when A2AR agonist-pretreated T cells are re-stimulated in the absence of this agent (139). Furthermore, agents that directly activate the cAMP pathway (233–235), as well as adenosine (122, 138, 232, 236, 237), have been shown to restrict stimulation-induced AKT activation (122, 232, 233, 238) and to induce stabilization of β-catenin, which restricts maturation toward terminally differentiated effector cells (239). Moreover, such agents can prevent FasL upregulation, thus averting FasL-mediated activation-induced cell death (AICD) (127, 138, 235, 237). Finally, such molecules abolish mitogenicstimulus-induced T cell proliferation, in a PKA-dependent manner (240), by downmodulating the transmission of TCR/ CD28- and IL-2 (241)-initiated signaling, as well as IL-2 production (126, 229, 231) and IL-2Ra expression (242).

Forskolin, cAMP analogs, PDE inhibitors (152, 243–245) and adenosine (188, 246–248) also diminish T cell adherence (152, 243, 246, 248) by down-modulating the expression levels of ICAM-1 (249, 250) as well as of the integrins α<sup>4</sup> (251, 252) and β<sup>2</sup> (251, 253), components of VLA-4 and LFA-1, respectively. Such agents also impair T-cell migration (188, 244, 245, 247) by inducing KCa3.1 inhibition (188, 189). In addition, cAMP-mediated signaling (230, 254, 255) or the presence of A2AR agonists (139, 168, 230, 231) diminishes T cell cytotoxicity, in a PKA-dependent manner (168, 230, 231), probably as a result of impaired TCR signaling, motility/adhesion, granule exocytosis (138), as well as due to decreased expression of FasL, Granzyme B (GzB), and perforin (127).

Lastly, cholera toxin (256), PDE inhibitors (257–259), forskolin (157) and A2AR agonists (126, 260) not only skew T cells toward the Treg lineage via induction of FoxP3 expression (126, 256–258, 260), but also enhance the capacity of Treg cells to suppress responder T cells (258–260), at least in part by upregulating CTLA-4 levels (157, 260). Thus, cAMP can potently diminish the differentiation and effector activities of CD4<sup>+</sup> and CD8<sup>+</sup> T cells, while promoting the differentiation toward Tregs, as well as their suppressive capacity.

# THE PLEIOTROPIC EFFECTS OF ADENOSINE IN THE TUMOR MICROENVIRONMENT

Along with T cells, many other cell types in the TME including other protective or suppressive immune infiltrates, tumorassociated fibroblasts, endothelial cells and cancer cells also express functional ARs (3, 261–266). Here we briefly describe the effects of adenosine-induced signaling on them (**Figure 2**).

# Dendritic Cells

The biology of DCs, specialized antigen presenting cells (APCs) and critical messengers between the innate and adaptive immune system, can be severely impaired by adenosinergic signaling. For example, it has been reported that adenosine binding to A2BR (267) halts the differentiation of monocytes to DCs (267, 268). In addition, adenosine averts inflammatory stimulus-induced DC activation (269), whereas A2AR (270) and A2BR triggering (267, 271, 272) diminishes the capacity of DCs to prime Th1 immune responses (267, 270, 271) but rather prompts DCs to skew naïve T cell differentiation toward Th2 (267, 271) and Th17 (272) lineages. Adenosine-treated DCs exhibit decreased expression or secretion of TNFα and IL-12 (268–271, 273) and enhanced production of IL-5 (270), IL-10 (267, 268, 270, 273), IL-6 (267, 272) and TGFβ (267). Moreover, such DCs are less motile due to chemokine receptor downregulation (274), and have a tolerogenic effect on the TME due to overexpression of TGFβ (267), IL-10, IDO-1 (267), arginase-2 (267, 275), as well as A2AR-mediated upregulation of PD-L2 (276). Finally, adenosine compels DCs to secrete the proangiogenic factors VEGF (267, 275) in an A2BR-dependent manner as well as IL-8 (267).

# Macrophages

Stimulation of adenosine receptors hinders the differentiation of monocytes to macrophages, probably through cAMP accumulation (277). Moreover, by engaging A1R (278), A2AR (278–282), A3R (281, 283) or setting off Gαs-paired ARs (284), adenosine reduces the pro-inflammatory activity of macrophages by dampening their ability to produce IL-12 (279), TNFα (278–280, 282, 283), macrophage inflammatory protein-1α

multiple potentially protective immune infiltrates including T cells, DCs, NK cells, macrophages and neutrophils, while enhancing the activity of immunosuppressive cell-types, such as MDSCs and Tregs. In addition, adenosine not only assists tumor cells in co-opting adjacent fibroblasts for support, but also induces the formation of new blood vessels. Adenosine also affects the capacity of some tumor cell-types to survive, proliferate, migrate and invade neighboring tissues (HPC, bone marrow-derived hematopoietic progenitor cells).

(MIP1α) (281), nitric oxide (278, 285) and superoxide (284). In addition, by triggering A2AR (282, 286, 287), A2BR (288, 289) or unidentified ARs, adenosine promotes an M2 polarization of macrophages by inducing upregulation of arginase-1 (288, 290), IL-10 (279, 286, 289) and VEGF production (282, 287).

#### NK Cells

A2AR stimulation by adenosine not only restricts the NK maturation (291), but also their capacity for stimulus-induced CD69 upregulation (292, 293), proliferation (291, 294) as well as IFNγ (292, 293) and TNFα (294, 295) production. Furthermore, largely via A2AR triggering, adenosine diminishes target cell killing by NK cells (292, 294, 296–298).

#### Neutrophils

Adenosine exerts a variety of inhibitory effects on neutrophils. For example, triggering of A2AR (299–303), A3R (304), nonspecified A2Rs (304–307) or ARs dampens their ability to adhere (299, 305, 308, 309), transmigrate (310), secrete TNFα and MIP1α (300, 306), degranulate (301, 302, 304, 311), perform Fc receptor-mediated phagocytosis (307) and produce superoxide (299, 301–303). Interestingly, others claim that A2AR and A2BR signaling has been shown to suppress VEGF production (310). Finally, A2AR stimulation prompts neutrophils to secrete higher levels of PGE2 (312).

### MDSCs

A2BR-mediated signaling boosts differentiation of bone marrow hematopoietic progenitors toward a tolerogenic myeloid-derived cell subset, the MDSCs (313). Moreover, A2AR activation promotes IL-10 production by MDSCs (314) and treatment with an adenosine analog results in increased expression of CD73 (313). Finally, it has also been shown that A2BR stimulation on MDSCs augments VEGF production (315).

#### Stromal Cells

Adenosine, along with critically contributing to the establishment of a tolerogenic TME, also enables tumors to subvert fibroblasts into supporting them and to induce formation of new blood vessels, processes essential to their growth and dissemination. CAFs, for example, are stromal cells that support tumors by secreting the pro-metastatic and angiogenic (316) chemokine CXCL12 (317), as well as the mitogenic (318) fibroblast growth factor 2 (FGF2) (319). Triggering of A2BR on the surface of CAFs boosts expression of both CXCL12 and FGF2 (320) whereas A2AR-induced signaling stimulates their proliferation (54). As previously mentioned, adenosine can stimulate VEGF secretion by multiple cell types found within the TME, which in turn promotes angiogenesis by supporting the survival, migration and proliferation of endothelial cells (321, 322). It has also been shown that A2AR (323) and A2BR (66) stimulation diminishes production of the anti-angiogenic factor thrombospondin-1 by endothelial cells. Furthermore, adenosine not only augments the rate of intra-tumoral nutrient delivery by inducing vasodilatation (324), but also hinders leukocyte extravasation (325) through downregulation of adhesion molecules, such as E-selectin (326, 327) VCAM-1 (326, 327) and ICAM-1 (327, 328) on the surface of endothelial cells, as well as by limiting vascular permeability (325, 327, 329–331) through A2BR activation (329–331). Finally, signaling initiated by triggering of A2AR (332, 333), A2BR (334–336) or non-specified ARs prompts endothelial cells to overexpress CD73 (334) as well as the proangiogenic factors VEGF (332, 333, 335, 336), IL-8 (335) and basic fibroblast growth factor (bFGF) (335, 336).

# Tumor Cells

Adenosine binding to ARs on the surface of cancer cells has a profound impact on their biology. For example, the triggering of A1R (337, 338), A2AR (54, 339, 340), and A3R (339, 341–343) induces a variety of cellular responses that augment cancer cell survival such as AKT and ERK1/2 stimulation, as well as Bad inactivation (342). Additional responses to AR signaling contributing to bolstered cancer cell survival include upregulation of Bcl2 (343), downregulation of p53 (338) and Bax (343) as well as aversion of caspase-9 (343) and caspase-3 (54, 343, 344) activation. Paradoxically, extracellular adenosine has also been demonstrated to cause cancer cell death either by setting off A1R (345, 346), A2AR (341, 347), A2BR (348, 349), and A3R (339, 350–354) or via induction of AMPK activation upon its cellular uptake and subsequent conversion to AMP (345).

Moreover, A1R (337, 355), A2AR (341, 356), A2BR (344, 357– 359), and A3R (343, 360) stimulation augments cancer cell proliferation through activation of PLC (356), protein kinase C-delta (PKC-δ) (356), AKT (356, 357), ERK1/2 (356–360), JNK (356, 358), and p38 (358). Furthermore, triggering of the ARs leads to upregulation of cyclins A (343), B (358), D (343, 358), E (337, 343, 358), estrogen receptor-α (355) as well as downregulation of the cell-cycle inhibitors p27 (337) and p21 (343, 358). Surprisingly, though, activation of A2BR (349) and A3R (341, 350, 353, 361–363) has also been reported to result in a potent cytostatic effect.

Motility (358, 359, 364–369) and invasiveness (358, 359, 367, 370) are additional features of cancer cells that are boosted upon engagement of A1R (364, 365), A2AR (366), A2BR (358, 359, 367, 368), and A3R (369, 370). In terms of mechanisms, signaling initiated by these receptors promotes filopodia formation (367) as well as expression of matrix metalloproteases (MMPs) (358, 359, 370) and FXYD5 (359), a cell membrane glycoprotein known to drive metastasis by reducing cell adhesion (371). In contrast, others claim that A3R triggering hinders the motility and invasiveness of cancer cells (372, 373). Finally, A2AR (374), A2BR (369, 375), and A3R (369, 375–377) stimulation on the surface of cancer cells promotes angiogenesis by boosting secretion of the pro-angiogenic factors VEGF (369, 375, 377), IL-8 (369, 375), angiopoietin 2 (376), and erythropoietin (374).

The contrasting consequences of triggering particular ARs, on the survival, proliferation or migration and invasiveness of tumor cells most probably occur due to the heterogeneity between cells and/or experimental settings employed to assess them. For instance, two different cancer cell lines of distinct tissue origin could have profoundly diverse AR expression profiles as well as different ability to transmit/terminate signaling initiated by these receptors. Moreover, they might have different capacity to produce adenosine, which once released into the medium can trigger ARs in an autocrine fashion. Finally, different concentrations used between experiments, as well as limited specificity of the AR agonists/antagonists, probably constitute additional factors contributing to the observed discrepancies.

#### TARGETING ADENOSINERGIC SIGNALING IN CANCER IMMUNOTHERAPY

Adenosine confers potent immunosuppressive as well as direct tumor-promoting effects in the TME. Thus, approaches to both blocking its generation and hindering binding to its receptors have become important areas of research (**Figure 3**). Indeed, extensive pre-clinical experimentation has firmly established that targeting the adenosinergic signaling on its own (**Table 1**) or in combination with emerging IMTs or established cancer treatments (**Table 2**) shows important promise and soundly supports the clinical evaluation (**Table 3**) of these concepts. Here we present an overview of such pre-clinical and clinical studies.

#### Blockade of Adenosine Generation

As previously described, CD73 is an nucleotidase that converts AMP, generated from CD39- or CD38/CD203-mediated

FIGURE 3 | Approaches for blocking adenosinergic signaling in the TME. The inhibitory effects of adenosine in the TME can be circumvented by administration of mAbs or small molecules that target enzymes involved in the catabolism of ATP and NAD, such as CD39,CD73 and CD38, as well as by pharmacologic antagonists of A2AR and A2BR to block adenosine-mediated signaling. Whereas multiple such mAbs and pharmacologic inhibitors/antagonists display antitumor activity within murine models of solid tumors (Tables 1, 2), depicted are only those currently evaluated in patients with solid tumor malignancies (Table 3). Finally, treatments that reduce the extracellular export of ATP, such as oxygenation to reverse hypoxia, can attenuate adenosinergic signaling.

catabolism of ATP or NAD respectively, to adenosine. Its central role in adenosine generation is underscored by the fact that CD73-deficient mice display drastically decreased interstitial levels of adenosine, not only at steady state, but also upon induction of trauma or hypoxia (409, 410). CD73 knock-out mice exhibit hindered tumor growth and metastatic spreading (378–380, 387) and mice inoculated with tumor cells lacking CD73 survive longer than mice inoculated with tumor cells expressing this ecto-enzyme (378, 388). Indeed, administration of anti-CD73 monoclonal antibodies (mAb) (368, 378–386) or of a CD73-specific pharmacologic inhibitor (378, 379, 381, 383, 384, 387– 389) impairs tumor growth (368, 378–382, 385, 387, 389) and metastasis (368, 379, 380, 384, 386) while increasing survival (378, 382, 384, 388). Of note, CD73 can also act as an adhesion/signaling molecule to promote metastasis in a catalytic-activity independent manner (386, 411, 412). Mechanistically, the aforementioned treatments have been shown to promote intra-tumoral accumulation of CD8<sup>+</sup>

T cells (381, 382, 385, 389), B cells (381) as well as of Th1- and Th17-associated cytokines (381) while decreasing the levels of intra-tumoral VEGF (383) and the presence of Tregs (389). Of note, even though metastasis can be modestly inhibited by anti-CD73 therapy in an immune-system independent fashion (368, 386), most of the antitumor effect of CD73 blockade is due to alleviation of A2AR-mediated immunosuppression (368).

No doubt encouraged by these pre-clinical studies, four anti-CD73 mAbs are currently being evaluated as monotherapies in small scale trials targeting a variety of solid tumors. In July 2015, MedImmune launched a first in-human trial (NCT02503774) evaluating the human anti-CD73 mAb Oleclumab, which allosterically prevents CD73 from assuming its catalytically active conformation (413). In June 2016, Bristol-Myers Squibb (BMS) launched a Phase I/IIa trial (NCT02754141) to assess the efficacy of BMS-986179, a human IgG2-IgG1 hybrid mAb that not only inhibits CD73-exerted AMP hydrolysis but also induces CD73 internalization (414). In April 2018, Corvus Pharmaceuticals TABLE 1 | Evaluation of adenosine-axis blockade in murine models of solid malignancies.


<sup>(</sup>Continued)


<sup>a</sup>Patient-derived tumor cell lines, NSCLC, Non-Small-Cell LungCancer. HNSCC, Head and neck squamous cellcarcinoma.

X > Y: X contributes more than Y to the anti-tumor effect of adenosine axis modulation.

initiated clinical evaluation (NCT03454451) of their humanized anti-CD73 mAb, CPI-006, which directly competes with AMP for the CD73 active site (415). Finally, in July 2018, Novartis listed a Phase I/Ib trial (NCT03549000) evaluating the efficacy of SRF373/NZV930, a human mAb that impedes CD73 activity via a currently undisclosed mechanism, and was pre-clinically developed by Surface Oncology before being exclusively licensed to Novartis for further clinical development.

CD39 also critically contributes to the generation of extracellular adenosine from ATP as evidenced by the fact that deficiency of this enzyme results in significantly decreased adenosine content in tissues, not only at steady state, but also upon ischemia induction (80). Similar to studies with CD73-deficient mice, tumor growth and metastasis are reduced in CD39-null mice (391, 416). In addition, intraperitoneal delivery of a CD39 inhibitor in immunocompetent mice reduces tumor growth rates (391). Administration of an anti-CD39 mAb increased the survival of immuno-deficient mice inoculated with patient-derived tumors (390), indicating that CD39 can also promote tumor growth or metastasis in an immune system independent manner. In terms of mechanisms, several studies have demonstrated that in vitro inhibition of CD39 activity by pharmacologic inhibitors (45, 47, 62) or blocking mAbs (45, 417, 418) results in enhanced functionality of T cells (45, 47, 62, 418) and NK cells (45, 47, 418), as well as decreased Treg-mediated suppression of T cell proliferation (47, 417). Even though restriction of CD39 activity in vitro conclusively alleviates adenosineinduced immunosuppression, a surprisingly small number of studies demonstrate effectiveness of this approach within tumor-bearing mice. Finally, while humanized mAbs targeting CD39, such as IPH52 (Innate Pharma) have been developed, clinical studies exploring CD39 blockade/inhibition have not been launched.

As previously mentioned, the concerted activity of CD38 and CD203a, can functionally replace CD39 toward the generation of extracellular adenosine. Further substantiating the soundness of CD38-blockade as a cancer treatment, immunocompetent CD38-null mice display reduced tumor growth (419) whereas tumors devoid of this ectonucleotidase grow slower both in immuno-competent (96) as well as in immuno-deficient mice (97). Indeed, administration of CD38 mAbs retards tumor growth (96, 420). Interestingly, tumors derived from anti-CD38 mAb-treated mice encompass more CD8<sup>+</sup> T cells and less Tregs and MDSCs (96). Moreover, increased fraction of CD8<sup>+</sup> T cells infiltrating these tumors display an effector memory phenotype while less of these cells are double positive for the exhaustion markers PD-1 and TIM3 (96). Three anti-CD38 mAbs, Daratumumab (Janssen Biotech), Isatuximab (Sanofi), and MOR202 (Morphosys) are being clinically evaluated. Daratumumab was FDA-approved in 2015 for treating multiple myeloma patients, while to date the most advanced testing of Isatuximab and MOR202 as monotherapies are respectively the Phase II trials NCT01084252, NCT02960555, and NCT02812706, as well as the Phase I/IIa trial NCT01421186. Of note, in addition to modulating the enzymatic activity of CD38, these mAbs also have the capacity to induce cytotoxicity through diverse mechanisms, such as induction of complement activation, Ab-dependent cellular cytotoxicity (ADCC) or phagocytosis, and programmed cell death (420). Albeit extensive clinical experience of utilizing the aforementioned mAbs against CD38 overexpressing hematologic malignancies, the recently launched trial NCT03473730 constitutes the first application of a CD38 specific mAb in patients with solid tumor malignancies.

Another approach for limiting the intratumoral interstitial adenosine is the oxygenation of the TME (293). As mentioned, hypoxia promotes build-up of extracellular adenosine at least by inducing upregulation of CD39 and CD73 as well as downregulation of adenosine transporters. Indeed, in pre-clinical models, respiratory hyperoxia (60% oxygen) lowers intra-tumoral adenosine levels (9), tumor growth rates (9), metastasis formation (293) and increases survival of tumorbearing mice (9, 293). Mechanistically, this treatment boosts MHC-I levels on the tumor-cell surface (9), the presence of CD8+, CD69+, or CD44<sup>+</sup> cells within the TME (293) and reduces the presence of Tregs (293) as well as the latter's capacity to express CD39, CD73, CTLA-4, or FoxP3 (293). Moreover, increased oxygenation of tumors not only averts angiogenesis through reduction of VEGF concentration (9), but also dampens expression of molecules associated with immune dysfunction, such as TGF-β, CD39, CD73, A2AR, A2BR and COX-2 (9, 293), the rate-limiting enzyme of PGE<sup>2</sup> biosynthesis, while increasing the mRNA levels of pro-inflammatory agents, such as IL-2, and IL-12a (293).

### Blockade of Adenosine Receptor Binding

Along with blocking adenosine production with small molecules or mAbs, another approach to inhibit adenosine-induced signaling is to directly block binding to its receptors A2AR and A2BR. Underscoring the potent protumoral effect of A2ARtrigerring, mice devoid of this receptor present reduced rates of tumor growth and metastasis, and in some instances tumors undergo complete rejection (38, 292, 400, 402). In addition, administration of pharmacologic A2AR antagonists recapitulates the anti-tumor effects of A2AR-deletion since it results to reduced primary tumor expansion (38, 54, 389, 392, 393, 396) and metastasis formation (292, 384, 393, 394, 397) ultimately leading to prolonged survival (384, 396). Mechanistically, tumors derived from A2AR-antagonist-treated mice are more heavily infiltrated by CD8<sup>+</sup> T cells (389, 392, 393) as well as NK cells (389, 392, 393) and encompass fewer Tregs (389, 392, 393). In addition, in vivo A2AR antagonism leads to increased expression of CD69 (393), T-bet (396), and 4-1BB (396) as well as production of IFNγ and TNFα (392, 396) by intra-tumoral CD8<sup>+</sup> T cells. Furthermore, this intervention increases the fraction of intra-tumoral NK cells producing GzB (292) and reduces the expression of PD-1, LAG3, FoxP3 and A2AR by tumor-infiltrating Tregs (392, 396). Interestingly, the A2AR antagonists ZM241385 and SCH58261 exhibit the capacity to curb primary tumor growth even in a T cell-independent manner (54). Notably, A2A antagonism in vivo increases activation induced cell death (AICD) of intra-tumoral T cells (395), a finding corroborating observations that cAMP-accumulation in the T cell cytosol averts terminal effector differentiation and AICD (421, 422). Three A2AR antagonists are currently being evaluated as single agents in Phase I/II trials to treat cancer patients bearing solid tumors. In particular, Corvus Pharmaceuticals, AstraZeneca, and Novartis have undertaken the clinical development of CPI-444 (NCT02655822), AZD4635 (NCT02740985), and NIR178 (NCT02403193, NCT03207867), respectively.

As for A2AR, genetic deletion of A2BR reduces tumor growth rate (399, 423) while A2BR−/<sup>−</sup> tumor cells display reduced metastatic potential (359, 367). Notably, administration of A2BR antagonists in tumor-bearing mice reduces tumor growth (315, 398, 399) and metastasis (292, 359, 368) eventually prolonging their survival (359). Mechanistically, antagonism of A2BR in vivo augments the intra-tumoral presence of CD8<sup>+</sup> T cells (315, 398), NKT (315, 398) as well as the mRNA levels of IFNγ and CXCL10 (399) and the concentration of TNFα, IFNγ, and GzB (398) in the TME. This intervention further results in decreased accumulation of MDSCc (315, 398) and IL-10 (398), as well as reduced levels of VEGF and angiogenesis (315). Based on encouraging preclinical results, Palobiofarma recently launched a dose escalation Phase I study (NCT03274479) administering TABLE 2 | Evaluation of concomitant adenosine-axis blockade in murine models of solid malignancies.


(Continued)

#### TABLE 2 | Continued


(Continued)


TDLN, tumor-draining lymphnode.

PBF-1129, a selective A2BR inhibitor, in patients with advanced Non-Small Cell Lung Cancer (NSCLC).

#### Combinatorial Treatment Approaches

Since multiple ecto-enzymes with redundant functions contribute toward extracellular adenosine production and both A2AR and A2BR triggering mediate the majority of adenosine's pro-tumoral effects, monotherapies may not be sufficient to block the adenosine-signaling axis. In addition, there is strong rationale for combination with IMTs, such as ICB of PD-1/PDL-1 or CTLA-4, as well as ACT, radiotherapy and chemotherapy, to further unleash the cytotoxic capacity of T cells, which, as will be discussed, can become highly sensitized to adenosine-mediated immunosuppression.

#### Combinations of Adenosine-Axis Blockade Agents

Concurrent mAb-mediated (418) or pharmacologic (47) inhibition of CD39 and CD73 failed to potentiate CD73 blockade-induced suppression of adenosine production by Tregs and ovarian cancer cell lines. These findings are corroborated by the observation that skin biopsies derived from CD39−/−CD73−/<sup>−</sup> mice have identical capacity to produce adenosine upon injury induction with counterpart biopsies derived from CD73−/<sup>−</sup> mice (424).

Alone the same lines, others addressed whether simultaneous blockade of CD73 and of A2AR would result in higher anti-tumor efficacy. Of note, CD73−/−A2AR−/<sup>−</sup> mice present superior tumor control as compared to single knockout mice (384). Moreover, tumors in A2AR-null mice express twice as much CD73 at their core when compared to tumors formed in wild-type mice (384). Indeed, dual therapy with an anti-CD73 mAb and an A2AR agonist confers superior tumor protection as compared to either one as a monotherapy (384). However, this additive effect is lost when CD73 is targeted with a pharmacologic inhibitor, thus underscoring the capacity of CD73 to promote tumor progression in a catalytic activity-independent manner (384). In light of these studies, Evotec and Exscientia have partnered to develop A2AR/CD73 bi-specific inhibitory molecules (425), whereas

NCT03454451, NCT03549000 as well as the Phase Ib/II clinical trial NCT03381274 sponsored by MedImmune all include solid tumor-bearing patient cohorts scheduled to be treated with combinations of an anti-CD73 mAb along with a pharmacologic A2AR antagonist.

#### Adenosine-Axis and PD-1 Blockade

Briefly, PD-1 is an immunosuppressive receptor that upon binding to its ligands, PDL-1 and PDL-2, dampens T-cell activity thereby enabling tumors to evade immune-destruction. Blockade of the PD-1-PDL-1/2 signaling axis results in durable complete responses in the clinic for a fraction of treated patients (1), and many pre-clinical and clinical studies have explored concomitant inhibition of adenosine production, or antagonism of A2AR and A2BR, to improve response rates.

It has been demonstrated that CD73<sup>+</sup> tumor cells are resistant to PD-1 ICB (401) and that simultaneous mAbmediated blockade of CD73 and PD-1 synergistically enhances tumor control and survival in mice (382, 385). Mechanistically, the dual therapy augments intra-tumoral CD8<sup>+</sup> tumorspecific T cells (382, 385) and IFNγ mRNA levels (382) as compared to single-agent treatments. Several clinical trials assessing anti-CD73 mAb treatment along with anti-PD-1 mAb (NCT03454451, NCT03549000) or anti-PDL-1 mAb (NCT02503774, NCT03773666, NCT03267589, NCT03334617) of advanced solid tumors are recruiting or underway. Intratumoral upregulation of CD38 and subsequent adenosine production was recently identified as a mechanism of acquired resistance to PD-1/PD-L1 blockade and mAb-mediated or pharmacologic inhibition of CD38 was shown to significantly improve the anti-tumor efficacy of an anti-PDL-1 mAb (96). In terms of mechanisms, tumors from mice receiving the combinatorial therapy displayed higher accumulation of CD8<sup>+</sup> T cells, effector memory CD8<sup>+</sup> T cells, ICOS<sup>+</sup> CD4<sup>+</sup> T cells and lower levels of MDSCs and Tregs as compared to tumors from single-agent treated mice (96).

The potential for synergy between the co-administration of A2R antagonists with anti-PD-1 mAb is underscored by the observations that PD-1 blockade enhances A2AR expression on tumor-infiltrating CD8<sup>+</sup> T cells (401), as well as that PD-1 blockade is more efficacious, in terms of increasing the survival of tumor-bearing mice, when these mice lack the A2AR (400). Vice versa, A2AR triggering on the surface of CD8<sup>+</sup> T cells derived from tumor tissue (382), tumor draining lymph nodes or spleen (396) promotes PD-1 expression suggesting that simultaneous PD-1 blockade would boost the anti-tumor efficacy of A2A antagonism. Indeed, several groups demonstrated that concurrent provision of PD-1 checkpoint inhibitors along with A2AR antagonists is more effective than single-agent treatments at reducing tumor growth rate (96, 396, 400, 401) and metastasis formation (394, 401), as well as at improving survival (394, 396, 401). Moreover, the combination enables increased production of IFNγ and GzB by CD8<sup>+</sup> tumor infiltrating T cells (401) while augmenting the intra-tumoral presence of NK cells (394). Five clinical trials for the treatment of solid-tumor patient cohorts with A2AR antagonists along with anti-PD-1 Ab (NCT02403193, NCT03207867) or anti-PD-L1 Ab (NCT02655822, NCT03337698, NCT02740985) are ongoing. Finally, dual therapy comprising A2BR antagonism and PD-1 blockade is superior to either monotherapy at decreasing metastasis and improving survival of tumor-bearing mice (359). However, no clinical trials have been launched to date to explore this combination in human cancer patients.

#### Adenosine-Axis and CLTA-4 Blockade

The blockade of CTLA-4, an immune checkpoint receptor predominantly expressed by T cells and which competes with the co-stimulatory receptor CD28 for binding to CD80/CD86 on the surface of antigen presenting cells (APCs), has also generated durable clinical responses in advanced cancer patients (1). Tumor-bearing mice receiving CTLA-4 blockade and pharmacologic (389) or Ab-mediated (382) inhibition of CD73 display superior tumor control (382, 389) and overall survival (382) than counterparts receiving single agent treatments. Mechanistically, these dual therapies are more effective than corresponding monotherapies at increasing the intra-tumoral presence of tumor-specific CD8<sup>+</sup> T cells (382), CD4+FoxP3neg T cells (389) as well as the levels of IFNγ (389) and of mRNA coding for IFNγ and T-bet (382). Likewise, concomitant provision of CTLA-4 ICB and antagonists of either A2AR (389) or A2BR (359) leads to decreased tumor growth (389) and metastasis formation (359), as well as to higher survival of tumor-bearing mice (359) when compared to single treatments. In terms of mechanisms, combining CTLA-4 ICB with an A2AR antagonist augments intratumoral CD8<sup>+</sup> T cell presence as well as IFNγ and GzmB levels (389).

#### Adenosine-Axis Blockade and Adoptive T Cell Therapy

There are two main approaches to ACT. Either autologous tumor-reactive T cells are expanded from tumor biopsies prior to patient re-infusion [i.e., tumor infiltrating lymphocyte (TIL) therapy], or peripheral blood T cells are gene-engineered to express a tumor-specific T cell receptor (TCR), or a so-called chimeric antigen receptor (CAR; a fusion protein that links scFv-mediated tumor antigen-binding with intracellular endodomains associated with T cell activation). Cancer patients are typically lymphodepleted prior to ACT, and following infusion they receive high doses of IL-2, both of which support T cell engraftment (426). TIL therapy has achieved robust and durable responses in advanced melanoma patients, while CAR therapy targeting CD19 has yielded unprecedented clinical responses against a variety of advanced, treatment-refractory B cell malignancies (118, 427, 428).

Synergy has been demonstrated between strategies limiting adenosine production blockade and ACT within tumorbearing mice. Indeed, ACT confers increased control of tumors lacking CD73 expression (388) and dual therapy of ACT and pharmacologic or mAb-mediated inhibition of CD73 was more robust than single treatments at augmenting tumor control and overall survival (378). Mechanistically, pharmacologic inhibition of CD73 potentiated the anti-tumor TABLE 3 | Clinical evaluation of adenosine-axis targeting in patients with solid tumors.


(Continued)

TABLE 3 | Continued


NSCLC, non-small-cell lung cancer.

\*Mentioned are schemes comprising at least one adenosine-axis modulator.

efficacy of ACT at least by boosting the homing of the adoptively transferred tumor-specific T cells at the tumor sites (378). Likewise, respiratory hyperoxia in mice increased the ability of adoptively transferred T cells to curb primary tumor expansion and metastasis formation by augmenting their capacity to accumulate in the TME and produce IFNγ (293).

Similarly, A2AR deficiency (402) or siRNA-mediated suppression of A2AR and A2BR expression (38) on the surface of adoptively transferred T cells leads to enhanced prevention of metastatic spreading (38, 402) and improved survival of tumor-bearing mice (38). Several groups have validated these observations by demonstrating that ACT and concomitant administration of A2AR antagonists is superior to single treatments in terms of decreasing tumor growth (135, 396), hindering metastasis formation (38, 402) and ultimately improving survival (135, 388, 396, 402). Interestingly, others claim that A2AR antagonism improves the efficacy of adoptively transferred CAR<sup>+</sup> T cells only if PD1 ICB is co-administered (135). In terms of mechanisms, concomitant A2AR antagonism not only increases intra-tumoral presence of adoptively transferred T cells (396) but also elevates their activation status. In particular, when A2AR antagonists were co-administered, tumor-derived, adoptively transferred or endogenous CD44<sup>+</sup> CD8<sup>+</sup> T cells, exhibit heightened expression levels of T-bet, 4-1BB, and CD69 (396) while demonstrating increased capacity to produce IFNγ and TNFα (135, 396, 402).

#### Adenosine-Axis Blockade Combined With Radiotherapy, Chemotherapy or Targeted Therapies

It is well documented that radiotherapy (RT) as well as several chemotherapeutic (CT) drugs have the capacity to induce ATP release (406, 429–433). Since such regimens also elevate the expression levels of CD39 (405, 407, 434) and CD73 (405, 407, 435–437), it is possible that the concentration of interstitial adenosine in the TME rises sharply upon application of these treatments. Therefore, several investigators have explored whether concomitant provision of agents targeting the adenosine axis increase the anti-tumor efficacy of RT or of various CT agents.

Indeed, mAb-mediated inhibition of CD73 increased the anti-tumor efficacy of RT (403, 404) and this synergistic effect was even more apparent upon concurrent CTLA-4-blockade (404). Mechanistically, CD73 inhibition increases the presence of CD8<sup>+</sup> T cells as well as of CD8α <sup>+</sup> or CD103<sup>+</sup> DCs within irradiated tumors while decreasing Tregs (403, 404). Moreover, concomitant CD73 blockade wasshown to increase the activation status of CD8<sup>+</sup> T cells and CD8α <sup>+</sup> DCs within irradiated tumors as evidenced by the elevated expression levels of CD69 and CD40, respectively (404). Likewise, concurrent mAb-mediated inhibition of CD73 (405) or pharmacologic blockade of CD39 activity (406) boosted the tumor control (405, 406) and survival (405) of mice treated with the CT drugs Doxorubicin (405), Paclitaxel (405), and Mitoxantrone (406). Of note, such dual therapies were shown to not only augment intra-tumoral presence of DCs (406) and tumor-specific CD8<sup>+</sup> T cells (405) but also the fraction of intra-tumoral CD4<sup>+</sup> or CD8<sup>+</sup> T cells producing IFNγ (406) as well as the levels of IFNγ in the TME (405, 406). In light of such observations, the clinical trials NCT03611556 and NCT03742102 are set to decipher the potency of CT regimens when provided in combination with the CD73 blocking Ab Oleclumab, supplemented or not by PD-1 blockade.

Along the same lines, others explored if direct antagonism of A2AR and A2BR would augment the antitumor effects of CT agents. Indeed, tumor-bearing mice treated with Doxorubicin (359, 405, 407), Dacarbazine (398), or Oxaliplatin (398, 407) in combination with A2AR (405), A2BR (359, 398), or dual A2AR/A2BR antagonists (407) displayed superior tumor control (398, 405, 407) or survived longer (359). Of note, tumors derived from mice treated with the combination of Dacarbazine and PSB1115, an A2BR antagonist, were more heavily infiltrated by CD8<sup>+</sup> T cells as well as NKT cells and contained higher levels of GzB than tumors derived from counterpart mice subjected to Dacarbazine monotherapy (398). Likewise, concomitant administration of AB928, a dual A2AR and A2BR antagonist, along with Doxorubicin or Oxaliplatin increased the intratumoral detection of tumor-specific CD8<sup>+</sup> T cells (407).

Finally, others have sought to decipher whether adenosine axis blockade enhances the anti-tumor efficacy of particular targeted therapies. For instance, it has been recently demonstrated that high expression levels of CD73 in tumors derived from breast cancer patients are associated with resistance to Trastuzumab, an anti-HER2/ErbB2 mAb, and that artificial CD73 overexpression promotes resistance to Trastuzumab-like therapy in immunocompetent murine models of breast cancer (408). Subsequently, the authors moved on to show that when such mice receive dual therapy comprising anti-CD73 and anti-ERB2 mAbs they exhibit inferior tumor expansion rate as well as reduced metastatic spreading and survive longer than counterpart mice treated with either single agent treatments (408). In terms of mechanisms, the combinatorial therapy significantly increases the intra-tumoral presence of CD8<sup>+</sup> and CD4+FoxP3neg T cells while decreasing MDSCs (408). In addition, melanoma patients harboring BRAF-mutant tumors exhibit a trend for elevated expression of CD73 whereas co-administration of an A2AR antagonist in mice bearing BRAF-mutant tumors increased the therapeutic benefit achieved either by BRAF inhibition or by the combination of BRAF and MEK inhibitors (393). Finally, CD73 and A2AR are overexpressed in NSCLCs harboring EGFR mutations (438) and even though preclinical studies demonstrating increased efficacy of concomitant inhibition of EGFR and A2AR are not currently publicly available, the clinical trial NCT03381274 includes a cohort of patients with advanced NSCLC that will receive both an EGFR inhibitor and an A2AR antagonist.

#### SUMMARY AND FUTURE PERSPECTIVES

Adenosine is critically involved in a range of physiologic processes including wound healing, and its levels are tightly regulated under homeostatic conditions. In solid tumors, however, adenosine concentration is significantly elevated, predominantly due to stress-induced ATP release coupled with the overexpression of nucleotidases, such as CD39 and CD73 that contribute to its catabolism. Primarily by engaging A2AR and A2BR, also overexpressed in the TME as a result of hypoxia and inflammation, adenosine diminishes the activity of protective immune infiltrates, such as T cells, NK cells and DCs, while boosting the inhibitory capacity of immunosuppressive subsets, including Tregs and MDSCs. For instance, A2AR and A2BR-induced cAMP accumulation within T cells blunts their differentiation, proliferation, cytokine production and target

#### REFERENCES


cell killing, predominantly through PKA activation. Along with establishing an anti-inflammatory and tolerogenic TME, adenosine also promotes blood vessel formation and assists tumors in subverting adjacent fibroblasts to further support tumor growth and metastasis.

Administration of small molecules or mAbs with the aim to block adenosine-signaling, either by limiting its production or its binding to ARs, has yielded important tumor control in various pre-clinical tumor models. Moreover, simultaneous blockade of adenosine production and receptor binding, achieved by an anti-CD73 mAb co-administered with an A2AR antagonist, for example, have demonstrated it synergy. Given the potent suppression of T cells by adenosine, it comes as no surprise that increases in tumor control and survival conferred by ICB (anti-PD-1 and anti-CTLA-4 mAbs) or ACT, is significantly enhanced by concomitant administration of agents countering the adenosine axis. Synergy of such adenosine axis modulators has further been shown with RT, as well as CTs, schemes known to promote immunogenic cell death (i.e., ATP is released), as well as with some targeted therapies.

While blockade of adenosine production and A2AR/A2BR antagonism are being tested in the clinic as monotherapies, increasing numbers of clinical trials combining adenosinesignaling blockade with IMTs or classic treatment approaches (i.e., RT, CT and targeted therapies) are recruiting and/or underway. Given the important responses achieved by a proportion of patients to immunotherapeutic-regimens, and the tremendous levels of immunosuppression mediated by adenosine, the development of existing or new agents targeting this axis, along with further testing of combinatorial strategies, is warranted. Indeed, targeting the adenosine axis holds great promise in the improved treatment of cancer patients.

# AUTHOR CONTRIBUTIONS

GC, MI, DA, and SV conceived the manuscript. DA, SV, and MI drafted the manuscript. GC, PR, CM-C, and CC reviewed the manuscript and provided feedback, and MI revised the manuscript. MI, DA, and SV made the figures, and DA assembled the tables.

# FUNDING

This project was supported by the Ludwig Institute for Cancer Research, the ISREC Foundation, an ERC Advanced Grant to GC (1400206AdG-322875), the Biltema Foundation, and a Kummer fellowship to DA.


kidney disease. Mol Aspects Med. (2017) 55:75–89. doi: 10.1016/j.mam.2017. 01.004


Oncoimmunology. (2016) 5:e1055444. doi: 10.1080/2162402X.2015. 1055444


phosphotyrosine and increased phosphoserine contents of PLC-γ1. J Biol Chem. (1992) 267:1496–501.


protein kinase A-mediated phosphorylation. FEBS Lett. (2012) 586:1631–7. doi: 10.1016/j.febslet.2012.04.033


with defective CD8<sup>+</sup> T-cell priming capacity. Immunology. (2013) 138:402– 10. doi: 10.1111/imm.12053


colon cancer HCT 116 and HT-29 cell lines. Arch Biochem Biophys. (2013) 533:47–54. doi: 10.1016/j.abb.2013.02.007


growth factor and interleukin-8 expression in human melanoma cells treated with etoposide and doxorubicin. Neoplasia. (2009) 11:1064– 73. doi: 10.1593/neo.09768


reduces tumor growth and metastasis. Cancer Res. (2017) 77:4684–96. doi: 10.1158/0008-5472.CAN-17-0393


reduced kidney weight. Pflügers Arch Eur J Physiol. (2006) 452:324– 31. doi: 10.1007/s00424-006-0045-x


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

Copyright © 2019 Vigano, Alatzoglou, Irving, Ménétrier-Caux, Caux, Romero and Coukos. 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.

# Overexpression of PDE4A Acts as Checkpoint Inhibitor Against cAMP-Mediated Immunosuppression in vitro

Klaus G. Schmetterer <sup>1</sup> , Katrin Goldhahn<sup>1</sup> , Liesa S. Ziegler <sup>1</sup> , Marlene C. Gerner <sup>1</sup> , Ralf L. J. Schmidt <sup>1</sup> , Madeleine Themanns <sup>2</sup> , Eva Zebedin-Brandl <sup>2</sup> , Doris Trapin<sup>3</sup> , Judith Leitner <sup>3</sup> , Winfried F. Pickl <sup>3</sup> , Peter Steinberger <sup>3</sup> , Ilse Schwarzinger <sup>1</sup> and Rodrig Marculescu<sup>1</sup> \*

*<sup>1</sup> Department of Laboratory Medicine, Medical University of Vienna, Vienna, Austria, <sup>2</sup> Center of Physiology and Pharmacology, Institute of Pharmacology, Medical University of Vienna, Vienna, Austria, <sup>3</sup> Center for Pathophysiology, Infectiology and Immunology, Institute of Immunology, Medical University of Vienna, Vienna, Austria*

#### Edited by:

*Katy Rezvani, University of Texas MD Anderson Cancer Center, United States*

#### Reviewed by:

*Chao Wang, Brigham and Women's Hospital, United States Doug Clayton Palmer, National Cancer Institute (NCI), United States*

\*Correspondence: *Rodrig Marculescu rodrig.marculescu@meduniwien.ac.at*

#### Specialty section:

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

> Received: *13 March 2019* Accepted: *16 July 2019* Published: *30 July 2019*

#### Citation:

*Schmetterer KG, Goldhahn K, Ziegler LS, Gerner MC, Schmidt RLJ, Themanns M, Zebedin-Brandl E, Trapin D, Leitner J, Pickl WF, Steinberger P, Schwarzinger I and Marculescu R (2019) Overexpression of PDE4A Acts as Checkpoint Inhibitor Against cAMP-Mediated Immunosuppression in vitro. Front. Immunol. 10:1790. doi: 10.3389/fimmu.2019.01790* Malignant cells acquire physiological mechanisms of immunosuppression to escape immune surveillance. Strategies to counteract this suppression could help to improve adoptive immunotherapy regimen. The intracellular second messenger cyclic AMP (cAMP) acts as a potent immunosuppressive signaling molecule in T-cells and is up-regulated by multiple tumor-relevant suppressive factors including prostaglandin E2 (PGE2), adenosine and the functions of regulatory T-cells. Consequently, we aimed to abrogate cAMP signaling in human T-cells by ectopic overexpression of phosphodiesterase 4A (PDE4A). We could show that retroviral transduction of PDE4A into T-cells led to efficient degradation of cAMP in response to induction of adenylate cyclase. Retroviral transduction of PDE4A into CD4<sup>+</sup> and CD8<sup>+</sup> T-cells restored proliferation, cytokine secretion as well as cytotoxicity under immunosuppression by PGE2 and A2A-R agonists. PDE4A-transgenic T-cells were also partially protected from suppression by regulatory T-cells. Furthermore, PGE2-mediated upregulation of the inhibitory surface markers CD73 and CD94 on CD8<sup>+</sup> T-cells was efficiently counteracted by PDE4A. Importantly, no differences in the functionality under non-suppressive conditions between PDE4A- and control-vector transduced T-cells were observed, indicating that PDE4A does not interfere with T-cell activation *per se*. Similarly, expression of surface markers associated with T-cell exhaustion were not influenced by PDE4A overexpression in long term cultures. Thus, we provide first *in vitro* evidence that PDE4A can be exploited as immune checkpoint inhibitor against multiple suppressive factors.

Keywords: immune tolerance, checkpoint inhibitor, adoptive immunotherapy, tumor immunosuppression, T-cell engineering

#### INTRODUCTION

One of the cardinal features of malignant processes is their ability to suppress anti-tumor immune responses, which allows them to escape the physiological tumor-immunosurveillance (1). To that end, malignant cells up-regulate mechanisms, which physiologically serve as regulators of adaptive immune responses. These measures include the overexpression of inhibitory ligands such as PD-L1, the secretion of immunomodulatory cytokines, metabolites (e.g., adenosine or kynurenines) and arachnidonic acid derivatives (e.g., PGE2) as well as the attraction, infiltration and accumulation of regulatory T-cells into the tumor stroma [reviewed in Vinay et al. (2)]. First successful clinical trials of melanoma patients with CTLA-4 and PD-1 inhibitors have highlighted the importance of these immune checkpoints for tumor growth as well as the potency of immune checkpoint blockade as therapeutic approach (3, 4) and currently numerous trials exploiting this principles are ongoing. Tumor-mediated immunosuppression also poses a major impediment in the development of adaptive immunotherapy protocols (5, 6). So far, most strategies to enhance the efficacy of T-cell therapy have aimed to improve tumor recognition by ectopic expression of tumor-antigen specific TCR or chimeric antigen receptors (CAR). In this respect, both approaches have shown great promise in various malignancies and recently first therapeutic applications of CAR T-cells have been approved by the FDA for the treatment of advanced ALL and B-NHL (7, 8) and have also been recommended by the EMA. However, it has become increasingly evident that other features of adoptively transferred T-cell therapies, such as trafficking, exhaustion, metabolic features and the response to suppressive modalities should also be targeted [reviewed in Sadelain et al. (9)]. Thus, recent studies on T-cell engineering have described modifications to overcome blockade of immunosuppressive signals in combination with antigen-recognition receptors (10–14). Advances in gene delivery systems have led to safe and clinically-approved protocols for the generation of engineered T-cells expressing multiple transgenes. Consequently, it can be envisioned that T-cells can be equipped with antigen-receptors as well as internal checkpoint inhibitors to guarantee optimal anti-tumor immunity. In this context, the definition of crucial immunoregulatory mechanisms, as well as measures to overcome their functions, should bear high therapeutic relevance.

Intracellular up-regulation of the second messenger cAMP acts as a common suppressive denominator for multiple extrinsic signals including PGE2 (15), adenosine (16), β-adrenergic agonists (17, 18), and histamine (19). Additionally, regulatory T-cells utilize cAMP to suppress effector T-cells either by production of adenosine (20, 21) or by direct transport through gap junctions into effector T-cells (22) [reviewed in Rueda et al. (23)]. The level of intracellular cAMP is regulated by the antagonizing actions of the adenylate cyclase (AC), generating cAMP, and phosphodiesterases (PDE), which degrade cAMP to 5′ -AMP. High cAMP levels in T-cells result in low activity and tolerogenic function (24). Along those lines, regulatory Tcells display high levels of cAMP due to up-regulation of AC9 (25) and down-regulation of PDE3B (26). Elevated cAMP levels promote distinct intracellular signaling events, whose net result is the suppression of effector functions such as proliferation, production of pro-inflammatory cytokines and cytotoxic activity of CD8<sup>+</sup> T-cells. Activation of protein kinase A (PKA) is one key pathway, which is triggered by cAMP. In turn, PKA activation initiates multiple intracellular signaling pathways including activation of the C-terminal Src kinase (Csk) and the CRE modulator (CREM) and its alternative splice product, the inducible cAMP early repressor (ICER) [reviewed by Mosenden and Tasken (24)]. PKA also activates PDEs, possibly initiating a negative feedback loop, which lowers cAMP levels and returns effector T-cell to a responsive state. Furthermore, studies indicate that cAMP also triggers PKA-independent intracellular signaling pathways that contribute to the modulation of T-cell activation (27, 28). Thus, cAMP constitutes a molecular hub between extrinsic modulators and intracellular signaling pathways.

Several studies suggest that manipulation of the cAMP-PKA axis can interfere with T-cell regulation. Pharmacological inhibitors of PDEs can enhance the regulatory capacity of Treg (29) while in contrast inhibition of AC or overexpression of PDE4C in CD4+CD25<sup>+</sup> Treg abrogates their suppressive function (30). Accordingly, PDE inhibitors have now been approved for the therapy of autoimmune and inflammatory diseases (31). On the other hand, cAMP-inducing signals strongly contribute to the immunosuppressive microenvironment of different tumors. Overexpression of the cyclooxygenase-2 (COX-2), which leads to the enhanced production of PGE2, has been reported for more than 80% of colorectal carcinomas (32). Similarly, COX-2 expression has been correlated to poor outcome in breast cancer patients (33). A comprehensive study of more than 500 human cancer tissue samples has established, that CD39 accompanied by increased adenosine production can be found in diverse types of solid as well as hematological malignancies (34). Expression of CD39 and /or CD73 have been found described as major mechanisms of immune tolerance in CLL (35, 36). Furthermore, the preferential infiltration of adenosine-producing CD39<sup>+</sup> Treg has been reported for solid tumors such as head and neck cancer (37), non-small cell lung cancer (38) and colorectal carcinomas (39) but also for hematological malignancies such as follicular lymphoma (40). Similarly, increased frequencies of adenosine-hydrolysing Treg were observed in peripheral blood of AML patients (41). Taken together, mechanisms which induce cAMP levels in tumorinfiltrating lymphocytes provide a major mechanism for the down-regulation of anti-tumor immune responses. Thus, cAMP can be considered as a highly relevant immune checkpoint.

In this proof of principle study, we aimed to restore T-cell activation in the presence of PGE2, adenosine or regulatory T-cells by ectopic overexpression of PDE4A. To that end, the human PDE4A cDNA was retrovirally transduced into CD4<sup>+</sup> and CD8<sup>+</sup> T-cells and their effector functions in the presence of the above described suppressive mechanisms were assessed in vitro.

#### MATERIALS AND METHODS

#### Ethical Considerations, Cell Isolation, and Culture

The study was approved by the local ethics committee of the Medical University of Vienna (EC number 1724/2012) and conducted according to the Declaration of Helsinki (1969, including current revisions) of the World Medical Association. After obtaining informed consent of study participants, peripheral blood mononuclear cells (PBMC) were isolated from healthy volunteers by standard Ficoll paque centrifugation. Total CD3<sup>+</sup> T-cells, CD4<sup>+</sup> and CD8<sup>+</sup> T-cells were isolated from PBMC using the respective T-cell Isolation Kits (Miltenyi Biotech, Bergisch Gladbach, Germany) according to the manufacturers' instructions. Purity was assessed by flow cytometric analyses using monoclonal antibodies against human CD3 (clone OKT3, eFluor 450 conjugated; eBioscience, San Diego, CA), human CD4 (clone OKT4, FITC conjugated; eBioscience) and human CD8 (OKT8, APC conjugated) and routinely found to be above 95%. All functional assays were performed in IMDM (GE Healthcare, Piscataway, NJ) supplemented with 10% fetal calf serum (GE Healthcare), 10µg/mL gentamycin and 1.25µg/mL amphotericin B (both Sigma Aldrich, St. Louis, MO).

#### Molecular Cloning

The cDNA encoding human PDE4A was amplified from a human T-cell cDNA library using the following primers: PDE4A for 5′ – CCCGCGAAGCTTGCCACCATGGAACCCCCGACCGTCCC – 3′ , PDE4A rev 5′ – CCCGCGGCGGCCGCTTTAGGTA GGGTCTCCACCTGACCC – 3′ (underlined sequences mark restriction enzyme sites). The cDNA was cloned into the pMMP-IRES-GFP vector using the restriction enzymes HindIII and NotI. The pMMP-FOXP3-IRES-GFP and the empty-control pMMP-IRES-GFP vector were described elsewhere (42).

#### Transfection of HEK-293 Cells

HEK-293 cells were transfected using the Ca2PO<sup>4</sup> precipitation method as described previously (43). In short, for the production of amphotropic T-cell transducing retrovirus supernatants, 30 µg of the pMD-MoMLV gag-pol, the envelope encoding pMD-GalV and transgene DNA in the pMMP-IRES-GFP vector were diluted in 900 µL ddH2O and 100 µL 2.5 M CaCl<sup>2</sup> were added and incubated for 5 min. Afterwards, 1 mL 2 × HBSS (Sigma Aldrich) was added and the mixture was spread on 293 cells at 10% confluency. On the next day, the mixture was removed and 10 mL fresh medium was added, and cells were cultured for 2 days two allow virus accumulation in the supernatant.

#### Retroviral Transduction

T-cells (5 × 10<sup>6</sup> /well) were stimulated in 6-well flat bottom plates with 5 × 10<sup>6</sup> anti-CD3/CD28 coated microbeads (Dynabeads, Invitrogen, Carlsbad, CA) and 300 U/ml IL-2 (Peprotech, London, UK) for 48 h. Retroviral transduction was performed by addition of cell-free retroviral supernatant in the presence of 8µg/ml polybrene (Sigma-Aldrich) followed by centrifugation at 900 g for 2 h. Twenty-four hours after transduction, cells were transferred to fresh medium containing 100 U/ml IL-2 and cultured for another 3–7 days. Typical transduction efficiencies ranged between 30 and 70%. Therefore, at the time point of the respective experiment, GFP<sup>+</sup> T-cells were stringently isolated by FACS-sorting on a FACS Aria II flow cytometer (Becton Dickinson).

#### Measurement of cAMP

Jurkat cells were incubated for 16 h with medium containing [ <sup>3</sup>H]-adenine (1 µCi/mL, PerkinElmer, Waltham, MA) to metabolically enrich the adenine nucleotide pool. Afterwards, cells were harvested, washed once with PBS and cells were resuspended in medium containing the phosphodiesterase inhibitor isobutylmethylxanthine (100µM; Sigma Aldrich). Subsequently, cells were stimulated with forskolin (30µM; sigma Aldrich) for 30 min to accumulate cAMP. After stimulation, cells were lysed in ice-cold 2.5% perchloric acid containing 0.1 mM cAMP for 30 min at 4◦C followed by neutralization of the lysis with 4.2 M KOH potassium hydroxide. ATP and cAMP were separated by sequential chromatography on columns containing Dowex 50-X4 (Sigma-Aldrich) and neutral alumina. Samples were then mixed with LSC-Universal Scintillation Cocktail (Roth, Karlsruhe, Germany) and the accumulated [3H]cAMP and [3H]ATP was measured by on a liquid scintillation counter (PerkinElmer).

#### Luciferase Assay

Jurkat E6 cells expressing an IL-2::Luciferase reporter construct (44) were either pre-treated for 30 min with the PKA inhibitors Rp-8-Br-cAMPS (200 nM) or H89 (1µM; both Sigma Aldrich) or transduced with an empty control-vector or the human PDE4A cDNA. 2 × 10<sup>6</sup> cells were activated with PMA (1 × 10−<sup>7</sup> M, Sigma Aldrich) and PHA (12.5µg/mL; Thermo Scientific, Waltham, MA). After 6 h, cells were lysed and luciferase activity was on a GloMax 96 well luminometer (Promega, Madison, WI).

#### Proliferation Assay, Co-cultures, and T-cell Cultures

1 × 10<sup>6</sup> FACS-sorted PDE4A- or control vector-transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells were activated with anti-CD3/anti-CD28 coated microbeads (beads to cell ratio 1:2) in the presence or absence of the indicated concentrations of PGE2 or the A2A-R agonist CGS21680 (both Sigma Aldrich) in 96-well flat bottom plates in triplicates. After 72 h, cells were pulsed with 1 mCi methyl-[3H] thymidine/well for an additional 18 h, and thymidine incorporation was measured on a PerkinElmer scintillation counter (PerkinElmer, Waltham, MA), as described (42).

For co-cultures, 5 × 10<sup>5</sup> FACS-sorted PDE4A- or control vector-transduced T-cells were activated with anti-CD3/anti-CD28 coated microbeads (beads to cell ratio 1:2) in the presence or absence of FOXP3-transgenic CD4<sup>+</sup> T-cells or FACSsorted CD4+CD25+CD127low thymic-derived regulatory T-cells at the indicated ratios. Proliferation was measured by thymidine incorporation as above. Values were corrected for proliferation of Treg in single culture.

In some experiments, transduced T-cells were FACS sorted for GFP+CD4<sup>+</sup> or GFP+CD8<sup>+</sup> cells and were activated with anti-CD3/anti-CD28 coated microbeads (beads to cell ratio 1:2) plus recombinant human IL-2 (10 U/mL). After seven and 14 days, cells were harvested, washed once in PBS and re-stimulated as above.

#### Cytokine Measurements

1 × 10<sup>6</sup> FACS-sorted PDE4A- or control vectortransduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells were activated with anti-CD3/anti-CD28 coated microbeads (beads to cell ratio 1:2) in the presence or absence of the indicated concentrations of PGE2. After 24 and 72 h supernatants were harvested and levels of IL-2 (24 h supernatants) and IFN-γ and TNF-α (72 h supernatants) were measured using specific ELISA (eBioscience) according to the manufacturers' recommendations.

#### Immunoblotting

1 × 10<sup>6</sup> FACS-sorted PDE4A- or control vector-transduced total T-cells were either left unstimulated or activated in the absence or presence of 200 nM PGE2 for 24 h using 5 × 10<sup>5</sup> anti-CD3/anti-CD28 coated microbeads. Cells were then harvested and lysed in RIPA buffer supplemented with protease inhibitors (Sigma). Cellular debris was removed by centrifugation at 25,000 × g at 4◦C for 15 min. Samples were normalized according to their protein content and were resolved by SDS-PAGE on 4– 12% gradient gels under reducing conditions (Life Technologies, Paisley, UK) followed by transfer onto PVDF membranes (GE Healthcare). Samples were then subjected to immunoblotting using the following antibodies: rabbit anti-S6 (clone 5G10), rabbit anti-p38 (clone D13E1), rabbit anti-ERK (polyclonal) and rabbit anti-Actin (clone D18C11; all New England Biolabs, Ipswich, MA). After incubation with a secondary anti-rabbit horse radish peroxidase-conjugated antibody, binding was visualized using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, Rockford, IL).

#### Flow Cytometry

All flow cytometry and FACS sorting experiments were conducted in accordance with the current guidelines (45). All analyses were performed on a FACS Canto II flow cytometer (Becton Dickinson) and analyzed using FlowJo software.

For intracellular measurement of PDE4A and cytokine expression, the Fix and Perm buffer set (An der Grub, Kaumberg, Austria) was used according to the manufacturers' recommendations. Cells were stained with a primary antihuman PDE4A antibody (clone 6B6; Abcam, Cambridge, UK) followed by staining with a PE-conjugated goat anti-mouse IgG antibody (Jackson Immuno Research; West Grove, PA). For measurement of intracellular cytokines, FACS-sorted Tcells were stimulated for 24 h (IL-2) or 48 h (IFN-γ and TNFα). During the last 6 h of culture Golgi-Stop reagent (Becton Dickinson, Palo Alto, CA) was added at a dilution of 1:1,500. Afterwards cells were harvested, fixed and permeabilized and stained with the respective antibodies against human IL-2 (PE or APC-conjugated; clone MQ1-17H12), IFN-γ (APCconjugated; clone 4S.B3) and TNF-α (PerCP-Cy5.5 conjugated; clone MAb11; all eBioscience). To determine phosphorylation of intracellular signaling proteins, FACS-sorted T-cells were activated for 24 h and processed as described previously. Cells were harvested and fixated in Fixation Buffer I (BD Phosflow, BD Biosciences) at 37◦C for 10 min. After washing in PBS + 0.5% BSA + 0.05% NaN3, cells were re-suspended in prechilled (−20◦C) Permeabilization Buffer III (BD Phosflow) and incubated on ice for 30 min. Afterwards, cells were washed twice in PBS + 0.5% BSA + 0.05% NaN<sup>3</sup> and stained with the respective antibodies (anti-phospho-S6RP S240; Alexa Fluor 647 conjugated; clone N4-41; anti-phospho-ERK T202/Y204; Pacific Blue conjugated; clone 20 A and anti-phospho-p38 T180/Y182; PE conjugated; clone 36/p38; BD Phosflow) or isotype-matched control antibodies for 60 min. For measurement of CD8<sup>+</sup> Tcell degranulation, control-vector or PDE4A-transduced CD8<sup>+</sup> T-cells were FACS-sorted 3 days after transduction and cocultured with BW target cells expressing a membrane-bound OKT3::scFv antibody (BW 3/2) labeled with 5µM eF450 CPD (eBioscience). Co-culture was performed in the presence or absence of 200 nM PGE2. Two hours after the start of coculture, Golgi-Stop reagent (Becton Dickinson, Palo Alto, CA) was added at a dilution of 1:1,500. After a total co-culture of 6 h, cells were harvested, surface stained with an eFluor670 conjugated anti-human CD107a antibody (clone eBioA4H3; eBioscience) and subsequently fixed and permeabilized as in the intracellular cytokine protocol and stained for intracellular expression of granzyme B (PE-conjugated; clone GB11; Becton Dickinson). Afterwards, CD107a and granzyme B expression on GFP+/eFluor450<sup>−</sup> T-cells were measured. For determination of surface expression of CD69 (eFluor450-conjugated; clone FN50), CD73 (eFluor450-conjugated; clone AD2), CD94 (APCconjugated, clone HP-3D9), CD244 (PE-conjugated, clone C17) PD-1 (PerCP-eFluor 710 conjuagted, clone J105), and TIM-3 (PE-conjugated; Rat IgG2a Clone #344823; R&D Systems, Minneapolis, MN) FACS-sorted T-cells were harvested from the respective cultures and washed twice in PBS + 5% BSA + 0.05% NaN<sup>3</sup> and stained for 30 min at 4◦C, followed by another wash step in PBS + 5% BSA + 0.05% NaN3.

# Cytotoxicity Assays

Purified human CD8<sup>+</sup> T-cells were transduced with either the PDE4A cDNA or an empty control-vector. Three days after transduction, GFP<sup>+</sup> T-cells were isolated by FACS-sorting and Tcells were co-cultured with 5 × 10<sup>4</sup> BW target cells expressing a membrane-bound OKT3::scFv antibody labeled with 5µM eF450 CPD (eBioscience) at the indicated ratios. After 6 h, cells were harvested, washed and 5 × 10<sup>4</sup> cell-counting beads and propidium iodide (final 100 ng/mL) were added. Viable BW target cells were quantified by flow cytometry by exclusion of GFP<sup>+</sup> T-cells and gating eF450+/PI<sup>−</sup> cells. Single culture of BW cells without transgenic T-cells served as control. Specific lysis was calculated according to the formula:

Number of viable BW cells (co-culture) per 10<sup>5</sup> counting beads/number of viable BW cells (control culture) per 10<sup>5</sup> counting beads.

#### Statistical Analyses

For multiple group comparisons, one-way ANOVA followed by Bonferroni correction was performed using GraphPad Prism version 5.0 for Windows (GraphPad Software, San Diego CA, USA). Two-group comparisons were performed using the Student's t-test. Data represent mean values + SD. Statistically significant values are denoted as follows: <sup>∗</sup>P < 0.05; ∗∗P < 0.01; and ∗∗∗P < 0.001.

FIGURE 1 | Overexpression of PDE4A counteracts cAMP mediated immunosuppression in Jurkat T-cells. (A) Jurkat IL-2P::Luc T-cells were pre-incubated with the PKA inhibitors Rp-8-Br-cAMPS (200 nM; left panel) or H89 (1,000 nM; right panel) and activated with PHA/PMA in the presence of the indicated concentrations of PGE2. After 6 h, cells were lysed and Luciferase activity was measured. Mean values ± SD from triplicate cultures from one representative experiment (*n* = 4) are shown. Squares: untreated cells, circles: cells treated with the respective inhibitor. (B) Jurkat T-cells were retrovirally transduced with either an empty control vector (left histogram) or the pMMP-PDE4A-IRES-GFP vector (right histogram). Following fixation and permeabilization of the cells, intracellular expression of PDE4A was measured using a mouse anti human PDE4A antibody followed by a PE-conjugated goat anti-mouse antibody. Histograms depict one representative experiment out of five. (C) Wildtype, control-vector transduced and PDE4A transduced Jurkat T-cells were pulsed with <sup>3</sup> [H]-adenosine overnight and adenylate cyclase activity was induced by addition of 30µM Forskolin. After 30 min, cells were lysed and the cAMP fraction was isolated by sequential chromatography and radioactivity was quantified on a scintillation counter. Mean values + SD from six individual experiments are depicted. (D) Control vector transduced (circles) or PDE4A transduced Jurkat IL-2P::Luc T-cells (squares) were activated with PHA/PMA in the presence of the indicated concentrations of PGE2. After 6 h cells were lysed and Luciferase activity was measured. Mean values ± SD from triplicate cultures from one representative experiment (*n* = 4) are shown. \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001.

#### RESULTS

# Pharmacological Inhibition of PKA Partially Restores IL-2 Production in Jurkat T-cells Under Suppression by PGE2

The PKA is one of the crucial signaling hubs for cAMP mediated immunosuppression. Thus, we first aimed to restore T-cell reactivity in the presence of PGE2 by use of the two well-defined PKA inhibitors Rp-8-Br-cAMPS and H89. In line with previous reports (46), we found that treatment of Jurkat T-cells with these inhibitors partially restored IL-2 promoter activity upon activation in the presence of suppressive concentrations of PGE2 (**Figure 1A**). This effect was especially pronounced at lower concentrations of PGE2 but a significant increase in IL-2 promoter activity was also found at highly suppressive concentrations (1,000 nM PGE2; n = 4; p < 0.01 for Rp-8-Br-cAMPS and p < 0.05 for H89 compared to mock-treated cells). However, neither inhibitor could completely abrogate the suppressive effects of PGE2.

#### Ectopically Expressed PDE4A in T-cells Efficiently Degrades cAMP Following Exposure to PGE2

Given that cAMP also triggers PKA-independent signaling pathways, we aimed to fully abrogate the suppressive effects of cAMP by ectopic overexpression of cAMP degrading phosphodiesterases. The human PDE4A cDNA was cloned into the retroviral pMMP-IRES-GFP vector, which guarantees highlevel overexpression with a strong correlation to expression of the GFP marker gene. Upon retroviral transduction into Jurkat T-cells, followed by isolation of GFP<sup>+</sup> cells by FACS-sorting, we found a robust expression of PDE4A, which was not present in control-vector transduced Jurkat cells (**Figure 1B**). To confirm functionality of the PDE4A transgene, we measured cAMP levels in untransduced, control-vector transduced and PDE4Atransduced Jurkat T-cell in response to the adenylate cyclase activator forskolin. As expected, a highly significant increase in cAMP levels could be observed in untransduced and controlvector transduced Jurkat T-cells, while PDE4A-expressing Jurkat cells showed only a slight increase in cAMP (**Figure 1C**). To further assess the functional impact of PDE4A overexpression, IL-2 promoter activity was measured in Jurkat T-cells following activation in the presence of PGE2. As above, activation of control-vector transduced Jurkat T-cells was strongly suppressed by PGE2 in a dose dependent fashion (**Figure 1D**). In contrast, PDE4A-overexpression led to a nearly complete restoration of activation. Even at high concentrations of PGE2 (200nM and 1,000 nM), IL-2 promoter activity reached 95.1 ± 5.1 and 93.3 ± 6.5% of promoter activity of the control (p = 0.57 and 0.48, respectively; n = 4; **Figure 1D**). Importantly, PDE4A and control-vector transduced Jurkat T-cells showed similar IL-2 promoter activity when activated in the absence of PGE2, indicating that PDE4A overexpression does not interfere with T-cell activation per se.

#### Activity of Ectopically Expressed PDE4A Counteracts the Suppressive Effects of PGE2 and A2A-R Agonists on T-cell Activation

In a next step, we assessed the effects of PDE4 overexpression on the activation of human peripheral blood T-cells. As early activation readouts we measured phosphorylation of the mTOR downstream target S6RP as well as the MAP kinases p38 and ERK following activation with agonistic anti-CD3/anti-CD28 antibodies in the absence or presence of PGE2. In FACS-sorted, control-vector transduced T-cells a significant downregulation of phosphorylation for all three molecules was found in the presence of PGE2 (p < 0.001, respectively; n = 4; **Figure 2A**). In contrast, FACS-sorted PDE4A-transduced T-cells showed equal up-regulation of ERK-, p38- and S6RP-phosphorylation in the absence and presence of PGE2 (**Figure 2A**). Under all conditions, total protein levels were unchanged (**Supplementary Figure 1**). Similarly, up-regulation of the early activation surface marker CD69 was strongly suppressed by PGE2 on control-vector transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells, while PDE4A-transduced T-cells showed a significant upregulation of CD69, especially in CD8<sup>+</sup> T-cells, in the presence of PGE2 (**Figures 2B,C**). In further experiments, proliferation of FACS-sorted controlvector transduced and PDE4A-transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells in the absence or presence of PGE2 or the A2A-R agonist CGS21680 was measured. Similar to the observations in Jurkat T-cells, overexpression of PDE4A in CD4<sup>+</sup> and CD8<sup>+</sup> T-cells did not alter their activation compared to their control-transduced counterparts under non-suppressive conditions (p = 0.751 for CD4<sup>+</sup> and p = 0.863 for CD8; n = 7; **Figures 3A,B**). A dose-dependent inhibition of proliferation by PGE2 was observed for control-transduced CD4<sup>+</sup> as well as CD8<sup>+</sup> T-cells following anti-CD3/anti-CD28 mediated activation. A significant decrease in proliferation was already observed at 50 nM PGE2 (p < 0.001; n = 7) and proliferation was suppressed by 62.7 ± 8.6% for CD4<sup>+</sup> and 73.5 ± 11.6% for CD8<sup>+</sup> T-cells in the presence of 200 nM PGE2 (p < 0.001; n = 7). In contrast, PDE4A-transgenic CD4<sup>+</sup> as well as CD8<sup>+</sup> T-cells were again completely resistant to the effects of PGE2, showing nearly equal proliferation levels in the presence as in the absence of PGE2 (**Figure 3A**). Similarly, culture of control-vector transduced T-cells in the presence of the A2A-R agonist CGS21680 strongly suppressed proliferation of CD4<sup>+</sup> as well as CD8<sup>+</sup> T-cells (reduction of proliferation by 63.1 ± 7.2% for CD4<sup>+</sup> and 57.7 ± 7.2% for CD8+; p < 0.001, respectively, n = 5). As above, overexpression of PDE4A also effectively abrogated suppression by the A2A-R agonist CGS21680 even under highly suppressive conditions (**Figure 3B**).

In order to further analyze T-cell function under these conditions, cytokine secretion levels from the corresponding cell culture supernatants were measured. In accordance with the proliferation experiments, control-vector transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells showed a pronounced and dosedependent reduction of the secretion of the effector cytokines IL-2, IFN-γ and TNF-α in the presence of PGE2. Again, the transduction of PDE4A completely abrogated the suppressive effects of PGE2 in these assays (**Figure 3C**). No significant difference in the cytokine secretion between PDE4A-transgenic T-cells and control-vector transduced T-cells was observed in the absence of PGE2 (**Figure 3C**). These observations were also confirmed by intracellular FACS analyses with control-vector and PDE4A-transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells (**Supplementary Figure 2**). In these experiments, we also found that PGE2 strongly reduced the number of polyfunctional IFN-γ <sup>+</sup>/TNF-α <sup>+</sup> CD8<sup>+</sup> Tcells. As above, FACS-sorted PDE4A-transgenic cells were fully resistant to suppression by PGE2 also in this read-out (**Supplementary Figure 3**).

increase in mean fluorescence intensity to the respective unstimulated cells. Mean values + SD from four independent experiments are depicted. (B) FACS-sorted control-vector transduced (gray filled histograms) or PDE4A (black line) transduced CD4<sup>+</sup> (left panels) or CD8<sup>+</sup> (right panels) human T-cells were either left unstimulated (dotted line and fine gray line) or were activated for 24 h in the absence (upper panels) or presence (lower panels) of 200 nM PGE2. After 24 h, CD69 expression was measured by flow cytometry. One representative experiment (*n* = 4) is depicted. (C) The percentage of CD69<sup>+</sup> control-vector (black bars) or PDE4A- (gray bars) transduced CD4<sup>+</sup> (left panels) and CD8<sup>+</sup> (right panels) T-cells either unstimulated or activated for 24 h in the absence (Medium) or presence of 200 nM PGE2 is depicted. Mean values + SD from four independent experiments are shown. n.s., not significant; \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001.

# PDE4A Overexpression Restores the Cytotoxic Function of CD8<sup>+</sup> T-cells Under Suppression by PGE2

One key anti-tumor function of the adaptive immune system is the removal of malignant cells by cytotoxic CD8<sup>+</sup> T-cells. Accordingly, we tested the impact of PDE4A-overexpression on the cytotoxic function of CD8<sup>+</sup> T-cells. As model target cells we used the mouse thymoma BW5147 cell line which was modified to stably express a membrane-bound OKT3 single chain fragment variable (mb-OKT3scFv) as surrogate T-cell ligand (further on referred to as BW). Following co-incubation of FACS-sorted control-vector transduced and PDE4A-transduced CD8<sup>+</sup> T-cells with the BW target cells, a strong up-regulation of the degranulation marker CD107a was detected (**Figures 4A,C**).

FIGURE 3 | Overexpression of PDE4A restores early proliferation and cytokine secretion in peripheral blood T-cells under suppression by PGE2 or adenosine-receptor agonists. FACS-sorted control-vector transduced (squares) or PDE4A transduced (circles) human peripheral blood CD4<sup>+</sup> (left panels) or CD8<sup>+</sup> T-cells (right panels) were activated with agonistic anti-CD3/anti-CD28 antibodies in the absence or presence of the indicated concentrations of PGE2 (A) or the A2A-R agonist CGS21680 (B). After 72 h, cells were labeled with [3H]-thymidine for another 18 h and thymidine incorporation was measured. Mean values ± SD from triplicate cultures from one representative donor (*n* = 6) are depicted. (C) 24 h (IL-2) or 72 h (IFN-γ, TNF-α) after activation supernatants were harvested and concentrations of the indicated cytokines were measured by ELISA. Mean values ± SD from triplicate cultures from one representative donor (*n* = 7) are depicted. n.s., not significant; \*\*\**p* < 0.001.

This effect was equally pronounced in both cell-types (mean fluorescence intensity increase 858 ± 72 for control-vector and 785 ± 107 for PDE4A; p = 0.15; n = 5). In the presence of 200 nM PGE2, CD107a up-regulation was nearly abrogated in control-vector transduced T-cells. Under these conditions, PDE4A-transgenic cells were able to mount a robust upregulation of CD107a, which amounted to 64.1 ± 9.5% of the level in the absence of PGE2 (p < 0.01; **Figures 4A,C**). The effector molecule granzyme B (GrzB) is essential for the cytotoxic function of CD8<sup>+</sup> T-cells. Accordingly, we also combined measurement of the surface levels of the degranulation marker CD107a with intracellular GrzB expression. The percentage of CD107a+/GrzB<sup>+</sup> cells was strongly down-regulated by PGE2 in control-vector transduced cells (10.3 ± 1.5 vs. 0.4 ± 0.2%; p < 0.01; n = 4). As above, PDE4A-transduced cells showed similar levels of CD107a+/GrzB<sup>+</sup> cells in comparison to control-vector transduced cells under non-suppressive conditions and were able to strongly induce CD107a+/GrzB<sup>+</sup> cells in the presence of PGE2 (9.9 ± 1.1 vs. 7.4 ± 0.8%; p < 0.05; n = 4; **Figures 4B,C**). Finally, we evaluated the lytic function of PDE4A-transduced CD8<sup>+</sup> T-cells in a FACS-based cytotoxicity assay. We found that both control-vector transduced and PDE4A-transduced CD8<sup>+</sup> Tcells were able to efficiently lyse BW cells in a dose-dependent manner. In the absence of PGE2, no significant difference in the cytotoxic capacity between the two cell types was observed (**Figure 4D**). Again, upon pre-incubation with 200 nM PGE2, a significant down-regulation in cytotoxicity was observed in control-vector transduced CD8<sup>+</sup> T-cells (p < 0.001 at all effector: target ratios; n = 4). In contrast, PDE4A-transduced T-cells were again fully resistant to the suppressive effects of PGE2 and showed a similar cytotoxic potency in the presence and absence of PGE2 (**Figure 4D** and **Supplementary Figure 4**).

#### PDE4A-transgenic T-cells Are Partially Resistant to Suppression by Regulatory T-cells

An important mechanism of Treg-mediated immunosuppression is conferred by an increase of cAMP in effector T-cells. Consequently, we hypothesized that overexpression of PDE4A in T-cells may also generate resistance to suppression by Treg. In a first proof of principle experiment, we co-cultivated FACSsorted control-vector transduced or PDE4A-transduced T-cells with FOXP3-transgenic regulatory T-cells. As expected a robust and dose-dependent inhibition of control-vector transduced effector T-cells was observed which amounted to 87.5 ± 9.8% at a Treg: Teff ratio of 1:1. In contrast to the experiments described above, PDE4A-overexpressing T-cells showed a similar reduction in proliferation under these conditions (89.6 ± 7.3%; p = 0.749 compared to control-vector transduced T-cells, n = 3; **Figure 5A**). However, PDE4A-overexpression resulted in enhanced activation at lower Treg to Teff ratios. This effect could already be observed at a Teff: Treg ratio of 1:2 (78 ± 6.4 vs. 62.4 ± 4.0% reduction of proliferation; p < 0.01) and was especially pronounced at ratios of 1:4 and 1:8 (56.3 ± 4.2 vs. 24.1 ± 5.7 and 37.5 ± 4.9 vs. 8.7 ± 3% reduction of proliferation, respectively; p < 0.001 and p < 0.01; **Figure 5A**). In

FIGURE 4 | Overexpression of PDE4A restores cytotoxic function of CD8<sup>+</sup> T-cells in the presence of PGE2. (A) FACS-sorted control-vector transduced (gray filled histograms) or PDE4A transduced (black line) human CD8<sup>+</sup> T-cells were co-cultured with BW cells expressing a membrane-bound OKT3::scFv at a 10:1 ratio in the absence (Medium) or presence of 200 nM PGE2. After 6 h, expression of CD107a/LAMP-1 was measured by flow cytometry. Dotted black line: unstimulated control; dotted gray line: unstimulated PDE4A. One representative experiment is depicted (*n* = 6). (B) Co-staining of surface expression of CD107a and intracellular expression of Granzyme B on control-vector transduced (upper panel) and PDE4A-transduced (lower panel) cells under the same conditions as in (A). One representative experiment is depicted (*n* = 4). (C) Cumulative data are depicted as difference to unstimulated cells at the indicated conditions. black bars: control-vector transduced; gray bars: PDE4A transduced; mean values + SD are shown. (D) Specific lysis of BW target cells by control-vector transduced or PDE4A transduced human CD8<sup>+</sup> T-cells in the absence or presence of 200 nM PGE2 at the indicated effector: target cell ratios. Mean values + SD from six independent experiments are shown. n.s., not significant; \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001.

suppression by regulatory T-cells. Control-vector transduced (squares) or PDE4A transduced (circles) human T-cells were co-cultured with FOXP3 transduced (A) or peripheral blood tTreg (B) at the indicated ratios. After 72 h, cells were labeled with [3H]-thymidine for another 18 h and thymidine incorporation was measured. Mean values ± SD from triplicate cultures from one representative donor (*n* = 4) are depicted. (C,D) Control-vector transduced (squares) or PDE4A transduced (circles) human T-cells were co-cultured with peripheral blood tTreg for 24 h and expression of IL-2 (C) and IFN-γ (D) was measured in the GFP<sup>+</sup> effector T-cells by intracellular flow cytometry. Mean values ± SD from three independent experiments are depicted. n.s., not significant; \**p* < 0.05; \*\*\**p* < 0.001.

accordance, a similar pattern was observed when control-vector transduced and PDE4A-transduced T-cells were co-cultured with peripheral blood CD4+CD25+CD127low tTreg from the same donor (**Figure 5B**). To further assess the impact of PDE4Atransduction in co-cultures with regulatory T-cells, we also measured cytokine production in the GFP<sup>+</sup> transgenic effector T-cells using intracellular flow cytometry. In contrast to the observations from the proliferation assays, intracellular levels of IL-2 and IFN-γ were less affected in PDE4A-transgenic Tcells than in control-vector transduced T-cells (**Figure 5C**). At a Treg: Teff ratio of 1:1, the percentage of IL-2 positive cells was reduced from 8.7 ± 0.3 to 3.9 ± 0.3% (p < 0.001, n = 3, **Figure 5C**) in control-vector transduced T-cells, while only a slight reduction was observed in PDE4A-transduced T-cells (8.6 ± 0.6 to 6.2 ± 0.4%; p < 0.01). At lower Treg: Teff ratios IL-2 production in PDE4A-transgenic T-cells recovered to levels under no suppression, while the percentage of IL-2 positive cells was still strongly reduced in control-vector transduced Tcells (4.8 ± 0.3 and 6.6 ± 0.5 vs. 6.8 ± 0.2% and 7.8 ± 0.6%

at 1:2 and 1:4 tTreg: Teff ratios; p < 0.001 and p < 0.01; n = 3). Similarly, IFN-γ levels in control-vector transduced T-cells were strongly suppressed by PGE2 from 7.0 ± 0.4 to 2.7 ± 0.5% at 1:1 tTreg:Teff ratios (p < 0.001; n = 3). In PDE4A-transduced T-cells only a slight reduction of IFN-γ was observed under these conditions (7.1 ± 0.3 to 6.1 ± 0.7%, p = 0.27; n = 3) which was not statistically significant. Along those lines IFN-γ production in PDE4A-transduced T-cells was completely restored at lower tTreg:Teff ratios (**Figure 5C** and **Supplementary Figure 5**).

#### PDE4A Counteracts Upregulation of Inhibitory Surface Molecules and Does Not Affect Exhaustion Upon in vitro Culture

Previous studies have established that PGE2 does not only directly suppress T-cell activation, but also leads to the upregulation of inhibitory molecules which could further potentiate an immunosuppressive environment (47). In line with these findings, culture of FACS-sorted control-vector transduced Tcells in the presence of PGE2 led to a significant up-regulation of the inhibitory ligand CD94 as well as the adenosine producing ectoenzyme CD73 on CD8<sup>+</sup> T-cells. In accordance with the functional experiments described above, PDE4A overexpression counteracted this effect of PGE2 and completely abrogated upregulation of these inhibitory molecules (**Figures 6A,B**).

Apart from tumor-endogenous immunosuppression, exhaustion of T-cells also poses an impediment for the efficacy of adoptive immunotherapies. Thus, we tested the effects of PDE4Aoverexpression on T-cell exhaustion upon repetitive stimulation using anti-CD3/anti-CD28 coated beads and recombinant human IL-2 as in vitro stimuli. As expected, prolonged culture of control-vector transduced CD4<sup>+</sup> and CD8<sup>+</sup> T-cells led to the up-regulation of the exhaustion-associated cell surface markers CD244, PD-1 and TIM-3. Importantly, overexpression of PDE4A did not alter the phenotype of cultured CD4<sup>+</sup> and CD8<sup>+</sup> T-cells in comparison to control-vector transduced T-cells (**Figure 6C** and **Supplementary Figure 6**).

# DISCUSSION

In this study we describe in a proof of principle in vitro study that overexpression of PDE4A in primary human effector CD4<sup>+</sup> and CD8<sup>+</sup> T-cells is sufficient to overcome the effects of cAMP, which acts as signaling hub for the suppressive function of soluble molecules such as PGE2 and adenosine as well as regulatory Tcells. PDE4A-transgenic T-cells were fully able to mount effector functions in the presence of PGE2 and adenosine and did not up-regulate PGE2-induced suppressive surface molecules such as CD73 and CD94. Furthermore, PDE4A overexpression also led to better activation of T-cells in the presence of Treg. Importantly, this manipulation did not affect T-cell functionality under nonsuppressive conditions. Taken together, we have shown that PDE4A is a safe and efficient immune checkpoint inhibitor, which efficiently disarms multiple suppressive entities. This knowledge could be essential for the improvement of adoptive T-cell therapies.

Adoptive T-cell therapies offer a promising directed approach to treat malignant diseases. Clinical trials have shown encouraging results especially in the therapy of hematological malignancies and first CAR T-cell therapies for refractory hematological diseases have been approved by the US Food and Drug Administration. While these developments highlight the enormous potential of adoptive T-cell therapy, not all trials have been successful so far and especially response rates in solid tumors have been poor (48). Consequently, further development of adoptive immunotherapy regimen is clearly warranted.

In principle, adoptively transferred T-cells have to fulfill at least four criteria to successfully target malignant cells: homing to the tumor, recognition of tumor cells, the mounting of effector functions under immunosuppressive conditions and abundant cytotoxicity. So far, most research has focused on the improvement of tumor-cell recognition and the concomitant full activation of the T-cells. In this regard, either tumor-antigen specific TCR or so called chimeric antigen receptors (CAR) have been introduced into patient T-cells (6). These antigenrecognition tools have been constantly refined, e.g., the use and recombination of different intracellular signaling domains derived from various co-stimulatory molecules and cytokine receptors (49) has led to the evolution of improved generations of CAR constructs. However, also strategies to optimize the other features described above have come into the spotlight in recent years (50).

In the light of the clinical efficacy of so-called checkpoint inhibitors it has become increasingly evident, that adoptive Tcell therapies, especially those targeted against solid tumors, might similarly profit from mechanisms which counteract immunosuppressive signals. In this regard, combination of CAR T-cells with checkpoint inhibitors has led to favorable results in pre-clinical models (51, 52) as well as clinical studies [reviewed in Yoon et al. (14)]. Generation of Tcells for adoptive transfer requires the ex vivo isolation, expansion and engineering of patient T-cells. It can be envisioned that molecules counteracting immunosuppressive signaling in T-cells, i.e., internal checkpoint inhibitors, can be additionally introduced during this process. In this regard, first attempts to target major immune checkpoints have been published. As early as 2002, Bollard et al. (53) could show that overexpression of a dominant-negative TGF-beta receptor strongly enhanced anti-tumor immunity in vitro, which was followed up by the same group in murine models (54). Similarly, approaches to modify PD1 signaling using overexpression of a dominant-negative PD-1 receptor (55), chimeric PD-1 fused to the intracellular signaling domain of CD28 (56), shRNAmediated knockdown (55) or CRISPR/Cas9-mediated knockout of PD-1 on adoptively transferred T-cells (57) has led to improved tumor recognition and clearance in murine models. Taken together these studies provide first important insights into the feasibility and the therapeutic potential of intrinsic immune checkpoints.

Malignant cells do not present as uniform cell-populations and among others the tumor-microenvironment and the acquisition of immunosuppressive mechanisms/immune

FIGURE 6 | Overexpression of PDE4A counteracts upregulation of inhibitory surface molecules under suppression by PGE2 but does not affect T-cell exhaustion in culture. (A) Control-vector transduced or PDE4A transduced human CD8<sup>+</sup> T-cells were activated in the absence (left panels) or presence (right panels) of 200 nM PGE2. After 7 days, surface expression of CD94 (upper lane) or CD73 (lower lane) were determined by flow cytometry. Dot blots from one representative experiment (*n* = 4) are depicted. (B) Statistical analysis of percentages of CD94<sup>+</sup> and CD73<sup>+</sup> cells of the indicated specimen, black bars: control-vector transduced; gray bars: PDE4A transduced; mean values + SD are shown. (C) Control-vector transduced (squares and triangles) or PDE4A transduced (dots and inverted triangles) human CD4<sup>+</sup> and CD8<sup>+</sup> T-cells were activated with agonistic anti-CD3/anti-CD28 antibodies and recombinant human IL-2 (10 U/mL) and restimulated every seven days. After each week of culture, surface expression of the exhaustion associated markers CD244, PD-1 and TIM-3 were measured by flow cytometry. Mean values ± SD from three independent experiments are depicted. n.s., not significant; \**p* < 0.05.

checkpoints may strongly vary between different tumor entities and even individuals affected from the same type of tumor (58). Increasing evidence exists that the landscape of immune-evasion is very complex and features many cellular and soluble entities as well as metabolic alterations (59). Among others, mechanisms which induce cAMP in tumor-infiltrating effector T-cells may play a prominent role (60). In this regard, the adenosine generating ectoenzymes CD39 (61, 62) and CD73 (62) as well as the A2AR (63) have been proposed as immune checkpoints and expression of these molecules in tumor tissue has been correlated with poor prognosis (64). Similarly, COX-2 expression leading to high PGE2 levels has been described in different tumors (32, 33). Several studies have aimed to assess the efficacy of blockade of these mechanisms in combination with CAR T-cells. In this context, most researchers have evaluated the application of small molecule inhibitors to the above described molecules (65). While these approaches have led to significantly enhanced CAR T-cell activity in murine models, several obvious pitfalls exist. First, targeting a single mechanism in a complex immunosuppressive environment might not suffice to restore full CAR T-cell activity. Furthermore, both adenosine as well as PGE2 have a multitude of physiological functions other than immunoregulation, which might be similarly affected by systemic application of the respective inhibitors. Consequently, other approaches have aimed to target intracellular signaling pathways downstream of cAMP, most prominently the PKA. In one study, addition of the PKA-inhibitor Rp-8-Br-cAMPS could significantly improve T-cell responses in the presence of A2AR agonists. However, in line with our observations, this treatment could not fully abrogate the suppressive effects of cAMP (46). In a recent study, overexpression of a peptide blocking binding of PKA to the scaffold protein Ezrin was shown to improve T-cell activation under suppression by both PGE2 as well as adenosine in vitro and in vivo (66). However, multiple studies have pointed out that PKA signaling is not solely responsible for the T-cell suppressive function of cAMP, and PKA-independent signaling via other molecules such as Epac have been described (27, 32, 67, 68). Consequently, we have chosen to target cAMP as it acts as the central hub linking different extrinsic immunosuppressive mediators to several downstream signaling pathways. In our approach, we followed the strategy first described by Klein et al. (30), who used overexpression of a PDE to degrade cAMP in tTreg. These studies have led to important insights into the relevance of cAMP in Treg function. However, the possibility to use this mechanism as checkpoint inhibitor in effector T-cells has not been considered so far. In our experiments, overexpression of PDE4A in primary human CD4<sup>+</sup> and CD8<sup>+</sup> T-cells efficiently protected them from cAMP-mediated immunosuppression. PDE4A-transgenic T-cells were fully resistant to the suppressive effects of PGE2 and A2AR agonists in all activation and effector readouts measured. This included proliferation, the secretion of pro-inflammatory cytokines, the lysis of target cells as well as the up-regulation of inhibitory molecules. Similarly, effector T-cell responses under suppression by regulatory T-cells were significantly improved by PDE4A overexpression. Generation of transgenic T-cells for adoptive therapy relies on the genetic manipulation of donor lymphocytes which may contain residual Treg. Considering the data from Klein et al. (30) and from our study, our approach would thus not only protect effector T-cells from the suppressive effects of potential bystander Treg, but also efficiently disarm these cells. Importantly, PDE4A-transgenic T-cells did not show increased reactivity in comparison to control-vector transduced T-cells under non-suppressive conditions, thus limiting the potential for increased immune mediated side effects during therapy. Importantly, overexpression of PDE4A also did not affect viability of the transgenic T-cells. Furthermore, the use of a human physiological enzyme should also minimize the potential immunogenicity of the introduced molecule, which could in principle limit the efficacy of the adoptively transferred T-cells. It must also be considered that 11 different PDE families have been characterized in human cells so far which all show different affinity and specificity toward cAMP (69). It remains to be assessed, whether overexpression of PDE4A is the optimal approach, or whether other PDEs show an even better profile as immune checkpoint inhibitors.

From these presented points it can be concluded that the described approach thus offers the possibility for direct translation into therapeutic settings. However, our observations also shed light on basic principles of T-cell biology. Indeed, the selective abrogation of cAMP signaling by the overexpressed PDE4A provides a model system to study basic biological principles of this pathway for T-cell activation under nonsuppressive and suppressive conditions. First, abrogation of cAMP signaling in CD4<sup>+</sup> and CD8<sup>+</sup> effector T-cells did not influence major effector functions under non-suppressive conditions. Thus, the induction of cAMP does not constitute an intrinsic feedback mechanisms that is required for physiological T-cell homeostasis, but rather a system which allows inhibition of T-cell activation in immunosuppressive environments generated by tolerogenic myeloid and lymphoid cells. This hypothesis is also supported in our long term in vitro culture experiments of the transduced T-cells. There, we found that PDE4A-transgenic T-cells were not more prone to up-regulate surface markers of T-cell exhaustion, demonstrating that cAMP signaling is not intrinsically involved in T-cell exhaustion upon repetitive activation. Furthermore, our experiments also provide intriguing insights into the interplay between tTreg and Teff and the importance of cAMP signaling in this context. A large body of literature exists which shows that Treg utilize different molecular mechanisms which subsequently trigger different intracellular signaling cascades in Teff to suppress the activation of the latter (70). Among them, the generation of extracellular adenosine by the ectoenzymes CD39 and CD73 (20, 21) as well as the direct transfer of cAMP via gap junctions (22) can lead to an increase in intracellular cAMP in Teff. Notably, in our experiments, the abrogation of cAMP signaling by overexpression of PDE4A could only partially overcome the suppression of Teff proliferation in co-culture with Treg. Especially at high Treg: Teff ratios, PDE4A-transgenic T-cells showed no or only slightly improved proliferation compared to control-vector transduced T-cells. The overexpression levels of PDE4A reached with our vector system were able to robustly abrogate cAMP levels even under high adenylate cyclase activity. Thus, it seems improbable that tTreg might overwhelm the transgenic PDE4A by abundant induction and transfer of cAMP into Teff. Consequently, our observations suggest that proliferation of Teff is suppressed by Treg using additional mechanisms other than cAMP. In contrast, production of cytokines such as IL-2 and IFN-γ is highly sensitive to cAMP increases but independent of other Treg mechanisms. It has been amply demonstrated that different T-cell effector functions such as proliferation and cytokine production are governed by different intracellular signaling pathways (71). Thus, it seems conceivable that suppression of these effector functions might also require different inhibitory signals. The described multitude of suppressive Treg mechanisms might serve to selectively suppress distinct Teff functions. Further studies to assess this intriguing hypothesis are clearly warranted and may help to define novel strategies how to fine-tune the effects of Treg.

In conclusion we here show in a proof of principle study that overexpression of a cAMP-degrading phosphodiesterase can fully protect effector T-cells from suppression by diverse soluble mediators such as PGE2 and adenosine in vitro. Furthermore, PDE-transgenic T-cells show improved responses in the presence of regulatory T-cells. Thus, our approach constitutes a novel internal checkpoint inhibitor against multiple immunosuppressive entities to improve adoptive T-cell therapy. It will be imperative to define further central signaling hubs of T-cell suppression and strategies to counteract their function. This is especially important in the light of recent data, that tumors can develop resistance to blockade of singular immune checkpoints by the use of other immunosuppressive entities (72). Thus, use of PDE4A also in combination with other internal immune checkpoint inhibitors could significantly improve the efficacy of adoptive T-cell regimen.

#### DATA AVAILABILITY

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

#### REFERENCES


#### ETHICS STATEMENT

The study was approved by the local ethics committee of the Medical University of Vienna (EC number 1724/2012) and conducted according to the Declaration of Helsinki (1969, including current revisions) of the World Medical Association. Blood was drawn after obtaining written informed consent of the study participants.

#### AUTHOR CONTRIBUTIONS

KS and RM planned the project and designed experiments. KS, KG, RS, MG, and LZ performed all described functional experiments and flow cytometric analyses. MT and EZ-B performed cAMP assays. JL and PS assisted with cloning of the PDE4A construct and provided reagents. DT and WP assisted with thymidine incorporation assays. IS assisted with flow cytometric analyses and provided reagents. KS and RM wrote the manuscript. All authors critically read and approved the manuscript.

#### FUNDING

This work was supported by grants from the Medical Scientific Fund of the Mayor of the City of Vienna (Grant 13040 and 15099), from the Austrian Society for Laboratory Medicine and Clinical Chemistry and from the Austrian Science Funds (FWF, projects P29654-B30 and SFB-F4609).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01790/full#supplementary-material


patients with chronic lymphocytic leukemia. Clin Lymphoma Myeloma Leuk. (2011) 11:367–72. doi: 10.1016/j.clml.2011.06.005


human T cells engineered to target NY-ESO-1 to control tumor growth after adoptive transfer. Clin Cancer Res. (2016) 22:436–47. doi: 10.1158/1078-0432.CCR-15-1070


**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 Schmetterer, Goldhahn, Ziegler, Gerner, Schmidt, Themanns, Zebedin-Brandl, Trapin, Leitner, Pickl, Steinberger, Schwarzinger and Marculescu. 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.

# CD8<sup>+</sup> T Lymphocyte and NK Cell Network: Circuitry in the Cytotoxic Domain of Immunity

Roman V. Uzhachenko<sup>1</sup> and Anil Shanker 1,2,3,4 \*

*<sup>1</sup> Department of Biochemistry, Cancer Biology, Neuroscience and Pharmacology, School of Medicine, Meharry Medical College, Nashville, TN, United States, <sup>2</sup> Host-Tumor Interactions Research Program, Vanderbilt-Ingram Comprehensive Cancer Center, Vanderbilt University School of Medicine, Nashville, TN, United States, <sup>3</sup> Vanderbilt Center for Immunobiology, Vanderbilt University School of Medicine, Nashville, TN, United States, <sup>4</sup> Vanderbilt Institute for Infection, Immunology and Inflammation, Vanderbilt University School of Medicine, Nashville, TN, United States*

Multiple effector layers in the immune system ensure an optimal temporal and spatial distribution of immune defense. Cytotoxic innate lymphoid natural killers (NK) and adaptive CD8<sup>+</sup> T lymphocytes (CTL) interact to elicit specific cytolytic outcomes. The CTL carry antigen-specific T cell receptors (TCR) to recognize cognate peptides bound with major histocompatibility complex class-I (MHC-I) or human leukocyte antigen (HLA) molecules on target cells. Upon TCR engagement with MHC-I:peptide at a threshold of avidity, T cell intracellular programs converge into cytolytic activity. By contrast, NK cells lack antigen-specific receptors but express a repertoire of highly polymorphic and polygenic inhibitory and activating receptors that bind various ligands including MHC and like molecules. A highly calibrated maturation enables NK cells to eliminate target cells with lowered or absent MHC-I or induced MHC-I-related molecules while maintaining their tolerance toward self-MHC. Both CTL and mature NK cells undergo membranous reorganization and express various effector molecules to eliminate aberrant cells undergoing a stress of transformation, infection or other pathological noxa. Here, we present the cellular modules that underlie the CTL–NK circuitry to maximize their effector cooperativity against stressed or cancerous cells.

Keywords: CD8 T cells (CTL), natural killer cells (NK), lymphocyte crosstalk, immune networks, cytolytic function, effector cooperativity, cancer, immunotherapy

# INTRODUCTION

The organization of immune cells into a social network (1) underscores the functional complexity inherent in its design to defend against any pathological noxa. Signals mediated by acting-at-a-distance molecules or juxtaposing intercellular contacts lead to formation of responsive modules necessary for the execution of effector functions. Therefore, organizing functional modules into networking units lets the immune system accomplish a broader task at the level of organism without disturbing organismal homeostasis (2).

Upon a pathological insult, both cellular and humoral immune responses develop through a typical Darwinian selection process. A multitude of cues guide this process: what is the nature of antigen? What is the dose of antigen? Is the antigen self or foreign? What is the appropriate magnitude of response to the antigen? When will the immune contraction phase start? These

#### Edited by:

*Nadia Caccamo, University of Palermo, Italy*

#### Reviewed by:

*Gabriella Pietra, University of Genoa, Italy Paolo Dellabona, San Raffaele Scientific Institute (IRCCS), Italy*

> \*Correspondence: *Anil Shanker ashanker@mmc.edu*

#### Specialty section:

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

Received: *23 March 2019* Accepted: *29 July 2019* Published: *13 August 2019*

#### Citation:

*Uzhachenko RV and Shanker A (2019) CD8*<sup>+</sup> *T Lymphocyte and NK Cell Network: Circuitry in the Cytotoxic Domain of Immunity. Front. Immunol. 10:1906. doi: 10.3389/fimmu.2019.01906*

**298**

questions guide the selection of cellular subsets and molecules to launch an appropriate immune response in a universe of diverse antigens. This cannot be tackled by homotypical nodes of lymphocytes acting in isolation. It requires cooperativity from different immune cell subsets.

A repertoire of membrane receptors along with a milieu of intracellular secretory molecules provide input signals to drive various transcriptional master regulators that commit lymphocyte subsets to send a certain array of outputs. Thus, lymphocytes scan the environment for information from other cells as their input and vice versa. In other words, immune cell subsets located in proximity can be organized into a unit of mutual receiver-sender modules. In terms of the information theory, this process may be represented as a communication channel for computation. Notably, immune cells respond to instructions from extracellular environment to exercise plasticity in choosing specific cell subsets to launch a dynamic immune response that fits ad hoc to input information (2).

In a biological world, different cell types form a stable circuit if they constitutively share information via exchange of molecules. For example, the platelet-derived growth factor (PDGF) secreting macrophages that exclusively express receptors for colony stimulating factor-1 (CSF1) form a stable two-cell circuit with PDGF receptor-expressing fibroblasts, which also supply macrophages with CSF1 (3). Similarly, cancer cells form a reverse Warburg circuit with cancer-associated fibroblasts (CAF) wherein they supply transforming growth factor-β and reactive oxygen species to CAF, triggering their glycolysis, and lactate production. In turn, CAF provide tumor cells with lactate, which is converted into pyruvate and utilized in mitochondrial metabolism necessary for tumor growth and proliferation (4, 5).

In a multicellular network, nodes of individual cells (modules) communicate with each other and act as one regulatory functional unit (analogous to an electrical circuit comprising a transistor, resistor, capacitor, or inductor with logic gates) to manifest the cumulative function of the aggregate of cells. We have called such interconnected modules a circuit in the multicellular network. Here, we discuss the cellular circuitry underlying the two cytotoxic lymphocyte subsets, CD8<sup>+</sup> Tcells (CTL) and natural killers (NK). Immunosurveillance and cytolytic activity toward transformed or infected cells may benefit from a cross-talk between the CTL and NK cells, which could be recruited at different stages of immune control.

#### CELLULAR CIRCUITRY UNDERLYING CTL FUNCTION

CTL recognize their targets via a wide repertoire of membraneexpressed T-cell receptors (TCR) present as an octameric complex of variable TCR-α and β chains with three dimeric signaling modules: CD3δ/ε, CD3γ/ε, and CD247ζ/ζ or ζ/η (6). TCR diversity in T-cells is generated by an integration of processes including somatic VDJ recombination, palindromic and random nucleotide additions, and extra-thymic peripheral TCR revision (7, 8). Each TCR complex recognizes a specific MHC/HLA:peptide (antigen) complex cross-presented by the professional antigen-presenting cells (APC) such as dendritic cells (DC), B-cells and macrophages, or presented directly by the target cells.

T-cell interaction with DC represents a classical two-cell circuit wherein an immunological synapse (IS) is maintained by the engagement of multiple pairs of DC-expressed ligands with T-cell-expressed receptors such as ICAM-1 (CD54):LFA-1 (CD11a-CD18 heterodimer), CD80/CD86:CD28, MHCpeptide:TCR, CD40:CD40L (CD154) (9). It is important to note that CD40:CD40L binding ushers DC to secrete IL-12, an instructive cytokine for T-helper-1 (Th1) development, whereas MHC-peptide:TCR and CD80/CD86:CD28 interactions trigger production of IL-2, a proliferative cytokine, by T-cell subsets (10). Moreover, by a prolonged IS, DC and CTL provide each other with survival signals. The costimulatory molecule CD28 engagement on CTL activates the PI3K/Akt survival pathway, and prevent anergy (hyporesponsiveness) by upregulating Bcl-xL and downregulating CD95L (10). The growing list of costimulatory receptors expressed on T-cells includes 4-1BB (CD137), OX40 (CD134), TNFRSF7 (CD27), ICOS (CD278), TNFRSF8 (CD30), LFA-2 (CD2), DNAM-1 (CD226), and NKG2D (CD314) among others (11–13). DC survival is associated with the stimulation of CD40:CD40L axis (9). Lately, the importance of specific Notch receptor-ligand interactions has also been demonstrated in the antitumor DC–CTL network (14–16).

For CTL activation, a three-cell circuitry has been proposed. Initial model assumed that a single DC can bind both CD4<sup>+</sup> T and CD8<sup>+</sup> T cells through the expression of MHC-II and MHC-I molecules, respectively. In this three-cell interaction, CD4<sup>+</sup> T while synapsed with DC supply IL-2 whereas DC provide co-stimulatory signals to CD8<sup>+</sup> T cells (17). Later, this model was modified to an alternative view in which DC sequentially interact with CD4<sup>+</sup> T and CD8<sup>+</sup> T cells, thus forming temporary bridges between the two T-cell subsets. After dissociation from the DC:CD4+T licensing coupling via CD40:CD40L interaction, the same DC presents antigen to CD8<sup>+</sup> T-cells in a dynamic DC:CD4+T:CD8+T serial interaction (18). Further, trogocytosis (intercellular transfer of membrane proteins) observed between DC and CD4<sup>+</sup> T supports the dynamic three-cell serial interaction exhibited by CD4<sup>+</sup> T acquiring MHC-I:peptide complexes from DC to present to CD8<sup>+</sup> T with concurrent provisions of instructive cytokines (IL-2, IL-12, etc.) and co-stimulation (19). Currently accepted dynamic three-cell interaction model proposes that during DC:CD4+T interaction, DC become licensed whereas CD4<sup>+</sup> T acquire MHC:peptide complexes and transform into primed CD4+T:DC clusters. CD8<sup>+</sup> T-cells then interact with CD4+T:DC cluster or licensed DC alone (20). Thus, trogocytosis and expression of wide variety of costimulatory molecules allow CD8<sup>+</sup> T-cells to flexibly find their interaction partners, and activate specific transcriptional programs to support the expression of proteins responsible for effector function. It is these differentiated mature CTL in the lymph nodes, which extravasate into the area of infected cells or tumor and reengage with the cognate MHC-I:peptide complexes to execute their effector programs. Since secretion of a large number of apoptosis-triggering molecules may be harmful for surrounding tissue, a synaptic contact with the tumor or infected cells that allows a polarized or membranous-vesicular or nanotube-guided delivery of cytolytic granules will avoid bystander off-target CTL cytotoxicity.

### CELLULAR CIRCUITRY UNDERLYING NK FUNCTION

NK are enigmatic in that they display intrinsic ("natural") ability to lyse tumor, infected or stressed cells without prior priming (21). Although initially considered an artifact, it became obvious by the 1970s that NK represent a population distinct from antigen-specific T-cells (22–25). Yet, their cytolytic activity continued to be overshadowed by large granular cytotoxic lymphocytes (26). Recent developments in transcriptomics finally defined their molecular identity as innate cytotoxic lymphocytes and established their distinct yet overlapping patterns with CD8<sup>+</sup> CTL (27–29). They have lately garnered interest due to their "on demand" NK-poiesis coordinated in space and time (30).

Unlike CTL, where diversity lies in rearranged TCRs specific to antigens, and tolerance to self is achieved by selective survival of developing thymocytes, NK express a diverse repertoire of germ-line encoded activating and inhibitory receptors to generate diversity and maintain tolerance. NK receptors belong to either the type-1 transmembrane proteins of the immunoglobulin superfamily (e.g., activating natural cytotoxicity receptors, NCRs), the immunoglobulin-like superfamily (e.g., human killer-cell immunoglobulin-like receptors, KIRs), or the C-type lectin-like receptor superfamily (e.g., CD94/NKG2A heterodimer and the multigenic murine Ly49). Binding partners for these receptors are classical or non-classical MHC-I molecules or MHC-I laden with foreign peptide or pathogenencoded molecules. Although initial expression of inhibitory and activating receptors on NK appears to be stochastic, an education process involving MHC-I alleles expressed by the host tissue determines the final repertoire of NK receptor expression (31). Several models of NK education have been proposed to balance the stimulatory and inhibitory signals and calibrate their reactive potential.

The "licensing and arming" models assume that NK education is dependent on the phosphorylation of immunoreceptor tyrosine-based inhibitory motif (ITIM) in intracellular domains of NK-inhibitory receptors upon binding with classical MHC-I. Subsequent downstream signaling triggers NK cell acquisition of effector functions (32). A "disarming model" postulates that in the absence of inhibitory signals NK stay in a sustained state of activation but they become hyporesponsive upon engagement with cognate inhibitory self-MHC class-I ligands (33). "Rheostat model" describes NK education as a process where magnitude of the integrated inputs from different inhibitory signals (MHC-I and non-MHC-I ligands) translate into the strength of effector output (34). "Tuning model" proposes a refinement of NK responsiveness to sudden modifications within the host environment in line with the discontinuity theory of immunity (35, 36). The balancing of the stimulatory and inhibitory signaling through multicellular receptor:ligand engagements calibrates NK activity in conjunction with the target cell recognition through lowered or absent self-MHC or HLA expression termed "missing-self or induced-self recognition" of aberrant cells undergoing transformation, infection or other pathology. Cellular stress induces ligands for NK-activating receptors, such as NKG2D, namely MICA, MICB, and UL16 binding proteins (ULBP), all of which are MHC-1-related molecules. Thus, NK display transcriptional pre-primed state, which may allow them to stay "alert" and mount effector responses rapidly after encountering targets.

### BIDIRECTIONAL CIRCUITRY BETWEEN CTL AND NK

Overlapping cytolytic abilities of CTL and NK warrant a close regulated collaboration. NK-expression of a wide range of chemotactic, synaptic, effector, and regulatory molecules allow NK to form multiple contacts with various cells, which also would affect CTL (37). NK express CXCR3 chemokine receptor directing them to lymph nodes, where they interact with DC and affect their maturation (38, 39). As mentioned above, DC form dynamic three-cell interaction circuit with CD4<sup>+</sup> T and CD8<sup>+</sup> T cells. NK would thus modify the dynamics of DC– CD4+T–CD8+T circuitry via a bi-directional cytokine exchange and cell-to-cell contacts. DC secrete IL-12, type-I IFN, TNF-α, and express MIC-A and B ligands for the NK-activating receptor NKG2D. In turn, NK support DC maturation through IFN-γ and TNF-α production. This promotes Th1 polarization and CTL responses (37, 40, 41). By NK-mediated target lysis, the content of antigens is increased for DC presentation to CTL (42). NK can also selectively kill DC based on their maturity status, and thus affect the strength of CTL response. It is known that mature DC express inhibitory ligands for KIR whereas immature DC lack these ligands, making the latter susceptible to NK-mediated lysis (39, 40).

Further, NK may acquire MHC-II and other molecules through trogocytosis and compete with DC for interaction with CD4<sup>+</sup> T-cells. Such NK do not activate T-cells, rather suppress them (43). Besides, activated T-cells often up-regulate ligands for NKG2D and DNAM-1, which make them prone to lysis by NK. Lysis of activated CD4<sup>+</sup> T-cells will compromise Tcell help to CD8<sup>+</sup> T-cells (40). Also, NK secrete IL-10, which inhibits CD8<sup>+</sup> T-cell proliferation (44). At the same time, to resist NK attacks, T-cells employ mechanisms such as express IFN-α receptor, upregulate MHC-I molecules (KIR ligands) and CD48 (2B4 ligand) or down-regulate NCR1 ligands (45). It has been hypothesized that potentially autoimmune CTLs (deprived of essential signals, such as type-I IFN) are eliminated by NK (40).

Moreover, CD8<sup>+</sup> CTL can also affect NK activation. Studies have demonstrated that CTL can activate intratumoral NK cells (46–48). This teamwork of CD8<sup>+</sup> T and NK prevents the development of antigen-deficient tumor escape variants. No role of a bystander CTL lysis of antigen-deficient tumor cells or tumor stroma following possible uptake of antigen

two classical antigen-presenting cells, dendritic cells (DC), and B-cells. The following cell subsets were considered: (1) activated CD56bright and CD56dim NK-cells; (2) naïve central-memory, effector-memory and EMRA CD8+T-cells; (3) activated myeloid and plasmacytoid DC; and (4) naïve and memory B-cells. Molecular partners in intercellular contacts formed by CD8+T-cells and NK (blue, CD8+CTL–NK), dendritic (red, CD8+CTL–DC), or B (green, CD8+CTL–B) cells were compared. Venn diagram represents common Boolean molecular couples (cross-sections, merged colors) as well as unique intermolecular interactions (single color) based on STRING database score > 0.7 for known and predicted protein–protein interactions.

from the lysed antigen-expressing tumor cells was found (49, 50). Rather, antitumor NK activity was observed when CTL were present in vicinity of NK. Gene profiling of NK from the tumor where antigen-specific CTL were present showed a strong expression of effector (Gzma, Gzmb, Prf1, and 4.1BB), tissue–migratory (Gpr33 and Ccr5), and signaling (Ifngr, Klhdc2, Eif3s6, Map3k6, Tnfrsf1b, Icos, Ly49G, and Nmi) transcripts. This suggested that tumor-infiltrating NK gene expression program was influenced by locally present CTL. Separation of antigendeficient variants (NK targets) from the antigen-expressing tumors (CTL targets) prevented CD8<sup>+</sup> T-cell help to NK (47). Cooperative CTL–NK interaction in tumor rejection especially under conditions of limited TCR diversity (46), involving NKG2D-mediated mechanisms, has been observed in multiple tumor models (51–56).

The human immune network proteomics resource containing a depth of >10,000 proteins (1) supports profound intercellular circuitry based on the sender (cytokines, membrane ligands) and receiver (receptors) molecules on different immune cells. Database analysis demonstrates that CD8+CTL–NK interface occupies intermediate position between CTL–DC and CTL– B-cell in terms of the number of participating immune stimulatory and inhibitory molecules (**Figure 1**). B-cells and DC appear classical interaction partners for CD8<sup>+</sup> T-cells for antigen presentation. A dynamic regulation within CTL–NK circuit is pointed by a spectrum of surface co-stimulatory and co-inhibitory molecules, such as ligand adaptor SLAM-associated protein (SAP), CD48, on CTL with receptor 2B4/CD244 on NK. This molecular interaction may protect CTL from NK lytic attack as well as inhibit or activate NK depending on the intracellular concentrations of CD48 and CD244 (57). Analysis also indicates that NK express stimulatory molecules from the Nectin-like family (CD155/poliovirus receptor) with potential to interact with CD266 (DNAM-1) on CTL. Further, expression of lymphotoxin-A and tumor necrosis factor superfamily-14 (TNFSF14/LIGHT) by NK and complementary TNFRSF1B and TNFRSF14/HVEM (herpes virus entry mediator) by CTL may serve as a substrate for CTL–NK compartmentalization in tumors or tumor-draining lymph nodes. These receptor-ligand couples are involved in the formation of secondary and tertiary lymphoid organs, ectopic lymphoid tissues necessary for antitumor T-cell immunity (58, 59). Finally, the expression of CD8A molecule by CTL and β2-microglobulin and HLA-B molecules by NK correlated with NK antigen presentation. APC-like properties were demonstrated for porcine NK cells (60). Notably, there is no molecular partner exclusive to the circuitry among CTL, NK, and DC (**Figure 1**).

The lytic function of CTL–NK effector circuit is fine-tuned by checkpoint inhibitory molecules from B7/CD28 family, CTLA-4 and BTLA, expressed on NK partnered with CD86 and TNFRSF1, respectively, expressed on CTL. CTLA-4 was found to be up-regulated with CD28 after NK stimulation with IL-2 while ligation of CTLA-4 with B7 molecules inhibited IFNγ production (61). Negative effect of blocking BTLA on NK has been reported (62). Further, while CTL express inhibitory ILT/CD85 member, LILRB2, NK express its interaction partner HLA-B. The up-regulation of LILRB2/ILT4 on NK may increase their activation threshold (63). Also, CD200-CD200R1 expressed by CTL and NK, respectively, may represent another inhibitory coupling controlling CTL–NK lytic circuitry (**Figure 1**). Suppressive CD200R1 cross-linking on NK was demonstrated for acute myeloid leukemia as one escape mechanism for tumor growth (64).

# CONCLUSIONS

Formation of negative and positive feedback loops in the CTL– NK lytic circuitry suggests that multiple functional modules are responsible for sensing and killing target cells undergoing cellular stress of infection, transformation or other pathological noxa. As we proposed earlier (54), antigen-specific adaptive immune cells provide tissue specificity and guide recruitment of innate effector cells to the site of tumor or other pathological insults. While the CTL subset, scans tissue to lyse targets in an antigen-specific manner, the NK subset recognizes and eliminates antigen-escape variants under the antigen-selective pressure of CTL.

The topology and the spectrum of lytic circuitry appears to be defined by the nature and dose of antigens and the context of tissue microenvironment. Thus, many bacterial or acute viral antigens elicit B-cell response and require the presence of Th2 cells to support antibody production. A persistent/chronic viral infection involves intracellular antigen processing and cross-presentation by DC and differentiation of CTL responses. However, availability of immune cells and their maturation/priming/activation status is important for the outcome. Recently, CD103<sup>+</sup> DC expressing basic leucine zipper ATF-like transcription factor-3 (Batf3) were found necessary for recruiting and activating CD8<sup>+</sup> T-cells in tumors (65, 66). Moreover, when DC decline in number, B-cells take over their role (67). Indeed, CD11b<sup>+</sup> B-cells with potent APC function border T-B cellular area in spleens (68). This is also supported by the observation that CD20<sup>+</sup> B-cells are co-present with CTL in ovarian cancer (69). Apparent proximity of CTL and NK inside solid tumors (47) offer multiple opportunities for their interactions. Accordingly, DC, CTL, and NK form a cellular network of functional circuitry, characterized by redundancy and degeneracy necessary for robust network properties. Such topological organization has obvious advantages for system flexibility: deficiency in individual elements of network will rearrange connections between remaining elements, thereby increasing functional robustness to perturbations. The proposition goes in line with the concept of co-respondence proposed for the immune system (2).

From an evolutionary viewpoint, activation of NK by CTL to eradicate aberrant or tumor cells may be considered a bet hedging strategy. Bet hedging is the ability of cells or organisms to diversify their phenotypes to increase future fitness at the expense of benefits in current situation (70). The latter is common in the prokaryotic and eukaryotic worlds. Antibiotic resistance in Mycobacterium tuberculosis (71) or cannibalism in Bacillus subtilis during spore formation (72) are examples of bet hedging in prokaryotes. In eukaryotes, a classical bet hedging is production of different size eggs by animals in a clutch (70). Multiple types of drug resistance and cellular heterogeneity in individual tumors are also indicative of increased tumor fitness (73, 74).

Thus, the CTL–NK circuitry may indicate a provision by which CTL curb the expansion of targets and their escape variants by recruiting NK cells. Examples are prevalent where impaired CTL–NK communication or lack of either partner results in the failure of CTL–NK circuitry and leads to disease progression. In lymphocytic choriomeningitis infection, at high viral doses, NK prevent excessive CTL response whereas at suboptimal lower doses, NK facilitate CTL-mediated lethality. Hence, a perturbed or imbalanced CTL–NK axis can cause severe immunopathology (75, 76). Consequently, nature appears to have evolved another layer of immune control. Immune cells are found in the brain (77–79), with evidence of modulation of anti-bacterial (80) and anti-tumor (81) immune responses by the brain's reward system. In this context, dynamics of CTL and NK cell circuitry and their lytic capacity need to be optimized with supplementation from a neurostimulatory moodenhancing neuronal circuitry in the central nervous system (81) to develop effective immunotherapies capable of out-pacing infections, cancer or other pathologies.

# DATA AVAILABILITY

Publicly available datasets were analyzed in this study. These data can be found here: http://www.immprot.org/.

# AUTHOR CONTRIBUTIONS

RU performed in silico analysis of immune network interactions (http://www.immprot.org/). AS and RU conceived, designed, and wrote the manuscript. Both authors read and approved the final manuscript.

# FUNDING

This work was supported by funds to AS by the following NIH grants: U54 CA163069, U54MD007593, SC1CA182843, and R01 CA175370. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. There was no role of the funding bodies in the design or writing of the manuscript. No writing assistance was utilized in the production of this manuscript.

# ACKNOWLEDGMENTS

The authors thank Zerick Dunbar, BS, MS and Maria Teresa Prudente de Aquino, MS, Ph.D for feedback on the manuscript.

# REFERENCES


CD8 T cell regulation during viral infection. J Exp Med. (2009) 206:2235– 51. doi: 10.1084/jem.20082387


anti-tumor response in acute myeloid leukemia. Leukemia. (2011) 25:792– 9. doi: 10.1038/leu.2011.1


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

Copyright © 2019 Uzhachenko and Shanker. 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 the Lymphocyte Functional Crosstalk and Regulation in the Context of Checkpoint Inhibitor Treatment—Review

#### Barbara Seliger\*

*Institute of Medical Immunology, Martin Luther University Halle-Wittenberg, Halle (Saale), Germany*

During the last decade, the dynamics of the cellular crosstalk have highlighted the significance of the host vs. tumor interaction. This resulted in the development of novel immunotherapeutic strategies in order to modulate/inhibit the mechanisms leading to escape of tumor cells from immune surveillance. Different monoclonal antibodies directed against immune checkpoints, e.g., the T lymphocyte antigen 4 and the programmed cell death protein 1/ programmed cell death ligand 1 have been successfully implemented for the treatment of cancer. Despite their broad activity in many solid and hematologic tumor types, only 20–40% of patients demonstrated a durable treatment response. This might be due to an impaired T cell tumor interaction mediated by immune escape mechanisms of tumor and immune cells as well as alterations in the composition of the tumor microenvironment, peripheral blood, and microbiome. These different factors dynamically regulate different steps of the cancer immune process thereby negatively interfering with the T cell –mediated anti-tumoral immune responses. Therefore, this review will summarize the current knowledge of the different players involved in inhibiting tumor immunogenicity and mounting resistance to checkpoint inhibitors with focus on the role of tumor T cell interaction. A better insight of this process might lead to the development of strategies to revert these inhibitory processes and represent the rational for the design of novel immunotherapies and combinations in order to improve their efficacy.

Keywords: T cells, tumor growth, tumor microenvironment, microbiome, inflammation, checkpoint inhibitors

### INTRODUCTION

It has been generally accepted that the development and progression of tumors is a result of an altered crosstalk between the tumor and the host immune system (1–3). The immune system not only suppresses tumor growth by destroying tumor cells or inhibiting their outgrowth, but also promotes tumor progression by either selecting for tumor escape variants or by establishing conditions within the tumor microenvironment (TME) and periphery that facilitate tumor outgrowth, which has been classified as a hallmark of cancer (4). These include an increased frequency of immune suppressive cells, metabolites, cytokines and soluble factors, hypoxia and acidic pH (5, 6). Further changes of the TME during neoplastic transformation are a selective ablation of immune effector cells and deletion or neutralization of cytokines, like interferon (IFN)-γ (7). Despite interferon (IFN)-γ exert pro-tumorigenic effects under certain circumstances dependent on the cellular and molecular context (8, 9), it represents a key mediator

Edited by:

*Anil Shanker, Meharry Medical College, United States*

#### Reviewed by:

*Udo S. Gaipl, University Hospital Erlangen, Germany Yona Keisari, Tel Aviv University, Israel*

> \*Correspondence: *Barbara Seliger barbara.seliger@uk-halle.de*

#### Specialty section:

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

> Received: *15 March 2019* Accepted: *12 August 2019* Published: *06 September 2019*

#### Citation:

*Seliger B (2019) The Role of the Lymphocyte Functional Crosstalk and Regulation in the Context of Checkpoint Inhibitor Treatment—Review. Front. Immunol. 10:2043. doi: 10.3389/fimmu.2019.02043* of immunosurveillance produced by natural killer (NK) cells and T cells known to promote cytotoxic activity of macrophages and enhance the expression of immune modulatory molecules on tumor cells (7). This results in the release of tumor associated antigens (TAA) for cross presentation by dendritic cells (DCs), which uptake and process these antigens into peptides then presented via the major histocompatibility complex (MHC) class I and class II molecules to CD8<sup>+</sup> and CD4<sup>+</sup> T cells, respectively. However, elimination of transformed cells can be incomplete due to a decreased tumor immunogenicity (10). This results first in an equilibrium state characterized by a balance between proliferation and killing of tumor cells by CD8<sup>+</sup> T cells thereby maintaining the tumor at a subclinical stage, followed by the generation of tumor cells, which are resistant to immune rejection due to constant selective pressure of the immune system (2). These immune escape mechanisms are associated with the loss or downregulation of TAA and/or HLA class I surface molecules or aberrantly expression of the non-classical HLA-G and HLA-E antigens as well as co-inhibitory molecules (**Table 1**). This might be at least partially mediated by the induction of oncogenic pathways (11, 12) and changes in the tumor cell metabolism (13, 14).

However, interventions, such as chemotherapy, radiotherapy (RT), physico-chemical, and thermal ablation can promote the release of TAA and might overcome the dominant immune suppressive pathways leading to an increased immunogenicity (15–18). Therefore, the combination of immunotherapies with other strategies offers novel opportunities to recover immune activity and increase their efficacy, which result in a better patients' outcome. Indeed, this approach is currently investigated in a number of experimental models and clinical trials (19).

Players involved in mounting anti-tumor immune responses include in particular cells of the adaptive immune system, which protect and/or control tumor outgrowth and the interaction of the host against viral/pathogen infections and neoplastic transformation. The therapeutic potential of host-vs.-tumor activity has been analyzed by various groups and is based on CD4<sup>+</sup> and CD8<sup>+</sup> T cell responses, which are part of the cancer immune cycle and significantly influence the clinical outcome of patients (20, 21). It is well-known that the initial antigenmediated activation of T cells is modulated by the engagement co-stimulatory signals with its ligands on antigen-presenting cells (APC). Under physiologic conditions, immune checkpoint pathways avoid auto-immunity by inducing inhibitory pathways important for maintaining self-tolerance thereby regulating the type and magnitude of T cell responses required to mount a proper anti-tumoral activity. During the last decade, a number of different inhibitory T cell and non-T cell iCP pathways have been well-characterized (**Table 2**) (31). The prototype is the cytotoxic T lymphocyte antigen 4 (CTLA-4; CD152), which competes with CD28 for the ligands CD80 and CD86, and antagonizes the T cell receptor (TCR) signaling (32–34). In addition, the interaction of the programmed cell death protein 1 (PD-1; CD279) with its ligands the programmed cell death 1 ligand 1 (PD-L1; CD274/B7- H1) and/or PD-L2 (CD273/B7-DC), negatively interferes with TCR signaling (35–38). Thus, immune checkpoints (iCPs) have either a stimulatory or inhibitory potential, the latter acting as "breaks" on the immune response. Recently, there exists evidence that inhibitory iCPs could be targeted by immune check point inhibitors (iCPIs) leading to an increased anti-tumoral response and patients' survival (39).

#### GENERAL STRATEGIES OF TUMORS TO FACILITATE TUMOR SUPPRESSION BY ANALYZING THE COMPOSITION AND FREQUENCY OF IMMUNE CELLS IN THE TUMOR MICROENVIRONMENT (TME) AND PERIPHERAL BLOOD (PB)

The impaired anti-tumoral immune response represents an important hallmark of solid tumors and hematopoietic malignancies and involves many distinct mechanisms at the tumor site, in the tumor microenvironment (TME) and in the peripheral blood (21). The generation of an inflammatory and immune suppressive milieu in the TME induces tumor escape mechanisms, such as downregulation of classical HLA class I antigens and an upregulation of HLA-G and -E as well as iCPs including e.g., PD-L1 in the TME and CTLA-4 in the lymphoid tissues leading to evasion of adaptive immune responses (40–42). Furthermore, an upregulation of other immune inhibitory molecules like PD-1, T cell immunoglobulin and mucin domain-3 (TIM-3), lymphocyte-activation gene 3 (LAG-3), 2B4, and T cell immunoglobulin and immunoreceptor tyrosinebased inhibitory motif (TIGIT) have been reported, which is accompanied by a reduced IFN-γ and TFN-α secretion of T cells (43–45). An altered TME is further characterized by an altered cellular composition and activity of tumor infiltrating immune cells. Next to a reduced frequency and activity of immune effector cells, such as CD8<sup>+</sup> T cells, NK cells, an increased frequency of immune suppressive cells, such as tumor associated neutrophils (TANs), myeloid-derived suppressor

**Abbreviations:** AML, acute myeloid leukemia; APC, antigen-presenting cells; Arg-1, arginine 1; cDC, classical dendritic cell; COX-2, cyclooxygenase−2; CRC, colorectal carcinoma; CTL, cytotoxic lymphocytes; CTLA-4, T lymphocyte antigen 4; DC, dendritic cell; ECM, extracellular matrix; EMT, epithelial to mesenchymal transition; FcγR, Fc-gamma receptor; FDA, Food and Drug Administration; FOXP3, forkhead box P3; gzmb, granzyme B; HLA, human leukocyte antigen; HNSCC, head and neck squamous cell cancer; ICOS, inducible T cell costimulatory; iCPI, immune check point inhibitors; iCP, immune checkpoint; IFN, interferon; IL, interleukin; IL-2Rα, interleukin-2 receptor chainalpha; LAG-3, lymphocyte activation gene-3; M1, type 1 TAM; M2, type 2 TAM; mAb, monoclonal antibody; MDSC, myeloid-derived-suppressor cell; MHC, major histocompatibility complex; MSI, microsatellite-instable; MSS, microsatellitestable; NK, natural killer; NO, nitric oxide; NSCLC, non-small cell lung cancer; OS, overall survival; PD-1, programmed cell death protein 1; pDC, plasmacytoid dendritic cell; PD-L1, programmed cell death 1 ligand 1; RCC, renal cell cancer; ROS, reactive oxygen species; RT, radiation therapy; SCLC, small cell lung carcinoma; TAA, tumor associated antigen; TAF, tumor associated fibroblasts; TAM, tumor-associated macrophage; TAN, tumor-associated neutrophil; TCR, T cell receptor; Teff, effector T cell; Tex, exhausted CD8<sup>+</sup> T cell; TGF-β, transforming growth factor-beta; TIGIT, T cell immunoglobulin and immunoreceptor tyrosinebased inhibitory motif; TIL, tumor infiltrating lymphocyte; TIM-3, T cell immunoglobulin and mucin domain containing protein-3; TLR-4, toll-like receptor 4; TMB, tumor mutational burden; TME, tumor microenvironment; Treg, regulatory T cell.

#### TABLE 1 | Immune escape mechanisms.


↑ *upregulation,* ↓ *downregulation.*

TABLE 2 | T cell associated inhibitory immune checkpoint pathways.


↑ *upregulation,* ↓ *downregulation.*

cells (MDSC), tumor associated macrophages (TAM), tumor associated fibroblasts (TAF), regulatory T cells (Treg), and stroma cells leading to a complex interaction network of heterogeneous immune and non-immune cell populations with overlapping and opposite functions (46). In addition, soluble factors, like the transcriptional growth factor (TGF)-β, interleukin (IL)-10, the vascular endothelial growth factor (VEGF)A, chemokines as well as metabolites, e.g., arginase, hypoxia, and low pH, have been identified to be responsible for the establishment of an immune suppressive TME (47, 48). Furthermore, a reduced frequency and impaired function of effector cells and capacity of dendritic cells to present antigen as well as an increased number of immune suppressive cells were also found in peripheral blood. Both an immune suppressive TME and a reduced immune function of PB are associated with a poor patients' outcome (49–52). There exists evidence that functional T cell responses could be missed by analyzing only PBL (53) and no TIL. Therefore, it is essential to determine the composition, organization and function of the TME and PB of individual patients as well as the tumor itself to predict potential anti-tumoral effects of antigen specific T cells, since this has been shown to have prognostic relevance and therapeutic implications (47, 54).

#### IMMUNE CHECKPOINT INHIBITORS AND PATIENTS' RESPONSE

Novel immunotherapeutic approaches have recently revolutionized the treatment of solid and hematopoietic tumors. The clinical success of monoclonal antibodies (mAb) directed against CTLA-4 and the PD-1/PD-L1 pathway was a breakthrough achievement (55–57). The anti-CTLA-4 mAb Ipilimumab was the first iCPI approved by the Food and Drug Administration (FDA) (58, 59) followed by the approval of the anti-PD-1 mAbs Pembrolizumab and Nivolumab in 2014 or 2016, respectively. The anti-PD-L1 mAbs Durvalumab, Atezolizumab, and Avelumab were FDA approved in 2017 after promising results in non-small cell lung cancer (NSCLC), urothelial carcinoma and Merkel cell carcinoma (60–63).

Despite the rapid progress of approvals for iCPI, the accumulated experience demonstrated that approximately only one-third of patients had a durable response upon single iCPI treatment. Thus, the majority of patients do not benefit from iCPI alone, which might be due to primary, adaptive and acquired resistance mechanisms (64). Therefore, a number of clinical trials using iCPIs across all tumor types using different combinations, e.g., chemotherapy, iCPIs, chimeric antigen receptor, hypermethylating agents, CDK4 inhibitors, RT and targeted therapies, are currently conducted. Some of these combinations have achieved response rates over 50% (57, 65, 66). Regarding the combination of RT with iCPI it is noteworthy that RT could not only stimulate immune responses, but could also exert immune suppressive effects (67, 68). In this context, scheduling of iCPI therapy is important for the therapeutic outcome in combination with RT (69, 70), which has been shown to shape the T cell receptor repertoire of TIL (71). However, there is still an urgent need to explore biomarkers to predict response to these treatments and to identify combinations of agents to improve treatment efficacy, overall survival (OS) of patients and mitigate toxicities of these treatment options.

# IMMUNE MODULATORY MOLECULES AND THEIR RELEVANCE FOR T CELL RESPONSES AND PATIENTS' OUTCOME Expression of Classical and Non-classical

HLA Class I Antigens This topic has been reviewed and discussed by various authors (72–75). HLA class I surface expression is frequently downregulated or lost in solid and hematopoietic tumors. These abnormalities have functional relevance, since they impair T cell recognition of tumors. Furthermore, HLA class I alterations have been associated with a worse patients' outcome and a reduced overall survival (OS) and play a role in the resistance to iCPI therapy. The underlying molecular mechanisms of impaired HLA class I surface expression are diverse and often associated with deficiencies in the expression of components of the antigen processing machinery (APM) and IFN pathways as recently summarized (72, 76). This could be due to either structural alterations or deregulation at the transcriptional, epigenetic or posttranscriptional level of these molecules (77). Furthermore, HLA class I expression has been associated with an increased density of tumor infiltrating lymphocytes (TIL) and an increased anti-tumoral T cell response (78).

Next to the impaired expression of HLA class I antigens, a frequent overexpression of HLA-G and/or –E was found in tumors of distinct origin, but not in adjacent normal tissues or in healthy controls. This was accompanied by a reduced T cell and NK cell recognition and a bad patients' prognosis (79–83). Soluble HLA-G levels (sHLA-G) were also frequently detected and inversely correlate to numbers of activated T cells suggesting that sHLA-G promotes tumor immune escape through activation of immune responses (84).

# Expression of Immune Checkpoints: Challenges and Pitfalls

Next to alterations of HLA class I antigens, high expression levels of co-inhibitory checkpoints, such as e.g., PD-L1 and B7-H4, in various tumor entities are often associated with the clinical outcome of cancer patients (85–88). An altered expression pattern of PD-L1 was found in primary and metastatic bladder tumors suggesting a dynamic nature of the TME (89). The capacity of the PD-L1 mediated immune suppression was inversely proportional to the antigenicity of the tumor (90). Since PD-L1 expression on both tumor and host's immune cells could lead to escape from immune surveillance, PD-L1 expression has been suggested as a biomarker for prediction of prognosis and response to iCPI (91). Its expression correlates with the adaptive immune resistance in several tumor types, including melanoma, NSCLC, Merkel cell carcinoma, breast cancer, mismatch-repair deficient tumors, and Hodgkin's lymphoma (91–93). However, PD-L1 expression does not reliably predict response to iCPI. In melanoma, tumor PD-L1 expression showed a significant correlation with response to five out of eight iCPI studies treating patients with anti-PD-1 mAb, while it did not predict response to anti-CTLA-4 therapy (94). Furthermore, some patients negative for PD-L1 expression can have a response to iCPI (92). In NSCLC, no association of PD-L1 expression with response has been reported with Nivolumab, while PD-L1 expression on at least 50% of NSCLC lesions almost doubled the response rate to Pembrolizumab from 19 to about 45% (95). In contrast to pre-treatment biopsies, tumor biopsies in early treatment phase obtained from metastatic melanoma patients treated sequentially receiving CTLA-4 and PD-1 iCPI showed high PD-1 and PD-L1 expression levels in responders (96). In NSCLC cells, the PD-L1 genomic locus amplification correlated with PD-L1 expression and anti-tumor responses (97, 98). Despite a significant heterogeneity was observed, higher levels of CTLA-4 and PD-L2 expression were found in melanoma patients, who benefit from CTLA-4 antibodies (99, 100). In contrast, PD-L1, PD-L2, and CTLA-4 expression did not correlate to anti-PD-1-responsiveness of melanoma patients (101). In addition to the discrepant results on the role of PD-L1 expression for prognosis and iCPI response, there exist some limitations regarding the analysis of the PD-L1 expression, including membranous vs. cytoplasmic expression, expression by multiple cell types in the TME, focal expression in tumor samples, changes in the expression during disease progression, upon radiation, chemotherapy, and epigenetic drugs and in particular the variability of laboratory techniques and anti-PD-L1 antibodies employed for immunohistochemistry (IHC) (102).

#### Somatic Mutations and Neoantigen Load

Increasing evidence demonstrated that mutations lead to the generation of neoantigens, which are presented by HLA class I molecules and can be recognized by CD8<sup>+</sup> cytotoxic T cells (CTL) (77). Thus, the tumor mutational burden (TMB) might be correlated with the level of response to T cell based immunotherapies. Indeed, a systemic review of melanoma patients showed that responses to iCPIs correlated with TMB, neoantigen load, and immune-related gene expression (103, 104). Microsatellite instable (MSI) colorectal carcinoma (CRC) has large mutational burdens, higher immune cell infiltration and higher response rates to PD-1 blockade (105). However, a high TMB does not always predict responders to iCPI therapy, which might be due to neoantigen heterogeneity and an extremely diverse array of somatic mutations (106, 107). It is noteworthy that T cell epitopes have a similarity to bacterial and viral antigens suggesting a cross reactivity of T cells to intestinal bacterial and viral antigens, which can also modulate the iCPI therapy (108). In addition, PD-L1 expression can be controlled by driver mutations and oncogenic signaling (109–112).

#### Immune Profiling Signatures and iCPI

Genetic and immune heterogeneity was found in melanoma responding to immunotherapy. Mutanome and individual genebased expression analysis demonstrated mesenchymal and T cell suppressive inflammatory or angiogenic tumor phenotypes, which were associated with innate anti-PD-1 resistance (113). Genes, which were higher expressed in non-responding pretreatment tumors, include molecules involved in epithelial to mesenchymal transition (EMT), immune suppression and chemotaxis of monocytes and macrophages (114). Interestingly, a dormant TIL phenotype characterized by an elevated TMB and intra-tumoral CD3 signal, elevated TILs with low activation and proliferation was associated with a favorable response to iCPI (115). The IFN signature is correlated with an improved prognosis and iCPI response or resistance to iCPIs (116). Furthermore, T cell diversification reflects antigen selection in the blood of patients on iCPI treatment (117). Recently, a 15-gene pre-treatment classifier model was identified to predict response to anti-CTLA-4 treatment (118).

# ROLE OF IMMUNE CELL SUBPOPULATIONS FOR TUMOR IMMUNITY AND THE EFFECT OF iCPI

#### Tumor-Infiltrating Cytotoxic Lymphocytes (CTL) and iCPI

The success of checkpoint blockade depends on the presence of TIL, particularly of CD8<sup>+</sup> CTL, in the TME. These CTL are located at the invasive tumor margin and intratumorally, and are negatively regulated by the PD-1/PD-L1-mediated adaptive immune resistance (119, 120). In metastatic melanoma, the presence of CTL at the tumor margin predicted better response to iCPI. Colon cancers with MSI are highly infiltrated with T cells, particularly with CTL, relative to microsatellite-stable (MSS) colon cancers (121–123). Members of the CCL and CXCL chemokine families have been associated with T cell recruitment to melanoma metastases (124, 125). Higher levels of CCL2, CXCL4, and CXCL12 have been noted in tumors responding to iCPI therapy (126).

So far, it is not clear whether CD8<sup>+</sup> effector memory cells might explain the durable response observed in many patients. Interestingly, brisk CTL infiltrates at time of progression in patients on iCPI treatment were observed suggesting an impaired activity of effector immune cells in the TME leading to therapeutic resistance. Despite CD8<sup>+</sup> T cell responses against tumor cells are well-understood, information about the role of CD4<sup>+</sup> T cell immunity in cancer is limited. Tumor specific CD4<sup>+</sup> T cells have a broad activity beyond the provision of helper signals to CD8<sup>+</sup> T cells (127, 128). CD4<sup>+</sup> T cells exhibit antitumor effects and Th1 T cells are involved in the killing of tumor cells by secretion of cytokines that activate death receptors on tumor cells and induce epitopes spreading. Furthermore, CD4<sup>+</sup> Th1 T cells can activate DC functions. The secretion of IL-4 from CD4<sup>+</sup> T cells could establish long-term memory immune responses and further recruit eosinophils and macrophages.

# Tumor-Infiltrating Regulatory T Cells (Tregs)

Tumor-infiltrating CD4<sup>+</sup> Tregs were frequently detected in the TME and suppress CTL activity leading to reduced anti-tumor T cell responses (129). They promote tumor growth by iCP expression (CTLA-4, PD-1 and others) as well as production of IL-10 and TGF-β. An increased frequency of Tregs correlates with disease progression and metastasis in both experimental models and humans (130). CTLA-4 blockade expands the Treg frequency and high levels of soluble CD25 interleukin-2 receptor chair-alpha (IL2Rα) has been correlated with resistance to anti-CTLA-4 therapy (131). This was confirmed by Treg depletion potentiating the iCPI therapy (132). It has been suggested that early recruitment of Tregs to the TME inhibits an effective tumor response and lack of response to iCPI. PD-1 blockade with Nivolumab attenuated the activated T cell phenotypes during the course of therapy, promoted CTL proliferation and resistance to Treg-mediated suppression by down-regulating the intracellular expression of FoxP3, while Tregs increased during disease progression (133). An increased ratio of CTL to Treg in tumor tissues has been associated with response to CTLA-4 and PD-1 blockade.

# Natural Killer Cells as Players for Innate Immune Responses

Natural killer (NK) cells are effector cells of the innate immune system and important players in mounting innate anti-tumoral immune responses by their ability to directly target and eliminate viral infections as well as neoplastic transformed cells (134). Under pathological conditions and during inflammation, the NK cell activation depend on the balance between inhibitory as well as activating signals, which determine the NK cell mediated cytotoxicity. In addition, NK cells are involved in other immune regulatory processes and could modulate adaptive immune responses, since they share characteristics with adaptive lymphocytes (134). They could also interact with mast cells and effect tumorgenesis due to the production of pro-angiogenic factors and thus play an important role alone or in combination with mast cells in the regulation of angiogenesis (135). There is increasing evidence that NK cells are involved in regulating metastatic dissemination. NK cells are often shown to reduce metastatic efficacy of tumor cell lines in vivo, while low NK cell activity is correlated with advanced disease and metastasis formation (136, 137). Furthermore, the presence of tumorinfiltrating NK cells is a positive prognostic marker for multiple tumors (138–140).

# Tumor-Infiltrating Regulatory Myeloid Cells

Tumor-infiltrating myeloid cells comprise MDSCs, tumorassociated granulocytes, TAMs and DCs, generate and promote both immunogenic and tolerogenic responses (141–143). MDSCs are heterogeneous immune-suppressive immature myeloid cells that can be divided into a polymorphonuclear subset and a monocytic subset. They support tumor growth, epithelial to mesemchymal transition (EMT) and predict poor prognosis of patients, but their role in tumorgenesis has still to be defined (144, 145). MDSCs exert their effects by producing immune suppressive factors, like arginine 1 (Arg-1) expression, nitric oxide (NO), cyclooxygenase-2 (COX-2), reactive oxygen species (ROS), and activate Treg via CD40–CD40L interactions (146– 148). In melanoma, elevated levels of CXCL17 were found, which recruits MDSCs and predicts non-responders to iCPI (149).

Tumor-associated neutrophils (TANs) and TAMs have been classified as an anti-tumor (type 1) or pro-tumor (type 2) phenotype. Pro-tumor effects of TANs include dampening of CTL response, increased angiogenesis, and modulation of cellular trafficking. Type 1 TAMs (M1) produce immune stimulatory cytokines, like IL-6, IL-12 and CXCL9, that promote recruitment of CTLs, while type 2 TAMs (M2) exhibit an immune suppressive signature and support tumor growth by release of angiogenic factors, like IL-10 and CCL22, matrix remodeling mediated by proteases, and by inhibition of CTL and DC activity (150–153). In addition, TAMs promote Tregs by inducing the skewing of blood-derived CD4<sup>+</sup> T cells toward an immunosuppressive phenotype due to their decreased production of effector cytokines, increased IL-10 production and enhanced expression of the co-inhibitory molecules PD-1 and TIM-3 (154, 155). However, the interaction between TA-specific CD4<sup>+</sup> Th1 cells and TAMs might shift the intra-tumoral M1/M2 ratio toward an M1 phenotype (155). PD-L1 expression of monocytes and TAMs promote immune evasion and correlate with disease progression in hepatocellular carcinoma. This might be mediated by a hypoxia inducible factor 1α induced increased expression of the receptor TREM1 in TMAs resulting in immune suppression mediated by Treg recruitment, which was associated with disease progression as well as resistance to anti-PD-L1 treatment (156). Fc-gamma receptors (FcγRs) expressed by M2 TAMs facilitate anti-tumor response to CTLA-4 inhibition through Treg depletion (157, 158). Tumor-infiltrating eosinophils promote infiltration of CTLs by polarization of TAMs and normalization of the tumor vasculature, and predict a better prognosis in colon cancer (159).

The heterogenic family of DCs, including classical (cDCs) and plasmacytoid DCs (pDCs), are antigen-presenting cells (APC) that prime and regulate CTL responses. Anti-viral immune responses rely heavily on pDC-derived type I IFNs, while pDCs in tumors exert immunosuppressive activities. In contrast, tumor-infiltrating cDC increase T cell activation in lung cancer and melanoma patients forming tertiary lymphoid clusters, which are associated with better outcomes (160, 161). Tertiary lymphoid clusters also correlated with improved survival in pancreatic cancer (162). The rare subgroup of CD103(+) (integrin αE)<sup>+</sup> DCs are strong stimulators of CTL and dependent on different transcription factors, like IRF8, Zbtb46, and Batf3. These CD103 cDCs (Batf3-cDC, cDC1) are also associated with CTL and increased OS for patients with breast, head and neck or lung cancer (163). In lung adenocarcinoma murine models, immunogenic chemotherapy (oxaliplatin-cyclophosphamide) has been reported to up-regulate toll-like receptor 4 (TLR-4) on tumor-infiltrating Batf3-cDCs, which leads to the recruitment of CTLs and sensitization to iCPIs (164).

# Gut Microbiome, Immune Cell Interaction—iCPI Therapy

T cells as members of the adaptive immune system are involved in gut homeostasis, inflammation, and carcinogenesis (165). Recently, association between microbiota profiles, cancer susceptibility, and responsiveness to cancer therapy has been suggested (166–168). Indeed, microbiota could modify the immune response and influence the response to chemotherapy and immunotherapy (169, 170). Emerging evidence has suggested that the cross-talk between the gut microbiome and immune cells plays a role in determining responses to iCPI therapy. Indeed, the composition of the gut microbiome has been associated with response to iCPI in pre-clinical models as well as in patients. For example, in murine melanoma, commensal Bifidobacterium has been reported to promote the efficacy of anti-PD-L1 therapy by augmenting the function of DCs leading to CTL priming and infiltration (171). Recent studies in melanoma, lung, and kidney cancer patients have demonstrated an association of commensal gut microbiome with response to iCPI (172). Baseline gut microbiota enriched with Faecalibacterium and other Firmicutes is associated with a better response (173). In melanoma patients responding to iCPI more abundant species included Bifidobacterium, Collinsella, Enterococcus, Clostridiales, Rominococcus and Faecalibacterium, while low levels of Akkermansia muciniphila were observed in epithelial cancers not responding to iCPI (174). Patients with a favorable gut microbiome had increased expression of cytolytic T cell markers and APM components, and an increased ratio of CD8<sup>+</sup> CTLs to FoxP3+CD4<sup>+</sup> Tregs. Furthermore, metagenomic studies revealed functional differences in gut bacteria in responders. These are characterized by an enrichment of anabolic pathways and an enhanced systemic and antitumor immunity in responding patients with a favorable gut microbiome as well as in germ-free mice receiving fecal transplants from responding patients (172). Thus, the modulation of the components in the gut microbiome can augment anti-tumor immunotherapy. However, there exist several challenges including optimal composition of the gut microbiome and the therapeutic strategy to achieve that composition.

#### Resistance to Checkpoint Inhibitors

Abnormalities of the HLA class I antigen and IFN signaling pathways often correlate with the development of resistances to various kinds of immunotherapies including iCPI treatment and adoptive cell therapy (ACT) (64, 175–178). These could be categorized into intrinsic and acquired immune resistance (177, 179, 180) and are associated with an altered tumor T cell interaction. An increased knowledge of these processes might lead to the reprogramming of the immunologically "cold" TME characterized by a low immune cell infiltration and low TCR diversity (181, 182) and an increased T cell function (183–185). Combining the modulation of the immune cell repertoire and the reduction of immune suppressive metabolites and cytokines of the TME and enhancement of T cell tumor interaction with iCPI and/or vaccinations or even targeted therapies are currently tested in diverse clinical trials (186, 187).

# CONCLUSIONS

There is strong evidence of an emerging role of T cell tumor interactions for the outcome of patients in general and regarding the efficacy of immunotherapies including iCPIs. The pathways involved in the regulation of the interaction between tumor and T cells are broad and highly dynamic. Tumors developed a plethora of adaptions leading to escape from counter-regulations of the immune system. This is mediated by ineffective T cell responses due to low tumor immunogenicity and the suppressive influence of the TME. The use of iCPI showed that the manipulation of inhibitory signaling pathways creates anti-tumoral immune responses. However, the efficacy of iCPIs is still limited. Thus, a better understanding of these processes might lead to the development of innovative therapies in order to reactivate T cell responses.

# AUTHOR CONTRIBUTIONS

BS designed the project and wrote the manuscript.

#### ACKNOWLEDGMENTS

I would like to acknowledge Maria Heise for excellent secretarial help.

#### REFERENCES


correlation with tumor-infiltrating immune cells and clinical outcome. Oncoimmunology. (2016) 5:e1235107. doi: 10.1080/2162402X.2016.1235107


melanoma patients treated with ipilimumab. Ann Oncol. (2017) 28:1368– 79. doi: 10.1093/annonc/mdx108


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

# Trauma Induces Interleukin-17A Expression on Th17 Cells and CD4+ Regulatory T Cells as Well as Platelet Dysfunction

#### Friederike Hefele1,2, Alexander Ditsch<sup>1</sup> , Niels Krysiak 1,3, Charles C. Caldwell 4,5 , Peter Biberthaler <sup>6</sup> , Martijn van Griensven<sup>1</sup> , Stefan Huber-Wagner <sup>6</sup> and Marc Hanschen1,6 \*

<sup>1</sup> Experimental Trauma Surgery, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany, <sup>2</sup> Division of Oncology and Hematology (CCM), Medical Department, Charité–Universitätsmedizin Berlin, Berlin, Germany, <sup>3</sup> Department of Trauma Surgery, Berufsgenossenschaftliche Unfallklinik Murnau, Murnau, Germany, <sup>4</sup> Division of Research, Department of Surgery, University of Cincinnati College of Medicine, Cincinnati, OH, United States, <sup>5</sup> Division of Research, Shriners Hospital for Children, Cincinnati, OH, United States, <sup>6</sup> Department of Trauma Surgery, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany

#### Edited by:

Raghvendra Mohan Srivastava, Memorial Sloan Kettering Cancer Center, United States

#### Reviewed by:

Anil Dangi, Duke University Medical Center, United States Krishna Beer Singh, University of Pittsburgh, United States

> \*Correspondence: Marc Hanschen marc.hanschen@mri.tum.de

#### Specialty section:

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

Received: 05 March 2019 Accepted: 23 September 2019 Published: 11 October 2019

#### Citation:

Hefele F, Ditsch A, Krysiak N, Caldwell CC, Biberthaler P, van Griensven M, Huber-Wagner S and Hanschen M (2019) Trauma Induces Interleukin-17A Expression on Th17 Cells and CD4+ Regulatory T Cells as Well as Platelet Dysfunction. Front. Immunol. 10:2389. doi: 10.3389/fimmu.2019.02389 Background: The organism's immune response to trauma is distinctively controlled, its dysregulation leading to severe post-traumatic complications. Platelets, CD4+ regulatory T cells (CD4+ Tregs) and T helper 17 (Th17) cells have been identified to participate in the post-traumatic immune response. Unfortunately, little is known about their exact role and potential interdependency in humans. Aims of this clinical trial were to phenotype the human immune response following injury and to identify risk factors rendering the host more susceptible to trauma induced injury.

#### Methods: This non-interventional prospective clinical trial enrolled patients following multiple trauma, follow up was conducted for 10 days. Peripheral blood CD4+ Tregs and Th17 cells were analyzed using flow cytometry to determine Interleukin 17A (IL-17A) expression. Hemostasis and platelet function were assessed with rotational thromboelastometry (ROTEM®). Subgroup analysis was conducted for the factors gender, age, and trauma severity.

Results and Conclusion: This is the first clinical trial to phenotype the immune response following trauma, focusing on platelets, and the adaptive immune response. We discovered a novel increased IL-17A expression on Th17 cells and on CD4+ Tregs following trauma and describe the kinetics of the immune response. The IL-17A response on CD4+ Tregs challenges the ascribed role of CD4+ Tregs to be solely counter inflammatory in this setting. Furthermore, despite a rising number of platelets, ROTEM analysis shows post-traumatic platelet dysfunction. Subgroup analysis revealed gender, age, and trauma severity as influencing factors for several of the analyzed parameters.

Keywords: T helper 17 cells, CD4+ regulatory T cells, trauma, hemostasis, IL-17A, flow cytometry, thromboelastometry

#### SUMMARY

IL-17A expression on CD4+ Tregs and Th17 cells in multiple trauma patients increases during the first 10 days after trauma with no significant changes in cell numbers detectable. Platelets of trauma patients show signs of dysfunction in thromboelastometry despite increasing counts.

#### INTRODUCTION

Unintentional injuries, especially road traffic injuries, remain to be the leading cause of death in the age group of 15– 29 years, outnumbering the fatalities that result from malaria, tuberculosis, and HIV/AIDS combined, according to the Global Health Estimates data of the World Health Organization (WHO) of 2016 (1). Over all age groups, road traffic injuries increased profoundly, ranking 8th in 2016 as compared to ranking 10th in 2000. The injury-associated mortality is commonly categorized into three groups according to the time of occurrence: within minutes or even seconds following trauma, during the first 24 h, and after several days. The first two groups account for ∼50% of all deaths (2). The main reasons for this early mortality are the injury itself and or in combination with pre-existing conditions and can therefore not be effectively prevented. The late mortality, on the other hand, is often caused by multiple organ dysfunction syndrome (MODS) or multiple organ failure (MOF), which are results not only of the trauma but predominantly of the patients' dysfunctional immune response (3). The post-traumatic immune responses can be categorized into proinflammatory and anti-inflammatory reactions, commonly referred to as systemic inflammatory response syndrome (SIRS) and compensatory antiinflammatory response syndrome (CARS) as well as mixed antiinflammatory response syndrome (MARS) (4). An imbalance between these pro- and anti-inflammatory immune responses is believed to be the cause for heightened susceptibility to infections and organ damage. While trauma itself and the associated early mortality cannot be undone, the late mortality due to the host's immune response following trauma is potentially modifiable. As the mechanisms of the injured bodies' immune system to react to severe trauma are still not fully understood and due to the fact that most knowledge has been gathered in animal models, this study aims to phenotype the human immune response in severely injured patients.

Due to recent findings in the literature, highlighting the role and interplay of CD4+ Tregs, Th17 cells, and platelets following trauma, our study attempts to address these key players. A decreased CD4+ Treg/Th17 cell ratio was recently discovered in a multiple trauma rat model, correlating inversely with disease activity (5). In our previous studies utilizing a murine burn injury model, we have been able to show impaired activation of intracellular pathway signaling in CD4+ Tregs after platelet depletion (6). Hence, we were the first to show reciprocal activation of CD4+ Tregs and platelets following trauma induced injury. The pathways responsible for the modulation of CD4+ Treg activation seem to be tumor necrosis factor receptor 2- and toll-like receptor 4- dependent (7).

Trauma is associated with a suppressed Th1 immune system phenotype and an expansion of CD4+ Tregs (8). Of interest, little is known of Th17 cells in the setting of trauma.

T helper 17 (Th17) cells are a recently discovered new lineage of T helper cells named after their production and expression as well as secretion of IL-17 (9). Their characteristic cytokine profile—IL-17, IL-17F, IL-9, IL-10, IL-21, IL-22, IFN-γ, and GM-CSF—makes them mostly proinflammatory cells (10–15). The transcription factor retinoid acid-related orphan receptor gamma t (RORγt) was revealed to be the key regulator in the development of this T helper cell subset, additionally a similar potential could be proven for retinoid acid-related orphan receptor alpha (RORα) (16, 17). In autoimmunity, the primary research field for Th17 cells, targeting the p40 subunit of IL-23 and IL-12 with Ustekinumab as well as the IL-17-receptor with Brodalumab has proven to be a highly efficient therapeutic option for psoriasis (18, 19). Aside from causing several autoimmune disorders, Th17 cells partake in microbial defense—they are integral in the clearance of extracellular pathogens like Klebsiella pneumoniae and Candida albicans (20, 21). No specific set of identifying markers has yet been agreed on for this cell type, however, several studies have used CD161 and chemokine receptor 6 (CCR6/CD196) to identify CD4+ lymphocytes as Th17 cells (22–24). In 2009, Brucklacher-Waldert et al. showed IL-17A surface expression identifies Th17 cells and correlates with its intracellular production (25). While Th17 cells have been mainly considered with autoimmune disease in the past, recent findings point out a potential interplay with platelets in the setting of burn and trauma (26). The mechanisms and role of platelet-Th17 interaction following trauma need yet to be characterized.

CD4+ regulatory T cells (CD4+ Tregs) have been established key players in the post-traumatic immune response, they contribute to the counterinflammatory reaction to severe injury (27). CD4+ Tregs were first described by Sakaguchi et al. as suppressors of T effector cell activation and proliferation in 1995 and were characterized as highly expressing CD25 (IL-2-receptor α chain) (28). Several other markers for this cell type have since been identified, the most commonly used being intracellular expression of transcription factor forkhead box p3 (Foxp3) and lack of surface-CD127-expression (29, 30). CD4+ Tregs play a crucial role in maintaining immunologic selftolerance and preventing excessive immune reactions to weak stimuli, preserving the delicate, and crucial balance between proand anti-inflammatory immune reactions (31, 32). However, they also display a significant potential for plasticity and adaptability: in certain settings, CD4+ Tregs can convert into Th17 cells (33). Furthermore, in 2009, a subset of IL-17-producing CD4+ Tregs was discovered by Voo et al. (34).

Platelets, for a long time only recognized for their pivotal role in the coagulation system, are nowadays established as a key component of the immune system (35–37). By releasing

**Abbreviations:** 95% CI, 95% confidence interval; APC, Allophycocyanin; B, Regression coefficient; EDTA, ethylenediaminetetraacetic acid; GEE, Generalized estimating equations; IL-17A, Interleukin 17A; ISS, Injury severity score; MCF, Maximum clot firmness; MFI, Median fluorescence intensity; p, Probability value; PE, Phycoerythrin; SE, Standard error; Th17, T helper 17; CD4+ Treg, CD4+ regulatory T cell.

cytokines and chemokines from their granules, they participate as mediators in host defense against pathogens. However, platelets also act as effector cells of the immune system by releasing bactericidal defensins (38). Platelets can even synthesize new molecules as they contain mRNA; furthermore, recent findings suggest that the type of mRNA they are equipped with varies depending on the state of the host—healthy or sick (39). As pointed out above, platelets have the capability to modulate the immune response following injury. The interaction with Th17 cell and CD4+ Tregs as well as its role following injury need yet to be characterized, studies in humans are mostly missing.

While animal models are necessary and crucial to elucidate new aspects of basic immunological pathomechanisms, there are considerable interspecies differences (40). In order to discover new therapeutic targets to combat the post-traumatic immune dysfunction, a more complete knowledge of the key players and mechanisms involved in the human post-traumatic response is integral.

Taken together, we conducted a clinical prospective noninterventional trial on patients following multiple trauma, utilizing serial blood analyses. The aims of this study were first to phenotype the post-traumatic immune response focusing on lymphocytes (Th17 cells, CD4+ Tregs) and platelets, and second to investigate the host's susceptibility for trauma by conducting subgroup analysis. We used flow cytometry and rotational thromboelastometry (ROTEM <sup>R</sup> ) to characterize the cells and their functionality.

We discovered a significant increase in IL-17A expression on both Th17 cells and CD4+ Tregs during the first 10 days after trauma. Our findings challenge the ascribed role of CD4+ Tregs to be solely counterinflammatory in the setting of trauma induced injury. In our thromboelastometric measurements we found an increase in maximum clot firmness (MCF) alongside with post-traumatic platelet dysfunction. Furthermore, assessment of gender, age, and trauma severity measured by the injury severity score (ISS) as possible influencing factors yielded significant differences in several of the analyzed parameters. The presented subgroup analysis supports the concept, that the immune response following trauma is shaped according to the susceptibility of the host.

#### MATERIALS AND METHODS

#### Patients

For our prospective non-interventional study, we recruited severely injured patients from ages 18 to 95 with multiple traumata defined by an injury severity score (ISS) ≥16. Exclusion criteria were pregnancy and imprisonment. The patients included in the study were brought to the emergency room of the Klinikum rechts der Isar of the Technical University of Munich no more than 12 h after their respective trauma between December 2014 and March 2017. Written informed consent was obtained from all patients or their relatives according to the patient's suspected will. The study follows the principles of the declaration of Helsinki with its novelizations of Tokyo 1975, Hongkong 1989, and Somerset West 1996. It was approved by the ethics review committee of the Technical University of Munich (reference number 5925/13) prior to starting the research.

#### Reagents

For flow cytometry, cells were stained in PBA buffer: PBS supplemented with bovine serum albumin and sodium azide (all by Sigma Aldrich, St. Louis, MO). For red blood cell lysis, we used Schwinzer solution: 1 l distilled water supplemented with 8.3 grams of ammonium chloride (Carl Roth GmbH + Co. KG, Karlsruhe, Germany), 1.0 g of potassium carbonate (Caesar & Lorentz GmbH, Hilden, Germany) and 0.1 g of ethylenediaminetetraacetic acid (EDTA) (Carl Roth GmbH + Co. KG, Karlsruhe, Germany). Fc-blocking agent (eBioscience, San Diego, CA) was used to prevent non-specific binding of staining antibodies. Surface staining was performed using allophycocyanin (APC)-labeled anti-CD4 (eBioscience, San Diego, CA), eFluor450TM-labeled anti-CD161 (eBioscience, San Diego, CA), FITC-labeled anti-CD196 (eBioscience, San Diego, CA), FITC-labeled anti-CD4 (eBioscience, San Diego, CA), eFluor450TM-labeled anti-CD25 (eBioscience, San Diego, CA), APC-Cy7-labeled anti-CD127 (eBioscience, San Diego, CA) and phycoerythrin (PE)-labeled anti-IL17A (LifeSpan BioSciences, Seattle, WA). MACSQuant <sup>R</sup> calibration beads (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) were used prior to measurements according to the manufacturer's instructions.

For thromboelastometric analysis we used star-tem, r ex-tem, in-tem, and fib-tem reagents (all by TEM International GmbH, Munich, Germany).

#### Sample and Data Retrieval

After hospital admission and inclusion of the patient in the study according to the criteria stated above, a series of nine blood draws was performed at several time points after trauma: the first one directly in the emergency room, next after 6, 12, 24, 48, and 72 h, furthermore after 5, 7, and 10 days. At each time point we collected citrated whole blood for thromboelastometric analysis of platelets and EDTA-treated whole blood for flow cytometrical lymphocyte analysis. Sarstedt S-Monovette <sup>R</sup> 3 ml with citrate 3,2% (1:10) and Sarstedt S-Monovette <sup>R</sup> 9 ml with K3 EDTA (both Sarstedt AG & Co. KG, Nümbrecht, Germany) were used for blood collection.

Furthermore, absolute platelet counts and demographic as well as clinical patient data (demographics, injury mechanism, and severity) were gathered.

#### Flow Cytometry

Flow cytometric analysis of T-cell subpopulations was performed on a MACSQuant <sup>R</sup> device (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). Samples were prepared as follows: within 15 min after blood draw, EDTA-treated blood was added to Schwinzer red blood cell lysis solution. After 15 min of incubation at 4◦C the cells were washed and buffered with PBA and plated on 96-well round bottom plates for staining. Fcblocking agent was added to prevent unspecific antibody binding.

After incubation we used APC-labeled anti-CD4, eFluor450TM-labeled anti-CD161, and FITC-labeled anti-CD196 for detection of Th17 cells. For CD4+ Treg detection we used FITC-labeled anti-CD4, eFluor450TM-labeled anti-CD25, and APC-Cy7-labeled anti-CD127. Cell activation was evaluated using PE-labeled anti-IL17A. After incubation, cells were washed and resuspended in PBA buffer for immediate flow cytometric analysis. Prior to measurements, the MACSQuant <sup>R</sup> was calibrated using MACSQuant <sup>R</sup> calibration beads according to the manufacturer's instructions.

The obtained data sets were analyzed using FlowJo Software (FlowJo LLC, Ashland, OR). After single cell selection, Th17 cells were defined as CD4+, CD161+, CD196+ cells, and CD4+ Tregs were defined as CD4+, CD25+, CD127<sup>−</sup> cells. IL-17A expression on Th17 cells was assessed by calculating the relative median fluorescence intensity (MFI) of PE on Th17 cells stained with PE-conjugated anti-IL17A antibody. To calculate the relative MFI, the MFI of PE on anti-IL17A stained cells was divided by the MFI of PE measured on Th17 cells not stained with any PE-conjugated antibody. IL-17A expression on CD4+ Tregs was assessed analogously.

#### Thromboelastometry

For thromboelastometric assessment of platelet function we used the ROTEM <sup>R</sup> delta device (TEM International GmbH, Munich, Germany) according to the manual. We examined platelet function after extrinsic activation of the coagulation cascade using star-tem reagent to recalcify the citrated blood and using r ex-tem reagent to activate platelets. For evaluation of plasmatic coagulation without influence of platelets fib-tem reagent was added to inhibit platelet function before starting the measurement.

We analyzed the parameter maximum clot firmness (MCF), measured in millimeters (mm) as the greatest amplitude of the reaction curve. For evaluation of the platelet contribution of the maximum clot firmness, the platelet MCF was calculated as the MCF (extem-fibtem) by subtracting the MCF measured in fibtem from the one measured in extem, as it has been described in literature before (41).

#### Statistics

The Statistical Package for the Social Sciences (SPSS) (IBM, Armonk, NY) was used to perform statistical analysis. In the descriptive statistical analysis values are given as mean ± standard deviation. Generalized estimating equations (GEE) with an exchangeable correlation matrix were applied to test for statistically significant changes over time, which was measured in days. Gender, dichotomized age (<55 years, ≥55 years), and ISS (<25 or ≥25) were calculated in as influencing factors. Values given are the regression coefficient (B), the standard error (SE), the 95% confidence interval (95% CI) and the probability value (p). p < 0.05 was considered significant.

#### RESULTS

#### Study Population

Twenty trauma patients were enrolled in the study, the detailed epidemiologic data are provided in **Table 1**. Of the included patients, 70% (14) were male. The mean patient age was 46.5 ± 18.7 years, in 65% (13) of all cases, the injuries were caused by TABLE 1 | Demographic patient data.


SD, standard deviation; AIS, abbreviated injury scale; ISS, injury severity score; ICU, intensive care unit.

road traffic accidents. The mean ISS was 28.4 ± 11.8, 55% (11) of the study cohort had an ISS ≥25. Injuries to the extremities including the pelvic girdle were the most common, affecting 85% (17) of the patients, followed by injuries to the chest, which affected 80% (16) of the study cohort. With a mean abbreviated injury scale (AIS) of 3.7 ± 1.0, injuries to the chest also were the most severe. 85% (17) of the patients survived. In the survivor group, the mean intensive care unit stay was 13.2 ± 16.6 days and 3.8 ± 2.0 surgical procedures were performed.

### Th17/CD4+ Ratio

To assess the number of Th17 cells, we looked into the percentage of Th17 cells (CD4+, CD161+, CD196+) to all CD4+ lymphocytes. This percentage did not change significantly over time (B = 0.001, SE = 0.001, 95% CI [0.000; 0.002], p = 0.094), which is also displayed in **Figure 1A**. However, males had higher Th17/CD4+ ratios than females (B = 0.021, SE = 0.009, 95% CI [0.003, 0.039], p = 0.024), while younger (<55 years of age) patients had lower ratios when compared to older (≥55 years of age) patients (B = −0.017, SE = 0.007, 95% CI [−0.031,−0.003], p = 0.017). No significant difference was visible when comparing lower (<25) to higher ISS (≥25) (B = −0.003, SE = 0.119, 95% CI [−0.026, 0.021], p = 0.825).

cytometry. Blood was drawn at nine time points: in the emergency room (0), after 6 and 12 h, and after 1, 2, 3, 5, 7, and 10 days. Each dot displays a single patient and time point. The fit lines are for illustrative purposes and do not represent the generalized estimating equation (GEE). (A) Th17 cells were defined as CD4+, CD161+, CD196+. There was no significant increase in the percentage of Th17 cells in CD4+ lymphocytes during the first 10 days after trauma (B = 0.001, SE = 0.001, 95% CI [0.000; 0.002], p = 0.094). (B) IL-17A expression on Th17 cells was assessed by calculating the relative median fluorescence intensity (MFI) of PE on Th17 cells stained with PE-conjugated anti-IL-17A antibody, dividing the MFI of PE on anti-IL-17A-stained Th17 cells by the MFI of PE on Th17 cells not stained with any PE-conjugated antibody. During the first 10 days after trauma, there was a significant increase in the IL-17A expression on Th17 cells (B = 0.056, SE = 0.017, 95% CI [0.022; 0.090], p = 0.001).

#### IL-17A Expression on Th17 Cells

Next, we analyzed the median fluorescence intensity of the PEconjugated IL-17A-antibody on Th17 cells to measure the IL-17A expression of these cells. As shown in **Figure 1B**, the IL-17A expression on Th17 cells increased over time (B = 0.056, SE = 0.017, 95% CI [0.022; 0.090], p = 0.001). However, no significant differences were seen when comparing the subgroups: males to females (B = −0.053, SE = 0.150, 95% CI [−0.346; 0.241], p = 0.725), younger to older patients (B = −0.011, SE = 0.167, 95% CI [−0.338; 0.316], p = 0.946), and lower to higher ISS (B = −0.002, SE = 0.101, 95% CI [−0.200; 0.197], p = 0.987).

#### CD4+ Treg/CD4+ Ratio

Analogously to the Th17 cells, the percentage of CD4+ Tregs to all CD4+ lymphocytes was calculated to measure the number of CD4+ Tregs. Contrary to the percentage of Th17 cells, the percentage of CD4+ Tregs (CD4+, CD25+, CD127−) to all CD4+ lymphocytes increased over time (B = 0.002, SE < 0.001, 95% CI [0.001; 0.003], p < 0.001), as shown in **Figure 2A**. Like with the Th17/CD4+ ratio, males had higher CD4+ Treg/CD4+ ratios compared to females (B = 0.008, SE = 0.003, 95% CI [0.003; 0.014], p = 0.004), while younger patients had lower ratios compared to older patients (B = −0.016, SE = 0.004, 95% CI [−0.024; −0.009], p < 0.001). No significant difference was to be seen when comparing lower to higher ISS (B = −0.007, SE = 0.004, 95% CI [−0.014; >0.000], p = 0.056).

#### IL-17A Expression on CD4+ Tregs

Again, just like on Th17 cells, we analyzed the median fluorescence intensity of the PE-conjugated IL-17A antibody to assess the IL-17A expression on CD4+ Tregs. This expression increased over time (B = 0.064, SE = 0.022, 95% CI [0.021; 0.107], p = 0.004), the according scatter plot is shown in **Figure 2B**. Again, like with the IL-17A expression on Th17 cells, no significant differences were detectable when comparing the subgroups: males to females (B = −0.170, SE = 0.309, 95% CI [- 0.777; 0.436], p = 0.582), younger to older patients (B = −0.222, SE = 0.311, 95% CI [−0.832; 0.387], p = 0.475), and lower to higher ISS (B = 0.108, SE = 0.195, 95% CI [−0.276; 0.491], p = 0.582).

#### Absolute Platelet Count

Next, we looked into the number of platelets, which increased over time (B = 22.178, SE = 3.523, 95% CI [15.273; 29.082], p < 0.001). This is illustrated in **Figure 3**, furthermore we saw higher platelet counts in patients with lower ISS compared to patients with higher ISS (B = 63.530, SE = 28.279, 95% CI [8.104; 118.955], p = 0.025). There were, however, no significant differences when comparing males to females (B =32.740, SE = 26.570, 95% CI [−19.337; 84.817], p = 0.218), and younger to older patients (B =11.333, SE = 16.680, 95% CI [−21.360; 44.026], p = 0.497).

#### MCF in Extem Measurements

With extem, the patients' hemostatic function after extrinsic activation of the coagulation cascade is measured (compare **Figure 4A**).

In the extem measurements, the MCF increased over time, as seen in **Figure 4B**, (B = 1.677, SE = 0.319, 95% CI [1.053; 2.301], p < 0.001), and males had higher MCFs compared to females (B = 7.220, SE = 2.366, 95% CI [2.583; 11.858], p = 0.002). There were no significant differences in the other subgroups, when comparing younger to older patients (B = −1.877, SE = 1.856, 95% CI [−5.516; 1.761], p = 0.312), or lower to higher ISS (B = 1.174, SE = 1.785, 95% CI [−2.324; 4.673], p = 0.511).

#### MCF in Fibtem Measurements

The fibtem panel allows the assessment of hemostatic capacity without platelets after extrinsic coagulation activation.

FIGURE 2 | Percentage of CD4+ Tregs in CD4+ lymphocytes and IL-17A expression on CD4+ Tregs in peripheral blood of multiple trauma patients, assessed via flow cytometry. Blood was drawn at nine time points: in the emergency room (0), after 6 and 12 h, and after 1, 2, 3, 5, 7, and 10 days. Each dot displays a single patient and time point. The fit lines are for illustrative purposes and do not represent the generalized estimating equation (GEE). (A) CD4+ Tregs were defined as CD4+, CD25+, CD127−. The CD4+ Treg/CD4+ lymphocytes percentage increased significantly during the first 10 days after trauma (B = 0.002, SE < 0.001, 95% CI [0.001; 0.003], p < 0.001). (B) IL-17A expression on CD4+ Tregs was assessed by calculating the relative median fluorescence intensity (MFI) of PE on CD4+ Tregs stained with PE-conjugated anti-IL-17A antibody, dividing the MFI of PE on anti-IL-17A stained CD4+ Tregs by the MFI of PE on CD4+ Tregs not stained with any PE-conjugated antibody. There was a significant increase in the IL-17A expression on CD4+ Tregs in the peripheral blood of patients during the first 10 days after trauma (B = 0.064, SE = 0.022, 95% CI [0.021; 0.107], p = 0.004).

In the fibtem measurements, the MCF increased over time as well, which is shown in **Figure 4C**, (B = 2.790, SE = 0.151, 95% CI [2.494; 3.087], p < 0.001) with a higher MCF in males compared to females (B = 3.948, SE = 1.551, 95% CI [0.909; 6.987], p = 0.011). We also measured a lower MCF in patients with a lower ISS compared to patients with a higher ISS (B = −3.493, SE = 1.126, 95% CI [−6.034; −0.952], p = 0.007). There were no significant differences when comparing younger to older patients (B = 0.346, SE = 1.658, 95% CI [−2.905; 3.596], p = 0.835).

#### Platelet MCF: MCF(extem)–MCF(Fibtem)

To assess the platelet contribution to the clot firmness, we subtracted the MCF measured in fibtem at each time point from the MCF measured in extem.

Contrary to the MCFs measured in extem and fibtem, we could see a decrease over time in this calculated MCF (B = −1.141, SE = 0.296, 95% CI [−1.721; −0.561], p < 0.001), which is illustrated in **Figure 4D**. Furthermore, younger patients had a lower calculated platelet MCF than older patients (B = −2.633, SE = 0.660, 95% CI [−3.926; −1.339], p < 0.001). Corresponding

the platelet MCF, which represents the platelet contribution to the overall MCF. (B) The MCF measured with the extem panel increased significantly during the first 10 days after trauma (B = 1.677, SE = 0.319, 95% CI [1.053; 2.301], p < 0.001). (C) A significant increase of the MCF measured with the fibtem panel was also detectable (B = 2.790, SE = 0.151, 95% CI [2.494; 3.087], p < 0.001). (D) However, the platelet MCF decreased significantly (B = −1.141, SE = 0.296, 95% CI [−1.721;−0.561], p < 0.001).

to the findings in the fibtem MCF, patients with a lower ISS had a higher calculated platelet MCF than patients with a higher ISS (B = 3.808, SE = 1.065, 95% CI [1.720; 5.896], p < 0.001). No significant differences were seen when comparing males to females (B = 1.776, SE = 1.006, 95% CI [−0.195; 3.746], p = 0.077).

#### Subgroup Analysis Summary

An overview of the influencing factors is given in **Table 2**.

Males had significantly higher Th17/CD4+ and CD4+ Treg/CD4+ lymphocytes ratios, furthermore the males' MCF measured in extem and fibtem was significantly higher than the females'.

Age influenced Th17/CD4+ and CD4+ Treg/CD4+ lymphocytes ratios as well: both were significantly lower in younger compared to older patients. Additionally, age had an impact on the platelet MCF, with a significantly lower calculated platelet MCF in younger compared to older patients.

The ISS affected platelets and coagulation: The absolute platelet count was higher in patients with lower ISS as compared to patients with higher ISS. The fibtem MCF was significantly lower and the platelet MCF significantly higher in patients with lower ISS than in patients with higher ISS.

# DISCUSSION

The immune response to trauma is a precisely regulated process with pro- and anti-inflammatory components; their imbalance or derailment can lead to fatal consequences and is associated with a high mortality due to sepsis and/or the development of MOF. While there have been made significant advances in treating the direct impact of trauma to the body, i.e., fractures, soft tissue, and organ injuries, no well-established therapeutic concept for the subsequent immune reaction exists to date. Furthermore, diagnostic measures to assess the immune response of the host following trauma are missing so far. A distinct understanding of the regulating mechanisms as well as the involved cells is crucial for the development of treatment strategies to not only manage immunologic complications, but prevent their occurrence entirely.



Positive regression coefficients (B) mean higher values in the male, younger age, or lower ISS group, respectively, while negative regression coefficients mean lower values in the male, younger age, or lower injury severity score (ISS) group.

Th17, T helper 17; CD4+ Treg, CD4+ regulatory T cell; MCF, maximum clot firmness; n.s., not significant.

In this study, we analyzed the peripheral blood derived from multiple trauma patients at nine specific time points, from admission to the emergency department until 10 days after trauma. This comparatively long observation period allowed us to detect even small alterations in the examined parameters. Furthermore, we were able to describe the development in trauma patients over time, opposed to comparing with a control group. For the characterization of CD4+ Tregs and Th17 cells, we applied flow cytometry, a well-established method for the fast screening of large numbers of cells. Its high-volume output facilitates the analysis of particularly small cell populations like Th17 cells and CD4+ Tregs. The resulting data are quantifiable and highly comparable. For the assessment of hemostasis and platelet function, on the other hand, we chose rotational thromboelastometry (ROTEM <sup>R</sup> ), which is applicable for clinical diagnostics as well as research purposes. Designed as a point of care technology device, it is resilient to errors and offers fast and comparable results.

In regards to Th17 cells in trauma patients, this study shows that while the number of these cells in peripheral blood does not increase significantly during the first 10 days after trauma (**Figure 1A**), their surface expression of IL-17A actually does (**Figure 1B**). Due to Th17 cells only being discovered recently in 2006 as a new subset of T helper cells, the research connecting this cell type with trauma is quite scarce. In 2013, Dai et al. found a decreased CD4+ Treg/Th17 cell ratio in a multiple trauma rat model compared to sham treatment 4 h after trauma. There were, however, no significant differences in the numbers of Th17 cells, which corresponds with our findings (5). Treating the animals with anti-IL-17A antibody, lung inflammation parameters improved, which points to a potentially harmful role for IL-17A producing Th17 cells after trauma (42). Another study from 2016 on trauma patients found elevated Th17/CD4+ Treg ratios in trauma patients who developed sepsis, furthermore the ratio of Th17 cells to CD4+ Tregs was skewed in favor of Th17 cells in non-surviving patients (43). This is once again in line with the view that Th17 cells might have a rather detrimental effect in a post-traumatic setting. A clinical trial on sepsis patients demonstrated elevated Th17 cell percentages as well as increased Th17/CD4+ Treg ratios in non-survivors with a positive correlation to the acute physiology and chronic health evaluation II (APACHE II) score (44). In another recently published clinical trial on intensive care unit (ICU) patients, elevated serum IL-17 levels have been shown to be a predictor for the development of sepsis, once more supporting the notion of the rather harmful role of Th17 cells (45).

Ever since the discovery of CD4+ Tregs in 1995 by Sakaguchi et al., they have been in the focus of immunologic research (28). In trauma, their role has been suggested to be mostly protective, as they are able to suppress excessive inflammation, therefore potentially inhibiting the development of severe inflammatory response syndrome (SIRS). CD4+ Tregs have been shown to be activated early after injury in a murine burn model (46). Furthermore, also using the murine burn model, we have earlier been able to demonstrate an interaction of CD4+ Tregs with platelets, this interaction being modulated by TNF-RII- and TLR4-dependent pathways (6, 7). In this study, regarding CD4+ Treg count, we saw an increasing number relative to all CD4+ lymphocytes during the observed time period (**Figure 2A**). However, a study on multiple trauma patients published by Serve et al. in 2018 showed a drop in the CD+ Treg/CD4+ lymphocyte ratio compared to healthy volunteers (47). This data might seem contradictory to our recent findings; however, their observation period was shorter with the last time point at 72 h, furthermore we have not compared our results to healthy volunteers yet.

The established concept of CD4+ Tregs as being solely immunosuppressive cells has recently been challenged, as several studies have shown a considerable amount of plasticity for CD4+ Tregs, especially toward the Th17 lineage (33, 34). Surprisingly, our data show that IL-17A expressing CD4+ Tregs occur after trauma and that their expression of IL-17A even increases during our observed time period (**Figure 2B**). Until now, CD4+ Tregs have been primarily associated with the compensatory anti-inflammatory response syndrome (CARS) following trauma, yet our data showing proinflammatory cytokine production by CD4+ Tregs in multiple trauma patients certainly necessitate a more differentiated look on these cells.

The detection of an interaction of CD4+ CD4+ Tregs with platelets in a murine burn model led us to investigate platelet function in this study on multiple trauma patients as well (6). Our thromboelastometric results show an increasing MCF both in the extem and fibtem measurements (**Figures 4B,C**); however, subtracting the fibtem MCF from the extem MCF, thereby calculating the platelet contribution or platelet MCF (41), we actually saw a decrease (**Figure 4D**). The rise in clot firmness we saw in extem and fibtem might be attributable to the substitution of fibrinogen and clotting factors trauma patients often receive, the decreasing platelet function on the other hand remains to be an interesting topic. The phenomenon of a post-traumatic platelet dysfunction despite a reassuring platelet count has also been described by Kutcher et al., moreover they identified a low Glasgow coma scale (GCS) as an independent predictor (48). A possible cause for the platelet dysfunction could be consumption after activation, in line with the development of disseminated intravascular coagulation, yet our data show an actual rise in numbers during the observed time period with a simultaneous dysfunction (**Figures 3**, **4D**). Therefore, consumption alone does not serve as a sufficient explanation. The platelet count increase might—at least partly—occur due to transfusions, and studies have shown that these transfused cells do forfeit some of their function as a result of platelet storage lesions (49). Lastly, it remains in question, as to what extent this observed dysfunction in hemostasis also concerns the immunologic role of platelets.

To facilitate a more nuanced evaluation of our data, we chose generalized estimating equations for the statistical analysis, as they provide the opportunity of including potential influencing factors. We chose to integrate gender, age, and trauma severity in our calculations, whereby age and trauma severity were dichotomized: "<55 years" and "≥55 years," "ISS < 25" and "ISS ≥ 25." The summarized results are displayed in **Table 2**. Conflicting data exists on the frequencies of Th17 cells in older compared to younger healthy individuals: there is evidence for higher (50) as well as for lower numbers in the elderly population (51). Our data shows a higher percentage of Th17 cells in older compared to younger multiple trauma patients. This offers an explanation on a cellular level for the typically worse outcome of elderly patients suffering multiple trauma, considering that higher numbers of Th17 cells and elevated IL-17A levels are generally associated with a higher mortality. In healthy individuals, CD4+ Treg frequencies increase during the aging process; evidence for this has been provided in animal as well as human studies (52, 53). We have now shown that this applies to trauma patients as well: we found elevated CD4+ Treg/CD4+ percentages in older compared to younger patients. To our knowledge, there is not a lot of data available on possible gender-associated differences in the number of Th17 cells and CD4+ Tregs. One study on patients suffering from acute myeloid leukemia found a higher number of circulating Th17 in male healthy controls compared to females (54), another study on healthy adults found higher CD4+ Treg numbers in males compared to females (55); both studies support our findings in trauma patients, where males had both higher Th17 cell and CD4+ Treg percentages. The higher platelet count in patients with a lower ISS we observed might be attributable to less hemorrhaging—with subsequent loss of cellular blood components in particular—occurring in patients with fewer injuries. Furthermore, the platelet MCF was also higher in patients with lower ISS, which led us to the assumption that the previously described post-traumatic platelet dysfunction might be positively correlated to trauma severity. The results of the platelet MCF comparison between age groups was unexpected: it was actually lower in younger compared to older patients, which could raise the suspicion that younger patients might be more prone to the development of a platelet dysfunction. On the other hand, a shift toward a more procoagulant state is generally agreed upon for the elderly population, leading to a higher risk for thrombotic complications (56). This development is not exclusively, but at least in part, caused by the platelets' increased activity, which might just be what we observed here.

A schematic summary of some of our results is also displayed in **Figure 5A**. Trauma seems to impact both Th17 cells and CD4+ Tregs as well as platelets. The mechanisms of this and the potential interaction between these cell types remain unclear at this point; nevertheless, quite some interesting findings have been published by other workgroups, all of which might offer starting points for further studies in this field. For once, a significant plasticity between CD4+ Tregs and Th17 cells has been shown (33), meaning that T cells can convert into one another; furthermore, platelets have been shown to induce inflammation through sCD40L (26). In previous work from our group, we were able to show a tumor necrosis factor receptor 2 and toll-like receptor 4- dependent activation of CD4+ Tregs by platelets in a murine burn model (7) (**Figure 5B**). Taken together, while these findings cannot yet explain all the post-traumatic proceedings we discovered, these pathways should certainly be studied further in order to better understand the post-traumatic immunologic reaction.

Limiting factors of this present patient study include the solely descriptive and therefore non-interventional design, meaning that unlike in most animal models, no causalities can be proven. Furthermore, only 20 multiple trauma patients have been included to date, resulting in relatively small subgroups in our subgroup analysis. Therefore, we have not been able to correlate our findings with the patients' outcomes yet. Additionally, we have not compared our findings in the post-traumatic immune response with healthy controls thus far. These issues will, however, be addressed by enrolling more trauma patients and recruiting healthy volunteers, which will allow for a more thorough analysis of our results and subgroups. Another point to take into consideration might be the lack of a specific set of surface markers for the unambiguous identification of Th17 cells to date. We chose the co-expression of CD4, CD161, and CD196 as our identifying set for Th17 cells, as this combination has been used in a number of studies with satisfying results (24, 57). Furthermore, we have not taken blood transfusions into account. Considering platelet storage lesions have been proven to take place, their exact implications albeit remaining to be fully understood, especially the platelet function could be influenced and altered by transfusions (49). Lastly, infection and autoimmune diseases may influence both the count of Th17 cells and CD4+ Tregs as well as their IL-17A expression, we did, however, not factor those in as potential confounders.

In conclusion, this present study is the first to characterize IL-17A expression on peripheral blood Th17 cells and CD4+ Tregs of multiple trauma patients. We were able to demonstrate that these cell types actually increase their IL-17A expression during the first 10 days following trauma. The discovery of a subset of CD4+ Tregs expressing IL-17A following trauma suggests that their role in the post-traumatic immune response is not only an anti-inflammatory one, CD4+ Tregs might show just the same plasticity that has been described in other settings. Regarding platelets and hemostatic function in multiple trauma patients, our study shows that despite an increase in the absolute number of platelets in the trauma patients' blood

over time, their function as measured thromboelastometrically actually decreases. These results support the notion of a platelet dysfunction occurring after trauma. Lastly, we were able to identify gender, age, and trauma severity as factors that influence and alter the analyzed parameters to various extents. Our data thereby adds to understand the hosts' susceptibility to trauma in dependence of gender, age, and trauma severity.

#### DATA AVAILABILITY STATEMENT

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

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the review committee of the Technical University of Munich with written informed consent from all subjects. All subjects gave written informed consent in

#### REFERENCES

1. World Health Organization. Global Health Estimates 2016 Summary Tables: Global Deaths by Cause, Age and Sex, 2000–2016. (2018). Available accordance with the Declaration of Helsinki. The protocol was approved by the review committee of the Technical University of Munich (reference number 5925/13).

# AUTHOR CONTRIBUTIONS

FH and MH conceptualized the study, designed the experiments, established the sample collection, processing protocol, analyzed and interpreted the data, and wrote the manuscript. FH, AD, and NK performed the experiments. MH, PB, MG, CC, and SH-W provided critical resources. FH, CC, SH-W, MG, and MH edited the manuscript. MH supervised the work.

#### ACKNOWLEDGMENTS

The authors thank Fritz Seidl, M.B.A., M.A. English translating and interpreting, for editing, and language revision of this manuscript.

online at: http://www.who.int/healthinfo/global\_burden\_disease/GHE2016\_ Deaths\_Global\_2000\_2016.xls?ua=1 (accessed February 1, 2019).

2. Lefering R, Paffrath T, Bouamra O, Coats TJ, Woodford M, Jenks T, et al. Epidemiology of in-hospital trauma deaths. Eur J Trauma Emerg Surg. (2012) 38:3–9. doi: 10.1007/s00068-011- 0168-4


neutrophil recruitment, and host defense. J Exp Med. (2001) 194:519– 27. doi: 10.1084/jem.194.4.519


**Conflict of Interest:** This report includes experimental work performed by FH in fulfillment of her doctoral thesis requirements.

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

Copyright © 2019 Hefele, Ditsch, Krysiak, Caldwell, Biberthaler, van Griensven, Huber-Wagner and Hanschen. 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.