Effects of Engineered Costimulation on the Function of T Cell Subsets

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Boroughs, Angela C. 2019. Effects of Engineered Costimulation on the Function of T Cell Subsets. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:41121316

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA Effects of Engineered Costimulation on the Function of T Cell Subsets

A dissertation presented

Angela Clare Boroughs

The Division of Medical Sciences

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Immunology

Harvard University

Cambridge, Massachusetts

October 2018

Dissertation Advisor: Dr. Marcela V. Maus Angela Clare Boroughs

Effects of Engineered Costimulation on the Function of T Cell Subsets

Abstract

Progress in clinical adoptive immunotherapy was initially hindered by the cells’ lack of antigen specificity, poor engraftment, and limited persistence in the host. The introduction of chimeric antigen receptors (CARs) into T cells has overcome these obstacles by re-directing polyclonal T cells to a specific antigen with an extracellular binding domain and by including costimulation domains that enhance engraftment and persistence. From this observation, we hypothesized that the overexpression of CARs in different T cell subsets fundamentally changes their biology.

CAR T cell products bearing either 4-1BB or CD28 costimulatory domains have been approved as therapies for leukemia and lymphoma. However, the engraftment kinetics, persistence, and toxicity profiles of CD28 versus 4-1BB CAR T cells are distinct. To obtain an in-depth understanding of the functional state of different types of CAR- modified T cells, we performed RNA sequencing on first- and second-generation CAR T cells both at rest and following CAR or endogenous T cell receptor (TCR) stimulation.

We describe a high-resolution view of the transcriptional differences between 4-1BB and CD28 containing CAR T cells, including variances in cytokine profiles, cytokine receptors, and metabolic pathways. These transcriptional profiles define CAR signaling pathways that ultimately determine CAR T cell fate.

iii Adoptive immunotherapy with regulatory T cells (Tregs) also holds promise in transplantation, graft-versus-host disease, and autoimmune diseases. Tregs are key modulators of inflammation and are important for peripheral tolerance. The challenges of using Tregs as adoptive immunotherapy mirror those for cancer: namely antigen specificity, engraftment, and persistence. We modified primary human Tregs with CARs bearing different costimulatory domains and performed rigorous analyses in vitro of the functional potential of Tregs. We showed that the presence of a CAR and the type of costimulation domain does not affect the Treg’s expression of Foxp3+. However, the costimulation domain does affect CAR-Treg cytokine production and surface marker expression. Furthermore 4-1BB costimulation decreases the suppressive function of

CAR-Tregs. In vivo experiments, demonstrated that CAR-Tregs traffic to antigen expressing sites and can suppress antigen specific effector T cell responses. Our findings support the use of CD28-based CARs for tissue-specific suppression in the clinic.

Acknowledgements

I am deeply grateful for the opportunity to study in Harvard’s immunology program. Reflecting on my time here, I am ever more aware of how indebted I am, especially to the many educators who believed in my ability to succeed. I have learned that writing a thesis and performing experiments are not nearly as difficult as believing that I can successfully accomplish them. So, I cannot thank enough, my family, friends, and mentors who have supported and encouraged me to reach this goal.

I would like to acknowledge the people who gave me precious first and second chances that were fundamental in shaping me as a student and scientist. Among those who built my confidence during my school years in South Africa, a few stand out. My great aunt Mary Hart spent many hours patiently guiding me through primary school homework. In 7th grade, my math teacher, Ms. Barrow, gave me a second chance to complete a test after I froze on the first try, not realizing what a turning point that would be for me. Two inspiring high school math teachers, Mrs. Rose Karam and Dr. Caroline Dickens, never doubted I could become an accomplished mathematician and scientist and gave me the ideal foundation to make that possible.

My arrival at Harvard was first made possible by the encouragement of Dr. Fyodor Urnov, an adjunct professor at UC Berkeley, who during one brief semester as an undergraduate exchange student not only convinced me to switch fields from applied mathematics to molecular biology but also suggested that I apply to Harvard’s Immunology Program. Finally, to the immunology faculty who offered me a place in this program: I am eternally grateful that you took a chance on a “wild card” from South Africa.

I would also like to acknowledge Dr. Ramnik Xavier for offering me a place in his lab, where I spent my first two years working with some of the most brilliant and inspirational scientists. I was fortunate to learn from Dr. Leigh Baxt, Dr. Nicole Desch, Dr. Dan Graham, Dr. Bernard Khor, Dr. Christine Becker, Jess Gracias, and Dr. Kara Conway. All of them helped me develop into a better scientist and have continued to be valuable sources of knowledge and dear friends since I moved on.

With the never-failing support of Dr. Shiv Pillai, I began my PhD research anew in the Maus lab, yet another transformative second chance. I don’t have adequate words to express how grateful I am that Dr. Marcela Maus took me under her wing at that opportune moment. She had just opened her lab at Massachusetts General Hospital, and I doubt she had planned on braving a third-year graduate student so soon. Under her guidance, I found not only a knowledgeable and inspiring mentor, but also someone whose enthusiasm, patience, and encouragement boosted my confidence and my love of science. Despite her incredibly busy schedule, Dr. Maus has always been available to support me in whatever way I needed. She has kept me focused and provided

invaluable direction, while allowing me the flexibility to follow scientific questions in my own way.

Every member of the Maus Lab, past and present, has helped me in some way, whether it be through critical thinking, technical assistance, or personal advice. I am thankful for their extraordinary friendship. I would especially like to thank Selena Lorrey, Lauren Riley, and Maria Cabral Rodriguez, the smart, enthusiastic, and oh-so-organized technicians who work tirelessly to keep the lab running smoothly. I was delighted when Rebecca Larson joined the lab as a fellow graduate student, brightening the lab with her enthusiasm and brilliance. The critical insights and advice on my research projects from my dissertation advisory committee, Dr. Laurence Turka, Dr. James Moon, and Dr. Shiv Pillai, have been invaluable.

Some of my most treasured moments at Harvard were spent as a teaching fellow in the undergraduate immunology course. Again, I am grateful to Dr. Shiv Pillai for this opportunity and to all the students who helped me develop as an educator.

My parents have been nothing but supportive, encouraging, and loving from before I can remember. From a young age, they instilled in me the wonders of learning, thinking, and teaching, not to mention the value of the arts and exercise that kept me somewhat balanced through my PhD. And I am thankful for the continuous encouragement from my diverse and incredible siblings, Meghan, Grace, and Patrick. I deeply miss my big South African family at home and thank them for the never-ending Snapchats and WhatsApp texts of encouragement that reminded me that there is life outside the lab.

Throughout my time at Harvard, I have been inspired by those family members and friends who have so bravely lived through cancer and other serious illnesses. I am blessed to have angels and a dragonfly watching over me and inspiring me to appreciate each day and to keep persevering in a field that is making new cures possible.

Finally, I am eternally grateful to my partner in dance and in life, my husband, Vaughan. He stuck with me through the two years we had to spend an ocean apart and then left his family and job in South Africa to restart in Massachusetts while I finished my degree. Throughout the highs and lows of this journey, Vaughan has been my rock, and I am grateful for the sacrifices he has made to be here for me while I reach toward my dream of a PhD in immunology.

Dedication

To Educators everywhere, who lift learners up, give second chances, and teach students to believe in themselves

vii

Table of Contents Effects of Engineered Costimulation on the Function of T Cell Subsets ...... i

Abstract ...... iii

Acknowledgements ...... v

Dedication ...... vii

Glossary ...... xii

Chapter 1. Introduction ...... 1

1.1 Costimulation and Chimeric Antigen Receptors ...... 2

1.1.1 CD28 signaling ...... 2

1.1.2 4-1BB signaling ...... 4

1.1.3 Costimulation domains beyond 4-1BB and CD28 ...... 6

1.1.4 An overview of CAR T cell therapy ...... 9

1.1.5 Chimeric antigen receptor constructs ...... 10

Figure 1.1 – First- and second-generation CAR constructs ...... 11

1.1.6 CAR vs. TCR signaling ...... 14

1.1.7 The effect of CD28 and 4-1BB costimulation on CAR T cells ...... 17

1.1.8 Tonic signaling in CAR T cells ...... 19

1.1.9 Cell therapy/clinical impact of CAR T cells ...... 20

1.2 Antigen-Specific T Regulatory Cells ...... 23

1.2.1 T cell subsets ...... 23

1.2.2 T regulatory cells ...... 27

1.2.3 Mechanisms of Treg suppression ...... 29 viii

1.2.4 Treg cytotoxicity ...... 32

1.2.5 Treg lineage stability ...... 33

1.2.6 Adoptive Treg cell therapy ...... 35

1.2.6.1 TCR transgenic Tregs ...... 37

1.2.6.2 CAR-Tregs ...... 37

Chapter 2: Transcriptional Atlas of First- and Second-Generation CAR T Cells

Reveals Distinct Effects of Intracellular Domains ...... 40

2.1 Introduction ...... 40

2.2 Results ...... 42

Figure 2.1 Generation of CAR T cells...... 44

Figure 2.2 EGFR CAR constructs ...... 45

Figure 2.3 Gene signature of ligand-independent signaling from the CD3z domain...... 47

Figure 2.4 Ligand-independent signature genes in EGFR CAR T cells ...... 48

Figure 2.5. DE genes between CAR BBz and 28z...... 49

Figure 2.6 BBz CAR T cells have increased fatty acid metabolism before activation ...... 50

Table 2.1 ...... 52

Figure 2.7 BBz and 28z DE genes...... 53

Figure 2.8 Increased HLA-DR surface protein on activated EGFR BBz CAR T cells ...... 54

Figure 2.9 BBz upregulates ENPP2, IL12RB2 and IL23 ...... 56

Figure 2.10 Cytokine secretion by stimulated EGFR CAR T cells...... 57

Figure 2.11 BBz CARs T cells express increased adhesion molecules ...... 58

Figure 2.12 4-1BB costimulation initiates early TH1 polarization program...... 59

Figure 2.13 CD28 costimulation increases PD1 expression in CAR T cells...... 60

Figure 2.14 28z CARs express increased anti-apoptotic genes ...... 61

2.3 Discussion ...... 62 ix

2.4 Materials and Methods ...... 67

2.5 Acknowledgements ...... 74

Chapter 3: Human Regulatory T Cells Modified with CD28 but not 4-1BB-Based

CARs Maintain Suppressive Function...... 75

3.1 Introduction ...... 75

3.2 Results ...... 77

Figure 3.1 Generation of CAR-Tregs...... 79

Figure 3.2 Treg phenotype and Foxp3 stability...... 80

Figure 3.3 Foxp3 expression is stable after transduction, bead expansion, and

re-stimulation...... 82

Figure 3.4 CAR-Tregs can be activated through their CAR or their TCR...... 85

Figure 3.5 4-1BB costimulation affects aspects of CAR-Treg phenotype...... 86

Figure 3.6 4-1BB costimulation decreases CAR-Treg suppressive function...... 89

Figure 3.7 BBz CAR-Tregs have increased basal IL-2 consumption and form aggregates in

culture...... 91

Figure 3.8 Foxp3+ CAR-Tregs degranulate and mediate target cell cytolysis ...... 94

Figure 3.9 Target specific cytolysis by CAR-Tregs...... 95

Figure 3.10 Target specific cytolysis by CAR-Tregs is independent of the CD19 scFv ..... 96

Figure 3.11 Target specific lysis by CAR-Tregs is perforin/ granzyme dependent...... 98

Figure 3.12 CAR-Tregs traffic to antigen expressing tissue in vivo...... 99

Figure 3.13 Skin xenograft model of CAR-Treg mediated suppression...... 102

Figure 3.14 In vivo skin xenografts model setup ...... 103

Figure 3.15 In vivo CAR-Tregs exhibit low amounts of tissue cytotoxicity and express

immunosuppressive cytokines...... 104

3.3 Discussion ...... 105 x

3.4 Materials and Methods ...... 110

3.5 Acknowledgements ...... 124

Chapter 4: General Discussion ...... 125

4.1 Differential effects of costimulation in CAR Teff cells are supported by the known

mechanisms of natural 4-1BB and CD28 T cell signaling ...... 126

4.2 The overlap of 4-1BB signaling in T cells and other cell types ...... 128

4.3 4-1BB and memory CD8+ T cells ...... 129

4.4 Ligand-independent signaling in CAR T cells ...... 131

4.5 Antigen-specific CAR-Treg cells...... 133

4.6 Indirect Treg suppression of Teff cells ...... 135

4.7 CAR-Treg mediated cytotoxicity ...... 136

4.8 4-1BB inhibition of CAR-Treg suppressive function ...... 139

4.9 Future Directions ...... 141

4.9.1 Further interrogating the transcriptome of CAR T cells ...... 141

4.9.2 Alternative models of tissue-specific suppression by CAR-Tregs ...... 144

4.9.3 Designing safer tissue antigen-specific CAR-Tregs ...... 145

4.10 Concluding Remarks ...... 146

Appendix ...... 148

References ...... Error! Bookmark not defined.

Glossary

28z – Second-generation CAR construct containing CD28 costimulation 4-1BB – Tumor necrosis factor receptor superfamily, member 9; TNFRSF9 4‑1BBL – 4-1BB ligand AAV – Adeno-associated virus AKT – Protein kinase B ALL – Acute lymphoblastic leukemia AP-1 – Activator protein-1 APC – Antigen presenting cell ATCC – American Type Culture Collection BBz – Second-generation CAR construct containing 4-1BB costimulation BLI – Bioluminescence imaging CAR – Chimeric antigen receptor CBG-GFP – Click beetle green luciferase and green fluorescent protein CCL3 – Chemokine (C-C motif) ligand 3 CCL4 – Chemokine (C-C motif) ligand 4 CCR7 – C-C chemokine receptor type 7 CD – Cluster of differentiation CD137 – 4-1BB/TNSFRSF9 CD27 – Cluster of differentiation 27 CD3z – T-cell surface glycoprotein CD3 zeta chain, CD247 CD39 – Ectonucleoside triphosphate diphosphohydrolase-1 CD45 – Protein tyrosine phosphatase, receptor type, C CD45RA – Isoform of CD45 on found on naïve cells CD45RO – Isoform of CD45 on found on memory T cells CD62L – L-selectin CD73 – Ecto-5′-nucleotidase CD80 – Costimulatory ligand B7-1 CD86 – Costimulatory ligand B7-2 CD95 – Fas receptor CFSE – Carboxyfluorescein succinimidyl ester CMA – Concanamycin A CNS – Central nervous system CNS2 – Foxp3 intronic element CRISPR – Clustered regularly interspaced short palindromic repeats CTL – Cytotoxic T lymphocytes CTLA4 – Cytotoxic T-lymphocyte-associated protein 4, CD152 DAG – Diacylglycerol DAPI – 4′,6-diamidino-2-phenylindole DC – Dendritic cell DE – Differentially expressed DNA – Deoxyribonucleic acid xii

EAE – Experimental autoimmune encephalomyelitis EGFR – Epidermal growth factor receptor EGFRvIII – Epidermal growth factor receptor variant III ERK – Extracellular signal-regulated kinases FACS – Fluorescence-activated cell sorting FAM – Fluorescein dye Fas – Fas receptor, also known as apoptosis antigen 1 and CD95 FasL – Fas ligand FBS – Fetal bovine serum Foxp3 – Forkhead box P3 GADS – GRB2-related adapter protein 2 GSEA – Gene set enrichment analysis GITR – Glucocorticoid-induced TNFR-related protein also known as TNFRSF18 GITRL – Glucocorticoid-induced TNFR-related ligand GM-CSF – Granulocyte-macrophage colony-stimulating factor GRB2 – Growth factor receptor-bound protein 2 GZMA –Granzyme A GZMB –Granzyme B h – Hours HEK293T – Human embryonic kidney cells 293T HEX – Hexachlorofluorescein HLA-A2 – human leukocyte antigens, serotype A2 ICAM-1 – Intracellular adhesion molecule 1 ICOS – Inducible T-cell costimulatory also known as CD278 IFNg – Interferon gamma IKZF2 – IKAROS Family Zinc Finger 2, gene encodes the helios protein IL – Interleukin IP.– Intraperitoneal IP3 – Inositol triphosphate ITAM – Immune receptor tyrosine-based activation motif IV. – Intravenous IgG – Immunoglobulin G JNK – c-Jun N-terminal kinase K562 – Erythroid-myeloid precursor cell line derived from chronic myeloid leukemia LAG3 – Lymphocyte-activation gene 3 LAP – Latency-associated peptide LCK – Lymphocyte-specific protein tyrosine kinase LCMV – Lymphocytic choriomeningitis virus LFA – Lymphocyte function-associated antigen MAPK – Mitogen-activated protein kinase MHC – Major histocompatibility protein MHCII – Major histocompatibility protein class II MLR – Mixed lymphocyte reaction MOI – Multiplicity of infection xiii

NFAT – Nuclear factor of activated T-cells NFkB – Nuclear factor kappa-light-chain-enhancer of activated B cells NK Cell – Natural killer cell NSG mouse– NOD scid IL2Rgammanull mouse OX40 – Costimulatory receptor also known as CD134 and TNFRSF4 PBMC – Peripheral blood mononuclear cell PDK1 – Pyruvate dehydrogenase kinase 1 PI3K – Phosphatidylinositol-4,5-bisphosphate 3-kinase PKCθ – Protein kinase C theta PMA – Tetradecanoylphorbol acetate PRF1 – Perforin 1 RASGRP – RAS guanyl-releasing protein 1 RNA – Ribonucleic acid RNAseq – RNA sequencing RT – Room temperature scFv – Single chain variable fragment SMAC – Supramolecular activation cluster SQ. – Subcutaneous SynNotch – Synthetic notch receptor T2A – Type of self-cleaving 2A peptide TBP – TATA-binding protein TCR – T cell receptor TGFb – Transforming growth factor beta TH – T helper cell TNP – 2,4,6-trinitrophenol TNBS – 2, 4, 6-trinitrobenzene sulfonic acid TNF – Tumor necrosis factor TNFRSF - TNF receptor superfamily TRAF – TNF receptor associated factor TSDR – Treg-specific demethylated region Tconv – Conventional T cells Teff – Effector T cells TH1 – T helper cells type 1 TH2 – T helper cells type 2 Treg – Regulatory T cell UT – Untransduced T cell VCAM-1 – Vascular cell adhesion protein 1 ZAP70 – Zeta-chain-associated protein kinase 70 z– First-generation CAR with a functional CD3 zeta signaling domain Dz – First-generation CAR with a truncated CD3 zeta signaling domain

xiv

Chapter 1. Introduction

T cells genetically modified to express a Chimeric Antigen Receptor (CAR) gain incredible target-specific cytolytic properties1. The original CARs were engineered to be expressed in effector T cells (Teff) in order to target virus-infected and cancerous cells2,3. A new T cell type, T regulatory cells (Treg), are being modified with CARs to investigate their use as “suppressors” of the immune system4.

The first generation of CARs contained an antigen-binding region combined with a T cell receptor based signaling domain (CD3z)5. The addition of costimulatory domains in

CAR T cells improved their persistence and proliferation, enabling clinical anti-tumor responses in cancer patients6. Each costimulation domain is thought to affect the CAR

T cell properties differently, though the mechanisms underlying these differences are not well understood7. Due to the incredible therapeutic potential of CARs in both Teff cells and Tregs, there is a large interest in understanding each component, including the costimulation domain to better improve CAR-modified cell therapies.

This dissertation seeks to advance the understanding of engineered costimulation domains and their effects on CAR T cells, from two different perspectives. In Chapter 2, we describe a transcription based approach to determine the RNA expression changes downstream of CAR T cells signaling with different costimulation domains. In Chapter 3, we detail the experiments performed to examine the effect of costimulation on the suppressive function and phenotypic stability of CAR-Tregs.

1 To give context about the costimulation domains used in CAR T cell therapies, this chapter will detail aspects of costimulation, CARs, and the biology of Teff and Treg cells.

1.1 Costimulation and Chimeric Antigen Receptors

T cell activation is described as a two-signal model. The first signal, which provides specificity, is from the T cell receptor (TCR) binding to its particular antigen in the form of a peptide bound to the major histocompatibility (MHC) complex. The second signal is from the interaction of costimulatory receptors with their ligands displayed by mature antigen presenting cells (APCs)8. In the absence of costimulation, T cells fail to mount an efficient effector response entering a state of anergy (non-responsiveness) or in some cases undergoing apoptosis9,10. Furthermore, it is understood that costimulatory molecules play an important role in regulating T cell activation, effector function, subset differentiation, and T cell survival11-13. Thus, costimulation is fundamental to the activity of T cells and therefore must also play an important role in CAR T cell therapy. This thesis focuses on the two costimulation domains of commercially available CARs, 4-

1BB, and CD283. This chapter will also briefly highlight other costimulation domains that are being actively investigated in new CAR iterations.

1.1.1 CD28 signaling

CD28 is a homodimeric glycoprotein and the best-characterized of the costimulatory receptors14. CD28 is constitutively expressed on all mouse T cells, 80% of human CD4+

T cells, and 50% of human CD8+ T cells14. The receptor signals when bound to its 2

ligands B7-1 (CD80) or B7-2 (CD86), which are expressed by activated APCs that include B cells, dendritic cells (DCs), and macrophages15.

Three cytoplasmic signaling motifs have been identified in the CD28 intracellular tail:

YMNM, PYAP, and PRRP16. Signal transduction through the CD28 intracellular tail involves the recruitment of kinases to phosphorylate the tyrosine-based signaling motifs followed by the recruitment of additional adaptor proteins and kinases to bind to the newly phosphorylated motifs and activate further downstream pathways through PI3K,

PDK1, PKCθ, GRB2, GADS, and RASGRP, among others. This results in the activation of NFAT, AP-1, and NF-κB family transcription factors14,17,18.

Activation of these transcription factors leads to T cell proliferation and the transcription of IL2 cytokine gene 17. In addition to amplifying TCR signaling, CD28 has unique roles because CD28 signaling promotes IL-4 production and can lead to the differentiation of

19 TH2 helper cells . CD28 signaling induces the upregulation of anti-apoptotic protein

Bcl-xL, which has been shown to be important for T cell resistance to activation-induced cell death20. Furthermore, signaling can increase glucose uptake and enhance glycolysis in activated T cells18,21.

In addition to promoting the Teff cell functions described, CD28 also serves as an important receptor for the promotion of the anti-inflammatory function of Treg cells.

CD28 signaling is important in generating maximum IL-10 responses from Tregs22.

CD28 has also been shown in some contexts to prevent spontaneous autoimmunity by promoting Treg function23.

CTLA4 is the coinhibitory receptor of CD2824. CTLA4 also binds to CD80 and CD86, but with higher affinity than CD28. Signaling through CTLA4 inhibits T cell activation via

CD28/TCR signaling. CTLA4 excludes CD28 from the immune synapse during TCR activation. Furthermore, CTLA4 can dampen CD28 costimulation through cell extrinsic effects by decreasing the available B7 on APCs25.

1.1.2 4-1BB signaling

Like CD28, 4-1BB (CD137 or TNFRSF9) is a costimulation receptor, however unlike

CD28 which is expressed on naïve T cells, 4-1BB is only upregulated after T cell activation. 4-1BB is a surface glycoprotein from the TNF receptor superfamilies

(TNFRSF)26. As with CD28, activating T cells through the TCR in combination with

4-1BB delivers costimulation to T cells. 4-1BB costimulation enhances IL-2 secretion, T cell proliferation, and the expression of anti-apoptotic proteins. The 4-1BB receptor is induced on activated natural killer (NK) cells, natural killer T cells, Tregs, and CD4+, and

CD8+ T cells. It is also constitutively expressed on monocytes, blood vessel endothelial cells, and DCs27-30. 4-1BB ligand (4-1BBL) is the only know ligand of 4-1BB, and it is expressed on activated APCs, including DCs, B cells, and macrophages31,32.

4-1BB/4-1BBL signaling is bidirectional, meaning that 4-1BBL also signals through its intracellular signaling domain in turn, activating the ligand-expressing APCs33.

The binding of 4-1BBL to 4-1BB causes clustering of multiple 4-1BB cysteine-rich cytoplasmic tails, which facilitate the recruitment of TRAF adapter proteins. 4-1BB is known to signal through TRAF1, TRAF2, and TRAF3, though the latter only applies in humans32. Signaling through TRAF proteins leads to activation of canonical and non-canonical NF-kB, p38 mitogen-activated protein kinase (MAPK), activator protein 1

(AP1), ERK, and NFAT. TRAF2, in particular, activates JNK and NF-kB34.

In mice, signaling through 4-1BB can act as an antigen-independent signal for cell proliferation. 4-1BB is also important for promoting long-lived memory T cell formation35.

36 T cells activated with 4-1BBL were shown to express more TH1 cytokines . This fits with data describing that triggering of 4-1BB on DCs can increase their secretion of IL-6

30 and IL-12, a TH1 polarizing cytokine . 4-1BB costimulation more effectively activates and expands CD8+ T cells than CD28 costimulation in human T cells37-39. The addition of 4-1BBL in human CD8+ T cell culture results in the upregulation of effector molecules, such as perforin, granzyme A, and IFNg32. Furthermore, on T cells, 4-1BB-promotes T cell adhesion to fibronectin by activating integrin-1b40.

4-1BB signaling has a number of effects outside of T cells. In NK cells, 4-1BB enhances

IFNg secretion and antibody-dependent cell-mediated cytotoxicity. On endothelial cells,

4-1BB agonist activation leads to the upregulation of adhesion molecules VCAM1 and

ICAM1. This will become relevant in Chapter 2 where we found adhesion molecules were upregulated in CAR T cells with 4-1BB costimulation.

Agonist CD137 antibodies can induce curative immune responses in mouse tumor models41. This has led to clinical trials using agonist anti-CD137 antibodies either as monotherapies or in combination with checkpoint blockade39 (PF-05082566,

NCT02253992). In human clinical studies, 4-1BB agonists increase the IFNg production by T cells42. Furthermore, anti-CD137 therapy reduces Treg infiltration into human tumors43.

The role of 4-1BB stimulation in Tregs is controversial. Some reports suggest that anti-4-1BB mAbs can inhibit Treg suppressive function, and others indicate that 4-1BB activation on Tregs induces their proliferation44,45.

1.1.3 Costimulation domains beyond 4-1BB and CD28

Costimulatory receptors have an important role in determining T cell fate, therefore considerable effort has gone into understanding how signaling through different costimulatory domains can modulate the phenotype of both Teff and Treg cells. Other costimulation domains are being considered to improve and diversify CAR T cell design and activity—including CD27, ICOS, GITR, and OX40 CARs46. Costimulatory receptors fall into either the immunoglobulin superfamily (IgSF), such as ICOS and CD28, or the

TNFRSF, which includes 4-1BB, OX40, and GITR, among others34. This section will

focus on the costimulation molecules that are being tested as intracellular signaling domains in CAR T cells.

1. ICOS

2. GITR

3. OX40

4. CD27

Inducible T cell costimulator, ICOS, is an IgSF costimulation domain and a member of the B7 family. ICOS binds ICOS ligand and signals by recruiting PI3K to its intracellular domain which enhances AKT downstream signaling. Activation through ICOS induces

IL-4, IL-10, and IL-21 cytokine secretion, but it does not induce IL-2 secretion to the same degree as CD28. In addition, ICOS has been shown to induce transcription factor,

BCL6 expression, enhancing T follicular helper (TFH ) cell differentiation, and therefore plays an important role in humoral immunity47. ICOS is also essential for the

48 development of human TH17 cells . In CAR T cells, ICOS has been shown to enhance

+ 49,50 the TH17/TH1 phenotype and improves CD4 T cell persistence . Furthermore,

ICOS-containing CAR T cells secrete a distinct cytokine repertoire that includes IL-17A,

IL-17F, and IL-2249.

Glucocorticoid-induced TNFR-related protein (GITR) is a member of the TNFRSF that signals through TRAF2 when bound to its ligand, GITRL51. In mice, GITR costimulation improved the quality of CD8+ T cell responses that are CD4+ T cell dependent, enabling

mice to resolve chronic LCMV infection52. Research has shown signaling through GITR in T regulatory cells abrogates their suppressive function. Similarly, mice develop spontaneous autoimmunity following administration of anti-GITR agonist antibodies51. In

CAR T cell cytotoxicity assays, GITR costimulation is as effective as CD28 and 4-1BB.

GITR costimulation is significantly better than a CAR designed with mutated GITR signaling domains, which indicates that costimulation through GITR is beneficial to CAR

T cell cytotoxicity53. Whether GITR affects any other aspects of the CAR T cell phenotype has yet to be determined.

OX40, otherwise known as TNFRSF4 or CD139, is also from the TNFRSF. Like 4-1BB, it signals through TRAF2, but in addition, TRAF3 and TRAF5 adaptor proteins are also known to associate with the receptor. T cells, NK cells, and neutrophils induce OX40 expression transiently after activation. OX40 signals when it binds its ligand OX40L and activates downstream canonical and non-canonical NFkB31. Signaling through OX40 promotes the expression of survival proteins Bcl-xL and Bcl-2 and is thought to be essential for the long-term survival of CD4+ T cells54,55. OX40 signaling has also been

56 shown to promote TH9 responses . Similar to 4-1BB, CAR T cells containing OX40 costimulation domains express higher IFNg and lower IL-2 levels than CD28 CAR T cells do. But in contrast to 4-1BB, OX40 was the most effective at preventing activation-induced cell death of the effector memory T cell population57.

OX40 signaling can prevent the induction of Tregs in the periphery and reduce

Tregs’ stability58. In contrast, others have reported that OX40 increases proliferation and maturation of thymic Tregs59. It is still unclear whether OX40 signaling benefits Tregs’ function.

CD27 is another member of the TNFRSF, and it signals when bound to CD70. Unlike

4-1BB, GITR, and OX40, whose expression is induced on T cells following TCR/CD28 activation, CD27 is constitutively expressed on naïve and memory T cells60. CD27 signaling in the absence of TCR/CD28 activation has been shown to induce CD8+ T cell proliferation and enhanced survival, but CD27 does not induce cytokine secretion or cytolytic activity61. Signaling through CD27 induces the binding of TRAF2 and TRAF5, which, like OX40, induces the canonical and non-canonical NF-κB pathways, as well as

AP-1 downstream of the JNK–MAPK pathway. CD27 signaling also promotes TH1 cell differentiation62. CAR T cells expressing a CD27 costimulation domain had enhanced survival, cytokine secretion, and proliferation compared to first-generation CAR T cells and had comparable in vivo tumor clearance when compared to 4-1BB and CD28 containing CAR T cells63.

1.1.4 An overview of CAR T cell therapy

CAR T cell therapy is the use of genetically modified T cells that are redirected to target cells expressing a specific antigen. In CAR T cell therapy, T cells are isolated from a patient by apheresis. T cells are then expanded ex vivo, usually using anti-CD3 and

anti-CD28 coated beads or cell-based artificial APCs, during which time T cells are genetically modified to incorporate the CAR construct64. CAR T cells are expanded in culture for several days, frozen, and shipped back to the clinic, where they can be thawed and administered to the patient as adoptive T cell therapy64,65.

There are factors that can influence the quality of the CAR product aside from the costimulation domain. For instance, the T cell starting population and the materials used to expand the T cells can affect the CAR T cell product66. Another element affecting the final modified cells is the DNA delivery system used. DNA is usually brought into the system using lentivirus or retrovirus, but there are other non-viral methods of DNA delivery such as piggyback transposons, and electroporation7,64,66-68. Finally, the T cell culture conditions—including the growth media, cytokines supplemented in the media such as IL-2, IL-15, or other polarizing cytokines, and the number of days the T cells are grown in culture before cryopreservation—can influence the potency of the final product66,69-71.

1.1.5 Chimeric antigen receptor constructs

Unlike B cells, T cells have the functionality to specifically lyse and destroy tumor cells.

However, T cells are restricted, in that they can only be activated by the appropriate peptide-MHC combination72. Chimeric antigen receptors combine the antibody-binding properties of a B cell with the cell-lysing properties of a T cell by fusing an extracellular antibody derived binding domain with the signaling domain of the CD3z chain from the T

cell receptor complex1. The receptor therefore redirects T cells to lyse antigen-bearing target cells in a manner that is independent of MHC. Therefore CAR T cells have an advantage by being unaffected by some of the tumor mechanisms of immune escape, such as downregulation of MHC class I and defective antigen processing73. Chimeric antigen receptors include the following modules (Fig 1.1):

1. Single-chain variable fragment (scFv).

2. Hinge region

3. Transmembrane domain

4. Costimulation domain

5. CD3z chain

CAR ζ CAR 28ζ CAR BBζ CAR ∆ζ Linker

scFv

Hinge

Transmembrane Domain

CD28 4-1BB

CD3ζ

CD3ζ CD3ζ

Figure 1.1 – First- and second-generation CAR constructs

The scFv is derived from the antibody variable regions of the light and heavy chain joined by a flexible linker. The scFv gives the CAR antigen specificity, directing the CAR

T cell to signal when it binds its cognate antigen. CARs have been further developed to 11

utilize different extracellular recognition modules, such as a camelid-derived VHH binding domain 74, Fn- or DARPin-based binding domains that are made of fibronectin or ankyrin repeat proteins respectively75,76. These domains are also made of natural ligands, which can be used to destroy cells expressing their cognate receptor77,78. The affinity and avidity of the scFv and the distance of the epitope from the target cell membrane can both affect the CAR T cell activity79-81. The immunogenicity of the scFv is an additional aspect important to the clinical efficacy of the CAR, as murine-derived or other foreign CAR-derived peptides could be rejected by the host immune system82.

The CAR hinge, with or without a longer ‘spacer’ region, links the scFv with the transmembrane domain. To date, transmembrane domains have been derived from IgG antibody subfamilies, CD8, and CD28 extracellular membrane proximal regions81,83,84.

The cysteine residues in the hinge region enable disulfide bonds between two CARs, allowing their dimerization85. The hinge region is important for the positioning of the binding domain. Depending on the scFv and target antigen, longer spacer regions may be necessary to give the CAR the flexibility required to interact with its epitope86. The optimal length of hinge-spacer region studied over several CAR constructs was scFv-target epitope specific87. Classically the added spacer is based on CH2-CH3, or

CH3 portion of the IgG domain. Incorporation of certain IgG-derived spacer-region domains increased tonic signaling84,88. Furthermore, in vivo Fc-FcgR interactions between the CAR- and FcgR-expressing cells influenced CAR T cell trafficking and abrogated tumor clearance83,84. The length and composition of the hinge-spacer region

is underappreciated, and though the region lacks signaling function, the hinge-spacer region is still a critical component of the CAR construct. Optimizing the hinge through both in vitro and in vivo studies is a valuable task to improve CAR T cell function89.

The transmembrane domain anchors the receptor in the membrane. The transmembrane domain is composed of hydrophobic alpha helix domains that generally derived from CD3z, CD4, CD8, or CD28 molecules73. Though the transmembrane module does not contain any signaling domains, it can influence ligand-independent

“baseline” signaling of the CAR receptor which we will detail later in Chapter 188.

Furthermore, simply swapping the CD3z transmembrane domain to a CD28 domain in the CD19-specific CAR improved the stability of the CAR on the cell surface73. An in- depth understanding of the effect of hinge regions on downstream costimulation and

CD3z signaling is still lacking.

The addition of a costimulation domain greatly improved CAR persistence compared to first-generation CAR T cells. First-generation CARs had incomplete T cell activation due to the lack of natural costimulation because tumors frequently do not provide ligands for costimulatory molecules on T cells. The costimulation domains promote anti-apoptotic proteins and affect T cell proliferation, polarization, and the generation of memory responses46,49,50,90. Different costimulation domains are being tested in CAR T cells, though 4-1BB and CD28 are the only domains included in the FDA-approved CAR

T cells53,57,91. In addition, new CARs described as “third-generation” combine multiple

costimulation modules into one CAR construct to investigate the potential of synergy between several costimulation domains 92. The position of the costimulation domain in the construct and the type of transmembrane domain used seem to affect the degree of signaling through the costimulatory domain46,53.

The CD3z chain contains three immunoreceptor tyrosine-based activation motif (ITAM) signaling modalities. Signaling through the ITAMs is mediated by tyrosine phosphorylation and recruitment of Zap70 kinases93,94. CD3z signaling is necessary for

CAR T cell degranulation and cytolysis, and the z module is generally the most distal from the membrane.

1.1.6 CAR vs. TCR signaling

The TCR dictates the antigen-specificity and activation of a T cell. Classical TCR signaling involves the formation of a highly organized TCR signaling complex that is initiated by the binding of the variable domains of the TCR a and b subunits with a short amino acid chain presented in the peptide binding cleft of the MHC on an APC 95. The

TCR forms a CD3 complex with a and b subunits as well as signaling subunits d, g, e, z96. TCR–peptide-MHC binding enables ITAMs to be phosphorylated in the d, g, e, z subunits by phosphokinase LCK95. The CD45 receptor, a tyrosine phosphatase, suppresses ITAM phosphorylation and is excluded from the TCR complex during activation. In concert with TCR activation, a supramolecular activation cluster (SMAC) is formed by shuffling membrane-bound molecules, excluding larger molecules, and allowing smaller ones, such as costimulatory CD28 and coreceptors CD4 or CD8, to 14

cluster in the interior synapse97. Together, an immune synapse between the APC and the T cell is formed, allowing for full T cell activation. Phosphorylation of the ITAMs leads to the recruitment and phosphorylation of Zap70, which begins the signaling cascade by activating adaptor proteins and signaling molecules such as effector enzyme PLCγ172.

Membrane phosphatidylinositol(4,5)biphosphate is hydrolyzed by PLCγ1 to produce inositol triphosphate (IP3) and diacylglycerol (DAG). DAG regulates a number of signaling molecules, and IP3 enables calcium flux. Calcium flux in the T cells leads to the activation of calcineurin phosphatase to activate NFAT, a transcription factor that, once activated, can enter the nucleus to upregulate genes important for T cell activation such IL2 98.

In contrast to the canonical TCR-peptide-MHC activation complex, CAR T cells are activated by recognition of an epitope on a surface protein or glycoprotein by the scFv or alternate binding domain. T cells expressing chimeric receptors are able to discriminate between antigen-expressing and non-expressing cells with negligible bystander cytotoxicity1. Upon antigen recognition, CAR intracellular signaling domains transmit activation and costimulatory signals to T cells through phosphorylation of the

ITAMs contained in the CD3z domain99. In concert with ITAM signaling, the costimulation domains will also signal, which recruits adaptor proteins and activates their downstream pathways. Downstream of ITAMs, signaling is thought to be similar in

many ways to TCR signaling, and CAR T cells also produce calcium flux, activate

NFAT, and upregulate IL-2 expression100,101.

Compared to the 10 subunits that make up a TCR which contains 10 ITAMs, the CAR is made of a single molecule that contains only 3 ITAMs in the intracellular domain.

Another key difference between CAR and TCR signaling is the organization of the immune synapse. The TCR synapse forms as an organized SMAC, with a classical bullseye structure made up of a central patch of LCK molecules, surrounded by actin clearance and then an LFA adhesion ring. In contrast, the CAR immune synapse forms a disorganized pattern of LCK microclusters with less actin clearance and no LFA adhesion ring102. Furthermore, the smaller CAR T cell synapse reached maximum signaling faster. It also more rapidly recruits lytic granules compared with the TCR. The differing immune synapse organizations may explain the more immediate degranulation by CAR T cells to lyse tumor targets102.

A difference between CARs and TCRs is that CAR affinities for antigen are much higher than the typical TCR affinities for peptide-MHC103. Though T cells can be activated through the TCR by a single peptide-MHC, generally the TCR-peptide-MHC interactions are low affinity within the 104–106 M−1 range104. The threshold for TCR signaling is hypothesized to be reached by a combination of coreceptors CD4 or CD8 binding to the

MHC to lower the threshold for TCR signaling and by the ability of an individual peptide-

MHC to bind multiple TCRs on a cell (serial-triggering) with a fast off-rate105. In contrast,

CAR scFvs typically bind their cognate antigen in the range of 106–109 M−1. While CARs have higher affinity for antigen than TCRs, they are not activated by cells with very low levels of surface antigen80.

Additionally, CARs differ from TCRs by the lack of temporal activation of costimulation.

In TCR signaling, TCR and CD28 costimulation signal early whereas, only later are other costimulation molecules such as 4-1BB and OX40 upregulated to signal.

Alternatively, CAR T cells signal only though their given costimulation domains and this signaling is always in combination with CD3z signaling. Finally, CAR T cells are thought to be independent of antigen-presenting cells as they do not require costimulation or

MHC to be activated. However, to what extent there is crosstalk between APCs and other immune cells with CAR T cells is not well known.

1.1.7 The effect of CD28 and 4-1BB costimulation on CAR T cells

CARs have been developed by independent groups to incorporate CD28 or 4-1BB costimulatory domains90,91,106. The addition of costimulation greatly enhanced signaling strength, cytokine secretion, and improved persistence and in vivo potency of the CAR

T cell6,107. However, each costimulation domain is thought to affect the CAR T cell properties differently. Both types of CAR constructs have demonstrated efficacy in B cell leukemias108,109 and lymphomas110,111, but their engraftment kinetics, persistence, and toxicity profiles are distinctive112.

CD28-based CARs undergo more rapid expansion but have not been found to persist past 30 days113,114. Alternatively, 4-1BB-based CARs have been detected in circulation 6 years after T cell infusion46,112. 4-1BB has been found to be more effective at promoting CD8+ cytotoxic T cell responses115. Furthermore, Kawalekar, O., et al. found 4-1BB-containing CAR T cells have higher frequencies of CD8+ central memory cell populations as determined by CD45RO+ CCR7+ markers, whereas CD8+ CD28- containing CAR T cells had an increased percentage of effector memory T cells7.

Additionally, 4-1BB-containing CAR T cells had increased maximal respiratory capacity and induced mitochondrial biogenesis. 4-1BB CARs utilize enhanced fatty acid oxidation to fuel their energy needs. In contrast, CD28-containing CAR activation induced a glycolytic metabolism signature that was further supported by higher glucose uptake and an increased extracellular acidification rate7.

Investigations of CAR T cells with alternate scFvs have identified additional effects caused by 4-1BB versus CD28 costimulation. For example, CAR T cells targeting

CD5—a molecule also expressed on T cells—found 4-1BB-containing CARs upregulate the adhesion molecule, ICAM, which was shown to increase fratricide due to increased cell-cell interactions between CAR T cells116. Additionally, in GD2 CARs, the scFv increased tonic signaling through CARs and the effect of this tonic signaling on CAR T cell phenotype was distinct depending on the type of costimulation117. In this case,

CD28 costimulation induces exhaustion in CAR T cells with extreme tonic signaling, whereas 4-1BB costimulation in the CAR construct can rescue this exhaustion.

However, in other 4-1BB-based CAR constructs, “tonic signaling” from 4-1BB costimulation can lead to Fas receptor upregulation on the cell surface, increasing

Fas-FasL mediated cell death68,117.

1.1.8 Tonic signaling in CAR T cells

The overexpression of any receptor such as a CAR on the surface of a cell has the potential to signal, usually at low levels, independent of antigen stimulation118. Antigen independent signaling, referred to as tonic signaling, has been described to varying degrees in CAR T cells88,117. Each signaling module of the CAR has the potential to signal independent of antigen binding. Tonic signaling can be enhanced by the high surface expression and clustering properties of scFvs. Moreover, the level and outcome of tonic signaling seems to be CAR construct dependent.

Frigault et al. reported antigen-independent proliferation and cytokine secretion by T cells modified with specific CAR constructs88. Mesothelin and c-Met CAR T cells, but not

CD19 (FMC63), display tonic signaling detectable by antigen independent proliferation.

Furthermore, CD28 signaling domain containing CARs, as well as IgG4 over CD8 transmembrane domain, had the highest level of tonic signaling, which correlated with decreased persistence and inferior anti-tumor properties in vivo. Long et al. described tonic signaling from CARs with a GD2 scFv, where both CD28 and 4-1BB second generation CARs displayed antigen-independent CD3z phosphorylation, but only CAR T cells with CD28 costimulation expressed high levels of exhaustion markers PD1, TIM3, and LAG3117. The spacer region added to PSCA-specific CARs increased ligand

independent CD3z phosphorylation, cell size, and CD25 expression and modified the T cell phenotype to be more “effector memory like” but did not induce exhaustion84.

Tools such as CRISPR allowed the CAR to be integrated into the TCR alpha (TRAC) locus, thus preventing expression of the surface TCR and enabling CAR expression off the TRAC promotor119. This led to reduced tonic signaling and improved T cell potency.

Tonic signaling from the 4-1BB costimulatory domain as opposed to tonic signaling from the CD3z domain has also been described by Gomes-Silva et. al68. In this case, 4-1BB tonic signaling increased surface Fas expression and therefore made T cells more susceptible to Fas-FasL mediated apoptosis68. Tonic signaling comes in a variety of

“flavors” leading to diverse effects on CAR T cell properties. The effect of CD3z chain versus costimulatory domains’ tonic signaling is still not well understood. In addition, to what degree the choice of scFv, hinge, and transmembrane domain and costimulation affect the degree of ligand-independent signaling remains to be fully understood.

1.1.9 Cell therapy/clinical impact of CAR T cells

CAR T cell therapy rose to FDA approval due to the stunning success of CD19-targeting second generation CARs to treat of acute lymphoblastic leukemia (ALL)109,113,114,120,121, chronic lymphoid leukemia122-125, and lymphoma122,126. New research with BCMA targeting CAR T cells for multiple myeloma illustrates the power of CAR T cells for hematological malignancies127. Antigen-loss tumor variants have been seen after CD19

treatment, and a wave of additional B cell antigen-targeting CARs are in development and being tested clinically5,128,129.

The success of CAR T cells in liquid tumors has not been mirrored in solid tumors71,130,131. The therapeutic effect of CAR T cells in solid tumors may be limited by the tumor microenvironment or by the lack of appropriate surface tumor specific antigens to target132. For example, EGFRvIII CAR T cells trafficked to the brain in glioblastoma models, and the cells were accompanied by a wave of Treg cells to the tumor site131. These EGFRvIII CAR T cells led to a decrease in target-expressing tumor cells, though the therapy was not curative.

As with all cancer therapies, CAR T cells do have toxic side effects5. Most of these side effects are on-target effects. Therefore, side-effects are reduced as target cells are destroyed5. This contrasts with chemotherapy, in which many of the adverse effects are off-target. B-cell aplasia in CD19 CAR T cell treated patients was a predicted and now well-described example of on-target toxicity of CAR T cell therapy120. Cytokine release syndrome presents with clinical symptoms of hypoxia, hypotension, and fevers that correlate with a huge elevation in cytokine serum levels, which occurs in a subset of patients treated with CD19 and BCMA CAR T cells109,133. High CAR T cell activity is thought to lead to systemic monocyte/macrophage over-activation, resulting in their release of IL-6. Cytokine release can be treated with corticosteroids and anti-IL-6 antagonists (tocilizumab) to limit severity109,133. Neurological toxicities are another dangerous adverse effect of CAR T cells5. There has been new evidence from 21

humanized mouse models that IL-1 secretion by monocytes may lead to neurologic toxicities, however our understanding is limited about this toxicity and potential treatments for patients134,135.

Efforts to understand CAR T cell toxicity and how it can be inhibited is critical to improving CAR designing for liquid and solid tumors. Furthermore, engineering costimulation and other domains to offer T cells the most suitable properties to be effective in the tumor microenvironment would be a major step towards the eradication of solid tumors by CAR T cells.

1.2 Antigen-Specific T Regulatory Cells

T cells can be classified into many subsets that are specialized for certain functions.

Chapter 3 of this dissertation focuses on the expression of CARs in the regulatory T cell subset. CD4+ Tregs are lymphocytes that regulate tolerance to self-antigens, maintain tissue homeostasis, and suppress excessive immune responses136. Since their discovery, Tregs have been shown to play an essential role in preventing autoimmune diseases. Mouse models and human xenogeneic models have established that antigen- specific Tregs are more potent for restoring peripheral tolerance than polyclonal Tregs.

Modifying Tregs with CARs is a way to provide Tregs with antigen specificity. To date, the effect of CAR expression on Treg stability and function has not been fully investigated. Furthermore, the role of costimulation domains and their effect on CAR-

Treg cell phenotype and function has yet to be determined. Chapter 3 of this dissertation investigates the phenotypic and functional stability of Tregs modified with different CAR-encoded costimulation domains. Below, I will introduce the topic of T cell subsets with a specific focus on Tregs, their stability, and mechanisms of suppression.

Finally, this introduction will describe the research to date on engineering Tregs as adoptive cell therapy.

1.2.1 T cell subsets

Immunologists have a long history of classifying immune cells into subsets based on surface markers or secreted factors that predict their specialized function. Of the ab

TCR expressing cells, matured T cells can first be divided simply into the CD8+ and

CD4+ subsets, based on the expression of the CD8 and CD4 coreceptors, which 23

preferentially bind to MHC class I and MHC class II respectively. CD4+ and CD8+ T conventional (Tconv) cells are generated during T cell development in the thymus. CD4+ natural—thymus derived—Tregs are another subset of cells that differentiate during T cell development in the thymus.

Once cells egress from the thymus, Tconv cells can differentiate further based on the antigenic and environmental stimuli received during the initial T cell priming and subsequent recurring memory responses. T cells are activated based on the accumulation of information from TCR-peptide-MHC signal strength, the cytokine milieu, costimulation, coinhibition and the metabolic environment to become effector T cells. In simplistic terms, activated CD8+ T cells are known as cytotoxic T lymphocytes (CTLs) and are important for the destruction of virally infected and cancerous cells. CTLs are cytotoxic and degranulate to release granules containing perforin and granzyme B to induce target cell apoptosis.

Human CD8+ and CD4+ T cells can each be further classified into naïve, effector, stem cell memory, effector memory, or central memory based on the expression of CCR7,

CD62L, CD95, and CD45RO/CD45RA137. Naïve cells include T cells that have never experienced peptide-MHC specific activation and express CD45RA and lymph node homing receptors CCR7 and CD62L. In the lymph nodes, naïve T cells can be activated by APCs to become Teff cells. Teff cells reduce their expression of CCR7 so they can leave the lymphatic system to respond to injury in peripheral tissues138. Some Teff cells

will resolve infection and die, while others will differentiate into memory T cell subsets139. Memory T cells usually express CD45RO and are defined by their long- lifespan and faster response compared to naïve T cells. These memory cells produce greater quantities of inflammatory cytokines and/or quickly kill infected cells in the case of a recurring immune assault. Memory T cells develop from Teff cells or after the initial priming of naïve T cells, in some cases. Central memory cells express CCR7, retaining them in the secondary lymphoid organs, whereas effector memory cells are thought to remain in peripheral sites. Therefore, central and effector memory T cells can patrol lymphoid and peripheral tissues respectively138. The generation of T cell memory is critical for the persistence of CAR T cells7. It is unclear whether effector memory or central memory cells make up the optimal CAR T cell population, but mouse and human models tend to suggest that central memory cells may be more effective at mounting multiple responses and persist better than effector memory cells140-143. As described in earlier sections, the type of costimulation affects the memory phenotype of CAR T cells.

Stem cell memory T cells are a more recently described population expressing

CD45RA, CD62L, CD95, and CCR7144. Stem cell memory T cells are being actively studied for CAR T cell therapy because of their increased persistence and ability to differentiate into multiple cell subsets137.

Beyond their memory subsets, during activation, naïve CD4+ T cells can be polarized into different subsets based on the cytokine milieu from activated APCs137. CD4+ T cells are also known as helper T cells because they support CD8+ T cell and B cell

responses by providing cytokines depending on the type of inflammation. The importance of CD4 help is illustrated in the absence of these T cells where memory

CD8+ T cells exhibit impaired cytokine production, decreased persistence, and reduced ability to lyse virus infected and tumor cells. The kind of “help” provided by a CD4+ T cell is determined by their interaction with APCs. APCs are stimulated through their innate receptors, such as toll-like receptors, that sense conserved pathogen-associated molecular patterns and damage-associated molecular patterns. APCs respond by secreting polarizing cytokines. Based on the cytokines secreted by the activated APCs,

+ CD4 T helper cells can polarize into TH1, TH2, TH9, TH17, TH22, pTregs, or TFH

145,146 (follicular helper) cells . TH1 cells are polarized by IL-12 and secrete higher levels of IFNg and TNFa compared to other T helper subsets. They express the TH1 transcription factor T-bet and are thought to be important for driving CTL and NK cell responses against viruses and intracellular bacteria. Furthermore, they have been described as important for anti-tumor responses due to their ability to support NK and

147 CTL cells that have the potential to recognize and kill tumor cells . TH2 cells are polarized by IL-4. They classically express the GATA-3 transcription factor and make

IL-4, IL-5, and IL-13 cytokines. TH2 cells are found in allergic responses and are

145 important for protection from helminth infections . TH2 cells are also critical for generating IgE humoral responses. TH17 cells are a newer subset, thought to play a role in defense against extracellular bacteria and fungi148 as well as pathogenic responses in many autoimmune diseases. TH17 cells are defined by the transcription factor RORgt and are polarized by IL-6 and TGFb. TH17 cells secrete several cytokines,

including IL-17 and IL-22. In Chapter 2, I present new research to show that the type of costimulation domain included in Tconv cells can influence T helper cell polarization in

CD4+ CAR T cells.

Finally, CD4+ Tregs are the suppressor cells of the immune system. They can be induced by TGFb and IL-2, either ex vivo (iTregs) or in the periphery (pTregs). They can also be generated during T cell development in the thymus (tTregs). The rest of this introduction will be dedicated to describing Treg cells and their properties.

1.2.2 T regulatory cells

CD4+ CD25high Foxp3+ T cells, called regulatory T cells (Tregs) are considered the primary mediators of peripheral tolerance. They work by suppressing the activation and proliferation of both CD8+ and CD4+ Tconv cells, which we will be referred to as Teff when they are used in the context of activated T cells for Tregs to suppress149. Tregs were initially defined by their high expression of the high-affinity IL-2 receptor subunit,

CD25150. It was later discovered that Foxp3 is an important transcription factor of the

Treg lineage and that Scurfy mice, with mutated Foxp3, develop spontaneous autoimmunity151. Furthermore, IPEX (immunodysregulation polyendocrinopathy enteropathy X-linked) syndrome, a human disease with fatal multi-organ autoimmunity, is caused by deficiencies in the human FOXP3 gene152. Tregs are therefore critical for tolerance to peripheral self-antigens, thereby preventing autoimmunity. They are also important for suppressing other aberrant immune responses such as in the case of inflammatory bowel disease. Treg depletion can increase colitis in mice and humans,

driven by excessive inflammation in reaction to intestinal microbes153. Conversely, the addition of Tregs can suppress allergy as well as prevent graft-versus-host disease

(GvHD) and organ transplant rejection149,154-156. Tregs can also have a deleterious role in limiting anti-tumor immune responses157.

Tregs are divided into two subsets based on the site where they are induced: thymus- derived (tTregs) and peripheral (pTregs). tTregs are positively selected in the thymus from developing T cells with a higher-affinity TCR for MHC/self-peptide. High TCR signaling during development induces the expression of Foxp3 to initiate the Treg polarization process and ultimately leads the generation of CD4+ CD25high Foxp3+

Tregs158. pTregs are induced in the periphery by the combination of TCR bound to peptide-MHCII in the presence of TGFb and/or other immunosuppressive extracellular cues159.

Tregs are found in high numbers in the lymph nodes, where they are activated to mediate suppression in their local environment or by trafficking to sites of inflammation, infection and tumors149,160,161. They express various homing receptors such as CCR4,

CCR6, CXCR4, and CXCR5 to control this trafficking and localization. The homing properties of Tregs are key to the compartmentalization of Treg-controlled immune responses.

In summary, Tregs are defined by transcription factor Foxp3, high CD25 expression, and lack of IL-2 production. They modulate immune responses in the lymph nodes or at sites of inflammation to suppress excessive immune responses, and prevent reactions to self-antigens thus maintaining tissue homeostasis.

1.2.3 Mechanisms of Treg suppression

Understanding the mechanisms of Treg suppression is of vital importance to optimize adoptive Treg therapy to promote immune tolerance. In vivo, the main duty of thymic and peripheral Treg cells is to migrate to sites of inflammation and suppress various effector B, NK, and T cell responses. In vitro, Tregs co-cultured with Teff cells can suppress Teff cell proliferation and cytokine secretion after APC/TCR stimulation162.

Tregs need to be activated in order to mediate cell suppression. However once activated, Tregs can inhibit responder T cells irrespective of whether they share the same antigen specificity, known as bystander suppression163.

Tregs exert immune suppression though multiple mechanisms that can occur in a contact-dependent or contact-independent manner164. Tregs can suppress Teff cells directly162 or indirectly through their effects on the maturation and phenotype of

APCs164. Tregs are known to mediate suppression by secreting inhibitory cytokines like

TGFb and IL-10165,166. They can also modulate APC and Teff phenotype through cell- surface-bound molecules such as CTLA4 and LAG3166 and they can suppress immune

cells by metabolic inhibition. Finally, Tregs can suppress cells directly through cytolysis of inflammatory Teff and NK cells.

Data suggests cytokine secretion is one mechanism by which Tregs modulate their microenvironment to promote tolerance. However, whether cytokine secretion is the main mechanism of suppression remains debated. Tregs express high levels of surface- bound TGFβ in complexes with LAP and GARP proteins. Membrane-tethered TGFβ can be further upregulated during TCR activation, and membrane-tethered TGFβ has been implicated in Treg suppression both in vitro and in vivo167,168. In a murine model of type-1 diabetes, TGFβ secretion by Tregs and TGFβ receptor signaling in CD8+ T cells were both necessary for the suppression of islet destruction by islet-reactive CD8+ T cells. Activated Tregs can also secrete IL-35 and IL-10. In vitro, neutralizing IL-10 antibodies are unable to prevent Treg inhibition. However, in vivo colitis models have demonstrated that IL-10 secretion is essential to the ability of Tregs to resolve intestinal inflammation. IL-35 is a heterodimeric cytokine induced by transcription factor Foxp3 and secreted by Tregs. IL-35 is required for Treg suppressive function in mouse in vitro co-culture assays and in vivo colitis models169. Cytokines IL-10, IL-35, and TFGβ are all mediators of cell suppression by Tregs, and these cytokines seem to function only in specific settings that warrant further investigation.

As mentioned previously, CD25, the IL-2 receptor alpha chain, is highly expressed on the surface of Treg cells. IL-2 is also important for the maintenance of Foxp3+ thymus-

dederived Tregs, as Tregs do not produce their own IL-2 cytokine and are unable to grow or survive without it170,171. Tregs can suppress Tconv cell proliferation by acting as an IL-2 sink165,172. It is thought that Tregs can reduce IL-2 availability and therefore limit

T effector cell responses. Furthermore, Treg suppression is abrogated when exogenous

IL-2 is added to the culture173.

In addition to manipulating the cytokine milieu, Tregs can affect the metabolic environment of Teff cells through the intracellular or extracellular release of adenosine nucleosides164. Ectoenzymes CD39 and CD73 are expressed in concert specifically on the surface of Treg cells174. These enzymes generate pericellular adenosine.

Accumulation of adenosine acts on the adenosine A2A receptor—expressed on activated Teff cells—and elicits immunosuppressive cellular responses in the Teff cells.

Thus, Tregs can affect extracellular nucleotide catabolism to dampen Teff cell responses174.

As mentioned previously, Tregs can also indirectly suppress Teff cells through their actions on DCs and other APCs. Tregs are constantly interacting with DCs in vivo175,176.

DCs can be conditioned to be more inflammatory APCs or more tolerance-inducing

APCs. Tregs can indirectly suppress Tconv cells and B cells through their modulation of

DC phenotype via various mechanisms including CTLA4-B7 interactions. CTLA4, mentioned earlier, is a coinhibitory receptor that is highly expressed on the Treg surface after activation. CTLA4 mediates suppression by binding CD80 and CD86 on the

surface of DCs. CTLA4 can remove B7 molecules from the surface of APCs, a process called trans-endocytosis, preventing Teff cell costimulation177. CTLA4-B7 interactions between DCs and Tregs can also induce the expression of indoleamine

2,3-dioxygenase (IDO), a molecule that causes the catabolism of tryptophan to kynurenine, leading to starvation of effector T cells and inhibition of their proliferation166.

Other mechanisms of cell-contact Treg-DC crosstalk is through Lag3 expressed on

Tregs, which can bind MHCII on immature DCs and inhibit their maturation into inflammation-promoting APCs178.

1.2.4 Treg cytotoxicity

Treg mediated cytotoxicity is the final known mechanism of Treg inhibition of effector cells. There is evidence that human and murine Tregs can mediate control of the immune response by directly killing activated Teff and NK cells. This evidence is relevant to the research described in Chapter 3, where we investigated whether CAR- modified Tregs would induce apoptosis in CAR antigen-expressing cells. Most data point to Treg cytolysis as being a granzyme/perforin-dependent mechanism of inducing apoptosis in inflammatory lymphocytes. CD3/CD46-stimulated human Tregs can induce perforin-dependent target cell apoptosis166. Furthermore, murine tumor-infiltrating Treg cells can be isolated and induce NK and CD8+ T cell lysis in a granzyme B and perforin- dependent manner179. Others have concluded that murine Tregs rely on granzyme A, rather than granzyme B, in order to induce perforin-dependent cytolysis of Teff cells.

They also found that granzyme A in Tregs was necessary to prevent GvHD in a murine model180. Bi-specific antibodies targeting CD3 and tumor antigen can redirect Tregs

through TCR engagement to destroy target-bearing tumor cells in a granzyme

B/perforin-dependent manner181. Finally, immunofluorescent staining microscopy data of human T cells with granzyme B and Foxp3 clearly show that Foxp3+ T cells express granzyme B. These researchers further showed that granzyme B leaking out of Treg granules during activation could cause self-inflicted apoptosis in the Tregs of transplant recipients undergoing rejection182. Though the exact molecules involved and their function are still under debate, there is clear evidence that Tregs have the cellular machinery to induce perforin/granzyme cytolysis in cells. Knowing the cytolytic potential of CAR-Tregs is important for designing safe clinical trials and therapies. The cytolytic potential of tissue-specific CAR-Tregs warrants further investigation, since tissue destruction by CAR-Tregs could be detrimental to the success of the adoptive cell therapy.

1.2.5 Treg lineage stability

The issue of Treg stability is relevant in the clinical setting of adoptive Treg transfer, when Tregs need to remain stable and suppressive while being expanded183. Treg instability could lead to their conversion into inflammatory Tconv cells184. Typically,

Tregs maintain their population by the proliferation of existing Tregs. Therefore, it is a necessity that during cell divisions their stability is ensured185. In arming Tregs to express a CAR for adoptive Treg cell therapy, it will become even more essential to maintain stable Tregs because conversion to Tconv cells would cause T cell mediated inflammation in the tissues requiring protection from excess inflammation. While many

reports tracing Foxp3+ T cells in mice have shown them to be exceptionally stable populations186, others have demonstrated Treg plasticity, with the loss of the Foxp3 and/or Treg function under certain inflammatory conditions184.

As previously discussed, Foxp3, a gene encoded on the X-chromosome, is critical for the generation of Treg cells. Ectopic expression of Foxp3 by retroviral transduction in

Tconv cells transfers some characteristics of Treg cells, including the upregulation of certain Treg suppressive genes and surface markers CD25 and CTLA4187. However, other Treg signature genes are not induced in Tconv cells expressing a Foxp3 transgene, illustrating that Foxp3 stability alone is not sufficient to maintain Treg function188. Furthermore, human Tconv cells transiently upregulate Foxp3 during activation without acquiring suppressive characteristics.

Epigenetic changes such as histone and DNA methylation are key contributors to the mediation of cell differentiation and lineage stabilization. Genome-wide analysis has identified several regions of differential DNA methylation between Treg and Tconv cells.

For example, gene loci FOXP3, CTLA4, and IKZF2 were identified as hypomethylated and highly accessible for transcription in Treg cells183. CNS2, also termed the Treg- specific demethylated region (TSDR), is specifically demethylated in thymus-derived

Tregs and is needed for their sustained expression of Foxp3 during division.

Maintenance of TSDR hypomethylation is the gold standard of tTreg stability.

The role of costimulation in Treg stability is still being investigated. In mice, CD28 stimulation is required for both thymic development and peripheral homeostasis of Treg cells. Mice with Tregs lacking CD28 have reduced Treg numbers and are more susceptible to autoimmune diseases23. CD137 agonist antibodies given to mice resulted in the conversion of Tregs into inflammatory cytokine secreting cells without losing

Foxp3 expression189, suggesting that costimulation may play a role in the conversion of

Tregs to Tconv cells. This result also indicated that Foxp3 stability alone does not confer Treg suppressive function. An aim of the research described in Chapter 3 is to determine whether engineered costimulation in Tregs can affect their stability.

1.2.6 Adoptive Treg cell therapy

Adoptive Treg cell therapy holds great promise for the treatment of autoimmune and inflammatory diseases, as well as for preventing GvHD after hematopoietic stem cell transplantation and preventing organ transplant rejection190,191. To date, these diseases are usually treated with strong immunosuppressive drugs that have severe, even lethal side effects such as opportunistic viral and bacterial infections192,193. Adoptive Treg therapy involves the isolation, ex-vivo expansion, and subsequent administration of

Tregs cells. Research has demonstrated that the presence of high numbers of suppressive Tregs correlates with tolerance to donor organs and prevention of

GvHD194,195. The therapeutic potential of adoptively transferred Tregs was demonstrated in preclinical murine models196-198, paving the way for human clinical trials using the

adoptive transfer of polyclonal Tregs for GvHD, organ transplants, and certain autoimmune diseases156.

Isolating a pure population of Foxp3+ Treg cells is a critical step in the development of successful adoptive Treg therapy. Generally, tTregs are enriched by fluorescence- activated cell sorting or magnetic bead-based selection190,199. Human Tregs can be enriched from bulk CD4+ T cells by their high expression of CD25. The Treg population can be further purified if sorted on cells that lack the IL-7 receptor (CD127)200. Finally, some Treg therapies are using CD45RA+ Tregs, which are thought to be naïve and have superior stability to bulk Tregs201. Bulk CD4+ T cells can also be induced to form iTregs using TCR activation in the presence of TGFb and IL-2 cytokines202. Inducing

Tregs from the more abundant CD4+ T cell pool is an attractive idea, however naturally derived Tregs are considered more stable and are therefore favored to date203.

Collecting Tregs by sorting CD25++ CD127- CD4+ or CD25+ CD127- CD45RA+ from a patient apheresis product does not isolate sufficient numbers of Tregs to have a therapeutic effect in most cases. Using anti-CD3/anti-CD28-coated expander beads in media supplemented with high IL-2 cytokine, ex vivo expansion is a critical step to produce the number of Tregs needed for adoptive transfer191.

In improving Treg therapy further, mouse models and human xenogeneic models have established that antigen-specific Tregs are more potent for restoring peripheral

tolerance than polyclonal Tregs, reducing the number of T cells needed to see clinical benefit204,205. Lower numbers of tissue-specific Tregs could reduce the risk of globally suppressing other host immune responses206. Recent research has investigated genetic engineering as a way to generate antigen-specific Tregs by using either transgenic

TCRs or CARs. The following subsections will briefly describe where these two lines of research stand to date.

1.2.6.1 TCR transgenic Tregs

The overexpression of known TCR a and b chains in Tregs can be used as an approach to engineer TCR specificity in Tregs, enabling their activation when the Treg binds the appropriate peptide-MHCII complex. This technology has been used in a murine experimental autoimmune encephalomyelitis (EAE) model to show that Tregs expressing TCRs specific to a myelin basic protein peptide bound to MHCII are more efficient than polyclonal Tregs at suppressing central nervous system inflammation207.

Human Tregs expressing transgenic TCRs cloned from replacement factor VIII-reactive

Teff cells can suppress anti-factor VIII T cell and B cell responses, suggesting a possible treatment for hemophilia A patients who have developed anti-FVIII inhibitory antibodies208.

1.2.6.2 CAR-Tregs

In Chapter 3, we investigate the role of costimulation in CAR-Tregs. CARs have an advantage over TCR-directed antigen specificity in that CARs are not MHC-restricted.

Therefore, a single CAR can be used for all patients of a disease group rather than being restricted to only those patients with a certain HLA type.

The first CAR-Treg was described by Elinav et al. in 2008, in which the researchers showed that murine TNP-specific CAR-Tregs traffic to the colon of mice and suppress

TNBS-mediated colitis. This model also highlighted the concept of bystander suppression by CAR-Tregs, whereby TNP-specific CAR-Tregs suppressed T cell damage in oxazolone-inflamed intestines if a small amount of TNBS was present to activate the CAR T cells209. With bystander activation, the choice of target antigens for

Tregs becomes very broad because the auto-antigen that causes Teff cell responses does not need to be the same antigen to which the CAR-Treg is targeted. This concept opens the door to treating localized inflammatory and autoimmune diseases where the exact antigen being targeted by Teff cells is unknown. After these first seminal findings, other researchers demonstrated that murine CAR T cells can mediate tissue-specific suppression of inflammation in various proof of principle studies, from EAE to allergic airway inflammation models210-212.

More recently, research has investigated the feasibility and activity of human CAR-

Tregs. MacDonald et al. produced one of the first papers on human CAR-Tregs using an HLA-A2 CAR to prevent xenogeneic GvHD. HLA-A2 CAR-Tregs were also used to prevent human skin xenograft rejection4,213. Replacement factor VIII-specific human

CAR-Tregs, described by Yoon et al., not only prevented T cell responses against factor

VIII replacement protein but were also able to suppress B cell responses in vitro, similar to what these same researchers found with FVIII-specific TCR transgenic Tregs214.

CAR-activated Tregs upregulate LAP, GARP, CTLA4, and CD394, but it is not known which of these molecules are relevant in CAR-Treg-mediated suppression. Second- generation CARs with CD28 costimulation were used in all of the murine and human

CAR-Treg papers described in this section. Very little research has explored alternative costimulatory domains in CAR-Tregs. In Chapter 3 we hypothesized that costimulatory domains in CAR modified Tregs would differentially affect the stability and suppressive function of CAR-Tregs.

Chapter 2: Transcriptional Atlas of First- and Second-Generation CAR T Cells

Reveals Distinct Effects of Intracellular Domains

Parts of this chapter are currently under revision for publication as:

Angela C. Boroughs, Nemanja D. Marjanovic, Ana P. Castano, Selena J. Lorrey,

Livnat Jerby, Robert Morris, Lauren S. Riley, Orit Rozenblatt-Rosen, Aviv Regev,

Marcela V. Maus. Transcriptional Atlas of First- and Second-Generation CAR T Cells

Reveals Distinct Effects of Intracellular Domains

2.1 Introduction

T cells engineered to express chimeric antigen receptors (CARs) targeting CD19 have produced impressive outcomes for the treatment of B cell malignancies114,125,215,216.

CARs have been developed by independent groups and have incorporated different costimulatory domains, such as CD28 or 4-1BB90,91,106. Both types of CAR constructs have demonstrated efficacy in B cell leukemias108,109 and lymphomas110,111, but the engraftment kinetics, persistence, and toxicity profiles are distinctive. Clinical and in vivo studies have found CD28-based CARs undergo a more rapid expansion with less persistence than 4-1BB-based CARs, which result in a progressive T cell accumulation with persistence up to 6 years46,112. Since CD28 and 4-1BB costimulatory molecules signal though different pathways, the output of these pathways can lead to distinct transcriptional effects which could impact CAR T cell functions16,26. Recently, in vitro studies have demonstrated that CD28z CAR T cells have enhanced glycolytic metabolism and induce an effector memory phenotype, whereas 4-1BBz CAR T cells

rely more heavily on fatty acid metabolism and result in a greater number of central memory T cells7. Despite the tremendous momentum of CAR T cell development, an unbiased and in-depth understanding of the functional state of different types of CAR- modified T cells is still limited.

Transcriptional profiles provide a unique opportunity to characterize cells comprehensively, as they highlight multiple facets of cell state and function217. We hypothesized that the sum of the gene expression changes affected by the different costimulatory domains in CARs drives the functional consequences of CAR T cell products. In this study, we performed RNA sequencing of first-generation and second- generation human CAR T cells with 4-1BB or CD28 costimulatory domains. We collected an atlas of more than 240 profiles, spanning all combinations of our four CAR constructs, expressed in either CD4+ or CD8+ human T cells. Data was gathered from resting cells and from cells stimulated through either their CAR or again through their endogenous TCR both at 4 and 24 h after stimulation.

Recent research has highlighted an effect from ligand-independent signaling on CAR T cell functional state. Therefore, we chose to perform differential gene expression and co-expression analysis on CAR T cells at rest and revealed a signature of CAR- mediated CD3z chain signaling that was independent of antigen stimulation. We also describe a high-resolution view of the transcriptional differences between 4-1BB- and

CD28- containing second-generation CAR T cells at rest and directly following antigen engagement, which mimics encounters with tumors in patients.

2.2 Results

Generating a transcriptional atlas of CAR T cells

To generate a transcriptional atlas of CAR T cells, we synthesized four different CAR constructs bearing different signaling domains (Fig 2.1A): a first-generation CAR (z) that contained only a CD3z signaling domain, two second-generation CARs, one with a

CD28 (28z) and the other with a 4-1BB (BBz) costimulatory domain, and a control CAR construct that contained a truncated, non-signaling CD3z chain (Dz). All CARs had the same single chain variable fragment (scFv) against CD19 with identical CD8 hinge and transmembrane domains. Fluorescent reporter gene mCherry was included downstream of the CAR construct to facilitate evaluation of T cell transduction. Lentiviral vectors were used to transduce T cells to express the CAR. We also synthesized CARs with an anti-EGFR scFv based on cetuximab (Fig 2.2). All of the components besides the EGFR scFv were kept the same as the CD19 CAR designs. These EGFR CARs were used to validate our key findings from the CD19 CAR T cells’ transcriptomic analysis and therefore determine the generalizability of our results to CARs targeting different antigens.

We isolated CD4+ and CD8+ primary T cells from human donor peripheral blood, and then we mixed these two populations in equal proportions prior to activation with anti-

CD3/anti-CD28 beads. The following day, bulk T cells were transduced to express one of four CD19-specific CAR constructs (z, 28z, BBz and Dz) or left untransduced (UT).

CAR T cells were expanded for one week and rested for another week. We stimulated the CAR T cells for either 4 or 24 h through their CAR using irradiated CD19+ Nalm6 leukemia cells or through their TCR using anti-CD3/anti-CD28 beads. Finally, we sorted

CAR+ CD4+ and CAR+ CD8+ T cells separately and performed RNA sequencing (Fig

2.1, B-C).

Principal Component Analysis (PCA) combined with linear modeling to account for donor variation showed that samples were grouped primarily according to the type of antigen stimulation, except for Dz CARs stimulated with CD19-expressing targets, which clustered with unstimulated cells (Fig 2.1D). These data confirm that Dz CARs are an appropriate negative control for CAR-mediated signaling. Indeed, the first principal component (PC), accounting for 45% of the variance in the data, reflected the type of antigen stimulation: unstimulated, stimulated through the CAR, or stimulated through the TCR (Fig 2.1D). We further noted a striking difference in the transcriptional changes when human CAR T cells were activated through their TCR compared to their CAR, consistent with data from mouse CAR T cells which showed that stimulation through the

CAR versus TCR induces distinct gene expression signatures218. Indeed, the bead- activated CAR T cells clustered further away from the unstimulated cells than the CD19- activated CAR T cells. Furthermore, there were more than 5,000 differentially expressed

(DE) genes in 28z CAR T cells after 24-h TCR activation compared to 24-h CAR

Figure 1 activation. The most highly upregulated of these DE genes were enriched in the gene ontology (GO) signatures for cytokine signaling, positive regulation of cell

A T2A ∆ζ LTR L Anti-CD19 ScFv TM ∆ζ mCherry LTR T2A ζ LTR L Anti-CD19 ScFv TM CD3ζ mCherry LTR T2A 28ζ LTR L Anti-CD19 ScFv TM CD28 CD3ζ mCherry LTR T2A BBζ LTR L Anti-CD19 ScFv TM 4-1BB CD3ζ mCherry LTR

B Sort CD4/CD8, Mix 1:1 Remove Sort for Bead expansion beads RNA-Seq

Donor CAR Leukopak transduction Stimulate

Day Day Day Day Day Day 0 1 2 7 14 15

24-h Bead CAR Condition C D ∆ζ 24-h beads UT ∆ζ ζ 28ζ 24-h CART stimulation 105 BBζ 4-h beads 0 86.4 96.0 ζ 4-h CART stimulation 104 UT No stimulation 20 103 102 101 10 24-h CAR 10–3 0 103 104 105 BBζ 28ζ 5 10 82.6 95.0 0 104

3 10 PC2: 12% variance 102 −10 No Stim 101 4-h Bead 4-h CAR

–3 3 4 5 –3 3 4 5

CD3 10 0 10 10 10 10 0 10 10 10 −20 mCherry −40 −20 0 20 40 PC1: 55% variance

Figure 2.1 Generation of CAR T cells. (A) Vector maps of CD19 CAR constructs. TM, hinge and transmembrane domain. L, leader sequence. (B) Experimental design, T cells were isolated from a leukopak and sorted on CD3+, CD8+ or CD4+, then mixed at 1:1 CD4-to-CD8 ratio, activated with anti- CD3/anti-CD28 beads and transduced with one of four constructs or left UT. Cells were expanded for 7 days then beads were removed and cells were rested for 7 days prior to reactivation with anti-CD3/anti- CD28 beads or irradiated Nalm6 cells for 4 or 24 hours. Samples were then sorted and sequenced. Data were collected for T cells from three human donors with technical duplicates. (C) Representative transduction efficiency of CAR constructs determined by mCherry expression and CD3 surface expression on day 13. (D) Principal component analysis of gene expression of all samples from three donors corrected for donor-specific variation.

Extended Data Fig 1

T2A ∆ζ LTR L Anti-EGFR ScFv TM ∆ζ mCherry LTR T2A ζ LTR L Anti-EGFR ScFv TM CD3ζ mCherry LTR T2A 28ζ LTR L Anti-EGFR ScFv TM CD28 CD3ζ mCherry LTR T2A BBζ LTR L Anti-EGFR ScFv TM 4-1BB CD3ζ mCherry LTR

Figure 2.2 EGFR CAR constructs. Vector maps of EGFR CAR constructs. TM, hinge and transmembrane domain. L, leader sequence.

proliferation, and TCR activation (FDR = 1.26e-18, 1.42e-15,3.29e-10, respectively; Gene

Set Enrichment Analysis (GSEA)). These findings suggest that in this experimental setup, TCR activation with anti-CD3/anti-CD28 beads results in greater signal amplification and activation of more distinct signaling pathways than CD19 CAR activation using irradiated Nalm6 cells.

Gene signature of ligand-independent signaling

Differential gene expression analysis of resting CAR T cells revealed a signature of

CAR-mediated CD3z chain signaling that was independent of antigen stimulation. We interrogated the differences among the resting CAR T cells prior to activation (time 0), and found that several DE genes were significantly induced in all functional CARs compared with Dz (Fig 2.3A). However, there were no DE genes between Dz and UT

CAR T cells. This observation indicates that the expression of a CAR bearing a CD3z chain resulted in a specific transcriptional signature (Fig 2.3B). This ligand-independent

signature from CD3z was enriched for genes involved in the response to cytokine stimulus (FDR = 1.51e-6; Gene Set Enrichment Analysis (GSEA)). The genes in the ligand-independent signature included CCL3 and CCL4, which are involved in monocyte recruitment, and GZMB (encoding granzyme B), a key cytotoxicity gene. We further validated a selection of the up- and downregulated genes from our signature by digital droplet PCR in EGFR-directed CAR T cells (Fig 2.4, A-G). Thus, the expression of a

CAR in a T cell can affect the T cell transcriptome and therefore potentially its phenotype prior to CAR-antigen engagement on a tumor cell. Furthermore, these changes are not dependent on the CD19 scFv.

Evidence of ligand-independent signaling by costimulatory domains

We also found that there were differentially expressed genes between BBz and 28z resting CAR T cells (Fig 2.5 and Appendix: Supplemental Table A2.1), reflecting differences in ligand-independent downstream signaling pathways between 4-1BB and

CD28 costimulatory domains. Therefore, not only the CD3z signaling domain but also the costimulatory domains can affect the CAR T cell transcriptional profile in the absence of antigen. Recently, in vitro studies have demonstrated that several days after

CAR activation, 28z CAR T cells have enhanced glycolytic metabolism, whereas BBz

CAR T cells rely more on fatty acid metabolism7. Consistent with these results, we found fatty acid oxidation genes were enriched in BBz versus 28z CAR T cells (Fig 2.6).

However, our results were found in CAR T cells at rest and 4 h post-antigen stimulation, illustrating that ligand-independent signaling from the costimulatory domain plays a role

in this process. Indeed, metabolic changes can be established in CAR T cells even prior to antigen stimulation.

Figure 2.3 Gene signature of ligand-independent signaling from the CD3z domain. (A) Overlap of differentially expressed genes upregulated in 28z, BBz and z CARs all compared to Dz at time 0. Circle areas represent the number of DE genes to scale in CD4+ or CD8+ T cells. We defined the overlapping genes DE in all functional CARs vs. Dz as the signature of CD3z chain ligand-independent signaling. (B) Heat map of the normalized expression of both up and downregulated genes (rows) across all samples at rest, 14 days after CAR transduction.

Extended Data Fig 2

A BIRC3 B GGT1 C GZMB D CTLA4 ** * * * 5 NS 1.5 NS 30 NS 0.6 NS * NS ** NS 4 1.0 20 0.4 3

2 0.5 10 0.2 1 Relative expression 0 0.0 0 0.0 BBζ ζ ∆ζ UT BBζ ζ ∆ζ UT BBζ ζ ∆ζ UT BBζ ζ ∆ζ UT

E SDC4 F SOCS2 G TNFRSF10A NS 1.0 NS 0.8 * NS 1.0 ** NS NS * NS 0.8 0.8 0.6

0.6 0.6 0.4 0.4 0.4

0.2 0.2 0.2 Relative expression 0.0 0.0 0.0 BBζ ζ ∆ζ UT BBζ ζ ∆ζ UT BBζ ζ ∆ζ UT

Figure 2.4 Ligand-independent signature genes in EGFR CAR T cells. EGFR CARs were sorted on CD8+ mCherry+ after 7 days of bead expansion and 7 days of rest. Digital droplet PCR of genes upregulated (A-E) and down-regulated (F, G) in our signature for tonic signaling from CD3z was performed. N=4 normal donors, mean and SEM plotted. Significance was determined with a paired ratio t- test comparing to Dz CAR T cells. Genes are expressed relative to an internal reference gene TBP. *p<0.05 **p<0.01

Extended Data Fig 3

Time 0 h CD8 Time 0 h CD4 8 15 CXCL10 JUN FDR < 0.05 GPA33 Not significant

JUNB 6 KRT1

10 EGR1 CIITA

-value) NTRK2 4 p

( MSC

10 VNN2

5 −log 2

0 0

−1 0 1 2 −1 0 1

Time 4 h CD8 Time 4 h CD4

JUNB C17orf61−PLSCR3 CXCL10 ENPP2 ENOX1 6 ENPP2 10 DDIT4 NTRK2 FILIP1L HLA−DQA2 -value) 4 UBD GJB2 p CIITA ( IL4I1 P2RY14

10 ENOX1 HLA−DOA 5 COL6A1 HLA−DRA BTN2A2 NTRK2

−log HLA−DRB5 HLA−DRB1 2 HLA−DPB1 DMD HLA−DMB HLA−DQB2

0 0

−1.0 −0.5 0.0 0.5 1.0 −1 0 1 log2 fold change log2 fold change

Figure 2.5. DE genes between CAR BBz and 28z. Volcano plot of log fold-change on the x-axis and –log10(pvalue) on the y-axis between CD19 BBz and 28z genes at 0 and 4 h post-Nalm6 activation in CD4+ and CD8+ cells. Genes with FDR<0.05 are plotted in red. Positive x-axis is upregulated in BBz vs. 28z.

Extended Data Fig 5

Time 0 h Time 4 h

FDR q-value = 0.053 FDR q-value < 0.001

CD4

Gene Rank Gene Rank BBζ 28ζ BBζ 28ζ

FDR q-value = 0.030 FDR q-value = 0.0075

CD8

Gene Rank BBζ Gene Rank 28ζ BBζ 28ζ

Figure 2.6 BBz CAR T cells have increased fatty acid metabolism before activation. GSEA of hallmark fatty acid metabolism genes in rank fold-change list of DE genes between CD19 BBz and 28z CAR T cells at 0 hours and 4 h post-CAR activation with irradiated Nalm6 cells.

Differentially expressed genes between CAR activated BBz and 28z CAR T cells

Here, we describe the first transcriptional dataset of BBz and 28z CAR T cells following both CAR and TCR-mediated activation. In both CD4+ and CD8+ T cells, there were far more DE genes with increased expression in BBz compared to 28z CAR T cells (Fig

2.5, Fig 2.7A, and Appendix: Supplemental Table A2.1). These upregulated genes spanned multiple cytokine and immune signaling pathways (Fig 2.7B). In addition, we found that many of the upregulated genes in BBz CAR T cells were HLA class II genes or genes encoding proteins involved in the regulation of HLA class II gene expression

(Fig 2.7C), which we validated at the protein level by flow cytometry of HLA-DR expression in both anti-CD19 (Fig 2.7D) and anti-EGFR (Fig 2.8) CAR T cells. HLA genes were induced in BBz versus 28z CAR T cells at all time points including at rest, indicating that ligand-independent signaling of the 4-1BB costimulatory domain may cause the upregulation of HLA Class II expression. ENPP2, or autotaxin, was the most significantly DE gene and was increased in all BBz samples both at rest and more so following stimulation (Fig 2.9A). ENPP2 encodes a phosphodiesterase, which hydrolyzes lysophospholipids to produce lysophosphatidic acid and is involved in chemotaxis and proliferation219. Finally, there was increased expression of genes involved in TNF and IFNg signaling in BBz CAR T cells versus 28z CAR T cells 24 h post stimulation (Table 2.1).

Table 2.1: Selection of enriched gene sets in MSigDB with computed overlaps of DE genes upregulated in BBz compared to 28z CARs in CD4+ and CD8+ samples combined. ‘k’ refers to the number of genes in the intersection of the query set with a set from MSigDB. ‘K’ refers to the number of genes in the set from MSigDB. ‘FDR q-value’ refers to the false discovery rate using Benjamini-Hochberg procedure to correct for multiple testing.

Table 2.1: Selected Overlap Results

# Genes in

Gene Set Name Overlap k/K p-value FDR q-value

GO_RESPONSE_TO_CYTOKINE 22 0.0308 1.01E-21 1.69E-17

GO_INTERFERON_GAMMA_MEDIATED_SIGNALING_PATHWAY 11 0.1571 3.74E-19 1.25E-15

GO_IMMUNE_SYSTEM_PROCESS 27 0.0136 1.37E-17 3.83E-14

GO_POSITIVE_REGULATION_OF_CELL_CELL_ADHESION 13 0.0535 3.54E-16 5.32E-13

GO_MHC_CLASS_II_PROTEIN_COMPLEX 7 0.4375 3.82E-16 5.32E-13

GO_REGULATION_OF_CELL_ACTIVATION 15 0.031 5.10E-15 5.34E-12

GO_LUMENAL_SIDE_OF_MEMBRANE 7 0.2121 1.39E-13 9.71E-11

KEGG_ALLOGRAFT_REJECTION 7 0.1842 4.08E-13 2.63E-10

GO_CLATHRIN_COATED_VESICLE_MEMBRANE 8 0.0988 1.54E-12 8.31E-10

GO_TRANS_GOLGI_NETWORK_MEMBRANE 8 0.0988 1.54E-12 8.31E-10

REACTOME_PHOSPHORYLATION_OF_CD3_AND_TCR_ZETA_CHAINS 5 0.3125 5.68E-11 1.53E-08

GO_REGULATION_OF_CELL_PROLIFERATION 18 0.012 6.38E-11 1.70E-08

HALLMARK_TNFA_SIGNALING_VIA_NFKB 9 0.045 7.39E-11 1.77E-08

Antigen stimulation through the CAR induced expression of distinct cytokines and

cytokine receptors in BBz CAR T cells compared to 28z CAR T cells. Specifically, IL21

in CD4+ T cells and IL21R and IL23R for both CD4+ and CD8+ cells were all upregulated

in BBz CARs versus 28z CARs (Fig 2.7E, 2.9B-C). In addition, BBz CAR T cell

activation caused a more sustained activation of IL12RB2 whereas as 4 hours, 28z

upregulated IL12RB2 to a similar, if not higher degree, than BBz CAR T cells.

Figure 3

A FDR < 0.05 Not significant

CD4 GJB2 10.0 CD8 ENPP2 UBD

7.5 NTRK2 SH3BP5 THY1 7.5 HLA−DRB1 HLA−DRA HLA−DQA2 GJB2 JUNB IER2 G0S2 C4orf26 CDK6 CIITA CXCL10 MX1 ADA DOHH JUNB

-value) ACSL1 NTRK2 5.0 DMD MSC ELL2 p LGMN HLA−DRB5 HLA−DOA DGAT2 TNFRSF8

( ANXA3 HLA−DRB6 FILIP1L 5.0 PDCD1 ANK3 ITPR1 IL4I1 10 EMC8 MPZL1 IFNG NOD2 HLA−DRB6 TMEM165 –log 2.5 2.5

0.0 0.0 28ζ −1 0 1 BBζ 28ζ −0.5 0.0 0.5 1.0 BBζ log2 fold change log2 fold change

B CD4 CD8 C CD4 (n = 56) (n = 17) 28ζ BBζ TF / DNA binding CIITA HLA II CD74 Chemokine HLA-DMB Cytokine / cytokine receptor HLA-DPB1 TPM Other HLA II HLA-DQA2 (row normalized) Metabolism HLA-DRB1 genes 1 Proliferation, survival apoptosis HLA-DRB5 Structural protein HLA-DOA 0 Adhesion molecule HLA-DRA –1 Other immune receptor / pathways HLA-DRB6

D E IL21R expression F 40 10,000 * ** * 28ζ 40 * BBζ CD4 ∆ζ CD8 30 UT 30 Donor 1 2 1000 3 20 20 CAR

28ζ IL21 (pg/ml) MFI of HLA-DR 10 BBζ 10 Transcripts per Million Transcripts

100 0 No stim No stim Nalm6 Nalm6 0 5 10 15 20 25 28ζ BBζ ζ ∆ζ UT CD4 CD8 stim stim Time(hours) CD4 CD8 Figure 2.7 BBz and 28z DE genes. (A) Volcano plots of differentially expressed genes between BBz (positive x axis) and 28z CARs 24 h post-Nalm6 activation in CD4+ (left) and CD8+ (right) T cells. Genes with FDR<0.05 are colored red. (B) Classification of significantly differentially expressed genes with FDR <0.1 at 24 hours between BBz and 28z CAR T cells using GO annotation. (C) Heat map showing normalized HLA II gene expression of CD4+ T cells from all three donors in 28z vs. BBz CAR T cells 24 h post-Nalm6 stimulation. (D) MFI of HLA-DR surface expression by flow cytometry with no stimulation and 24 h after Nalm6 stimulation. N=3 normal donors, mean and SEM plotted. (E) IL21R expression in BBz (blue) and 28z (red) CAR T cells with Nalm6 stimulation. (F) IL-21 cytokine levels 24 h after bulk CAR T cells were stimulated with Nalm6. N=3 normal donors, mean and SEM plotted. P-values were determined using a paired Student’s t-test between BBz and 28z. *p<0.05, **p<0.01

Extended Data Fig 4

A B 20,000 ** 25,000 **

20,000 15,000

15,000 10,000 10,000

MFI of HLA-DR 5000 MFI of HLA-DR 5000

0 0 28ζ BBζ ∆ζ 28ζ BBζ ∆ζ

Figure 2.8 Increased HLA-DR surface protein on activated EGFR BBz CAR T cells. Mean fluorescence intensity (MFI) of HLA-DR surface expression by flow cytometry of EGFR CAR T cells after 20 h of stimulation with irradiated U87 cells gated on CD3+, mCherry+ and (A) CD4+ or (B) CD8+ cells. N=3 normal donors, mean and SEM plotted. *p<0.05, **p<0.01.

We confirmed that antigen stimulation of BBz CAR T cells produced more soluble IL-21 cytokine than 28z CARs in both anti-CD19 and anti-EGFR CAR T cells (Fig 2.7F, 2.10A, respectively).

We noted greater expression of the adhesion genes ICAM1 and VCAM1 in BBz CAR T cells (Fig 2.11). This expression is interesting since 4-1BB/4-1BB ligand interactions are known to induce the expression of adhesion molecules on endothelial cells and 4-1BB activation in T cells increases their adhesion to fibronectin40,220. We frequently noted increased cell aggregations in BBz CAR T cells in culture compared to 28z CAR T cells.

Costimulatory domains affect early T helper cell polarization in CD4+ CAR T cells

IL12RB2 is known to be upregulated in TH1 polarizing cells. Therefore, we hypothesized that there would be an increase in TH1 polarization in BBz CAR T cells. We performed

GSEA analysis to investigate whether there was an enrichment of known early

221 polarizing TH1 genes that were upregulated in BBz versus 28z CAR T cells . Indeed,

+ TH1 polarizing genes were enriched in CD4 BBz CAR T cells before and following antigen stimulation (Fig 2.12, A-B). CD4+ BBz CAR T cells had increased expression of

TH1 transcription factor genes EGR1 and TBX21 and an increase in the TH17 transcription factor gene RORC (Fig 2.12C). Conversely, 24 h after CAR activation, the significantly upregulated DE genes in 28z compared to BBz CAR T cells (FDR<0.1)

222 were enriched for known TH2 early polarizing genes . We experimentally confirmed a significant increase in TH2 cytokine production, IL-4, and IL-5, by 28z CARs with scFvs against both CD19 and EGFR (Fig 2.12, E and 2.10, B-C, respectively). Thus, the type of costimulatory domain plays a role in the initiation of T helper subset polarization.

28z CAR T cells upregulate PD1 and anti-apoptotic genes following activation

Notably, the most upregulated gene in 28z versus BBz CARs was PDCD1 (Fig 2.13A), which encodes the inhibitory receptor PD1. We validated this observation using flow cytometry to confirm that surface expression of PD1 was higher in 28z CARs compared to other CD19 CARs (Fig 2.13B). Finally, 28z CAR T cells had increased expression of anti-apoptotic genes and decreased pro-apoptotic genes after CAR activation (Fig 2.14,

A-C), which may contribute to their increased ability to proliferate after initial antigen stimulation. We believe that the transcriptional differences we found add to a greater understanding of how antigen stimulation drives CAR T cells to different cell fates.

A ENPP2 expression CD4 CD8 30 Donor 1 2 3 20 CAR 28ζ BBζ

10 Transcripts / kilobase million Transcripts

0.0

B IL12RB2 expression

CD4 CD8 60 Donor 1 2 3 40 CAR 28ζ BBζ

20 Transcripts / kilobase million Transcripts

C IL23R expression

CD4 CD8

7.5 Donor 1 2 3

5.0 CAR 28ζ BBζ

2.5 Transcripts / kilobase million Transcripts

0.0 0 5 10 15 20 25 Time (hours)

Figure 2.9 BBz upregulates ENPP2, IL12RB2 and IL23. Normalized gene expression in CD4+ and CD8+ in BBz and 28z CD19 CAR T cells with irradiated Nalm6 stimulation over time. (A) ENPP2- autotaxin, (B) IL12RB2- IL-12 receptor subunit and (C) IL23- IL-23 receptor. N=3 normal donors plotted individually.

Extended Data Fig 8

A B

800 ** 80 *

600 60

400 40 IL-4 (pg/ml) IL-21 (pg/ml) 200 20

0 0 28ζ BBζ ∆ζ UT 28ζ BBζ ∆ζ UT C 500 *

400

300

200 IL-5 (pg/ml) 100

0 28ζ BBζ ∆ζ UT

Figure 2.10 Cytokine secretion by stimulated EGFR CAR T cells. Bulk CD4+/CD8+ EGFR CAR T cells were stimulated for 24 h with irradiated U87. (A) Soluble IL-21, (B) IL-4 and (C) IL-5 were measured. N=3 normal donors, mean and SEM plotted. * p<0.05 **p<0.01

Extended Data Fig 9

A ICAM1 expression 80 CD4 CD8 Donor 60 1 2 3 CAR 40 28ζ BBζ

20 Transcripts / kilobase million Transcripts

B VCAM1 expression 10.0 CD4 CD8 Donor 7.5 1 2 3 CAR 5.0 28ζ BBζ

2.5 Transcripts / kilobase million Transcripts

0.0

0 5 10 15 20 25 Time (hours)

Figure 2.11 BBz CARs T cells express increased adhesion molecules. Gene expression of (A) ICAM1 – Intercellular Adhesion Molecule 1– and (B) VCAM1 – Vascular Cell Adhesion Molecule 1– in CD19 CAR T cells with Nalm6 stimulation over time. N=3 normal donors

Figure 4

A B 0.7 0.6 NES = 1.96 0.6 0.5 NES = 1.96 NES = 2.09 FDR q < 0.001 0.5 0.4 FDR q < 0.001 0.4 FDR q < 0.001 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 Gene Rank –0.1 Gene Rank –0.1 BBζ Enrichment score (ES) BBζ 28ζ 28ζ Enrichment score (ES)

C D CD4 300 * 28ζ BBζ STAT4 STAT1 200 TBX21 IL12RB2 TPM TH1 GLUL (row normalized) ENPP2 100 1

DMD IL-4 (pg/ml) P2RY14 0 IFNG –1 0 TH17 RORC 28ζ BBζ ζ ∆ζ UTD

Figure 2.12 4-1BB costimulation initiates early TH1 polarization program. GSEA of early polarizing TH1 signature genes in rank fold-change list of DE genes between BBz and 28z CAR T cells at (A) 0 h and (B) 24 h post-CAR activation with irradiated Nalm6 cells. (C) Heat map of known TH1 helper cell polarizing genes in CD4+ T cells from all three donors in 28z vs. BBz CARs 24 h post-Nalm6 stimulation. and (D) IL-4 levels after 24 h Nalm6 stimulation of bulk CD4+ and CD8+ CAR T cells. N=3 normal donors, mean and SEM plotted. * p<0.05 **p<0.01

Extended Data Fig 10

A PDCD1 expression

CD8 Donor 20 1 2 3 CAR 28ζ 15 BBζ

10 Transcripts / kilobase million Transcripts

0 5 10 15 20 25 Time (hours)

B 2000 *

1500

1000 MFI-PDI

500

0 28ζ BBζ UT

Figure 2.13 CD28 costimulation increases PD1 expression in CAR T cells. (A) Gene expression of PDCD1 (encoding the PD1 protein) in CD8+ CD19 CAR T cells with Nalm6 stimulation over time. (B) PD1 surface expression across 3 normal donors 24 hours after Nalm6 activation. N= 3 normal donors, mean and SEM plotted. *p<0.05 paired t-test.

Extended Data Fig 11

A BAD expression

CD4 CD8 70 Donor 1 60 2 3 CAR 50 28ζ BBζ

Transcripts / kilobase million Transcripts 30

B BCL2A1 expression 250 CD4 CD8

200 Donor 1 2 3 150 CAR 28ζ 100 BBζ

50 Transcripts / kilobase million Transcripts

C BCL2L11 expression CD4 CD8 Donor 1 40 2 3 CAR 28ζ BBζ 20 Transcripts / kilobase million Transcripts

0 0 5 10 15 20 25 Time (hours)

Figure 2.14 28z CARs express increased anti-apoptotic genes and decreased pro-apoptotic genes immediately after activation. Gene expression of (A) BCL2A1 – BCL2 Related Protein A1– (B) BCL2L11- BCL2 Like 11– and (C) BAD – BCL2 Associated Agonist Of Cell Death– in CD19 CAR T cells with Nalm6 stimulation over time. N=3 normal donors.

2.3 Discussion

CAR T cells using either CD28 or 4-1BB intracellular signaling domains can be very effective cancer immunotherapies. Indeed, CD19-directed CAR T cells with either of these costimulatory domains have been approved for the treatment of B cell lymphoma and leukemia. Nevertheless, several investigators have noted different proliferation kinetics, toxicity and persistence profiles between second-generation CAR T cells with differing costimulation. Therefore, we generated a transcriptional atlas of first- and second-generation human CD19-specific CAR T cells after signaling through their CAR for 0, 4, or 24 h to investigate whether transcriptional differences in response to different

CAR construct signaling may underlie the distinct phenotypes observed.

Until now, the effects of driving constitutive CAR expression in resting human T cells have not been defined at the transcriptional level. Here, we identified a transcriptional signature present in functional resting CAR-modified T cells, indicating the presence of ligand-independent transcriptional activity from the CD3z chain in CAR T cells. We believe that these transcriptional differences are subtler than the ligand-independent effects that have been described in c-Met and GD2-directed CAR T cells88,117, since those CAR cells had ligand-independent cytokine secretion and proliferation, whereas our CAR T cells do not. Furthermore, we did not identify any enrichment for T cell exhaustion signatures in our CAR T cells at all time points. Though the tonic signaling phenomenon may be partly dependent on the CAR construct scFv, transmembrane domain, and signaling domain88, we found that EGFR CARs replicated our tonic 62

signature gene expression findings in all tested cases. This finding suggests that the signature may be present in T cells transduced with CARs containing different scFvs and could be used as a measure of the degree of tonic signaling from a particular CAR construct in future studies.

Though not well understood, ligand-independent signaling from the constitutive expression of a costimulatory domain (CD28/4-1BB) could also impact CAR T cell phenotype, cell fate, and persistence. It is unknown whether ligand-independent signaling plays a role in the previously discovered differential CAR T cell phenotypes between CD28- and 4-1BB-containing CAR T cells, such as long-term persistence in

4-1BB-based CAR T cells110,111.

Here we focused on established CAR constructs that closely resemble those in clinical use. However, there were some deviations from the commercially used axicabtagene ciloleucel to our 28z since the FDA-approved CD28-based CAR T cell products use retroviral vectors to transduce T cells. Aside from the type of vector used, the only other substantial difference from products in clinical use to ours is that our 28z CAR contains a CD8 transmembrane domain rather than the CD28 transmembrane domain. We chose to keep all components of our CAR T cells identical aside from the costimulation domains, in order to specifically interrogate the differences attributable to changes in signaling domains. Our BBz construct design is essentially identical to the tisagenlecleucel product.

The results of the differential expression analysis between BBz CARs and 28z

CARs found that most of the significant genes were upregulated in BBz CARs compared to 28z CARs. The increased response in BBz CAR T cells could indicate that

CD28 has a reduced role in gene upregulation in response to activation. Alternatively, it is possible that because all the T cells had previously received CD28 stimulation as part of the transduction process with anti-CD3/anti-CD28 beads, additional signaling through this domain has less of a differential effect in the T cells. Another possibility is that the kinetics of CAR activation is different between the BBz and 28z constructs and that the peak of genes upregulated after CD28 signaling would be seen at a different time point.

Our data indicate many differentially expressed cytokines and cytokine receptors in BBz compared to 28z CARs after 4 and 24 h of CD19-CAR signaling. For example, we found that IL12RB2, which encodes a subunit of the IL12 receptor, was significantly upregulated in BBz CARs at 24 hours. IL-12-secreting CARs with CD28 costimulatory domains are being developed by various groups, but these data suggest that BBz CARs may be more sensitive to the additional IL-12 223. Furthermore, IL21 and IL21R were upregulated in BBz CAR T cells. IL-21 secretion from CD4+ cells is known to support the formation of memory CD8+ T cells, which could be important in the production of a lasting anti-tumor CD8+ immune response and might partially explain the increased persistence of BBz CARs224.

We also found a significant enrichment in TH1 polarizing genes in BBz CAR T cells and TH2 polarizing genes in 28z CAR T cells. The fact that the cytokine profiles differ based on the costimulatory domain used in a CAR may provide a basis for refining the treatment of cytokine release syndrome based on the CAR product causing the

+ + syndrome. Furthermore, TH1 CD4 T cells are known to be important for CD8 T cell activity147,223. From this data, we concluded that the type of costimulatory domain plays an early role in the polarization of T helper subsets, which could affect the quality of help provided to CD8+ T cells as well as their direct anti-tumor potential.

Beyond CAR therapy, these data have potential to improve our understanding of the role of costimulation in T cells. For example, we discovered the brisk upregulation of

MHC Class II on T cells with 4-1BB signaling to much higher levels than typical MHC

Class II expression following TCR-mediated activation. This finding could be exploited by combining CAR T cells with tumor vaccines to promote epitope spreading via antigen presentation to CD4 T cells. The increased expression of PD1 on the surface of 28z

CAR T cells may indicate enhanced sensitivity of this type of CAR construct to checkpoint antibodies targeting the PD1/PD-L1 axis. Together, these data expand our understanding of how novel antigen receptor expression affects the gene expression profile, functional state, and ultimately, the fate of engineered human T cells.

In conclusion, our results reveal a signature for ligand-independent signaling from the

CD3z chain that is present in both first and second-generation CAR T cells. Our data also define several differences between BBz and 28z CAR T cells, with most of the

gene expression differences in pathways related to cytokine and cytokine receptor expression and T cell polarization. These results indicate that ligand-independent and ligand-dependent signaling from CARs bearing different costimulatory domains can have a profound effect on both CD4+ and CD8+ CAR T cells. This knowledge should enhance CAR therapies by providing greater insight into the selection of costimulatory domains for specific cancers.

2.4 Materials and Methods

Generation of CAR constructs and CAR T cells

CD19 and EGFR-specific CARs were synthesized and cloned into a third-generation lentiviral plasmid backbone under the regulation of a human EF-1α promoter (GenScript

USA Inc). Replication-defective lentiviral vectors were produced by four plasmids co- transfected into HEK293T cells using TransIT-2020 transfection reagent (Mirus).

Supernatants were collected 24 h and 48 h after transfection and filtered. Virus was concentrated by ultracentrifugation. Vector was harvested and stored at -80 °C. Healthy donor leukopaks were obtained from the Blood Transfusion Services at Massachusetts

General Hospital under an IRB-exempt protocol. CD4+ and CD8+ T cells were negatively selected using RosetteSep Kits with a Ficoll gradient (Stemcell

Technologies). Enriched T cells were stained in PBS with 2% FBS for 30 min at room temperature with CD3-BV605, CD4-FITC, CD8-APC-Cy7 (clones OKT3, OKT4 and

SK1-BD Pharmingen). Cells were resuspended in HBSS supplemented with 25 mM

HEPES and 1% FBS and stained with 4′,6-diamidino-2-phenylindole (DAPI) prior to sorting. Cells were purified by CD3+CD4+ or CD3+CD8+ using fluorescence-activated cell sorting (FACS) with a BD FACSAria III (BD Biosciences). CD4+ and CD8+ T cells were mixed at a 1:1 ratio prior to expansion.

Target cell lines

The human embryonic kidney cell line 293 (HEK392T) and Nalm6 cell lines were purchased from American Tissue Culture Collection (ATCC). Both cell lines were

expanded in RPMI-1640 with 1x L-GlutaMAX and 25mM HEPES (Gibco, Life

Technologies) and supplemented with 10% heat-inactivated fetal bovine serum (FBS,

Gibco Life Technologies). U87 cell line was from ATCC and expanded in EMEM with

10% FBS. U87 cells were passaged with 0.05% trypsin (Gibco, Life Technologies).

Target cells were irradiated with 10 000 rads and frozen in FBS with 10% DMSO to be thawed prior to stimulation of CAR T cells.

T cell in vitro expansion and transduction for RNA sequencing

T cells were plated in a 24-well plate at 1 million cells/ml in RPMI-1640 with 1x L-

GlutaMAX and 25 mM HEPES (Gibco, Life Technologies) supplemented with 10% FBS and IL-2 (20 IU/ml, Peprotech). Anti-CD3/CD28 beads (Dynabeads, Invitrogen) were added with a 3:1 bead-to-cell ratio. T cells were cultured for one day and then transduced with one of the four lentiviral constructs at an MOI of 5. Cells were counted and maintained at 5e5 cells/ml with IL-2 replaced every 2 days during bead expansion and resting period. Beads were removed on day 7 using a magnet and cells were rested for a further 7 days. T cells were checked for mCherry expression and purity by flow cytometric analysis on day 13. On day 14, each CAR T cell population was divided into

5 wells at 5e5 CAR T cells/ml in a 6-well plate to be either left unstimulated or stimulated with Nalm6 or anti-CD3/CD28 beads (1:1 T cells-to-target/bead ratio) for 4 or 24 h prior to staining and sorting. Cells were stained with CD3-FITC (UCHT1-BioLegend), CD4-

BV786 (SK3, BD Biosciences), CD8-APC-H7 (SK1, BD Pharmingen), and CD69-APC

(FN50, BioLedgend). DAPI was added before FACS. 5,000 cells were sorted in

technical duplicates on CD3+, CD4+, or CD8+ and mCherry+ with a MoFlo Astrios EQ cell sorter (Beckman Coulter) and resuspended in lysis buffer to be bulk sequenced at the Broad Institute.

Flow cytometry

All antibodies were purchased from BioLegend unless otherwise stated. The following antibodies were used: HLA DR-PacBlue (clone L243), CD4-FITC (clone OKT4), CD8-

APC-H7 (clone SK1, BD Pharmingen), PD1-BV711 (clone EH122H7). Cells were stained for 30 min in the dark at 4 °C and washed twice in PBS with 2% FBS. We added

DAPI for cell viability directly before running flow. We collected events on a Fortessa x-

20 (Becton Dickinson) and analyzed the data with FlowJo software (Tree Star).

RNA sequencing

CAR T cells were collected using a FACS machine, resuspended at 200 cell/μl in lysis buffer composed of Buffer TCL (QIAGEN 1031576) plus 1% 2-mercaptoethanol (Sigma

63689) and immediately frozen at -80 °C.

For preparation of RNAseq libraries, cells were thawed (1000 cells per sample) and purified with 2.2x RNAClean SPRI beads (Beckman Coulter Genomics) without a final elution 225. The RNA captured beads were air-dried and processed immediately for cDNA synthesis. SMART-Seq2 protocol was carried out as previously described226 with minor modifications in the reverse transcription step (M.S.K. and A.R., in

preparation). Each PCR was performed in a 25 μl reaction with 15 cycles for cDNA amplification. We used 0.25 ng cDNA of each sample and one-fourth of the standard

Illumina NexteraXT reaction volume in both the fragmentation and final PCR amplification steps. Up to 30 libraries were pooled per 1 nextseq run (~ 500 million reads), and sequenced 50 × 25 paired-end reads using a single kit on the NextSeq500

5 instrument.

Initial Read alignment and QC

BAM files were converted to merged, demultiplexed FASTQ files. Reads were mapped to the UCSC hg19 human transcriptome using Bowtie 227, and transcript-per-million

(TPM) values were calculated with RSEM v1.2.8 in paired-end mode228.

Samples passed QC if the number of aligned reads greater than 1e7, if the percent of reads mapped above 30%, and if the percent rRNA was in the range of 10-30%. We calculated the correlation between technical duplicates and further interrogated any duplicates with an R2 of less than 0.9. Duplicates were then averaged for downstream analysis.

PCA and Differential Gene Expression

Differentially expressed genes were identified using the DEseq2 R package after correcting for the effect of different patient donors. Statistical p-values were corrected for multiple hypotheses testing using the Benjamini & Hochberg (1995) method. Genes

with an FDR (q-value) less than 0.05 were considered significant unless otherwise stated. PCA plots were constructed by DESeq using linear model batch corrected data from LIMMA R package229. The contribution of each covariate on the principle components was calculated using the SWAMP R package (Martin Lauss 2017). Row normalized (complete linkage, average linkage, ward method) heat maps were constructed using the gene expression (TPM) data generated by RSEM. Genes were classified using their Gene ontology annotation 230,231 and the gene cards database

(www.genecards.org).

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was performed at each time point against a gene

221 list of early polarizing TH1 genes with the Deseq2 generated DE gene lists that were ranked by log2-fold changes. Analysis was performed using an R version of the GSEA code provided by the Broad Institute. The identification of gene sets that were enriched in significantly DE genes (FDR<0.05) between any two CARs was determined by GSEA software232 run by the Broad institute. This computed the overlaps between our gene set and gene sets in MSigDB233.

Cytokine detection of stimulated T cells

For cytokine release assays, T cells were stimulated in a 96-well plate with 100,000 Teff cells/well combined with irradiated target cells at a CAR T cell-to-target ratio of 1:1 for

Nalm6 targets and 2:1 for U87 targets. Supernatants were harvested after 24 h and

frozen at -80 °C. Supernatants were analyzed for cytokine levels using FLEXMAP 3D® platform from Lumina Instrumentation (Thermo Fisher Scientific) according to manufacturer’s instructions with a panel of the following cytokines: IL-1β, IL-2, IL-4, IL-5,

IL-6, IL12p70, IL-13, IL-18, IFN-γ, GM-CSF, TNF-α, IL-10, and IL-21. Plates were read using xPONENT Software 4.1. All samples were measured in technical triplicates and with N=3 normal donors. Triplicates measured were averaged before graphing with

Prism (Graphpad software).

Digital Droplet PCR

EGFR CAR T cells were transduced, expanded and rested for 7 days. We collected 5e5 cells by FACs and resuspended cells in 350 μl RLT buffer with 1% 2-mercaptoethanol.

RNA was extracted and purified using RNAeasy kit (Qiagen) and cDNA was generated from 270 ng of RNA/20μl reaction using iScript Reverse Transcription supermix (Bio-

Rad). Digital Droplet PCR was performed using ddPCR supermix with no dUTPs (Bio-

Rad) with a QX200 Droplet Digital PCR (ddPCR™) System (Bio-Rad) platform for quantification. Droplet generation, PCR, and detection of positive droplets were performed according to manufacturer’s instructions (Instruction Manual, QX200™

Droplet Generator, Bio-Rad).

The cycling protocol was performed according the manufacturer’s instructions with a

57° C melting temperature. Human TBP was used as the reference gene in each reaction (HEX fluorophore: TBP PrimePCR™ ddPCR™ Expression Probe Assay:

Unique Assay ID: dHsaCPE5058363 (Bio-Rad)). The following FAM fluorophore

primer probes were used (IDT PrimeTime Std® qPCR Assay).

Gene Primer/Probe

CTLA4 PrimeTime Primer 1: CGG ACC TCA GTG GCT TTG

PrimeTime Primer 2: TTC ATC CCT GTC TTC TGC AA

PrimeTime Probe: /56-FAM/CG CCA GCT T/Zen/T GTG TGT GAG TAT GC/3IABkFQ

GZMB PrimeTime Primer 1: CAG AGA CTT CTG ATC CCA GAT

PrimeTime Primer 2: TCC TGA GAA GAT GCA ACC AAT

PrimeTime Probe: /56-FAM/CC CGC CCC T/Zen/A CAT GGC TTA TCT /3IABkFQ/

SOCS2 PrimeTime Primer 1: GAT ATT GTT AGT AGG TAG TCT GAA TGC

PrimeTime Primer 2: GGA GCT CGG TCA GAC AG

PrimeTime Probe: /56-FAM/AA AGA GGC A/Zen/C CAG AAG GAA CTT TCT TGA /3IABkFQ/

SDC4 PrimeTime Primer 1: GGT ACA TGA GCA GTA GGA TCA G PrimeTime Primer 2: GCA GCA ACA TCT TTG AGA GAA C

PrimeTime Probe: /56-FAM/CC ACG ATG C/Zen/C ACC CAC AAT CAG A/3IABkFQ/

GGT1 PrimeTime Primer 1: TTC AGG TCC TCA GCT GTC A

PrimeTime Primer 2: TGG CTG ACA CCT ACG AGA C

PrimeTime Probe: /56-FAM/CC GCC TGG A/Zen/T GTC CTT CAC AAT CT/3IABkFQ/ TNFRSF1 0A PrimeTime Primer 1: GTC CAT TGC CTG ATT CTT TGT G

PrimeTime Primer 2: GTC AGT GCA AAC CAG GAA CT

PrimeTime Probe: /56-FAM/AT TCT GCT G/Zen/A GAT GTG CCG GAA GT/3IABkFQ/ BIRC3 PrimeTime Primer 1: GTA GAT GAG GGT AAC TGG CTT G

PrimeTime Primer 2: GGT GTT GGG AAT CTG GAG ATG

PrimeTime Probe: /56-FAM/CC TTG GAA A/Zen/C CAC TTG GCA TGT TGA /3IABkFQ/

Statistical Analysis

Unless specifically indicated, comparison between different groups was conducted with two-tailed, paired Student’s t-tests. Unless otherwise stated, P values below 0.05 were considered significant. Statistical analysis was performed with Prism (Graphpad software).

2.5 Acknowledgements

I would like to acknowledge Ana P. Castano (Maus Lab, Massachusetts General

Hospital) and Marcela V. Maus (Massachusetts General Hospital) who designed the

CD19 z, BBz and 28z CAR constructs. Ana P. Castano assisted with the remaining

CAR constructs described. I would also like to thank Marcela V. Maus, Aviv Regev

(Broad Institute), Orit Rozenblatt-Rosen (Regev Lab, Broad Institute), and Nemanja D.

Marjanovic (Regev Lab, Broad Institute) assisted in outlining the RNA sequencing portion of this study. Nemanja D. Marjanovic prepared the RNA libraries. Livnat Jerby

(Regev Lab, Broad Institute) ran the alignment and gene quantification pipeline. Selena

J. Lorrey (Maus Lab, Massachusetts General Hospital) and Lauren S. Riley (Maus Lab,

Massachusetts General Hospital) assisted with lentiviral vector production and in vitro analysis. Robert Morris (Cancer Center, Massachusetts General Hospital) provided bioinformatics analysis guidance. I would like to acknowledge the Klarman Cell

Observatory for sequencing support and the Broad Institute flow cytometry core for their assistance sorting cells. Marcela V. Maus was closely involved with all parts of this study.

Chapter 3: Human Regulatory T Cells Modified with CD28 but not

4-1BB-Based CARs Maintain Suppressive Function.

Angela C. Boroughs, Rebecca C. Larson, Bryan Choi, Amanda Bouffard, Lauren S.

Riley, Erik Schiferle, Anupriya S. Kulkarni, Curtis L. Cetrulo, David Ting, Shadmehr

Demehri, Bruce R. Blazar, Marcela V. Maus.

3.1 Introduction

Tregs are lymphocytes that function to suppress excessive immune responses, regulate tolerance to self-antigens, and maintain tissue integrity136. Adoptive transfer of Tregs in preclinical mouse models has demonstrated therapeutic potential in solid organ transplantation197, GvHD 196, and a range of autoimmune diseases198. Pioneering clinical trials using infusions of polyclonal Tregs cells are in development, but have shown modest efficacy in these diseases, with the most advanced clinical data occurring in the setting of GvHD156. Current hypotheses for these limitations have included the lack of antigen specificity to enable trafficking to the target organ, insufficient numbers of Tregs to achieve adequate suppression, and poor persistence of the infused Tregs204-206,234. These limitations are essentially the same ones that applied to the use of conventional adoptive immunotherapy in cancer, prior to the advent of genetically-modified T cells. Genetic re-direction of Tconv cells with CARs has enabled polyclonal peripheral blood T cells to be redirected to a specific tumor-associated antigen2,3. The addition of a costimulation domain (e.g. CD28 or 4-1BB) in second- generation CAR T cells further improved CAR-Tconv cell persistence and anti-tumor

efficacy. The CD28 and 4-1BB costimulation domains are known to influence Tconv cell phenotype, and they differ in kinetics, persistence, and toxicity profile in patients as well7,117.

Transduction of Tregs with CARs may enable the production of large numbers of antigen-specific Tregs with enhanced proliferative capacity and persistence. However, there has been no systematic investigation of which costimulation domain best maintains the phenotype and function of CAR-Tregs while increasing their proliferation, survival, and persistence. Others have used CD28-based CARs in Tregs and have demonstrated antigen-specific immunosuppressive activity with CAR-Tregs directed to

Factor VIII and HLA-A2 in pre-clinical xenogeneic models of anti-VIII responses to

Factor VIII replacement for hemophilia and GvHD, respectively4,213,214. In Tregs, CD28 costimulation is known to be important for their activation22; in a pre-clinical model,

CD28-based CAR-Treg (HLA-A2 - 28z) persisted less than 3 weeks in NSG mice4. Data on the role of 4-1BB activation in T regulatory cells is more conflicting, with some results indicating 4-1BB activation can improve Treg expansion236, whilst others have found

4-1BB signaling inhibits Treg suppression235,236.

CARs are powerful synthetic constructs that may alter the baseline activity of Tregs as a result of constitutive expression of CD3z or costimulation under a strong promoter119, or may modify antigen-specific functions due to the high binding affinity of the extracellular domains which are typically based on antibodies. CAR transduction of conventional

CD4+ T cells (generally thought of as “helper” T cells with little cytotoxic potential) has been shown to confer them with cytolytic function against antigen expressing targets218,237, which would be undesirable for a CAR-Treg product that is meant to protect a target tissue. We hypothesized that transduction of Tregs with CARs bearing different intracellular domains, either CD3z alone or in tandem with CD28 or 4-1BB, would result in differential Treg phenotypic stability, persistence, and immunosuppressive function.

3.2 Results

Isolation of Tregs and transduction with CARs bearing different signaling domains.

Healthy donor leukopaks were enriched for a CD25+ population by magnetic bead positive selection. Tregs were further purified by FACS sorting based on expression of the surface markers CD4mid CD25++ CD127low (Fig 3.1A). In parallel, we also sorted

CD4+ CD25low Tconv cells to use as controls from the same donor in each experiment

(Fig 3.1A). The sorted human Tregs were confirmed to express intracellular Foxp3 (Fig

3.1B and Fig 3.2A) and were de-methylated at the Treg specific demethylation region

(TSDR) of the X chromosome (Fig 3.1C). Sorted Tregs also expressed higher levels of the phenotypic markers that differentiate resting Tregs from Tconv, including CD39 and the latency associated peptide (LAP), which is part of the latent TGFb complex. Surface expression of CTLA4 and LAG3 was below detection level, as expected for non-activated Tregs (Fig 3.2B).

We synthesized five different anti-CD19 CAR constructs in a lentiviral backbone

(Fig 3.1D): a control CAR construct that contained a truncated, non-signaling CD3z chain (Dz), a first-generation CAR (z) which contained CD3z as the only signaling domain, two second-generation CARs; one with a CD28 (28z) and the other with a

4-1BB (BBz) costimulation domain. Finally, we synthesized a 28z second-generation

CAR that expressed transgenic Foxp3 (28z-Foxp3) following a cleavable T2A element.

All CARs had the same scFv against CD19 and incorporated identical CD8 hinge and transmembrane domains. An mCherry fluorescent reporter gene was included downstream of the CAR construct after a T2A element to facilitate evaluation of CAR transduction efficiency. Immediately after sorting Tregs and Tconv cells, we activated them with anti-CD3/anti-CD28 beads, and 24 h later, we transduced T cells with lentiviral vector carrying the CAR constructs. CAR-Tregs were expanded with beads for

7 days followed by bead removal, then rested for another 7 days in media containing

300 IU/ml rhIL-2. CAR-Tregs were phenotyped and used in various functional assays ~

2 weeks from initial isolation (Fig 3.1E). Tconv and Treg cells showed similar transduction efficiencies with CAR vectors at the same multiplicity of infection (MOI) (Fig

3.1F). We confirmed that T cells modified with the CAR-28z-Foxp3 construct did indeed express transgenic Foxp3, even when transduced into Tconv cells (Fig 3.2C).

Figure 1 A B

Treg

Tconv CD8-APC-Cy7 CD25-PE

CD4-BV510 CD4-BV510 Count Foxp3

C Treg Gate 100

40 Tconv Gate 20

CD25-PE

CD25-PE 0 CD127-BV711 CD127-BV711 % mean methylation

reg conv T T D T2A ∆ζ LTR L Anti-CD19 ScFv TM ∆ζ mCherry LTR

T2A ζ LTR L Anti-CD19 ScFv TM CD3ζ mCherry LTR

T2A 28ζ LTR L Anti-CD19 ScFv TM CD28 CD3ζ mCherry LTR

T2A BBζ LTR L Anti-CD19 ScFv TM 4-1BB CD3ζ mCherry LTR T2A T2A 28ζ-Foxp3 LTR L Anti-CD19 ScFv TM CD28 CD3ζ FOXP3 mCherry LTR

E F 1.Stain and sort Tconv BBζ Treg BBζ 2.CD3/CD28 Remove 87.1 91.8 bead stim Beads

Day -1 0 1 7 14

Donor Transduce with Phenotyping/ Leukopak CARs Functional Assays Count mCherry

Figure 3.1 Generation of CAR-Tregs. CD4+ T cells were isolated from human donor PBMCs and enriched for CD25+ cells using positive selection. (A) Sorting gates for Tregs: CD4mid, CD25++ and CD127low and Tconv: CD4+, CD25low. Gates drawn on cells prior to enrichment (B) Foxp3 (clone PCH101) intracellular stain after sort. (C) Methylation status across the Treg Specific Demethylation Region (TSDR) on T cell populations after fluorescence-activated cell sorting from female-donor leukopaks. N=2 female donors, mean and SEM plotted. (D) Vector maps of CD19 CAR constructs. TM, hinge and transmembrane domain. L, leader sequence. (E) Experimental design. (F) Representative transduction efficiency of BBζ CAR constructs determined by mCherry expression.

Supplemental Figure 1

A B 80 *** 10000 Treg ) 60 Tconv

40 *** 1000 % of CD4+ 20 Foxp3 (Raw MFI

100 0 reg T conv T CD39+ LAP+ CTLA4+ LAG3+ C Tc 28ζ Tc 28ζ-Foxp3 Tr28ζ Tr 28ζ -Foxp3 Count Foxp3- APC

D E 100 D0

10000 D14 ) D0 80 CAR Stim

D14 60 CAR Stim 1000 TCR Stim 40

Foxp3 (Raw MFI 20 % mean methylation

100 0 ζ ζ ζ ζ ∆ζ ζ ζ ζ ζ c r r c ∆ζ r c UT T T T r 28 T r T T T r BB c UT T r 28 r BB T ζ Foxp3 T T T ζ Foxp3 r 28 r 28 T T F G H 100 100 CAR Stim * TCR Stim 80 100 80

80 60 9 60 60 %LAP 40 40

40 % CD3 20 20 20 % mean methylation 0 0 0

∆ζ ζ ζ ζ ζ ζ ζ ζ ζ UT c r c r r r T ∆ζ T T ∆ζ T T T c UT r r 28 r BB c UT r r 28 r BB T T T T T T T T Figure 3.2 Treg phenotype and Foxp3 stability. Flow cytometry of Tregs and Tconv cells stained after sorting and before activation, gated on CD4+ cells. (A) Intracellular staining of Foxp3 displayed as mean florescence intensity (MFI). (B) Flow cytometry staining of surface markers LAP, LAG3, CD39 and CTLA4. N>3 human donors, mean and SEM plotted. (C) Tconv cells and Tregs were transduced to express 28z or 28z-Foxp3 at an MOI of 5. Two weeks after transduction, T cells were measured for Foxp3 expression by intracellular staining and flow cytometry. Representative donor. (D) Foxp3 intracellular staining by MFI of T cells gated on live, CD3+, CD4+ cells after sorting (D0), after bead expansion and rest (D14) and on 9 days after stimulation with irradiated CD19-K562 stimulation (CAR Stim) or irradiated anti-CD3 K562 (TCR Stim). N=6 human donors. (E) Methylation status of the IKZF2 (Helios) promotor and (F) TSDR from sorted mCherry+ cells using direct bisulfite modification and pyrosequencing. N=3 human donors. Blue bars for D0 represent UT Tregs and UT Tconv directly after sort. Surface expression of T cells gated on of CD3+ CD4+ CAR+ stimulated with K562 cells for 9 days followed by G) LAP and H) CD39 staining measured by flow cytometry. N=3 normal donors. Mean and SEM plotted. *pval<0.05, paired t-test between CAR stim Tr 28z and CAR stim Tr BBz. UT – Untransduced, Tc – Tconv cells, Tr – Treg 80

CAR-Tregs display Foxp3 stability in culture irrespective of their CAR signaling domains.

CAR-modified Tregs were analyzed for the expression of Foxp3 and methylation of the

Treg-specific demethylation region (TSDR), CTLA-4 promoter, and Helios promoter. We expected that stable Tregs would maintain high expression of Foxp3, and remain demethylated at the TSDR and Helios and CTLA4 promoter loci. We chose to analyze clinically relevant time points where CAR-Tregs would presumably be harvested/infused

(day 14), and following antigen encounter, either through their TCR or their CAR (day

23). The antigen encounter stimulation was performed by a 9-day co-culture of

CAR-Tregs with irradiated K562 cells transduced to express either membrane-bound anti-CD3 scFv (OKT3) to stimulate the TCR or CD19 to stimulate the CAR, respectively.

We chose these ‘resting’ time points rather than immediately following bead activation or antigen encounter because many Treg-associated markers, including both CD25 and

Foxp3, are expressed on activated human Tconv cells238. Intracellular Foxp3 staining demonstrated that CAR-Tregs remain Foxp3+ irrespective of the CAR costimulation domain, and expression was maintained at day 14 and following antigen stimulation through their CAR or TCR at day 23 (Fig 3.2D and Fig 3.3A). Similarly, demethylation of the TSDR locus remained stable after isolation (day 0), initial transduction (day 14) and following antigen stimulation through the CAR (day 23) (Fig 3.3B). We also found that the mean methylation of CTLA4 (Fig 3.3C) and IKZF2 (Fig 3.2E), encoding the Helios transcription factor, was lower in all CAR-Tregs compared to CAR-Tconv cells at day 0.

The methylation status at these three loci remained stable through transduction 81

Figure 2 A B Foxp3 TSDR 100 D0 100 D14 80 CAR Stim 80 TCR Stim 60 60

40 40

% Foxp3+ CD25+ 20

20 % mean methylation

0 0

ζ ζ ζ ζ ζ ζ ζ ζ c ∆ζ r c ∆ζ r T r T T r T r 28 r BB c UT T r 28 r BB c UT T T T T T T T ζ Foxp3 ζ Foxp3 r 28 r 28 T T C D CTLA4 Surface CTLA4 100 100

* 80 80 + 60 60

40 40 % CTLA4 20

% mean methylation 20

0 0

ζ ζ ζ ζ ζ ζ ζ ζ c ∆ζ r c ∆ζ r T r T r 28 T r T r 28 c UT T T r BB c UT T T r BB T T ζ Foxp3 T T r 28 T

Figure 3.3 Foxp3 expression is stable after transduction, bead expansion, and re-stimulation. (A) Intracellular staining of Foxp3+CD25+ cells gated on CD3+, CD4+, live cells after sorting (D0), bead expansion and rest (D14) and on day 23, 9 days after the addition of irradiated K562-OKT3 (TCR Stim) or K562-CD19 (CAR Stim). (B) TSDR methylation day 0 post sort (D0), day 14 and day 23 (9 days after irradiated K562-CD19 stimulation). N=2 female donors. (C) Methylation status of CTLA4 promotor using direct bisulfite modification and pyrosequencing. N=3 human donors. (D) Surface CTLA4 expression 9 days after TCR or CAR stimulation with irradiated K562 cells. N=3 human donors. Mean and SEM plotted. Blue bars for D0 represent UT Tregs and UT Tconv right after sort. Tc – Tconv cells, Tr – Treg

(day 14) and re-stimulation (day 23) and was independent of the CAR construct. UT

Tregs behaved identically to Dz CAR Tregs, including by their methylation of the TSDR

(Fig 3.2F); for clarity, we chose to display only Dz CAR Tregs throughout this study.

We analyzed CAR-Tregs for surface expression of classic Treg functional markers

CTLA4, LAP and CD39 following stimulation through their CAR or TCR. We found that 82

antigen stimulation through CD28-based CARs induced significantly higher CTLA4 expression than signaling through the TCR in CAR-Tregs (Fig 3.3D), as had been previously described in HLA-A2-directed CD28-based CAR Tregs4. There were no statistically significant differences in CTLA4 expression when signaling through first-generation CARs or BBz second-generation CARs compared to TCRs in Tregs (Fig

3.3D). We also found that 28z CAR-Tregs had higher LAP expression than BBz

CAR-Tregs when stimulated through either their CAR or TCR (Fig 3.2G). CD39 expression was less variable across different kinds of CAR-Tregs (Fig 3.2H), and there was no significant difference between stimulation through the CAR and TCR. From these results, we concluded that the expression of a CAR and the type of costimulation domain does not affect Foxp3 stability or the methylation status of CTLA4 and IKZF2 promotors. However, transduction and activation with CARs can impact the expression of Treg phenotypic surface markers CTLA4 and LAP, with 28z increasing the expression of both markers compared to BBz and Dz in ‘rested’ CAR-Tregs.

CAR-Tregs can be activated through their CAR and through their TCR.

Tregs need to be activated to become suppressive163,239. Similar to Tconv cells, functional Tregs upregulate CD69 when activated. However, unlike Tconv cells, Tregs do not secrete inflammatory cytokines but instead secrete suppressive cytokines such as IL-10240. Tregs also upregulate specific functional markers such as LAP following activation. LAP is a protein that associates with TGFb, a pleotropic cytokine that can inhibit Teff cell proliferation and IL-2 secretion241. LAP expression on Tregs correlates

with TGFb secretion242. We measured the expression of activation markers in Tregs transduced with various CARs after stimulation. CAR-Tregs were activated through their

CAR or TCR for 20 h in vitro, using irradiated K562 cells as above. Like CAR-Tconv, all

CAR-Tregs upregulated CD69 in response to both CAR and TCR activation, except UT

Tconv and Dz CAR-Tregs which as expected, were not activated by K562-CD19 (Fig

3.4A). In contrast to CD69 expression, only Tregs and not Tconv cells expressed high amounts of LAP after CAR or TCR activation (Fig 3.4B). Interestingly, CAR-Tregs expressed higher amounts of LAP at rest compared to control Tregs (Dz), suggesting that there is a mild constitutive CAR signaling effect in the absence of antigen stimulation. Expression of 4-1BB is also an activation marker in Tconv and in Treg cells243, and 4-1BB was upregulated after stimulation through Treg CARs or TCRs.

Notably, BBz CAR-Tregs expressed higher levels of surface 4-1BB both at baseline and after activation (Fig 3.5, A-B).

We analyzed the cytokine profiles in the supernatants of CAR-Tregs activated for 24 h with K562-CD19 or K562-OKT3. CAR-Tconv cells produced high amounts of inflammatory cytokines (IL-2, TNFa, IFNg) in response to CAR activation, whereas

Tregs produced minimal, if any, inflammatory cytokines (Fig 3.4, C-E). Cytokines were not detected in resting Tconv or Treg supernatants (data not shown). We noted, however that CAR stimulation in 28z and BBz CAR-Tregs induced low but significant

Figure 3 A B 100 100 No Stim

80 80 CAR Stim TCR Stim + 60 + 60

40 % LAP 40 % CD69

20 20

0 0 ζ ζ ζ ζ ζ ζ ζ ζ ζ ζ c ∆ r r T r T r BB c ∆ T c UT T r 28 T c UT T r r 28 r BB T T T T T T D C 1.5 1.5

1.0 1.0 (normalized)

0.5 F 0.5 ** ** IL-2 (normalized) TN

0.0 0.0 ζ ζ ζ ζ ζ ζ ζ ζ ζ ζ r c ∆ r c ∆ T T r T r BB c UT T r r 28 r BB c UT T r 28 T T T T T T T E F 1.5 10 **

d) 8 d) 1.0 6

(normalize 4 γ 0.5 N IL-10 (normalize

IF ** ** 2

0.0 0 ζ ζ ζ ζ ζ ζ ζ ζ ζ ζ c ∆ r r T r T c ∆ T c UT T r 28 r BB c UT T r r 28 r BB T T T T T T T G H 8 8 CAR Stim TCR Stim s

s Tr ∆ζ NS 6 6 Tr ζ Tr 28ζ 4 4 ** Tr BBζ 2 2 Population Doubling Population Doubling 0 0 0 2 4 6 8 0 2 4 6 8 Days post stim Days post stim

Figure 3.4 CAR-Tregs can be activated through their CAR or their TCR. Cells were left unstimulated (No Stim) or stimulated with irradiated K562-CD19 (CAR stim) or K562-OKT3 (TCR stim) at a 1:1 ratio for 20 h followed by surface staining of (A) CD69 and (B) LAP measured as the % positive after gating on live, mCherry+, CD4+ T cells except in the case of the UT groups which were only gated on live, CD4+ cells. Supernatants were saved 20 h after T cell - K562 co-culture to measure (C) IL-2, (D) TNFa, (E) IFNg and (F) IL-10. Cytokine levels were normalized to Tc UT stimulated with OKT3-K562. N>3 human donors, mean and SEM plotted. ** pval<0.01, paired t-test between CAR vs. TCR stim for C,D and between BBz and 28z Treg groups in F. T cells were stimulated on day 14 after bead expansion and one week of rest (Time-point 0) at a 1:1 ratio with irradiated K562s expressing (G) CD19 or (H) anti-CD3. Live cell numbers were counted every two days and expressed as the log2 fold-change from the starting cell number. For G and H, significance calculated by a paired t-test at day 8 between Tr Dz and Tr 28z, mean and SEM plotted. **pval<0.01. Tc – Tconv cells, Tr – Treg, NS – not significant 85

Supplemental Figure 2 ** A ** 100 * No Stim 80 CAR Stim

+ TCR Stim 60

% 4-1BB 40

0 ζ ζ ζ ζ ∆ζ r c r T c UT T T r 28 r BB T T T 5 No Stim B 10 CAR Stim ** * TCR Stim NS I 104 4-1BB MF

103

ζ ζ ζ ζ c ∆ζ r T r T c UT T r 28 r BB T T T C 3 **

2 * ***

0 violet low (normalized)

-1

ζ ζ ζ ∆ζ r r T r 28 r BB T T T D 100 Teff ζ

80 Teff BBζ 60

40 % Proliferating 20

1:1 1:3 1:9 0:1 Treg:Teff

Figure 3.5 4-1BB costimulation affects aspects of CAR-Treg phenotype. Cells were left unstimulated (No Stim) or stimulated with irradiated K562-CD19 (CAR Stim) or K562-OKT3 (TCR Stim). 4-1BB (CD137) surface expression of live CD4+ CAR+ cells (A) as a % or (B) raw MFI. N=3 human donors, mean and SEM plotted. (C) Treg proliferation after violet cell trace-labeling and activation with irradiated K562-CD19 cells over 3 days. The number of mCherry+ proliferating (violet low) cells was normalized to the number of mCherry+ violet low cells in the unstimulated condition. N=3 human donors, mean and SEM plotted. pval *<0.05 pval **<0.01. paired t-test to Tr Dz . (D) 28z CAR-Tregs in MLR with CFSE labeled first- and second-generation CAR-Teff cells and irradiated Nalm6 target cells. Representative donor. Tc – Tconv cells, Tr – Treg 86

levels of IFNg and TNFa compared to the first-generation z CAR-Tregs or CAR-Tregs activated with K562-OKT3. We also found that CAR-Tregs produce higher amounts of

IL-10 in response to CAR stimulation than Tconv cells, and that 28z CAR-Tregs make significantly more IL-10 than BBz Tregs (Fig 3.4F). Encouragingly, CAR-Tregs also proliferated in response to either CAR or TCR stimulation, except for in the case of Dz

CAR-Tregs which, as expected did not proliferate in response to CAR stimulation (Fig

3.4, G-H and Fig 3.5C). Altogether, we conclude that CAR-modified regulatory T cells can be activated through their CAR with either CD28 or 4-1BB costimulation and proliferate while maintaining their Treg identity by their surface phenotype and cytokine profile.

4-1BB costimulation reduces CAR-dependent Treg suppressive function.

Next, we examined whether Tregs maintained their suppressive function toward Teff after being modified to express different kinds of CARs. (Note that to avoid confusion, all T cells to be suppressed by Tregs either in vitro or in vivo will be referred to as Teff cells while all T cells sorted as CD4+ CD25low with the purpose of being directly compared to Tregs will be referred to as Tconv cells). To measure CAR-activated Treg suppression specifically, we needed a system whereby Teff cells could be activated without using TCR stimulation. We decided to use Teff CD4+ cells transduced to express a first-generation anti-CD19 z CAR as the cells to be suppressed, and irradiated CD19+ Nalm6 cells as the antigen presenting cells, in mixed lymphocyte reactions (MLR). We specifically selected a first-generation CAR-Teff because CD19 z

CAR-Teff cells were easier to suppress than second-generation CAR-Teff cells (Fig

3.5D). MLRs were performed by titrating violet-labeled CAR-Tregs with constant numbers of CFSE-labeled Teff and Nalm6 target cells. Analysis of the CFSE dilution of

Teff cells confirmed that all functional CAR-Tregs could inhibit proliferation, except for the negative control Dz CAR-Tregs. However, 28z and z CAR-Tregs inhibited the proliferation of Teff cells to a greater degree than BBz CAR-Tregs (Fig 3.6A). We also measured inflammatory cytokines in supernatants of the MLRs, expecting that

CAR-Treg would inhibit the secretion of these cytokines by CAR-Teff. For these experiments, we used anti-CD19 CAR 28z Teff because they secrete the greatest amounts of cytokines. Again, we found that BBz CAR-Tregs were not as efficient as 28z and z CAR-Tregs at preventing the secretion of TNFa, GM-CSF, IL-2, and IFNg from

Teff cells (Fig 3.6, B-E, respectively). In summary, CAR-Tregs can suppress Teff cells after CAR antigen-specific activation, but the inclusion of a 4-1BB costimulatory domain in the CAR results in reduced CAR-mediated suppression.

Overexpression of CARs in T cells has been shown to result in ligand independent constitutive signaling68. To determine whether the decrease in suppressive capacity seen in the BBz CAR Tregs was caused by constitutive signaling through the 4-1BB costimulation domain or only by antigen-specific CAR signaling, we repeated the MLR experiments but with CFSE-labeled, autologous, resting Teff cells with anti-CD3/anti-CD28 beads for pan-T cell activation. Again, we found that CARz and

CAR 28z Tregs were able to suppress CARz Teff, whereas CAR Dz could not and CAR

Figure 4

A B 120 800 100

600 Tr ∆ζ 80 l Tr ζ 60 pg/m 400 Tr 28ζ

* F 40 Tr BBζ TN % Proliferating 200 20

0 0

1:1 1:3 1:9 0:1 1:1 1:3 1:9 0:1 Treg:Teff Treg:Teff

C D E

30 4000

l 1500 l l 3000 20 1000

2000 pg/m N IF 10 IL-2 pg/m 500 GM-CSF ng/m 1000

0 0 0 1:1 1:3 1:9 0:1 1:1 1:3 1:9 0:1 1:1 1:3 1:9 0:1 Treg:Teff Treg:Teff Treg:Teff

F G CAR Stim TCR Stim

120 120

100 100

g 80 80

60 60

40 40 % Proliferatin % Proliferating 20 20

0 0

3:1 1:1 1:3 1:9 3:1 1:1 1:3 1:9 Treg:Teff Treg:Teff

Figure 3.6 4-1BB costimulation decreases CAR-Treg suppressive function. (A) MLRs of CFSE labeled CD19 z CAR-Teff cells with different ratios of violet-labeled CAR-Tregs. After 3 days, CFSE dilution was measured on mCherry+ Teff cells to calculate % proliferation as the % of CFSElow in the condition of the % of CFSElow cells with no Tregs present. N=5 human donors. Mean and SEM plotted, pval *<0.05, paired t-test performed for 1:1 Treg-to-Teff ratio only. Supernatants were collected in the same MLR set up except with 28z CAR-Teff as the CFSE labeled cell. After 24 h cytokine levels of (B) TNFa, (C) GM-CSF, (D) IL-2 and (E) IFNg were measured. Representative donor, repeated with N=3 donors with technical triplicates. MLRs comparing Treg suppression of Teff cell proliferation after activation though (F) the CAR (CFSE labeled CD19 z CAR-Teff, irradiated Nalm6 targets, 1:2 Teff-to-target cell ratio) or (G) the TCR (CFSE labeled naïve T cells, anti-CD3/anti-CD28 beads, 10:1 cell-to-bead ratio). After 4 days, CFSE dilution was measured. Representative donor of N=3 with technical triplicates. Mean and SEM plotted. Repeated with N=3 human donors. Note: all CAR T cell groups were used at 50% transduction (mCherry+) for these assays. Tr – Tregs 89

BBz Tregs suppressed proliferation poorly (Fig 3.6F). In contrast, Tregs transduced with any CAR but stimulated through the TCR displayed equal ability to suppress Teff cells

(Fig 3.6G). We conclude that antigen specific signaling from CARs containing 4-1BB signaling domains inhibit Treg suppressive function.

We next sought to explore the mechanism of reduced immunosuppression in BBz

CAR-Tregs. Tregs can suppress Teff proliferation through a variety of mechanisms, including the secretion of inhibitory cytokines TGFb and IL-10 and the consumption of

IL-2165,172,173. Since activated BBz CAR-Tregs secrete less IL-10 than 28z CAR T cells, we wanted to investigate whether IL-10 was contributing to Teff cell suppression by

CAR-Tregs. Suppression assays were repeated with IL-10 blocking antibodies; however, we saw no effect of blocking IL-10 on Treg suppression (Fig 3.7A). Next, we wanted to investigate whether IL-2 consumption upon activation was different among various CAR-Tregs. We measured IL-2 in the media of CAR-Tconv, or CAR-Tregs after

40 h in culture in the absence or presence of antigen (irradiated CD19+ Nalm6 cells).

Without stimulation, BBz CAR-Tregs consumed more IL-2 than Tconv or other

CAR-Tregs (Fig 3.7B), suggesting that BBz Tregs are more metabolically active at baseline, which is also consistent with the increased CD69 expression that we found in rested BBz CAR-Tregs (Fig 3.4A). However, BBz CAR-Tregs did not significantly increase their consumption of IL-2 after antigen stimulation, which may partly, but not fully explain their reduced immunosuppressive capacity, given that the consumption of

IL-2 was similar to activated CD28z CAR-Tregs. In contrast, 28z CAR-Tregs consumed

less IL-2 than BBz CAR-Tregs at rest, but they consumed equal or more IL-2 after antigen stimulation (Fig 3.7B). Treg-mediated suppression has also been shown to require cell-cell contact in vitro244. We observed that BBz CAR-Tregs formed large cell aggregates, unlike any of the other CAR-Tregs (Fig 3.7, C-DSupplemental). Potentially, thisFigure 3

A B

100 1200 No Stim CAR Stim

1000 80 ) 800 * 60 ** 600

40 IL-2 (ng/ml

% Proliferating Tr 28ζ 400

20 Tr 28ζ +anti-IL-10 200

0 0 ζ ζ ζ ζ ζ ∆ζ r 3:1 1:1 1:3 1:9 0:1 c r T c UT T c BB T r 28 r BB Treg:Teff Media T T T T

C Tr BBζ Tr 28ζ

D Tr BBζ Tr 28ζ

Figure 3.7 BBz CAR-Tregs have increased basal IL-2 consumption and form aggregates in culture. (A) MLR with or without IL-10 blocking antibody at different ratios of CD19 28z CAR-Tregs to CD19 z CAR Teff cells with irradiated Nalm6 cells as targets. Representative donor with technical triplicates. (B) IL-2 consumption assay. IL-2 cytokine levels after Tregs and Tconv cells expressing different CARs were incubated with the same initial concentrations of IL-2 for 40 hours either with or without the addition of irradiated Nalm6 cells. Representative donor with technical triplicates, mean and SEM plotted. N=4 human donors. pval *<0.05 pval **<0.01. Significant pval were found for each donor tested. (C) Photograph of T cells in culture flasks. (D) Representative images of CAR Tregs in culture resting, taken with a 40x objective with bright light or an RFP detecting light cube. Tc – Tconv, Tr – Treg

aggregation of BBz CAR-Tregs leads to dysfunctional interactions between CAR-Tregs and Teff cells, thereby affecting BBz CAR-Treg suppression. In summary, at baseline

BBz CAR-Tregs consume more IL-2, express higher CD69, and aggregate with each other, but none of these findings alone seem to explain the mechanism by which 4-1BB

CARs abrogate antigen-mediated immunosuppression.

CAR-Tregs degranulate and induce target cell cytolysis in vitro.

CAR-modified CD4+ Tconv cells can kill antigen-expressing targets cells at similar efficiency to the classic cytotoxic CD8+ T cell218. In addition to their ability to suppress proliferation of Teff, Tregs can induce apoptosis in Teff through granzyme B mediated cytolysis179,245. We therefore sought to determine whether CD19 CAR-Tregs would gain similar cytolytic function and induce apoptosis in cells expressing CD19. We measured cytotoxicity by titrating CAR-Tconv or CAR-Tregs with CD19+ Nalm6 cells. We found that all functional CAR-Tregs killed their target cells, irrespective of their costimulation domain, but with a significantly lower efficiency than CD4+ Tconv cells (Fig 3.8A and Fig

3.9A). To confirm that target specific cytolysis was not due to contamination of

Foxp3- Tconv cells in the Treg cultures, we performed a flow-cytometry based degranulation assay, such that we could examine the degranulation of CAR+ cells that did or did not express intracellular Foxp3. We also included Tregs transduced with the

CAR-28z-Foxp3 construct in these assays. We observed clear degranulation by both

Tconv and Treg CAR T cells when incubated with CD19+ Nalm6 cells (Fig 3.8B), as well as robust degranulation when incubated with PMA/ionomycin (Fig 3.9B) and no

degranulation in the absence of CAR or TCR stimulation (Fig 3.9C). Notably, by dual staining, we confirmed that antigen stimulation resulted in degranulation of Foxp3- 28z

CAR-Tconv cells and Foxp3+ 28z CAR-Treg cells (Fig 3.8C). The addition of transgenic

Foxp3 did not prevent CAR-Treg degranulation, further establishing that degranulation was not due Foxp3- contaminating cells (Fig 3.8, B-C).

CAR-Tregs directed to HLA-A2 did not have significant cytotoxicity towards target cells in vitro4,213. We hypothesized that the cytotoxicity seen in our CAR-Tregs was dependent on the high affinity CD19 scFv binder. To test this hypothesis, we generated a first-generation CAR z using an scFv against EGFRvIII (Fig 3.10A), using a low-

246,247 affinity scFv compared to the CD19 scFv (EC50 of ~6ng versus ~100ng) . We confirmed that first generation EGFRvIII CAR-Tregs can suppress antigen-specific

CAR-Teff cells (Fig 3.10B). Once again, we found that the CAR-Tregs could kill target cells (Fig 3.10C) and degranulate in the presence of EGFRvIII (Fig 3.10D) but not in the presence of CD19 (Fig 3.10E). Finally, we wanted to investigate whether sorting ‘naïve’

Tregs based on the positive expression of CD45RA would prevent target-specific cytolysis by Tregs. Though CD45RA+ CAR-Tregs displayed less cytolytic activity than bulk Tregs, they still had higher target cell specific lysis when incubated with Nalm6 cells compared to Dz Tregs and UT Tconv cells (Fig 3.10F). Therefore, we concluded that neither lower affinity antigen nor using naïve cells can prevent CAR-mediated Treg cytotoxicity in vitro.

Figure 5 A B 100 Tc 28ζ 100 Tr 28ζ 80 Tc UT 80 CAR+ Tr ∆ζ + UT 60 60

40 40 % Specific Lysis % CD107a 20 20

0 0 ζ ζ ∆ζ r 3:1 1:1 1:3 c 28 r 28 T 10:1 1:10 T T ζ Foxp3 r 28 CAR:Target T C Tc 28ζ Tr 28ζ Tr 28ζ-Foxp3

45.9 2.4 2.8 23.9 4.8 49.5 CD107a Foxp3

D E * * 30 *** 10 ** 20 s 1 10

0.1 0 % specific lysi GZMB (relative expression) 0.01 -10 ζ ζ ζ ∆ζ r r T T r 28 T r BB Media T ζ ζ Z-AAD ζ Con-A rUT Media r 28 r 28 r 28 T T T T

Figure 3.8 Foxp3+ CAR-Tregs degranulate and mediate target cell cytolysis. (A) Luciferase based killing assay using Nalm6 CBG-GFP cells incubated with CAR T cells at varying ratios for 16 hours. (B) Degranulation assay of CAR T cells calculated as the percentage CD107a+ cells of total mCherry+ or mCherry- cells per condition over 6 hour co-culture with live Nalm6 cells. N=3 human donors. Mean and SEM plotted. Data was from non-fixed cells. (C) After a 6 hour degranulation assay with Nalm6, T cells were fixed, permeabilized and stained for intracellular Foxp3. Representative donor flow plots of CD107a vs. Foxp3. Repeated with N=3 human donors (D) GZMB expression was measured by ddPCR and expressed as a relative ratio to internal control gene (TBP). cDNA was isolated from Tregs stimulated with irradiated Nalm6 for 24 h and sorted on CD4+ mCherry+ cells, mean and SEM plotted. N=3 human donors. (E) Luciferase based killing assays of Nalm6 CBG-GFP with granzyme/perforin inhibitors CMA or Z-AAD-CMK (1:3 Tcell-to-target ratio, 16-hour incubation time) added to media. Representative donor, mean and SEM of triplicates plotted. Repeated with N=3 human donors, *pval<0.05 **pval<0.01 ***pval<0.001 paired t-test. Tc – Tconv cells, Tr – Treg 94

Supplemental Figure 4 A B Tc 28ζ CAR+ 100 Tc BBζ 100 UT Tc ζ 80 80

s Tr ∆ζ +

ysi 60 L Tr ζ 60 40 Tr 28ζ Tr BBζ 40 20 % CD107a

% Specific Tc UT 0 20

-20 0 ζ 3:1 1:1 1:3 ζ ∆ζ 10:1 1:10 r c 28 r 28 T CAR: Target T T ζ Foxp3 r 28 C T 100 CAR+ 80 UT + 60

40 % CD107a 20

0 ζ ζ ∆ζ r c 28 T T r 28 Foxp3 T ζ r 28 T

Figure 3.9 Target specific cytolysis by CAR-Tregs. (A) Luciferase based killing assay using Nalm-6 CBG-GFP cells incubated with CAR T cells at varying ratios for 16 hours. Degranulation assay of CAR T cells calculated as the percentage CD107+ cells of CAR+ (mCherry+) or UT T cells per well over (B) 2 hours with PMA/ionomycin stimulation or (C) no stimulation. N=3 normal donors mean and SEM plotted. Tc – Tconv cells, Tr – Treg

Supplemental Figure 5 A T2A EGFRvIIIζ LTR L Anti-EGFRvIII ScFv TM CD3ζ mCherry LTR

B C

100 EGFRvIII Tc ζ 120 CD19 Tr ζ

100 80 Tc UT EGFRvIII Tr ζ 80 60 EGFRvIII Tr ζ 60 CD19 Tr ζ 40

% Proliferating 40 CD19 Tr ∆ζ % Specific lysis 20 20 0 0

3:1 1:1 1:3 1:9 10:1 3:1 1:1 1:3 1:9 Treg:Teff CAR:Target

D 80 50 E CAR+ CAR+ CAR-

40 + 60 CAR- + 30 40 20 % CD107a

% CD107a 20 10

0 0 ζ ζ ζ ζ ζ ζ ζ ζ

CD19 Tc CD19 Tr CD19 Tc CD19 Tr EGFRvIII Tc EGFRvIII Tr EGFRvIII Tc EGFRvIII Tr F 100 Tr 28ζ 80 Tc ζ s TcUT ysi 60 L TrN ∆ζ 40 TrN 28ζ 20 % Specific 0

-20

3:1 1:1 1:3 10:1 1:10 CAR: Target

Figure 3.10 Target specific cytolysis by CAR-Tregs is independent of the CD19 scFv. (A) Vector map of EGFRvIII CAR construct. TM, hinge and transmembrane domain. L, leader sequence. (B) MLR of CFSE labeled EGFRvIII z CAR-Teff with different EGFRvIII z CAR Treg ratios stimulated with irradiated U87-EGFRvIII cells. After 4 days CFSE dilution was measured to calculate % proliferation as the % CFSElow mCherry+ Teff cells compared to the % of CFSElow mCherry+ Teff cells with no Tregs present. Representative donor. (C) Luciferase based killing assay using U87-EGFRvIII CBG-GFP cells incubated with z CAR-Tregs and z Tconv cells with CARs against CD19 or EGFRvIII at varying ratios for 16 hours. Representative donor. (D) Degranulation assay of CAR T cells in media with CD107a antibody and Befeldin A. 6 hour stimulation with U87-EGFRvIII target cells or (E) U87-CD19 target cells. Gated on CD3+ mCherry+ (CAR+) or mCherry- (UT T cells) within a sample. Representative donor. (F) Luciferase based assay using Tregs sorted on CD45RA+ (naïve) Treg cells and Nalm6-CPG-GFP Target cells. Representative donor with technical triplicates. Mean and SEM plotted. Tc – Tconv cells, Tr – Treg, TrN – Naïve Tregs 96

CAR-Treg cytotoxicity is dependent on granzyme B.

Given the robust antigen-specific degranulation of CAR-Treg cells, we hypothesized that the mechanism of CAR-Treg cytotoxicity was via release of cytotoxic granules. We performed digital droplet PCR to determine the expression levels of GZMB, GZMA and

PRF1 (encoding proteins granzyme B, granzyme A and perforin respectively) in CD19

CAR-Tregs and Tconv cells after a 24-h stimulation with CD19+ target (Nalm6) cells. We found that GZMB was specifically upregulated in functional, activated CAR Tregs compared to Dz Tregs (Fig 3.8D), though, this GZMB expression was not as high as the expression in activated CAR-Tconv cells (Fig 3.11A). GZMA was not induced by CAR activation, and PRF1 expression was similar across CAR Tregs and Tconv cells (Fig

3.11, B and C, respectively). Addition of perforin/ granzyme B pathway inhibitors with either Concanamycin A, a perforin inhibitor, or Z-AAD-CMK, a granzyme B specific inhibitor, reduced both Treg (Fig 3.8E) and Tconv cytotoxicity (Fig 3.11D). In conclusion,

CAR-modified Tregs display a low but significant level of target cell-specific lysis in vitro that is at least partly dependent on the granzyme B/perforin pathway.

In vivo models of CAR-Treg trafficking, suppression and cytotoxicity.

To test whether CAR-Tregs traffic specifically towards their CAR-target in vivo, we injected NSG mice with CD19- or CD19+ U87 solid tumor-cell lines subcutaneously (SC) into the left and right flank, respectively. Tumor-cell lines expressed click beetle green and were tracked by bioluminescent imaging (BLI) in the mice. One week later (day 0),

Supplemental Figure 6

A ** B ** * 40 ** 60 * n * 35 30 25 40 20 5 4 20 3 2

1 GZMA relative expression GZMB relative expressio 0 0

ζ ζ ζ ζ ∆ζ ∆ζ r UT r r UT c 28 T T r 28 c 28 T r r 28 T T T T T C D 8 80 **** n ** ** 6 60

4 40

2 %Specific Lysis 20

PRF1 relative expressio 0 0 ζ ∆ζ ζ r UT c 28 T r T T r 28 Media T ζ ζ Z-AAD ζ Con-A c UT Media c 28 c 28 c 28 T T T T Figure 3.11 Target specific lysis by CAR-Tregs is perforin/ granzyme dependent. (A) GZMB, (B) GZMA and (C) PRF1 expression from cDNA isolated from CD19 CAR-Tconv and CD19 CAR-Treg cells stimulated with Nalm6 for 24 hours and sorted on CD4+ mCherry+ cells. Gene expression was measured by digital droplet PCR and expressed as a relative ratio to TBP (T cell internal control gene). N = 4 normal donors. (D) Luciferase based killing assays of Nalm6 cells were repeated with CD19 28z-Tconv or UT- Tconv plus granzyme/ perforin inhibitors CMA and Z-AAD-CMK (1:3 Tconv to target ratio, 16-hour incubation time). UT-Tconv cells shown as the negative control for antigen-specific cytotoxicity. Mean and SEM plotted, representative donor, measured in triplicates. *pval<0.05, **pval<0.01, ****pval<0.0001. N=3 human donors. Tc – Tconv cells, Tr – Treg

we injected the mice intravenously (IV.) with either CD19 28z CAR-Tregs or EGFR 28z

CAR-Tconv as an internal control. Mice were administered IL-2 intraperiotenally (IP.) three times a week from day 0 to support the injected Tregs. Tumors were harvested at day 14 and examined histologically for tumor necrosis, and the presence of CAR-T cells

(Fig 3.12A). We found CD19 28z CAR-Tregs trafficked specifically to the U87-CD19+ expressing tumor, whereas we did not see trafficking to the U87-WT tumor as detected by immunohistochemistry (IHC) for CD3 or mCherry (Fig 3.12B). In contrast to EGFR

Tconv cells, which trafficked to both left and right tumors since both U87 based

cells-lines expressed EGFR (data not shown). CAR-Tregs did not decrease tumor BLI in

U87-CD19+ or U87-CD19- (Fig 3.12C). Compared to EGFR CAR-Tconv cells, which decreased luminescence of both U87-CD19+ and U87-CD19- tumors (Fig 3.12D). We conclude that CAR-Tregs traffic to sites of antigen but their cytotoxicity is not potent Figure 6 enough to lyse proliferating tumor cells in vivo.

A B

U87 - CD19 U87 WT

2e6 CD19-28ζ CAR-Treg or EGFR-28ζ CAR-Tconv H&E 4x U87-CD19 U87-WT

Day -7 0 4 7 14 CD3 4x

Tumor BLI BLI BLI BLI SC. Tumor mCherry 4x harvest

mCherry 20x

C D CD19-28ζ CAR-Treg EGFR-28ζ CAR-Tconv 10 10 10 U87 WT 10 U87 WT 109 U87-19 109 U87-19

108 108 I I 107 7

BL 10 BL

106 106

105 105

104 104 0 4 7 0 4 7 12 12 Day Day

Figure 3.12 CAR-Tregs traffic to antigen expressing tissue in vivo. (A) Experimental outline of U87 tumor model for CAR-Treg trafficking. (B) H&E, CD3 and mCherry staining of U87-CD19 and U87-WT tumors from mice treated with CD19 28z CAR-Tregs and IL-2. Tumor BLI of left and right flank tumors of mice treated with (C) CD19-28z Tregs and (D) EGFR-28z Tconv cells. Each line represents either a U87-CD19- (green) and U87-CD19+ (blue) tumor in an individual mouse. Tr – Treg

We hypothesized that if Tregs were to cause low levels of target specific destruction, it may be more relevant in a non-tumor model, where the target cells do not proliferate at the rapid rate. We therefore used a skin xenograft model where NSG mice were grafted with human abdominal skin that endogenously expresses EGFR (Fig

3.13A). We generated CAR-Tregs and CAR Teff cells (bulk CD4+ and CD8+ T cells) with a CAR containing an EGFR scFv and a 28z signaling domain (Fig 3.14A), and confirmed that EGFR 28z CAR-Tregs also had low but measurable cytotoxic activity against EGFR+ tumor target cells in vitro (Fig 3.14B). After the skin grafts had healed, we injected EGFR 28z CAR-Tregs or EGFR 28z CAR-Teff or equal ratios of the two intravenously. CD19 28z CAR-Tregs were injected as a negative control. Mice that received Tregs alone were injected with IL-2 as above. Grafts were monitored and photographed over two weeks and the grafts were harvested for histology at day 14 (Fig

3.13A).

The grafts of the mice that received Teff cells alone had reduced in size and showed clear signs of depigmentation compared to the xenografts of mice that received both

CAR-Teff cells and CAR-Tregs (Fig 3.13B and Fig 3.14C). However, in mice treated with the same number of Teff CARs but in combination with EGFR 28z Tregs, the graft had not changed in size and there was no observable skin depigmentation (Fig 3.13B).

As expected, by H&E (Fig 3.13C) and CD3 IHC (Fig 3.13D), all the mice that were treated with EGFR CAR T cells – both Tregs and Teff – had clear lymphocyte infiltration into the skin grafts. Graft rejection was obvious by histopathology in the Teff treated

100

graft, with dense lymphocyte infiltration, spongiosis, and exocytosis of the epidermal layer. In some areas, the epithelium had been destroyed and there were signs of epithelial apoptosis, dyskeratosis and keratinolysis (Fig 3.13, C-D). However, T cell infiltration alone did not dictate graft rejection in this model. We noted that Tregs remained closer to the dermal/epidermal junction, unlike the Teff cells that infiltrated the epidermal layer (Fig 3.13, C-D). The signs of epithelial barrier destruction that were seen in the CAR-Teff alone group were reduced when EGFR CAR-Tregs were administered with Teff cells, indicating that in vivo the CAR-Tregs could suppress Teff cell tissue destruction (Fig 3.13, C-D). Furthermore, the Treg suppressive function was dominant over Teff rejection. We could not detect CD19-CAR Tregs in the skin graft.

Further IHC with CD8, Foxp3 and mCherry antibodies confirmed that the EGFR CAR T cells infiltrating the grafts were indeed mCherry+ (Fig 3.13D). Foxp3 IHC staining revealed that the mice treated with EGFR CAR-Tregs alone or in combination with Teff cells had nuclear Foxp3 staining of some of the cells whereas the Teff-only treated grafts had only rare Foxp3+ cells in the graft (Fig 3.13D and Fig 3.14D).

Finally, although there were no signs of overt graft rejection by photographs or histology analysis of EGFR 28z CAR-Tregs alone, TUNEL staining showed minimal but observable keratinocyte death, which was also evident in the CAR-Treg/CAR-Teff combination (Fig 3.15A). There was no discernible keratinocyte death or TUNEL

101

Figure 7

A 2e6 CAR injection

Day -40 0 9 14

Human CAR T cell Photo Photo skin grafted IV injection Fix and paraffin Photo embed skin

B C H&E 4x Teff EGFR 28ζ Tr CD19 28ζ Tr EGFR 28ζ Teff EGFR 28ζ + Tr EGFR 28ζ Tr CD19 28ζ

Day 0 100um

Tr EGFR 28ζ

100um Day 9

Teff EGFR 28ζ

100um Day 14

Teff EGFR 28ζ + Tr EGFR 28ζ

100um

H&E 10x D CD3 4x CD8 10x mCherry 10x Foxp3 10x

Tr CD19 28ζ

100um 100um 100um 100um 100um

Tr EGFR 28ζ

100um 100um 100um 100um 100um

Teff EGFR 28ζ

100um 100um 100um 100um 100um

Teff EGFR 28ζ + Tr EGFR 28ζ 100um 100um 100um 100um 100um

Figure 3.13 Skin xenograft model of CAR-Treg mediated suppression. (A) Experimental outline Skin xenograft model. (B) Images of skin grafts, representative mouse. Repeated with N=3 donor grafts and N=3 donor T cells, 1 mouse per group (C) H&E histology of sections from grafts 2 weeks after Treg were injected taken with a 4x objective lens. (D) H&E and IHC staining of human CD3 (4x), human CD8 (10x), mCherry (10x) and human Foxp3 (10x), representative images. Tr – Treg

102

Supplemental Figure 7

A T2A EGFR-28ζ LTR L Anti-EGFR ScFv TM CD28 CD3ζ mCherry LTR

B C

100 Tc EGFRV 28ζ Tr EGFRV 28ζ 100 Tr CD19 28ζ t

s 80 Tc UT 80 ysi L 60 Tr CD19 28ζ 60 Tr EGFR 28ζ 40 40 Tc EGFR 28ζ Tc EGFR 28 + % Specific 20 ζ

% area of initial graf 20 Tr EGFR 28ζ 0 0 0 5 10 15 3:1 1:1 1:3 10:1 Day CAR:Target

100um

Figure 3.14 In vivo skin xenografts model setup. (A) Vector map of EGFR 28z CAR construct. TM, hinge and transmembrane domain. L, leader sequence. (B) In vitro luciferase-based killing assay using U87 CBG-GFP cells incubated with CAR EGFR 28z Tregs and Tconv cells at varying ratios for 16 hours. Representative donor, mean and SEM of technical triplicates. (C) Size of graft measured as a percentage of the size prior to CAR T cell injection N=2 skin graft donors and donor T cells. (D) Immunohistochemistry of nuclear Foxp3 staining in xenograft of EGFR 28z Tregs treated mouse (20x objective lens).

103

Figure 8

Teff EGFR 28ζ Tr CD19 28ζ Tr EGFR 28ζ Teff EGFR 28ζ + Tr EGFR 28ζ

100um 100um 100um 100um

TGFB1 10x IL10 10x PRF1 10x GZMB 10x

Tr CD19 28ζ

100um 100um 100um 100um

Tr EGFR 28ζ

100um 100um 100um 100um

Teff EGFR 28ζ

100um 100um 100um 100um Teff EGFR 28ζ + Tr EGFR 28ζ

100um 100um 100um 100um

Figure 3.15 In vivo CAR-Tregs exhibit low amounts of tissue cytotoxicity and express immunosuppressive cytokines. (A) Tunel staining (10X objective lens) and (B) RNAscope of TGF-β1 (10X), IL-10 (10X), PRF1 (10X) and GZMB (10X) of skin xenografts from mice after treatment with CAR-Tregs and CAR-Teff cells, representative images. Tr – Treg

104

staining in the grafts recovered from CD19 CAR-Treg treated mice. We conclude that

CAR-Tregs can mediate low but measurable cytotoxicity against tissues expressing the target antigen. To determine CAR-Treg functions in the grafts, we used RNAscope to detect IL10 and TGFB1 (encoding immunosuppressive cytokines IL-10 and TGFb respectively) as well as GZMB and PRF1. We found RNA expression of IL10 and

TGFB1 in the grafts with Treg alone, this expression was greatly increased in grafts from mice administered the combination of EGFR CAR-Tregs and CAR-Teff cells (Fig

3.15B). In contrast, GZMB and PRF1 were expressed at high levels in Teff alone treated grafts, but at a lower level in the Teff with Treg-treated grafts (Fig 3.15B). We did not detect significant levels of PRF1 or GZMB expression in EGFR CAR-Treg alone or CD19 CAR-Treg treated mice. We therefore conclude that CAR-Tregs bearing

CD28z signaling domains can traffic to target tissues in vivo and exert functional immunosuppression, despite low levels of target-directed cytotoxicity.

3.3 Discussion

Antigen-specific or tissue-specific Tregs have enormous therapeutic potential for GvHD, generating transplant tolerance and preventing site specific T cell and B cell mediated autoimmunity. Although the effects of CAR-encoded costimulation domains have been described in CAR-Tconv cells, the varying effects of these domains have not been described in Tregs. We found that the costimulation domains we tested do not affect the stability of transcription factor Foxp3 in Tregs. We confirmed that Tregs suppress antigen-specific T cell proliferation and cytokine secretion but here, the costimulation domain could affect the degree of this suppression. Previous research has shown that

105

the addition of 4-1BBL abrogate Treg suppression in standard co-culture assay and this was largely believed to reflect a role for 4-1BB signaling in allowing Teff cells to become resistant to suppression rather than impairing Treg function248. However, our data suggests this effect may occur both through suppression of Treg function and by providing additional costimulation to Teff cells. Others have shown that 4-1BB costimulation increases Treg expansion249. It is possible that the ability of a Treg to proliferate does not correlate with the ability of a Treg to suppress immune responses, as our CD19-BBz Tregs are able to proliferate to the same degree as CD19-28z CAR

Tregs after CD19 stimulation. Another possibility is that there is a difference in signal strength and kinetics between physiological 4-1BBL-receptor signaling in Tregs versus

CAR-mediated 4-1BB costimulation, and that this can lead to differing effects of 4-1BB on Treg suppression.

We found other indicators that CD28 costimulation was more beneficial to Tregs in terms of maintaining Treg phenotypic functions. For example, CAR BBz Tregs expressed less LAP and CTLA4 than 28z CAR-Tregs, and, CAR-mediated activation led to greater production of IL-10 in 28z CAR-Tregs compared to BBz CAR-Tregs. In addition, BBz CAR-Tregs consumed more IL-2 at baseline, but IL-2 consumption did not increase to the same degree as 28z CAR-Tregs when activated through their CAR.

Finally, BBz CAR Tregs form visible cell aggregations in culture which may affect their interactions with Teff cells during in vitro co-culture assays. Tregs are hypothesized to suppress Teff cells through the combination of several of these mechanisms165,166. We find 4-1BB containing CAR-Tregs have reduced ability to produce several molecules 106

involved in Treg suppressive mechanisms and moreover may be more metabolically activate at baseline. In addition, BBz CAR-Tregs seem to display dysfunctional cell-cell interaction kinetics. The effect of each one of these differences between 28z and BBz alone may not have an observable effect on Treg function, but in combination, results in reduced antigen-specific suppression of Teff cells by BBz CAR-Tregs.

Another aspect of Treg suppression to consider, is that Tregs suppress Teff cells both directly and indirectly through their regulation of antigen presenting cells. Our research focused on the direct effect of Treg suppression on Teff cells, though it would be interesting to see if the same differences between CD28 and 4-1BB costimulation could be observed when indirect suppression by Tregs is also considered.

Since Tregs can only suppress after activation, antigen-specific Tregs should only exert immunosuppressive functions when in close proximity to their cognate antigen. We found that CD19 CAR-Tregs and EGFR CAR-Tregs both traffic to antigen-expressing tissues in vivo. Activation of Tregs induces “bystander” suppression of other inflammatory cells in the vicinity206. Therefore, we postulate that it is not necessary to know the antigens activating auto-reactive T cells in order to broadly suppress inflammatory immune cells at a tissue site of inflammation.

The common occurrence of B-cell aplasia highlights the damage that CAR T cells can inflict on normal tissues that express the target antigen120. Therefore, as CAR-Tregs move towards the clinic, understanding the potential for tissue destruction is critical.

107

Furthermore, finding ways to improve the safety of CAR-Tregs is going to be important.

One way to do this would be to design CARs against secreted or soluble antigens, such that the CAR cannot form a synapse with a cell that can be lysed. Other possibilities could include modifying CAR-Tregs further by knocking out genes important for their cytotoxicity, such as GZMB. It is interesting that others have not seen the cytotoxicity we found in our in vitro assays, though Treg- mediated lysis of antigen presenting cells is thought to be one mechanisms by which Tregs induce tolerance250. Perhaps one reason for this observation is that we chose to culture Tregs in the absence of rapamycin, and rapamycin inhibits granzyme B production251.

Analagous to the use of conventional CAR T cells as adoptive immunotherapy for cancer, we and others have found that transduction of Tregs with lentiviral vectors coding for CARs endows Tregs with new antigen specificity and the ability to culture and expand large numbers of antigen-specific Tregs. We identified three novel findings in the field of CAR Treg therapy: (1) incorporation of the 4-1BB signaling domain in CAR-

Tregs reduces their immunosuppressive function, whereas first generation or CD28 costimulation maintains phenotypic stability and immunosuppressive function; (2) CAR-

Tregs can traffic to the sites of antigen expression in vivo and exert functional immunosuppression against large numbers of Teff cells that would otherwise destroy the target tissue; and (3) CAR-Tregs may exert minimal levels of antigen-specific cytotoxicity on their own, suggesting that a cautious approach is warranted in clinical trials of CAR-Tregs directed to life-sustaining tissues. Nevertheless, CAR transduction

108

of Tregs overcomes several obstacles previously inherent to Treg therapy, and holds promise for antigen-specific immune suppression in a variety of settings.

109

3.4 Materials and Methods

Sorting, Expanding and Transducing T Cell Populations

CD4+ T cells were negatively selected from healthy donor leukopaks using RosetteSep

Kits with a Ficoll gradient (Stemcell Technologies) and enriched for CD25+ cells using

CD25-PE antibody staining followed by anti-PE microbead selection according to the manufacturer provided protocol (Miltenyi). Non-CD25-enriched, CD25 enriched and

CD25 depleted T cells were stained separately at 4e6 cells/100µl, for 30 minutes at 4°

C, in PBS with 2% heat-inactivated fetal bovine serum (FBS, Gibco Life Technologies) with 50µl /100µl brilliant violet staining buffer (BioLegend), 5µl /100µl CD127-BV711, 5µl

/100µ CD4-BV510 and 2.5µl /100µl CD8-APC-H7 (clones CD3-OKT3 and CD4-OKT4,

BioLegend and CD8-SK1, BD Pharmingen). Stained cells were washed and resuspended in HBSS supplemented with 25mM HEPES and 1% FBS with

4′,6-diamidino-2-phenylindole (DAPI) prior to sorting. Tregs were purified from CD25 enriched cells by fluorescence-activated cell sorting (FACS) for live, CD4+, CD8-,

CD25++ CD127low and Tconv cells were sorted from CD25-depleted cells by live, CD4+

CD8- CD25 low into 50% FBS in PBS. Non-enriched cells were used to draw the CD25++ sorting gates which we defined as the level of CD25 expression where a shift to slightly lower for CD4 staining could be seen. 2e5 sorted Tregs and Tconv cells were stained for surface levels of CD39, LAP, LAG3 and CTLA4 and intracellular Foxp3 and analyzed by flow cytometry.

110

Tregs and Tconv cells were then expanded with anti-CD3/anti-CD28 beads (Dynabead

Human Treg expander or Dynabead Human T-activator beads respectively, Gibco) in

CTS™ OpTmizer™ T Cell Expansion Serum Free Media (Thermo Fisher Scientific) supplemented with 2% human serum (Access Cell Culture LLC), 1x GlutaMAX™

(Thermo Fisher Scientific), Penicillin-Streptomycin 100 U/ml (Thermo Fisher Scientific) and recombinant human IL-2 (Peprotech, 300 IU/ml and 20 IU/ml respectively). One day post-sort, T cells were transduced at an MOI of 5 with lentivirus carrying one of the 5

CAR constructs with a humanized scFv that binds human CD19 and intracellular domains: 4-1BBz, CD28z, z and Dz. The CD28z-Foxp3 construct expressed transcription factor Foxp3 which was inserted behind a T2A sequence in the plasmid before the second T2A element followed by mCherry.

T cells were expanded for one week with beads and then de-beaded and rested for another week. Media was added every 2-3 days to maintain cells at a concentration of

1e6-2e6 T cells/ml. IL-2 was replaced every 2-3 days. Assays were performed on fresh

(never-frozen) Tregs on day 14-15 in supplemented OpTmizer™ with no IL-2 or in

RPMI-1640 with 1x GlutaMAX and 25mM HEPES (Gibco, Life Technologies), supplemented with 10% FBS and 100 U/ml penicillin-streptomycin (R10) as stated. For all assays, percent transduction was measured by determining the percentage of mCherry+ T cells by flow cytometry analyzed on a BD Fortessa x-20. All CAR T cells were normalized to the same % CAR positive by adding appropriate numbers of

111

expanded UT Treg or Tconv cells. All experiments were performed with CAR transduction above 50%.

Cell Lines

Human embryonic kidney 293 (HEK293T), K562, U87 and Nalm6 cell lines were purchased from American Tissue Culture Collection (ATCC). HEK293T, K562 and

Nalm6 cells were expanded in R10. U87 cells were grown in Eagles minimum essential media (ATCC) supplemented with 10% FBS and 100 U/ml penicillin-streptomycin.

Nalm6 and U87 cell lines were lentivirally transduced to express the click beetle green luciferase and green fluorescent protein (GFP) under control of the EF-1a promoter.

Nalm6-CBG-GFP cells were then single-cell sorted to establish a clonal population. U87 cells lines were bulk sorted for a CBG-GFP+ population. Lentiviral particles, produced to express EGFRvIII, human CD19 or OKT3 scFv under the control of the EF-1α, were used to generate U87-CD19, U87-EGFRvIII K562-CD19 and K562-OKT3 by lentiviral vector transduction and single cell sorting for the K562 cells or bulk sorting for the U87 cells for positive populations by FACs. U87 and HEK293T cells were passaged using

0.25% and 0.05% Trypsin-EDTA (Thermo Fisher Scientific) respectively. Target cells were irradiated with 10 000 rads and frozen in FBS with 10% DMSO to be thawed prior to stimulation of CAR T cells. Cell lines were tested for mycoplasma contamination every 3 months.

112

Construct Generation and Lentivirus Production.

CD19, EGFRvIII and EGFR-specific CARs were synthesized and cloned into a third-generation lentiviral plasmid backbone under the regulation of a human EF-1α promoter (GenScript USA Inc). Lentiviral vector was made as described in Section 2.4.

Flow Cytometry Reagents and Analysis

Fluorescent anti-CD3 (OKT3), anti-CD4 (OKT4), anti-CD8 (SK1), anti-CD69 (FN50), anti-LAP (TW4-2F8), anti-CD137 (4B4-1) and anti-CTLA4 (L3D10) antibodies were purchased from BioLegend, while anti-CD25 (2A3), anti-CD127 (HIL-7R-M21) and anti-CD107a (H4A3) fluorescent antibodies were purchased from BD Biosciences. Cells were surface stained in 2% FBS PBS for 30 minutes at 4° C and DAPI or 7AAD was added prior to running samples if no other live dead stain was used. For Foxp3 intracellular staining anti-Foxp3 (PCH101) antibodies were purchased from eBioscience. Blue, Aqua or Violet LIVE/DEAD fixable dyes (Thermo Fisher) were used to stain dead cells in PBS for 20 minutes at 4° C before surface staining and fixation.

After surface staining T cells were fixed according to eBioscience Foxp3 transcription factor staining kit’s recommended protocol. Briefly, cells were fixed and permeabilized for 45 minutes and then washed in permeabilization buffer and blocked with rat serum

2µl/100ul for 15 minutes at room temperature (RT). Fixed cells were stained for 30 minutes with 2µl of Foxp3 antibody/100µl at 4° C. Fluorescence was measured on a BD

Fortessa x-20 and data were analyzed using FlowJo (Tree Star).

113

TCR versus CAR Re-stimulation

Assays were performed in supplemented OpTmizer media for all experiments unless otherwise mentioned. For methylation, phenotypic and proliferation analysis of long-term activated CAR T cells, 1e6 CAR-Treg or CAR-Tconv cells were stimulated with irradiated K562 cells expressing surface anti-CD3 scFv (OKT3) or CD19 at a 1:1 T cell-to-target ratio in a 12-well plate. Cells were maintained in culture at a concentration of 5e5-2e6 cells/ml. T cells were counted using a LUNA-FL dual florescence cell counter

(Logos Biosystems) and analyzed by flow cytometry every 2 days for 8 days to account for live, K562 cells in the culture whilst documenting the T cells’ expansion. For phenotypic analysis, surface and intracellular staining for markers CD39, CTLA4, LAP and Foxp3 were measured by flow cytometry pre- (day 14) and 9 days post-K562 stimulation (day 23). For methylation analysis, T cells were also sorted by mCherry+ and

CD3+ and then frozen at -80° C for DNA methylation analysis.

For proliferation assays measured by violet dilution, cells were washed in PBS and labeled with 1 ml of cell trace violet dye at a concentration of 1µM in PBS for 10 minutes at room temperature on day 14. Labeling was stopped by incubating cells for 1 minute with 1ml of FBS and for a further 10 minutes, after the addition of 10ml of R10. Tregs were then washed three times before resuspending in OpTmizer media supplemented

300 IU/ml of IL-2 at 1e6 CAR T cells/ml. 100 000 cells were plated per well with 100 000 irradiated Nalm6-target cells and incubated at 37° C for 3 days. 100µl of OpTmizer was added on day 2. T cells were washed before staining with CD3-APC for 30 minutes at

114

4° C. DAPI was added prior to performing flow cytometry with a high through-put plate reader on a Fortessa x-20. The percentage of violet low cells proliferating was calculated and represented as a ratio normalized to the number of violet low cells in the non-stimulated condition.

For short-term activation assays (24-hour), Tregs were grown in reduced IL-2 media (20 units/ml) from day 12 to day 14. On day 14, 1e5 T cells/well were activated in a 96-well round-bottom plate with either no target cells or at a 2:1 T cell-to-target ratio with irradiated K562-CD19 or K562-OKT3 cells in a final volume of 200µl/well in technical triplicates. Cells were incubated at 37° C for 24 h. Triplicate wells were pooled and stained for CD3, CD4, CD69, LAP and 4-1BB (CD137) and surface expression was measured by flow cytometry.

DNA Methylation Analysis

5e5 Tregs and Tconv cells were sorted by CD3 and mCherry expression except in the case of UT T cells which were sorted on CD3 only. Sorted cells were then washed in

PBS and snap-frozen before shipping to EpigenDx for methylation analysis. The methylation status of CpG motifs across the Foxp3 TSDR, CTLA4 and IKZF2 loci was assessed by targeted next-generation bisulfite sequencing using the EpigenDx Human

Foxp3 methylation panel. Percent methylation at each CpG site was averaged and then represented as an average from 2-3 human donors/group, only female donors were used for TSDR methylation analysis.

115

Cytokine Detection by Luminex

For cytokine release assays, Tregs or Tconv cells were stimulated in a 96-well round-bottom plate with 100, 000 CAR T-cells/well combined with irradiated K562-OKT3 or K562-CD19 target cells at a CAR T cell-to-target ratio of 2:1 in a total volume of

200µl. Supernatants were harvested after 24 hours and frozen at –80°C. 50µl of supernatant was measured by the Luminex platform as described in Section 2.4

Suppression Assays

Tregs were violet-labeled as described for TCR versus CAR re-stimulation and used in mixed lymphocyte reactions (MLRs) with CFSE-labeled Teff cells (following the same protocol for violet labeling, CFSE cell trace, Invitrogen, 1µM staining concentration).

Tregs were titrated in a 96-well plate with 5e4 Teff cells/well and 1e5 irradiated target cells/well in R10. Cells were left in an incubator at 37° C for 3 days unless otherwise mentioned with 100µl of media added on day 2. In the case where beads were used for the proliferation assay, anti-CD3/anti-CD28 beads (Dynabead: Human T-activator beads) were used at a 1:10 bead-to-Teff cell ratio. To analyze MLR suppression assays, cells were stained with CD3-APC for 30 minutes at 4° C and DAPI was added prior to flow cytometry on a Fortessa x-20 with a high throughput plate reader. The percentage of proliferating cells in any condition(x) was calculated as the % of CFSE low cells of the total mCherry+ CFSE+ (Violet- ) Teff cells in condition(x) of the number of

CFSE low cells of mCherry+ CFSE+ Teff cells in the no Treg condition. Experiments

116

were run in technical triplicates with N≥3 normal human donor T cells. Inhibition of cytokine secretion was measured from the supernatant of the MLRs described above in technical triplicates. Cytokines were measured from 50µl of supernatant using the

Luminex assay described above. For IL-10 blocking assays, IL-10 antibody (LEAF purified, clone JES3-19F1, BioLegend) was added to wells at a final concentration of 10

µg/ml.

IL-2 Sink Assay

Tregs or Tconv cells were plated at 1e5 CAR T cells/well with technical triplicates in a

96-well round-bottomed plate and stimulated with 1e5 irradiated Nalm6 or no stimulation. We also included a media only condition where no T cells were added. All cells were in OpTmizer with a final starting concentration of 50 IU/ml IL-2 in 200µl. Cells were incubated for 40 hours at 37° C after which, supernatants were frozen at -80° C.

Cytokine levels were measured from 50µl of supernatant via Luminex (described in

Section 2.4).

Cytotoxicity Assays

Luciferase based killing assays were performed by titrating each CAR construct, Tconv or Tregs in a 96-well plate and then adding 2e4 CBG-GFP expressing target cells/well

(Nalm6, U87, U87-CD19 or U87-EGFRvIII). Cells were lysed after 15 h in culture and the live target cells were quantified by BLI after the addition of D-Luciferin. Percent

!"# %&'()% *)++, &+-.) / !"# *-.01%1-. 2 specific lysis was calculated for any condition (x) as . !"# (%&'()% *)++, &+-.))

117

For the inhibition of granzyme/perforin pathways, we used granzyme-perforin axis inhibitor, concanamycin A (CMA, 200 nM; Sigma) and granzyme B-specific Inhibitor I

(Z-AAD-CMK, 50 μM; Calbiochem). Inhibitor concentrations were chosen based on previously published studies investigating cytotoxicity by granzymes181. Prior to use, each inhibitor was found to have insignificant effects on the viability of Tregs following an 18 h incubation at 37° C as assessed by LUNA-FL dual florescent counter. In inhibitor assays, all samples including the target-alone samples were incubated with the inhibitor to account for the effects of the inhibitor on tumor cell viability. Assays were run in technical triplicates.

Degranulation Assays

Degranulation of Tregs and Tconv cells was calculated by incubating 3e5 T cells in a

24-well plate with live, Nalm6 at a 1:1 Target-to-T cell ratio for 6 hours at 37°C in media with CD107a (7µl/well) and befelden A (BD golgiplug 1µl/ml). PMA ionomycin (cell stimulation cocktail, eBioscience) was used at 1X concentration in culture for 2 hours.

Cells were stained with live/dead fixable stain according to manufacturer’s instructions

(Invitrogen) followed by surface staining for CD3 and CD4. To determine the Foxp3 expression levels of degranulating cells, we fixed and permeabilized cells and stained for Foxp3 as described in the flow cytometry methods. Both unfixed, surface only stain as well as fixed and permeabilized cells were analyzed by flow cytometry.

118

Digital Droplet PCR

CD19 CAR T cells were transduced, expanded for 7 days and rested for 7 days. 1e6 cells were stimulated with 1e6 Nalm6 cells over 24 h and then stained with CD4 antibody. 5e5 cells were collected by FACs and resuspended in 350μl of lysis buffer with

1% 2-mercaptoethanol. RNA isolation, cDNA synthesis and droplet generation were performed as described in Section 2.4

The PCR cycling protocol was according the manufacturer’s instructions with a 57°C melting temperature. Human TBP was used at as the reference gene in each reaction, (HEX fluorophore : TBP PrimePCR™ ddPCR™ Expression Probe Assay:

Unique Assay ID: dHsaCPE5058363 (Bio-Rad)). The following FAM fluorophore primer probes were used:

GZMB: PrimeTime Std® qPCR Assay unique assay ID Hs.PT.58.26439821.g (IDT)

GZMA: PrimePCR™ PCR Primers unique assay ID dHsaCPE5047756 (BioRad)

PFR1: PrimePCR™ PCR Primers unique assay ID dHsaCPE5030232 (BioRad)

H&E, immunohistochemistry and TUNEL staining

Tissue was collected and fixed in 10% formalin for 24 h followed by standard paraffin embedding. 5µm paraffin-embedded tissue sections on glass slides were baked at 60°

C for 30 minutes, followed by deparaffinization in xylene and rehydration in graded alcohol into water. After washing with TBS/Tween 20, antigen retrieval was performed by boiling the slides in 10mM Sodium Citrate buffer pH=6.0 for 30 minutes. Endogenous

119

peroxidase activity was quenched with Dual Endogenous Enzyme Block (DAKO) for 5 minutes. After washing, tissue sections were incubated with 1:100 dilution of mCherry rabbit polyclonal antibody (Abcam, ab183628) or 1:400 dilution of CD3 rabbit polyclonal antibody (Dako A0452) in 1% TBS/BSA inside a humidified chamber 1 hour at room temperature. After washing, slides were incubated with HRP labelled anti-Rabbit

Polymer (Dako) 30 minutes at RT. After washing the DAB+ reagent (DAKO) was added with monitoring for 5-10 minutes. After washing counterstain was done using Harris type

Hematoxylin. Slides were briefly dehydrated and then mounted with Histomount solution

(Life Technology, 008030).

For TUNEL stain, slides were baked and deparaffinization and rehydrated as above.

After washing with TBS/Tween 20 proteinase K (DAKO) was added to slides at a dilution of 1:10 in Tris buffer. TUNEL staining was performed using the S7100 |

ApopTag® Peroxidase In Situ Apoptosis Detection Kit (Millipore) according to manufactures instructions. DAB substrate and DAB Enhancer (DAKO) were used for detection followed by hematoxylin counterstain. Slides were hydrated and mounted as for IHC.

Automation RNA in situ hybridization (ISH) Assay

Automated RNA-ISH assay was performed using the RNAscope 2.5 LS Reagent

Kit-Brown from Advanced Cell Diagnostics (ACD) (Catalogue No.322100) on the

120

BondRx platform. 5-μm sections of formalin-fixed paraffin embedded tissue were mounted on Surgipath X-tra glass slides, baked for 1 h at 60° C, and placed on the

BOND RX for processing. On the BOND RX, the staining protocol used was the ACD

ISH DAB Protocol. The RNA unmasking conditions for the tissue consisted of a 15- minute incubation at 95° C in Bond Epitope Retrieval Solution 2 (Leica Biosystems) followed by 15-minute incubation with Proteinase K which was provided in the kit. Probe hybridization was done for 2 hours with RNAscope probes which were provided by

ACD. The probes used for this study were, 2.5 LS TGF-β1 Probe (Catalogue No.

443488); 2.5 LS PRF1 Probe (Catalogue No. 550288); 2.5 LS GZMB Probe (Catalogue

No. 550328) and 2.5 LS IL-10 Probe (Catalogue No. 550348). The RNA-ISH assay uses highly specific, branched DNA technology in which signal amplification is implemented to detect target mRNAs within the FFPE tissue section via a series of sequential hybridization steps in which the probe binds to the target mRNA. Subsequent binding of the preamplifer, amplifier and alkaline phosphatase-labelled probe molecules creates a signal amplification structure which can then be visualized with the

3,3’-Diaminobenzidine (DAB) as a chromogen to form a brown dot which can then be visualized using a standard bright field microscope.

In Vivo Mouse Models

For in vivo experiments, bulk T cells (CD4+ and CD8+) were negatively selected using T cell enrichment RosetteSep Kits with a Ficoll gradient (Stemcell Technologies). Teff cells were grown in R10 supplemented with 20 IU/ml IL-2. T cells were expanded and

121

transduced as described above for Tconv cells. Tregs were sorted and expanded in

OpTmizer with 300 IU/ml IL-2 as previously described, for 7 days with beads followed by 7 days of rest. T cell groups were normalized to the same percentage mCherry+ on day 14.

For U87 tumor models, mice were injected on day -7, subcutaneously with 6e5 U87

CBG-GFP on the left flank and 6e5 U87-CD19 GBG-GFP on the right flank. 2e6 CAR T cells or the same cell number of UT T cells were injected IV. on day 0 with 5 mice per group. In the groups specified recombinant human IL-2 (Peprotech) was administered

IP. at 8 µg/mouse 3 times weekly. Tumor burden was regularly monitored using an Ami spectral imaging apparatus and analyzed with IDL software v. 4.3.1 following an IP. injection of D-Luciferin substrate solution (30 mg/mL) 2 times a week. Animals were euthanized as per the experimental protocol. U87 tumors were removed on day 14 post

CAR injection for paraffin embedding.

For skin xenograft models, human skin was obtained from Massachusetts General

Hospital according to IRB protocol. On the same day, skin samples were harvested at the depth of the dermal boundary using a standard DermaBlade (Medline) and washed in RPMI media supplemented with 10% FBS. Xenograft procedure was performed as outlined in animal experimental protocol approved by the Institutional Animal Care and

Use Committee. Mice were handled using standard aseptic technique. Mouse skin on the dorsum that was approximately twice as large as the graft was removed, immediately followed by creating micro wounds to the fascia using forceps to aid skin

122

engraftment. The human donor graft was then secured to the mouse with silk ligature followed by sterile dressing. After 7-10 days the sutures were removed and antibiotic ointment was applied to the grafts 3 times a week until grafts were fully healed. Once healed, approximately 30-60 days’ post-surgery, mice were injected IV. (day 0) with 2e6

CAR-Tregs, 2e6 CAR Teff cells or both (4e6 CAR T cells total). In the groups specified

IL-2 was administered IP. at 8ug/mouse 3 times weekly. Grafts were photographed 3 times a week. Mice were euthanized on day 14 and tissue was harvested and fixed for paraffin embedding. Graft surface area was measured as from photographs using

SketchAndCalc™ software.

Statistics

Data are presented as means ± SEM as stated in the figure legends. Unless otherwise noted, groups were compared using a two-tailed paired student’s T test. All statistical analyses were performed with Prism software version 7.0 (GraphPad).

Study Approval

Healthy donor leukopaks were obtained from the Blood Transfusion Services at

Massachusetts General Hospital under an IRB-approved protocol. Human skin was obtained from Massachusetts General Hospital according to an IRB-approved protocol.

All subjects provided written informed consent for the use of their discarded tissues for research. Tumor and xenograft procedures were performed as outlined in animal experimental protocols approved by the Institutional Animal Care and Use Committee.

123

3.5 Acknowledgements

I would like to acknowledge Rebecca C. Larson (Maus Lab, Massachusetts General

Hospital) who assisted with some of the in vitro studies and Treg sorts, and Amanda

Bouffard (Maus Lab, Massachusetts General Hospital) who performed all animal injections and imaging and monitoring. Erik Schiferle taught us how to do the skin xenograft models (Demehri Lab, Massachusetts General Hospital) and Dr. Curtis L.

Cetrulo (Massachusetts General Hospital) provided us with the donor skin for these models. Amanda Bouffard and Bryan Choi (Maus Lab, Massachusetts General

Hospital) assisted with grafting donor skin onto mice. Lauren S. Riley (Maus Lab,

Massachusetts General Hospital) assisted with lentiviral production, Dr. Shadmehr

Demehri (Massachusetts General Hospital) provided the clinical assessment of xenograft H&E staining and photographs. I would like to acknowledge the

Massachusetts General Hospital Pathology Core for performing the IHC histology and the Specialized Histopathology Core at Charlestown Navy Yard, Massachusetts

General Hospital for performing the TUNEL staining on the skin graft sections. We would also like to thank the Massachusetts General Hospital Flow Cytometry Core for the sorting services. We would like to thank David Ting (Massachusetts General

Hospital) and Anupriya S. Kulkarni (Ting lab, Massachusetts General Hospital) for performing RNAscope on the xenograft tissues. Bruce R. Blazar (University of

Minnesota) and Marcela V. Maus (Massachusetts General Hospital) provided invaluable discussions and Marcela V. Maus was closely involved with all parts of this study.

124

Chapter 4: General Discussion

A deep understanding of the effects of costimulation domains on CAR T cells will enable further rational engineering to improve adoptive T cell therapies. Logical CAR design and further T cell engineering has the potential to arm T cells with the appropriate properties to eliminate cancer cells even within harsh tumor microenvironments or, in the case of CAR-Tregs, provide long-lasting, tissue-specific, immune suppression. Now that patients are getting CAR T cells as drugs, and these genetically modified T cells can persist in a human for a several years, knowing how CARs can impact T cell biology is important. In this dissertation, we showed that the effect of engineered costimulation is not only complex but also cell-type specific. We demonstrated that there are transcriptional differences in conventional CAR T cells bearing distinct costimulation domains that can ultimately influence a T cell’s fate. The effect of costimulation on a

T cell’s phenotype was further highlighted in CAR-Tregs, where 4-1BB costimulation inhibited the suppressive properties of CAR-Tregs.

In Chapter 2, we identified hundreds of transcriptional differences between BBz and 28z

CAR T cells in CD4+ and CD8+ T cell subsets. We found that some of the differences in expression between T cells modified with different CAR constructs exist even before antigen-directed CAR activation, highlighting a clear transcriptional effect from the expression of a CAR in a T cell. Even more striking were the transcriptional effects following activation through a CAR. We compared the T cells’ transcriptional responses to CAR signaling through BBz versus 28z both 4 and 24 h after the addition of antigen- 125

expressing Nalm6 cells. These transcriptional profiles define antigen-dependent and antigen-independent CAR signaling pathways that ultimately determine CAR T cell fate.

4.1 Differential effects of costimulation in CAR Teff cells are supported by the known mechanisms of natural 4-1BB and CD28 T cell signaling

As we described in Chapter 1, there are differences between the CAR and TCR signaling properties. Research is limited in detailing to what degree the effects of different costimulation domains in a CAR reflect the known effects of natural ligand- costimulatory receptor signaling in T cells. We found that the comparison of CD28 versus 4-1BB costimulation in CAR T cells did recapitulate some known aspects of costimulation biology in unmodified T cells. We independently identified many properties caused by 4-1BB or CD28 costimulation in CAR T cells that support past literature describing the effect of natural CD28 and 4-1BB signaling. For instance, years of research into CD28 costimulatory pathways has led an understanding of its unique role in promoting IL-4 production, leading to the enhanced differentiation of TH2 helper cells19. Our unbiased gene-set enrichment analysis of 28z versus BBz CAR signaling identified a significant enrichment of genes involved in the early TH2 polarization of

CD4+ T cells. These genes were upregulated in CD4+ 28z CAR T cells compared to

BBz CAR T cells. We also found that there were increased IL-4 protein levels and IL4

RNA expression in 28z CAR T cells. We identified significant differences in the expression of BCL2-related genes such as BCL2A1, BCL2L11, and BAD, which were modulated in such a way as to inhibit T cell apoptosis. This is in alignment with what is

126

known about CD28 costimulation and its role in inducing the expression of anti- apoptotic genes to improve T cell resistance to activation-induced cell death20. Previous literature has found CD28 signaling can increase glucose uptake and enhance glycolysis in activated T cells18,21. This supports research, by other groups on the role of costimulation in CAR T cells, which confirmed increased glucose uptake and glycolysis in 28z CAR T cells7.

Past research has shown that activating 4-1BB in CD4+ T cells using 4-1BBL induces T

36 + cell polarization to TH1 cells . Indeed, TCR/4-1BB-activated CD4 T cells were shown to express more TH1 cytokines such as IFNg and TNFa. We found that BBz CAR T cells had increased the expression of a statistically-significant number of TH1 polarizing genes compared to 28z CARs by GSEA of the ranked gene-expression differences between BBz and 28z. Furthermore, we revealed IFNg and TNFa signaling pathway genes were significantly upregulated in BBz CAR T cells.

As mentioned previously, 4-1BB and 4-1BBL are known to be also expressed on dendritic cells. The triggering of 4-1BB on DCs is recognized to increase their secretion

30 of IL-12, a TH1 polarizing cytokine . Interactions between 4-1BB and 4-1BBL induce bidirectional signaling252. 4-1BBL signaling in monocyte-derived DCs induces DC maturation with the classic upregulation of costimulation ligands such as B7 molecules and antigen-presenting MHC class II, similar to TNFa and IFNg signaling in DCs. Like

DCs stimulated through 4-1BB, 4-1BBL intracellular signaling in DCs increases the

127

secretion of TH1-inducing cytokines IL-12 and IFNg and decreases secretion of IL-10.

Therefore, 4-1BBL-matured DCs also preferentially induce TH1 responses. This data illustrates the synergistic effect of 4-1BB and 4-1BBL signaling between APCs and T cells253. In summary, the effects caused by CAR-mediated CD28 and 4-1BB costimulation that we and others7 have uncovered recapitulate many of the findings on the effects of natural CD28 and 4-1BB signaling in T cells. If one thinks of 4-1BB and

4-1BBL signaling across both APCs and the T cells that express the 4-1BB receptor, we get a more holistic picture of the 4-1BB - 4-1BBL response as one that promotes TH1

+ development. This supports our findings of increased TH1 polarization in CD4 BBz CAR

T cells.

4.2 The overlap of 4-1BB signaling in T cells and other cell types

In addition to T cells, 4-1BB is known to be expressed on various cells types27-30. 4-1BB signaling was shown to induce HLA class II upregulation on the surface of dendritic cells254. On endothelial cells, 4-1BB agonist activation increased the surface expression of adhesion molecules VCAM1 and ICAM1. Our transcriptional data found BBz CAR T cells upregulated many HLA class II molecules as well as both VCAM1 and ICAM1.

Whether these genes are expressed after natural 4-1BBL/4-1BB interactions in T cells is an interesting consideration. Perhaps the supraphysiological signaling through 4-1BB in CAR T cells can upregulate certain genes known to be induced by 4-1BB signaling in other cells types. We are still investigating whether these genes, when upregulated in T cells, influence T cell properties. For instance, it would be clinically relevant to

128

understand whether VCAM1 and ICAM upregulation by BBz influence T cell adhesion and trafficking to tumor sites and whether HLA class II molecules on BBz CAR T cells are involved in presenting tumor cell antigen or activating endogenous CD4+ T cell responses.

4.3 4-1BB and memory CD8+ T cells

In addition to its polarizing properties, 4-1BB is known to be important for promoting the formation of long-lived memory T cells35. While we did not specifically investigate this, others have reported increased central memory in CD8+ BBz CAR T cells7. Our data suggests a role for 4-1BB in generating more effective CD4+ T cell help for CTLs. CD4+

T cells are critically important for CD8+ T cell function, illustrated by the observation that when CD4+ T cells are absent, memory CD8+ T cells exhibit impaired cytokine production, decreased persistence, and reduced ability to lyse virally infected cells255-

257. This reliance on CD4+ T cell help remains important for CAR T cells. For example, human CD8+ CAR T cells activated with anti-CD3/anti-CD28 beads and IL-2 do not expand nearly as efficiently in isolation as they do when CD4+ T cells are added in the culture with them 37,258. Therefore, the quality of CD4+ T cell help is likely to significantly affect CD8+ CAR T cell responses.

4-1BB signaling is known to more effectively activate CD8+ T cells than it activates

CD4+ T cells38,39. Furthermore, research has shown that 4-1BB CAR T cells improve the proliferation of CD8+ T cells7. However, it has not been well understood to what degree

129

this is caused by 4-1BB signaling through CD8+ T cells alone versus the indirect effect of 4-1BB improving CD4+ T cell help — or a combination of these two effects. We believe that our data suggests a model in which 4-1BB signaling in CAR T cells affects the quality of CD4+ T cell help provided to CD8+ T cells. We have already mentioned that 4-1BB induces a program of early TH1 polarization. TH1 cells are known to be the key cells supporting NK and CTL activity259. Furthermore, we found that BBz CD4+ T cells express cytokine IL-21 and that both BBz CD4+ and CD8+ CAR T cells upregulated

IL-21R. Previous research has shown that IL-21R signaling can activate granzyme B and perforin expression in CD8+ T cells260. IL-21 signaling has been shown to be critical to CD8+ T cell memory formation255. Much research has gone into the effect of 4-1BB in

CD8+ CAR T cells. Here, we propose that some of this effect may be through an indirect mechanism of 4-1BB in CD4+ cells, enabling them to become more efficient at supporting CD8+ T cells’ granule production and cytolysis as well as the induction of a

CD8+ memory program.

Future experiments could expand first-generation CD8+ T cells in co-culture with either

28z or BBz CD4+ with antigen-expressing cells. Other experiments could look at the effect of these combinations of first-generation CD8 with second-generation CD4 in in vivo tumor models. Both in vitro and in vivo combination studies would aim to determine the memory formation and functionality of CAR-z CD8+ cells through several lines of investigation. These could include: flow cytometry for classic memory surface markers, intracellular cytokine staining to look at recall cytokine responses by CD8+

130

cells, and isolation of CD8+ T cells to investigate target-specific cytolysis. Finally, additional RNAseq or other transcriptional profiling methods could be used to look for defined signatures of CD8+ T cell memory. By removing the effect of CAR-costimulation in CD8+, these experiments could illuminate the role of CAR costimulation domains in affecting the quality of CD4+ T cell help for the formation and expansion of memory

CD8+ T cells.

4.4 Ligand-independent signaling in CAR T cells

To better understand the transcriptional effect of ligand-independent signaling in CAR T cells, we determined the DE genes between our functional CARs and Dz CAR T cells at rest. We defined a signature of ligand-independent signaling from the CD3z chain of

CAR T cells as the intersection of the DE genes from each of the three functional CAR

T cells (z, BBz, 28z) compared to Dz. We used this criteria since the CD3z signaling domain was the only component that all three functional CAR T cells shared and that Dz

CAR T cells lack.

Frigault et al. and Long et al. reported that antigen-independent signaling, or tonic signaling, from CAR T cells with specific CAR constructs ultimately led to T cell exhaustion88,117. This tonic signaling was extreme and was accompanied by antigen- independent CAR T cell proliferation and cytokine secretion. Others have described ligand-independent CD3z phosphorylation from different CAR constructs that do not induce T cell exhaustion. For instance, PSCA-specific CARs with added spacer-hinge

131

regions increased cell size and CD25 expression but do not induce exhaustion84. We detected a transcriptional signature of ligand-independent CD3z signaling in CAR T cells previously described as having undetectable tonic signaling88. The accumulation of these described effects suggests that tonic signaling comes in a variety of “flavors,” leading to diverse effects on CAR T cell properties.

The TCR on a T cell is thought to signal constitutively at a very low level, sometimes described as “tickling” the T cell throughout its life261. Initially, the ability of a developing

T cell’s TCR to recognize self-peptide MHC enables their positive selection and is necessary for the maturation of T cells in the thymus262. Low-affinity self-recognition is also understood to play a role in the survival of naïve T cells when other survival factors are limited263. It is unknown how similar ligand-independent signaling by CARs is to

TCR “tickling,” but the concept of TCR tickling raises the possibility that a low amount of ligand-independent signaling may be beneficial or at least neutral to a CAR T cell’s survival. Perhaps only extreme tonic signaling that results in antigen-independent cytokine secretion and proliferation can induce exhaustion. Furthermore, such exhaustion may only result from certain costimulation domains, since 4-1BB containing

CARs—even with extreme tonic signaling—have not been described as exhausted117.

Another potential explanation is that exhaustion in GD2 28z CARs is from tonic signaling of the CD28 domain rather than from CD3z tonic signaling and a lack of

4-1BB, though this has not been investigated. A valuable line of investigation would be to see whether the combination of 4-1BB signaling with 28z in a third-generation GD2

132

CAR can reduce this exhaustion phenotype, as that would point to 4-1BB as preventing exhaustion rather than CD28 tonic signaling causing exhaustion.

The ligand-independent CD3z signaling signature that we found suggests some functional implications of the expression of a CAR in a T cell. For instance, ligand- independent signaling upregulated the expression of CCL3 and CCL4. Macrophage inflammatory protein 1 family members (CCL3 and CCL4) are chemokines that recruit antigen-presenting cells to sites of inflammation264. We speculated that ligand- independent signaling in a CAR T cell may be able to recruit monocytes and macrophages to the CAR T cells as soon as they are administered to a patient. The role that these chemokines may have in cytokine release syndrome could be an interesting route of investigation. Furthermore, we found that ligand-independent signaling from

CAR T cells upregulates GZMB, encoding granzyme B—the serine protease most commonly found in the granules of CTLs that is responsible for tumor cell cytolysis. This finding could also affect CAR T cell anti-tumor function. In conclusion, the data point to differential effects of tonic signaling that could be due to different domains but may also be due to the degree of tonic signaling. More research is needed to tease apart the effect of low versus high amounts of tonic signaling from each signaling module in a

CAR construct to better understand how to engineer potent CAR T cells in the future.

4.5 Antigen-specific CAR-Treg cells.

There are several disease situations in which tissue-specific local suppression of an aberrant immune response would be more appropriate than global immune 133

suppression. In Chapter 3, we showed that Tregs traffic in vivo specifically to sites where their target-antigen is being expressed. We also showed that CAR-Tregs are potent suppressor cells when activated through their CAR. We hypothesize that a “halo” of tissue-specific protection could be generated around an organ of choice, such as a pancreas or liver, by CAR-Tregs trafficking to an antigen-expressing site and being activated locally by that antigen to then suppress the local effector response. This effect could be extremely useful for tolerance to organ transplants, prevention of type 1 diabetes, and resolving inflammatory bowel disease, multiple sclerosis, and autoimmune hepatitis among others.

Though Tregs need to be activated in order to mediate cell suppression, once activated, they can inhibit responder T cells irrespective of whether they share the same antigen specificity, known as bystander suppression163. Based on what is known about Tregs’ mechanisms of suppression, we predict that once activated at a site of inflammation,

CAR-Tregs may affect local Teff cells’ activity by releasing extracellular adenosine nucleosides that shift the Teff cell metabolic environment to one that dampens immune activation164. Activated CAR-Tregs could also act as an IL-2 sink, limiting IL-2 availability for Teff cells, and we showed that CAR-Tregs can secrete immunosuppressive cytokine IL-10. We would also predict that CAR-Tregs can make additional inflammatory cytokines IL-35 and TGFb, all of which promote Teff cell suppression at a local site.

134

4.6 Indirect Treg suppression of Teff cells

We further predict that the power of adoptive CAR-Tregs will go beyond direct Treg suppression of Teff cells. Tregs can suppress Tconv cells directly as we just described or indirectly through their effects on the maturation and phenotype of APCs164. Tregs are constantly interacting with dendritic cells in vivo175,176, and during this process,

APCs such as DCs can be conditioned to be more inflammatory or more tolerance- inducing. We hypothesize that redirected CAR-Tregs could polarize macrophages and dendritic cells to be more immunosuppressive. This immune tolerance could propagate further, since toleragenic APCs can induce local T cells to polarize into iTregs265 with the aim of tipping the balance of the tissue environment from inflammatory to tolerogenic and reestablishing tissue homeostasis without affecting the host immune system’s ability to fight infections.

Our models focused on the direct suppression of Teff cells by Tregs. However, it would also be interesting to investigate indirect mechanisms of Treg suppression, which would require experiments involving more components of the immune system. In our in vitro experiments, MLRs were used to test Treg suppression on T cells using only irradiated target tumor cells or anti-CD3/anti-CD28 beads for T cell activation. In vivo, we used

NSG mice, which do not have a functional myeloid compartment. Any remaining myeloid cells in the NSG mice would be murine-derived, thus limiting crosstalk between

T cells and innate cells to only molecules that are highly conserved across species.

PBMCs were not injected with CAR T cells in our skin-graft models, since PBMCs

135

cause GvHD in NSG mice. Therefore, it is unlikely that our models were influenced by

Tregs mediating indirect suppression through their effect on APCs. However, there is a chance that Langerhans cells and conventional DCs were carried over within the skin graft and that these APCs could be influenced by CAR-Treg cells that traffic to the skin graft. Even if the Tregs did influence APCs to be immunosuppressive, many of the tolerance promoting effects of Tregs such as decreasing the expression of MHC and costimulation on DCs are unlikely to affect CAR-Teff cells because the CAR enables the

Teff cells to be independent of DC presented MHC and costimulation for their activity.

In Chapter 3 we found that IL-10 is differentially expressed between CAR stimulated

BBz and 28z Tregs. We did not find any effect of IL-10 on Treg direct suppression of

Teff cells when exogenous IL-10 blocking antibody was added in vitro. However, IL-10 signals through the IL-10R, which is expressed at far higher levels on DCs and macrophages. Furthermore, most literature agrees that IL-10 affects Teff cells through indirect mechanism via IL-10’s effects on APCs only. In future experiments, which we will describe in a later sub-section, we plan to investigate the potential indirect modes of

T suppression.

4.7 CAR-Treg mediated cytotoxicity

Human and murine Tregs can directly kill activated monocytes, DCs, Teff and NK cells as a mechanism of Treg-mediated control of the immune response166. There is good evidence to support that Foxp3+ cells co-express perforin and granzyme B and can

136

degranulate upon TCR activation182. It is well known that the modification of T helper cells to express a CAR resulted in tumor lysis capabilities similar to that of a CTL218.

Therefore, it is not unreasonable that the same would happen to CAR modified Tregs since they already have the cellular machinery to induce cell apoptosis via granzyme and perforin. CD4+ Tregs bind peptide-MHC class II, which is usually expressed on

APCs and not regular tissues, which only express MHC class I. Tregs do not naturally form synapses with tissue cells, but a CAR gives them this ability. We found that CAR-

Tregs kill tumor cells that are expressing the target by antigen-specific cytolysis. We confirmed this finding with three different scFvs, all against different antigens. Indeed,

CAR-Tregs kill in vitro by a target-antigen-dependent mechanism. This cytotoxicity could not be eliminated by a lower affinity scFv. In vitro, Foxp3+ Tregs degranulate in a precise antigen-specific manner. We also showed that activated CAR-Tregs upregulate granzyme B compared to Dz Tregs. Furthermore, granzyme B inhibitors could reduce

CAR-Treg mediated cytolysis. In vivo, CAR-Tregs trafficked to EGFR-expressing skin grafts, but in this case, CAR-Tregs did not induce large amounts of epithelial cell apoptosis, though we did detect very low levels of cell death by TUNEL staining in the

EGFR-CAR Treg compared to the irrelevant antigen CD19-CAR Treg. This finding could be difficult to detect perhaps because Tregs do not have the same serial killing properties as Tconv cells. Alternatively, there could be higher levels of Treg cytolysis, but the epithelial cell layer regenerates at a rate that is faster than the rate of Treg induced apoptosis. Another possible explanation is that epithelial cells are in some way

137

more resistant to Treg-induced cytolysis than the tumor cells we used in our in vitro models.

CD8+ T cells can release their granules, kill an infected cell, then move to a new target and kill again, often referred to as “serial killing.” It has yet to be determined if Tregs have the capability to induce cell cytolysis in multiple cells. Our data suggest that the level of cytolysis is low in vivo. If the turnover rate of the tissue is high, such as in the gastrointestinal epithelium266, low levels of Treg-mediated tissue apoptosis may be tolerable. But in tissue settings of low cell-turnover and minimal regeneration, such as in the CNS267, even low amounts of toxicity by Tregs could be devastating.

CAR synapses are different from TCR synapses102. CAR synapses are smaller, less- organized clusters, but they induce cell degranulation faster. We do not know what mechanism inhibits natural Treg cytolysis after synapse formation. There are data suggesting that Tregs interact with APCs in short interactions compared with the longer more stable interactions of APCs with Teff cells. However, the CAR and the nature of the synapses that CAR-Tregs form with target cells may affect the type and duration of interactions between target cells and Tregs, potentially increasing the chance of Treg- mediated target-cell apoptosis.

There are a few reports of cytotoxicity assays using CAR-Tregs against target- expressing cells4,213. All of them observe lower cell killing by CAR-Tregs compared to

138

Tconv cells, which is reflected in our data as well. This led to these papers concluding that CAR-Tregs do not show signs of target-cell lysis. However, unlike previous studies, we have included comparisons of target cell cytolysis by functional CAR-Tregs to three types of controls: untransduced, Dz, and target non-specific. The controls enabled us to determine that CAR-Tregs do indeed cause target-specific cytolysis, albeit at a low level. It is possible that this explains why the effect of CAR-Treg-mediated cytolysis has not been previously reported. Therefore, we find evidence that Treg cells sorted by

CD4+, CD25++, CD127low display a low but significant level of target cell-specific lysis that is — at least partly — dependent on the granzyme B/perforin pathway. However, this cytotoxicity is more pronounced in vitro than in vivo. These findings warrant further research and consideration as CAR-Treg therapies move to the clinic.

4.8 4-1BB inhibition of CAR-Treg suppressive function

The role of 4-1BB stimulation in Treg literature is controversial. Some reports suggest that anti-4-1BB antibodies can inhibit Treg suppressive function, while others indicate that 4-1BB activation on Tregs induces their proliferation44,45. Although 4-1BB has been shown to support the division and survival of Tregs, in mice 4-1BB is not necessary for

Treg development, since 4-1BB-knockout mice have normal Treg numbers45,268. CAR- mediated 4-1BB costimulation inhibited Treg suppression in our in vitro MLRs. This was not due to the lack of CD28 costimulation, since z CAR-Tregs were able to suppress

Teff responses just as well as 28z CAR-Tregs, thus implying that 4-1BB directly inhibits

Treg function. In line with our data, others have found that stimulation of 4-1BB reduces

139

Treg suppressive function both in vitro and in GvHD mouse models248. Indeed, one study even found 4-1BB agonist antibody could reprogram Tregs to eliminate immunogenic tumor cells189. Another research paper found that anti-4-1BB inhibits iTregs’ suppressive function44. Together, this data may not be as contradictory as it seems. Perhaps 4-1BB can induce Treg proliferation and inhibit Treg suppressive function for the duration of high 4-1BB costimulation. Then, during the resolution of inflammation, 4-1BBL levels decrease, and the now-larger numbers of Tregs can suppress the remaining Teff cells.

The studies describing 4-1BB inhibition of Treg function also claim to find no effect on

Foxp3 stability by 4-1BB signaling. This result aligns with our finding that 4-1BB costimulation in CAR-Tregs does not affect Foxp3 stability. The data here illustrate that

Foxp3 stability does not solely dictate Treg suppressive functions. Our data and others support the model in which 4-1BB may induce unique negative signals into Tregs that counteract their suppressive mechanisms without affecting Foxp3 stability. To date, no one has determined a direct mechanism of reducing Treg function through 4-1BB signaling. Our results indicated that 4-1BB containing CAR-Tregs have reduced ability to produce several molecules involved in Treg suppressive mechanisms and moreover may be metabolically different at baseline to 28z and z CAR Tregs. In addition, BBz

CAR-Tregs displayed dysfunctional cell-cell interaction kinetics. Though no one of these differences alone seems to account for the loss of 4-1BB suppression in vitro, maybe

140

the combination of these differences results in reduced antigen-specific suppression of

Teff cells by BBz CAR-Tregs.

4.9 Future Directions

Though we have characterized several aspects of costimulation in CAR Teff and Treg cells, many questions about the effect of engineered costimulation in T cells remain.

This last subsection will cover a selection of the many avenues of investigation that remain in order to develop a deeper understanding of costimulation and mechanisms of

Treg suppression.

4.9.1 Further interrogating the transcriptome of CAR T cells

In Chapter 2, we compared 28z and BBz T cells by their differentially expressed genes as a result of CAR signaling after 4 or 24 h with CD19-CAR stimulation. Fewer than 10 genes were significantly upregulated by 28z CAR T cells. This result was unexpected given the profound impact of the addition of CD28 costimulation in enabling T cell proliferation and tumor eradication compared to first-generation CAR T cells. A possible explanation for the small number of upregulated genes is that CD28 has been described as mostly amplifying the TCR signaling response, and perhaps CD28 has a reduced role in the upregulation of alternative pathways that can lead to the upregulation of different genes in response to activation. Alternatively, it is possible that because all of the T cells had previously received CD28 stimulation as part of the classically used

141

transduction and expanding process with anti-CD3/anti-CD28 beads, additional signaling through this domain had less of a differential effect on the T cells.

Though bead-expanded, lentivirally transduced CAR T cells more fully recapitulate the state of CAR T cells before they signal in response to antigens in patients, to complement these findings, it would be interesting to use electroporation methods to deliver DNA encoding the CAR. Unlike lentiviral transduction, electroporation does not require T cell activation. In this way, T cells could be activated ex vivo for the first time through their CAR, avoiding TCR/CD28 activation. If we were to execute this experiment, we would perform RNAseq 4 and 24 h after stimulating the electroporated

CAR T cells. We would repeat the differential gene expression analysis between 28z versus z or 28z versus BBz with this data to see if it aligned with the previous experiments, or if elecroporated 28z CAR signaling induces the upregulation of multiple different genes and pathways. This approach is different from the studies by Kawalekar et al. because we would be using RNAseq to study transcriptional effect of CARs in T cells and because we would be using a shorter activation (hours versus weeks) before sorting the cells for sequencing7. The addition of this proposed analysis with the analyzed data described in Chapter 2 from bead-expanded, lentivirally transduced CAR

T cells would allow us to interpret a more complete picture and to better isolate the transcriptional effects of CD28 costimulation on CAR T cell signaling.

142

Another way we could enhance our transcriptional data is through single-cell RNA sequencing. It cannot be assumed that of the population of CAR T cells we sequenced, all cells were homogeneous and at the same stage of activation. Single-cell RNA sequencing would allow us to interrogate the heterogeneity of CAR T cell activation. We would be able to determine the activation stages across a population of CAR T cells activated through their CAR construct. We could then look at differentially expressed genes between only highly activated BBz versus 28z CARs to see if this provides clearer transcriptional differences. In Chapter 2, we determined that 4-1BB

+ costimulation initiated TH1 polarization in CD4 T cells. Single-cell sequencing could be used to understand whether this polarization is represented by a shift in the all of the

+ CD4 CAR T cells toward a more TH1-like state, or if 4-1BB signaling simply increases the percentage of TH1 cells in a more heterogeneous population of various helper T cell subsets.

In addition to further sequencing of 4-1BB and CD28, we hope to investigate the transcriptional effects of other costimulation domains such as ICOS, OX40, and GITR on CAR T cell state, as they make their way to the clinic. New research is teasing apart the role of individual signaling motifs in CD28 by mutating the various motifs269. It would be interesting to take a transcriptional approach to look at the downstream effect of these mutations.

143

We would also like to use this sequencing data as a starting point for understanding the differences between FDA-approved CAR therapies used in the clinic. It would be exciting to do RNA sequencing on CAR T cells signaling from responding patients with

CD19 axicabtagene ciloleucel 28z CAR product versus the tisagenlecleucel BBz CAR product. We could investigate whether we can find new differences and/or identify similar expression profiles in patient-derived signaling CAR T cells compared to in vitro derived differential gene profiles.

4.9.2 Alternative models of tissue-specific suppression by CAR-Tregs

We have designed and produced an adeno-associated virus to induce liver-specific expression of human CD19 and click beetle green using a mouse liver-specific promotor. We plan to deliver this AAV to induce mouse livers to express CD19 and click beetle green. We will inject CD19 CAR-Tregs into mice with CD19+ livers and follow the liver toxicity using liver function tests and measuring luminescence over time after the administration of luciferase substrate.

Mice injected with human PBMCs develop GvHD with skin inflammation, splenomegaly, and lymphocyte infiltration into the liver, which leads to observable hepatocellular necrosis270. We propose to test Treg tissue-specific suppression by injecting CD19 28z

CAR-Tregs or control, non-specific antigen CAR-Tregs. We would then inject human

PBMCs into mice with liver-expressing hCD19-CBG. We will look at GvHD toxicity to determine if we can create a “halo” of suppression by CD19 CAR-Tregs around the

144

liver. We hypothesize that the liver will not show GvHD hepatocellular necrosis when

CD19 CAR-Tregs are administered but other signs of GvHD will still develop, whereas all GvHD pathologies will be observed in mice given control CAR-Tregs, demonstrating that CAR-Tregs can mediate tissue site-specific suppression. This model will also enable us to look at the Treg effects on APCs, since whole PBMCs will be administered into the mice rather than CD3+ cells in isolation.

4.9.3 Designing safer tissue antigen-specific CAR-Tregs

To improve the safety of CAR-Tregs, in future research we plan to test different CAR designs that will minimize toxicity caused by cytolysis of antigen-expressing cells by activated CAR-Tregs. Therefore, CAR -Tregs will have tissue specificity without the potential to destroy antigen-expressing cells. We propose two avenues of further research. The first is to knock out essential molecules in the granzyme pathway using

CRISPR/Cas9 and determine whether Tregs without cytotoxic function can still suppress via other mechanisms but without lysing target-expressing cells. The second avenue is to develop a combinatorial CAR system with synthetic Notch (synNotch) receptors such as the one described by Roybal et al.271. SynNotch receptors can be attached to scFvs as intracellular domains. Upon ligand recognition, synNotch receptors are cleaved to release a transcription factor from the cytoplasmic tail that can enter the cell nucleus and transcribe a second CAR receptor. In our proposed CAR-Treg setting, the first scFv would be tissue-specific and would have a synNotch intracellular domain, directing the Treg to a tissue site and inducing cleavage of a synthetic transcription

145

factor from the synNotch receptor. This transcription factor would upregulate a second- generation 28z CAR with an scFv that binds an APC-specific marker such as CD14. In this way, Tregs can traffic to the tissue site to start the activation process without forming a synapse that includes CD3z ITAM signaling when bound to tissue cells.

Tissue antigen-to-CAR activation will induce the expression of a fully functional CAR that can signal when bound to a monocyte, macrophage, or DC. If the Treg kills a few inflammatory APCs at the site of inflammation, we do not predict that it will increase inflammatory disease, and it may even help reduce disease by eliminating APCs that are mediating pro-inflammatory Teff cell responses. Furthermore, the APC-binding CAR would enhance Treg interactions with APCs. If they do not kill the APC, this could be a way to generate Treg suppression by influencing the APC’s phenotype to promote tolerance. The use of synNotch receptors and CRISPR/Cas9 are exciting new approaches to improving the safety and functionality of tissue-specific CAR-Tregs.

4.10 Concluding Remarks

Together, the body of work presented in this dissertation advances our knowledge of engineered costimulation and its effect on Teff and Treg cells. We identified transcriptional differences between BBz and 28z CAR T cells, which led to the discovery of differences in cytokines and cytokine receptors produced by CAR T cells modified with different costimulatory domains. Furthermore, we showed a signature of early TH1 polarization in CD4+ BBz CAR T cells, in contrast with 28z CARs, which contained a signature for TH2 polarization. In CAR-Tregs, antigen-dependent 4-1BB signaling

146

reduced the immunosuppressive functions of Treg cells, and signaling through 28z

CARs increased the Tregs’ ability to produce IL-10 and to express coinhibitory receptor

CTLA4.

We hope that these findings will be transformative in creating better adoptive cell therapies, either for fighting cancer or for establishing immune tolerance.

147

Appendix

Supplemental Table A2.1. Differentially expressed genes in BBz versus 28z CAR T cells at 0 , 4 and 24 h after the addition of irradiated Nalm6 cells. Padj value < 0.1

(Benjamini & Hochberg method, calculated by DESeq) CD-combined – CD4+ and CD8+

T cells samples were combined for the analysis. Higher expression of BBz than 28z assigned positive logFC. BaseMean – averaged gene transcript count across samples.

LogFC – log (base2) fold change. padj – adjusted p-value

148

Supplemental Table 2.1

CD combined – no stimulation CD combined – 4 hours CD combined – 24 hours Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj IL12RB2 277.90 0.88 2.23E-09 2.83E-05 ENPP2 112.75 1.66 9.27E-16 1.36E-11 GJB2 276.83 1.40 1.74E-12 2.21E-08 JUN 71.93 1.41 4.92E-09 3.12E-05 ENOX1 38.64 1.51 9.76E-14 7.14E-10 NTRK2 90.55 1.37 7.89E-11 5.02E-07 EGR1 135.61 1.25 2.89E-08 1.22E-04 DDIT4 1679.48 1.06 2.31E-08 1.13E-04 JUNB 1048.37 1.03 3.34E-10 1.42E-06 CORO7-PAM16 16.32 1.25 2.31E-07 7.30E-04 JUNB 1418.50 0.78 8.58E-08 2.51E-04 DGAT2 28.53 1.26 9.61E-10 3.05E-06 ARID5A 628.90 0.96 4.36E-07 1.10E-03 CIITA 408.87 1.04 7.67E-08 2.51E-04 AMICA1 1487.21 -0.75 6.68E-09 1.44E-05 WNT5B 15.10 1.11 1.67E-06 3.52E-03 DMD 176.23 1.09 1.03E-07 2.51E-04 MSC 279.07 1.23 6.80E-09 1.44E-05 CDKN1A 437.24 0.81 2.24E-06 4.05E-03 GJB2 357.82 1.06 1.63E-07 3.41E-04 SH3BP5 220.86 1.06 9.49E-09 1.72E-05 JAKMIP1 472.08 0.76 2.58E-06 4.08E-03 ARHGAP10 110.52 1.00 3.80E-07 6.94E-04 ELL2 334.11 1.02 1.10E-08 1.75E-05 ENPP2 71.10 1.12 3.20E-06 4.48E-03 HLA-DQA2 590.33 1.03 4.88E-07 7.94E-04 DNAJC6 34.01 1.18 2.32E-08 3.28E-05 JUNB 488.86 1.00 3.54E-06 4.48E-03 GNA15 859.52 0.74 8.49E-07 1.19E-03 IL12RB2 417.28 0.60 7.84E-08 9.97E-05 CHRNA6 55.84 1.04 5.98E-06 6.88E-03 EGR1 180.54 0.92 9.38E-07 1.19E-03 OAS3 789.59 0.59 8.63E-08 9.98E-05 C1orf56 243.41 -0.84 7.38E-06 7.79E-03 JUN 102.01 0.93 9.78E-07 1.19E-03 G0S2 64.03 1.12 1.51E-07 1.60E-04 FAIM3 685.92 -0.94 1.21E-05 1.13E-02 LOC100129034 47.25 -0.87 2.39E-06 2.52E-03 HLA-DQA2 529.13 1.05 1.95E-07 1.91E-04 FOS 172.24 1.06 1.25E-05 1.13E-02 POU2F2 157.41 -0.81 2.41E-06 2.52E-03 DMD 132.17 1.09 2.15E-07 1.95E-04 MPZL1 119.80 0.94 1.34E-05 1.13E-02 VOPP1 625.77 0.64 2.69E-06 2.62E-03 HLA-DRB6 356.69 0.98 3.47E-07 2.94E-04 VNN2 112.96 -0.97 1.80E-05 1.42E-02 TPM4 4616.52 0.46 3.06E-06 2.80E-03 FUOM 112.00 -0.70 3.97E-07 3.16E-04 MPP7 47.07 -0.99 1.99E-05 1.47E-02 E2F1 69.21 0.76 3.79E-06 3.26E-03 HLA-DRA 6969.62 0.88 4.52E-07 3.19E-04 EVI2A 832.55 -0.55 2.09E-05 1.47E-02 PLAUR 77.17 0.91 4.84E-06 3.94E-03 IL4I1 105.56 1.07 4.42E-07 3.19E-04 DMD 32.85 1.02 2.27E-05 1.52E-02 IL23R 42.63 0.93 5.99E-06 4.62E-03 ENPP2 168.50 0.98 7.84E-07 5.24E-04 CRMP1 6.60 0.96 2.47E-05 1.52E-02 CA2 24.93 0.91 7.21E-06 5.27E-03 P2RY14 47.06 0.95 1.05E-06 6.69E-04 IRF8 19.52 1.01 2.87E-05 1.77E-02 BCL2A1 523.29 -0.84 9.50E-06 6.62E-03 C4orf26 148.48 0.94 1.74E-06 1.04E-03 C4orf26 139.77 0.88 2.94E-05 1.77E-02 HLA-DPB1 2124.15 0.84 1.22E-05 7.57E-03 ADCY1 88.52 1.02 1.80E-06 1.04E-03 GCA 153.30 0.77 3.34E-05 1.92E-02 HLA-DRB5 1112.66 0.88 1.24E-05 7.57E-03 MPZL1 210.77 0.87 2.06E-06 1.14E-03 BATF3 87.94 1.00 3.60E-05 1.98E-02 FILIP1L 154.94 0.89 1.18E-05 7.57E-03 PDE4DIP 808.99 0.64 2.52E-06 1.34E-03 EGR2 38.66 0.99 4.03E-05 2.13E-02 DNAJC6 32.64 0.88 1.31E-05 7.69E-03 LAIR1 305.55 -0.64 2.86E-06 1.45E-03 EGR3 9.42 0.95 6.07E-05 3.07E-02 ATHL1 480.50 -0.74 1.41E-05 7.95E-03 IL23R 36.02 0.99 3.31E-06 1.62E-03 SH3YL1 103.50 -0.85 6.51E-05 3.09E-02 UBAC1 143.89 0.64 1.82E-05 9.50E-03 NFE2L3 95.63 0.76 4.37E-06 2.06E-03 GIMAP2 848.34 -0.49 6.59E-05 3.09E-02 NR5A2 26.90 0.87 1.79E-05 9.50E-03 ADA 481.83 0.87 5.05E-06 2.29E-03 NLN 65.85 0.72 6.84E-05 3.09E-02 NTRK2 323.52 0.88 1.95E-05 9.84E-03 ITPR1 721.74 0.78 5.25E-06 2.30E-03 RPS29 6613.67 -0.43 8.80E-05 3.80E-02 HLA-DRB6 383.18 0.86 2.20E-05 1.07E-02 HLA-DRB5 1052.81 0.88 6.56E-06 2.76E-03 STMN3 66.07 -0.90 9.00E-05 3.80E-02 LZTFL1 26.59 -0.87 2.40E-05 1.13E-02 TMEM165 258.26 0.90 6.79E-06 2.76E-03 LAIR1 436.11 -0.64 9.35E-05 3.82E-02 BTN2A2 160.92 0.73 2.48E-05 1.14E-02 HLA-DPA1 5123.01 0.80 6.95E-06 2.76E-03 ENOX1 14.01 0.80 1.05E-04 3.98E-02 UBE2F 1143.94 -0.50 2.74E-05 1.21E-02 PDE4A 94.22 -0.79 7.50E-06 2.89E-03 ICAM1 275.35 0.61 1.06E-04 3.98E-02 ENPP1 57.66 -0.65 2.83E-05 1.22E-02 HLA-DPB1 2070.98 0.80 9.49E-06 3.55E-03 ANKRD33B 43.61 0.87 1.10E-04 3.98E-02 ANKRD33B 91.63 0.81 3.08E-05 1.29E-02 HLA-DRB1 3596.31 0.82 1.40E-05 4.93E-03 PARP3 147.64 -0.72 1.14E-04 3.98E-02 LRRC32 48.70 -0.85 3.42E-05 1.32E-02 ZFAND5 577.02 0.85 1.39E-05 4.93E-03 ITPRIPL1 123.14 0.72 1.18E-04 3.98E-02 HLA-DRA 7237.39 0.83 3.43E-05 1.32E-02 MINA 699.67 0.69 1.46E-05 5.01E-03 ING4 479.95 -0.69 1.20E-04 3.98E-02 LHFP 45.16 0.86 3.43E-05 1.32E-02 RALB 847.42 0.59 1.51E-05 5.06E-03 ARHGAP10 101.10 0.81 1.21E-04 3.98E-02 HLA-DRB1 3891.74 0.81 4.16E-05 1.56E-02 PRKCDBP 158.64 0.86 1.72E-05 5.61E-03 ZNF672 242.59 0.76 1.23E-04 3.98E-02 ZNF704 25.26 -0.84 4.70E-05 1.63E-02 TMEM178B 48.02 0.91 1.90E-05 6.04E-03 PRDM1 607.02 0.69 1.35E-04 4.28E-02 TXLNG 306.77 -0.48 4.80E-05 1.63E-02 DGCR6L 550.79 -0.50 1.99E-05 6.08E-03 RPL39 7265.01 -0.40 1.40E-04 4.32E-02 ADA 518.05 0.82 4.69E-05 1.63E-02 ARHGEF10 26.81 0.91 2.01E-05 6.08E-03 GJB2 52.49 0.92 1.44E-04 4.33E-02 GCSAM 67.90 0.84 4.61E-05 1.63E-02 ANK3 227.89 0.84 2.31E-05 6.30E-03 FILIP1L 92.12 0.92 1.57E-04 4.53E-02 C4orf26 149.58 0.77 4.91E-05 1.63E-02 TNFRSF8 332.10 0.87 2.36E-05 6.30E-03 ATHL1 309.06 -0.79 1.60E-04 4.53E-02 CTH 32.00 0.83 5.16E-05 1.68E-02 EHD4 306.14 0.68 2.24E-05 6.30E-03 FOXP1 697.19 -0.73 1.61E-04 4.53E-02 ADRBK1 355.88 0.51 6.60E-05 2.02E-02 ARID5A 1030.46 0.78 2.30E-05 6.30E-03 MAPKAPK5-AS1 481.43 -0.49 1.74E-04 4.79E-02 G0S2 83.09 0.80 6.61E-05 2.02E-02 IL21 9.62 0.78 2.36E-05 6.30E-03 BBS2 151.93 -0.75 1.79E-04 4.81E-02 HLA-DPA1 5084.66 0.81 6.61E-05 2.02E-02 SPECC1 70.92 0.71 2.38E-05 6.30E-03 ALPK2 10.74 0.75 1.84E-04 4.84E-02 CD74 26227.53 0.69 6.87E-05 2.05E-02 CIITA 530.74 0.81 2.58E-05 6.56E-03 AMICA1 1918.97 -0.62 1.95E-04 4.99E-02 IL18RAP 81.80 -0.82 7.16E-05 2.09E-02 CTTNBP2NL 49.54 0.90 2.55E-05 6.56E-03 CDCP1 85.56 -0.84 1.99E-04 4.99E-02 ULBP2 14.34 0.75 7.87E-05 2.21E-02 GCSAM 83.51 0.89 2.94E-05 7.33E-03 HBEGF 17.84 0.90 2.01E-04 4.99E-02 F8 54.32 0.75 7.85E-05 2.21E-02 SH2D1A 1698.72 -0.52 3.59E-05 8.79E-03 SULT1B1 30.29 -0.90 2.08E-04 5.02E-02 HLA-DOA 266.79 0.81 8.15E-05 2.25E-02 JUN 167.92 0.73 4.21E-05 1.01E-02 LIF 34.30 0.88 2.10E-04 5.02E-02 ARNTL2 170.24 0.64 8.73E-05 2.30E-02 BIRC3 2844.67 0.61 4.53E-05 1.07E-02 CDK6 769.45 0.72 2.14E-04 5.03E-02 RNF19B 63.25 0.69 8.61E-05 2.30E-02 EMC8 165.14 0.86 4.71E-05 1.09E-02 C16orf54 1320.99 -0.51 2.28E-04 5.25E-02 IL4I1 218.73 0.76 8.79E-05 2.30E-02 ARHGAP10 124.94 0.82 6.51E-05 1.48E-02 EVI2B 1752.86 -0.44 2.53E-04 5.62E-02 TMEM178B 38.27 0.74 9.02E-05 2.32E-02 C15orf48 42.86 0.83 6.72E-05 1.50E-02 MINA 529.77 0.68 2.60E-04 5.62E-02 ODC1 476.86 -0.66 1.15E-04 2.91E-02 FBXO4 179.45 -0.56 7.41E-05 1.62E-02 SLC16A3 201.90 0.71 2.65E-04 5.62E-02 NEK6 93.44 -0.76 1.27E-04 3.13E-02 KLHDC2 337.67 0.66 8.26E-05 1.78E-02 LOC728875 80.09 -0.74 2.68E-04 5.62E-02 TBL1X 180.42 -0.72 1.28E-04 3.13E-02 HAGHL 23.92 -0.81 8.53E-05 1.81E-02 CIITA 414.44 0.84 2.73E-04 5.62E-02 LINC00176 22.63 0.77 1.48E-04 3.55E-02 UPP1 342.62 -0.68 9.27E-05 1.93E-02 PIK3IP1 1148.82 -0.68 2.79E-04 5.62E-02 MED12L 44.55 0.78 1.51E-04 3.55E-02 RNF19B 51.76 0.80 9.69E-05 1.96E-02 GNA15 322.20 0.85 2.82E-04 5.62E-02 DBNDD2 33.37 0.78 1.56E-04 3.62E-02 RNASE6 39.67 0.83 9.61E-05 1.96E-02 CTTNBP2NL 29.00 0.87 2.83E-04 5.62E-02 HBEGF 53.04 0.78 1.62E-04 3.70E-02 TNIP2 209.83 0.73 1.09E-04 2.16E-02 HLA-DQA2 611.02 0.85 2.84E-04 5.62E-02 HLA-DQB2 79.07 0.78 1.66E-04 3.74E-02 BIK 15.66 -0.82 1.13E-04 2.22E-02 ABLIM1 762.65 -0.70 3.07E-04 5.91E-02 TSHR 16.64 0.75 1.70E-04 3.77E-02 SCML4 93.00 -0.70 1.20E-04 2.31E-02

149

Supplemental Table 2.1 - (Continued)

CD combined – no stimulation CD combined – 4 hours CD combined – 24 hours Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj RRN3P1 44.58 -0.81 3.12E-04 5.91E-02 FSCN1 29.41 0.70 1.74E-04 3.80E-02 USP48 370.64 0.57 1.34E-04 2.54E-02 LINC00599 23.86 -0.86 3.16E-04 5.91E-02 BACH2 229.50 -0.75 1.95E-04 4.11E-02 P2RY11 22.58 -0.81 1.42E-04 2.65E-02 IL16 2572.00 -0.47 3.17E-04 5.91E-02 MMD 150.55 0.72 1.93E-04 4.11E-02 MATN4 10.13 0.67 1.51E-04 2.75E-02 P2RY14 39.72 0.87 3.28E-04 6.01E-02 CTTNBP2NL 40.69 0.77 1.97E-04 4.11E-02 NCALD 356.73 0.58 1.51E-04 2.75E-02 PRKCQ-AS1 318.17 -0.73 3.32E-04 6.01E-02 RNF167 436.03 0.44 2.23E-04 4.60E-02 NFKBIE 63.75 0.66 1.55E-04 2.77E-02 ADCY1 50.15 0.86 3.54E-04 6.30E-02 GPR132 292.35 0.69 2.38E-04 4.84E-02 CCDC88A 61.23 0.65 1.72E-04 3.04E-02 GPA33 74.95 -0.83 3.60E-04 6.33E-02 AMICA1 1481.74 -0.40 2.42E-04 4.84E-02 LOC10013289138.70 0.77 1.75E-04 3.06E-02 TNFSF10 1417.90 0.66 3.74E-04 6.48E-02 ADAT2 165.94 -0.59 2.48E-04 4.91E-02 LHFP 27.32 0.80 1.84E-04 3.16E-02 FAM200B 115.05 -0.61 4.06E-04 6.93E-02 GNPDA1 761.07 -0.70 2.61E-04 5.09E-02 MINOS1 779.99 -0.40 1.96E-04 3.28E-02 TCEA3 21.95 -0.83 4.11E-04 6.93E-02 ZNF502 11.75 -0.75 2.68E-04 5.10E-02 COL6A5 11.28 0.66 1.93E-04 3.28E-02 TTC39C 805.09 -0.66 4.48E-04 7.39E-02 CXCR6 905.34 -0.64 2.66E-04 5.10E-02 HLA-DQB2 73.00 0.78 1.98E-04 3.28E-02 TNFRSF8 78.55 0.85 4.49E-04 7.39E-02 BCL2L11 491.95 -0.74 2.87E-04 5.39E-02 KCNA3 320.90 0.68 2.07E-04 3.38E-02 MEGF6 58.84 -0.80 4.57E-04 7.42E-02 PP7080 156.01 -0.53 3.12E-04 5.78E-02 SLBP 999.60 0.71 2.13E-04 3.43E-02 ANKRD37 28.11 0.84 5.50E-04 8.81E-02 C10orf54 980.19 -0.64 3.25E-04 5.94E-02 MTSS1 149.43 0.63 2.20E-04 3.49E-02 NTRK2 46.83 0.75 5.76E-04 9.11E-02 OSM 1149.27 0.70 3.45E-04 6.23E-02 PAX8 14.26 -0.78 2.23E-04 3.50E-02 RALB 527.07 0.67 5.90E-04 9.19E-02 ANK3 328.00 0.68 3.60E-04 6.42E-02 FAS 835.36 0.38 2.37E-04 3.68E-02 SNHG6 583.78 -0.51 5.95E-04 9.19E-02 EPDR1 16.65 0.69 3.65E-04 6.44E-02 DDHD2 210.16 0.55 2.41E-04 3.70E-02 ANXA2R 272.59 -0.65 6.04E-04 9.21E-02 MINA 821.25 0.44 3.83E-04 6.60E-02 IL21R 468.09 0.65 2.70E-04 4.09E-02 PTBP1 866.25 0.57 6.28E-04 9.47E-02 PON2 98.42 0.70 3.83E-04 6.60E-02 PIK3C2B 184.00 0.53 2.81E-04 4.20E-02 MIR155HG 213.88 0.74 6.40E-04 9.54E-02 FOXP1 905.57 -0.68 3.92E-04 6.65E-02 C9orf16 472.77 -0.47 2.86E-04 4.23E-02 SOCS3 76.68 0.80 6.76E-04 9.85E-02 ELL2 440.84 0.63 4.00E-04 6.65E-02 HIVEP1 216.76 0.60 2.94E-04 4.29E-02 ZC4H2 155.84 -0.65 6.78E-04 9.85E-02 P2RY14 60.83 0.73 3.99E-04 6.65E-02 GPR132 317.39 0.54 3.24E-04 4.68E-02 SERINC5 1598.98 -0.64 6.95E-04 9.85E-02 WWTR1 25.58 0.73 4.15E-04 6.82E-02 WNT5B 16.54 0.66 3.29E-04 4.70E-02 SLC7A5 216.87 0.80 6.96E-04 9.85E-02 ANXA3 31.11 0.60 4.25E-04 6.90E-02 NDFIP2 742.96 -0.46 3.42E-04 4.73E-02 FASN 135.35 0.67 7.01E-04 9.85E-02 ENPP3 6.23 -0.58 4.40E-04 7.01E-02 PLK3 100.05 0.68 3.42E-04 4.73E-02 CYB5A 241.31 0.64 7.12E-04 9.87E-02 DDX4 7.56 0.60 4.38E-04 7.01E-02 NOD2 82.53 0.76 3.40E-04 4.73E-02 SDC4 208.05 0.56 7.18E-04 9.87E-02 USP18 108.69 0.67 4.76E-04 7.50E-02 UBE2J1 306.15 0.55 3.63E-04 4.85E-02 PLAGL2 162.77 0.56 7.29E-04 9.92E-02 ZDHHC9 129.72 -0.63 4.95E-04 7.71E-02 PNKD 405.60 -0.44 3.70E-04 4.85E-02 BAG1 122.34 -0.45 5.02E-04 7.74E-02 NCOA5 178.15 0.65 3.68E-04 4.85E-02 KIF1A 8.18 0.58 5.23E-04 7.97E-02 BATF3 210.02 0.76 3.66E-04 4.85E-02 TBKBP1 17.34 -0.71 5.46E-04 8.09E-02 VCAM1 21.59 0.67 3.62E-04 4.85E-02 KIAA1671 93.00 0.65 5.40E-04 8.09E-02 EGR1 403.52 0.70 3.74E-04 4.85E-02 ADCY1 48.55 0.71 5.47E-04 8.09E-02 IRF4 500.28 0.68 3.87E-04 4.98E-02 TMEM189 328.90 0.38 5.81E-04 8.51E-02 EVC 29.55 0.73 4.04E-04 5.14E-02 BAD 195.51 0.50 6.76E-04 9.80E-02 RUNX2 409.34 0.60 4.09E-04 5.15E-02 MTSS1 160.55 0.70 6.94E-04 9.96E-02 IL31RA 7.71 0.69 4.16E-04 5.17E-02 ZNRF1 64.38 0.68 4.19E-04 5.17E-02 KDSR 932.93 0.68 4.59E-04 5.61E-02 IGFLR1 620.25 -0.45 4.84E-04 5.73E-02 SEPW1 1141.31 -0.41 4.81E-04 5.73E-02 IFIH1 182.04 0.63 4.91E-04 5.73E-02 JMY 84.20 0.73 4.87E-04 5.73E-02 LOC10050666844.96 -0.66 4.78E-04 5.73E-02 ETV6 228.14 0.51 5.09E-04 5.88E-02 DENND4A 237.26 0.56 5.32E-04 6.09E-02 RGL4 144.59 -0.54 5.47E-04 6.11E-02 GLUL 2097.26 0.68 5.48E-04 6.11E-02 NOMO3 43.60 0.73 5.41E-04 6.11E-02 CD74 27339.03 0.58 5.53E-04 6.11E-02 ZDHHC3 298.89 0.44 5.62E-04 6.12E-02 NOTCH2 853.97 0.53 5.63E-04 6.12E-02 MAF1 368.42 0.71 6.22E-04 6.49E-02 CXCL10 208.18 0.61 6.20E-04 6.49E-02 MLLT3 398.50 -0.36 6.19E-04 6.49E-02 HMSD 25.09 0.72 6.08E-04 6.49E-02 ZNF704 8.42 -0.72 6.22E-04 6.49E-02 INSIG1 1252.70 0.67 6.36E-04 6.57E-02 TACO1 143.99 0.57 6.61E-04 6.76E-02 TRIM14 485.46 0.58 6.75E-04 6.76E-02 TARSL2 56.38 0.69 6.70E-04 6.76E-02 PON2 102.68 0.61 6.75E-04 6.76E-02 RPL37A 7903.48 -0.32 7.08E-04 6.92E-02 SLC25A10 110.25 -0.56 7.04E-04 6.92E-02 RGMB 21.71 0.69 6.98E-04 6.92E-02 TTC39C 569.24 -0.57 7.25E-04 7.04E-02 AKIRIN1 412.99 0.66 7.37E-04 7.04E-02

150

Supplemental Table 2.1-(Continued)

CD combined – 24 hours Gene baseMean LogFC p value padj FAM173B 205.91 -0.52 7.32E-04 7.04E-02 CLPTM1 190.39 0.55 7.58E-04 7.17E-02 ANXA11 844.72 0.63 7.62E-04 7.17E-02 FBXO32 92.00 0.63 7.79E-04 7.28E-02 GET4 231.33 0.65 7.91E-04 7.34E-02 RCN2 268.54 0.71 8.00E-04 7.37E-02 ALDH4A1 60.23 -0.60 8.37E-04 7.66E-02 CD58 861.00 0.36 8.66E-04 7.70E-02 LYSMD2 117.14 0.71 8.63E-04 7.70E-02 NFKBIA 2292.72 0.62 8.58E-04 7.70E-02 MKNK1 417.36 0.46 8.59E-04 7.70E-02 TMEM121 10.57 0.69 8.98E-04 7.88E-02 PROSER1 368.14 0.46 8.96E-04 7.88E-02 CIRBP 919.69 -0.48 9.26E-04 7.96E-02 MTDH 663.91 0.49 9.15E-04 7.96E-02 PPP1CC 799.37 0.48 9.24E-04 7.96E-02 PIR 64.21 0.65 9.64E-04 8.12E-02 APOBR 94.19 -0.52 9.56E-04 8.12E-02 B3GNT2 207.39 0.66 9.64E-04 8.12E-02 DECR1 876.66 -0.36 9.88E-04 8.27E-02 MAP3K6 29.88 0.68 9.97E-04 8.29E-02 TAF4B 26.05 0.69 1.03E-03 8.47E-02 PCED1B 173.45 -0.48 1.04E-03 8.55E-02 OGFOD3 133.69 0.54 1.05E-03 8.55E-02 C1orf228 60.18 -0.68 1.19E-03 9.61E-02 DNAJC5B 7.46 0.55 1.20E-03 9.67E-02 SLC25A22 119.57 0.58 1.21E-03 9.70E-02 BCL2L11 141.49 -0.67 1.25E-03 9.87E-02 RPL21P28 6107.53 -0.37 1.29E-03 9.87E-02 TMOD1 25.25 0.63 1.29E-03 9.87E-02 CDKN2A 120.19 -0.64 1.28E-03 9.87E-02 LRP8 142.29 0.54 1.29E-03 9.87E-02 MLLT4 66.25 0.64 1.27E-03 9.87E-02 ADAP1 23.82 -0.68 1.26E-03 9.87E-02 JAK1 1257.98 0.52 1.31E-03 9.96E-02 IFI44L 39.53 0.68 1.32E-03 9.96E-02 MROH8 23.32 -0.67 1.33E-03 9.99E-02

151

Supplemental Table 2.1 - (Continued)

CD4+ no stimulation CD4+ 4 hours CD4+ 24 hours Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj JUN 60.51 1.63 2.17E-08 2.99E-04 C17orf61-PLSCR3 41.84 -1.79 1.11E-13 1.35E-09 GJB2 244.40 1.55 1.49E-09 1.97E-05 GPA33 114.19 -1.53 5.94E-08 4.08E-04 ENPP2 97.77 1.77 4.63E-13 2.81E-09 UBD 37.13 1.49 5.44E-09 3.59E-05 KRT1 120.23 -1.45 6.42E-07 2.95E-03 FILIP1L 180.31 1.32 1.01E-08 4.10E-05 NTRK2 114.68 1.43 5.02E-08 2.21E-04 EGR1 123.62 1.30 6.20E-06 2.13E-02 HLA-DQA2 556.29 1.31 8.41E-08 2.55E-04 THY1 29.23 1.26 5.32E-07 1.55E-03 CIITA 333.26 1.22 9.77E-06 2.69E-02 UBD 81.00 1.22 1.36E-07 3.30E-04 HLA-DQA2 518.42 1.28 5.87E-07 1.55E-03 UBD 34.30 1.12 2.49E-05 5.71E-02 CIITA 367.08 1.20 2.75E-07 4.76E-04 HLA-DRA 6566.62 1.10 9.17E-07 2.02E-03 KLHL23 19.00 1.11 4.37E-05 7.78E-02 GJB2 308.61 1.25 2.62E-07 4.76E-04 G0S2 63.95 1.29 1.18E-06 2.23E-03 SCD 428.59 1.00 4.52E-05 7.78E-02 P2RY14 53.02 1.19 1.13E-06 1.71E-03 CXCL10 235.90 1.25 2.43E-06 4.01E-03 HLA-DOA 258.09 1.17 5.35E-05 8.19E-02 IL4I1 236.25 0.95 4.75E-06 6.41E-03 IER2 1658.94 0.92 2.84E-06 4.16E-03 ALPK2 16.74 1.06 6.57E-05 9.04E-02 HLA-DOA 210.18 1.08 9.61E-06 1.09E-02 CIITA 486.06 1.10 3.80E-06 5.01E-03 CXCL10 162.12 1.11 7.88E-05 9.85E-02 ENOX1 42.28 0.95 9.92E-06 1.09E-02 DOHH 120.11 1.18 5.65E-06 6.78E-03 HLA-DRA 6849.48 1.06 1.21E-05 1.23E-02 ADA 519.36 1.08 6.44E-06 7.09E-03 NTRK2 446.21 1.05 1.36E-05 1.27E-02 MSC 269.09 1.16 1.02E-05 1.03E-02 HLA-DRB1 3546.08 1.01 2.44E-05 1.98E-02 JUNB 1127.04 1.01 1.48E-05 1.39E-02 COL6A1 30.67 0.93 2.37E-05 1.98E-02 DMD 131.95 1.14 1.78E-05 1.42E-02 DMD 201.87 1.00 3.82E-05 2.58E-02 CDK6 1226.39 0.95 1.64E-05 1.42E-02 BTN2A2 155.42 0.93 3.79E-05 2.58E-02 HLA-DRB1 3456.58 1.05 1.83E-05 1.42E-02 HLA-DPB1 2102.74 0.99 3.61E-05 2.58E-02 HLA-DOA 266.72 1.11 2.16E-05 1.50E-02 HLA-DMB 313.07 0.96 7.63E-05 4.67E-02 SH3BP5 243.81 1.06 2.11E-05 1.50E-02 HLA-DRB5 1032.55 0.97 7.70E-05 4.67E-02 LGMN 325.32 -1.04 4.34E-05 2.86E-02 HLA-DQB2 79.53 0.96 8.54E-05 4.94E-02 ACSL1 94.77 1.04 4.77E-05 3.00E-02 JUN 83.06 0.94 9.46E-05 5.22E-02 ANXA3 49.93 1.06 5.00E-05 3.00E-02 GCSAM 93.84 0.92 1.50E-04 7.59E-02 HLA-DRB5 1038.39 1.02 5.67E-05 3.25E-02 HLA-DPA1 4787.23 0.93 1.47E-04 7.59E-02 EMC8 172.54 1.05 7.30E-05 3.90E-02 DDIT4 1468.62 0.88 1.58E-04 7.69E-02 FILIP1L 89.49 1.05 7.39E-05 3.90E-02 HLA-DRB6 350.45 0.91 1.88E-04 8.78E-02 PDCD1 177.81 -0.95 7.77E-05 3.94E-02 C7orf55-LUC7L2 28.17 0.81 1.97E-04 8.87E-02 ANK3 269.28 0.87 9.09E-05 4.28E-02 BCL2A1 446.67 -0.87 2.15E-04 9.30E-02 HLA-DRB6 344.95 1.01 8.83E-05 4.28E-02 KRT7 48.31 0.90 2.36E-04 9.88E-02 IFNG 770.38 1.02 9.49E-05 4.32E-02 MPZL1 230.37 0.96 1.12E-04 4.65E-02 TMEM165 255.98 0.98 1.13E-04 4.65E-02 NOD2 117.11 1.01 1.13E-04 4.65E-02 DGAT2 26.29 0.93 1.30E-04 5.18E-02 AKIRIN1 410.23 0.94 1.52E-04 5.47E-02 ELL2 329.55 0.95 1.47E-04 5.47E-02 MATN4 14.59 0.89 1.52E-04 5.47E-02 SREBF2 321.25 0.86 1.57E-04 5.47E-02 INSIG1 1187.28 0.88 1.57E-04 5.47E-02 BATF3 159.07 0.98 1.67E-04 5.63E-02 HLA-DPB1 2085.87 0.90 1.83E-04 5.84E-02 MAF1 408.02 0.98 1.86E-04 5.84E-02 HLA-DPA1 4953.28 0.89 1.81E-04 5.84E-02 ADCY1 116.14 0.99 1.97E-04 6.06E-02 NFKBIA 2113.32 0.87 2.27E-04 6.82E-02 JUN 161.21 0.88 2.37E-04 6.89E-02 P2RY14 42.63 0.97 2.40E-04 6.89E-02 ANXA11 949.22 0.82 2.86E-04 7.99E-02 COTL1 3174.02 0.90 2.91E-04 7.99E-02 HMHA1 1559.34 0.74 3.03E-04 8.15E-02 IL23R 32.41 0.95 3.12E-04 8.25E-02 GCSAM 103.55 0.94 3.21E-04 8.30E-02 ZFAND5 664.12 0.91 3.41E-04 8.66E-02 IL21 14.28 0.83 3.87E-04 9.62E-02 ACADVL 1075.91 0.83 3.97E-04 9.69E-02 IL21R 430.22 0.78 4.18E-04 9.85E-02 SLBP 1001.06 0.84 4.13E-04 9.85E-02

152

Supplemental Table 2.1 – (Continued)

CD8+ No Stimulation CD8+ 4 hours CD8 24 hours Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj Gene baseMean LogFC p value padj CXCL10 153.97 2.25 2.17E-15 3.94E-11 JUNB 1517.36 0.97 3.16E-08 5.80E-04 ENPP2 165.05 1.27 6.54E-11 8.48E-07 JUNB 508.83 1.66 1.71E-12 1.55E-08 CXCL10 222.56 1.20 9.65E-08 8.85E-04 GJB2 308.95 1.03 7.10E-08 4.61E-04 NTRK2 28.44 1.52 4.05E-08 2.45E-04 ENOX1 35.18 1.06 4.02E-07 2.46E-03 C4orf26 142.26 1.04 1.38E-07 5.95E-04 MSC 285.77 1.54 5.87E-08 2.67E-04 ENPP2 128.55 1.02 2.06E-06 9.43E-03 MX1 271.76 0.95 5.44E-07 1.76E-03 VNN2 133.36 -1.29 5.03E-07 1.83E-03 DDIT4 1901.41 1.01 4.50E-06 1.65E-02 NTRK2 67.96 0.95 1.98E-06 5.14E-03 NTRK2 202.63 1.02 5.71E-06 1.75E-02 JUNB 980.64 0.84 3.62E-06 7.82E-03 GCSAM 42.18 0.92 2.56E-05 6.44E-02 TNFRSF8 328.21 0.90 6.23E-06 1.16E-02 IL5 88.58 -0.94 2.81E-05 6.44E-02 DGAT2 30.68 0.76 7.26E-06 1.18E-02 ELL2 340.54 0.83 9.24E-06 1.33E-02 IL4I1 106.36 0.85 1.26E-05 1.63E-02 ITPR1 809.06 0.71 2.25E-05 2.62E-02 HLA-DRB6 370.24 0.80 2.43E-05 2.62E-02 GCSAM 64.54 0.74 6.71E-05 6.70E-02 ADCY1 62.65 0.76 7.58E-05 7.02E-02 HLA-DQA2 542.96 0.77 8.19E-05 7.08E-02 HLA-DRA 7408.60 0.67 1.12E-04 9.08E-02 ANK3 189.91 0.76 1.29E-04 9.86E-02

153

References

1. Eshhar Z. Tumor-specific T-bodies: towards clinical application. Cancer Immunol Immunother. 1997;45(3-4):131-136.

2. Eshhar Z, Bach N, Fitzer-Attas CJ, et al. The T-body approach: potential for cancer immunotherapy. Springer Semin Immunopathol. 1996;18(2):199-209.

3. June CH, O'Connor RS, Kawalekar OU, Ghassemi S, Milone MC. CAR T cell immunotherapy for human cancer. Science. 2018;359(6382):1361-1365.

4. MacDonald KG, Hoeppli RE, Huang Q, et al. Alloantigen-specific regulatory T cells generated with a chimeric antigen receptor. J Clin Invest. 2016;126(4):1413- 1424.

5. June CH, Sadelain M. Chimeric Antigen Receptor Therapy. N Engl J Med. 2018;379(1):64-73.

6. Milone MC, Fish JD, Carpenito C, et al. Chimeric receptors containing CD137 signal transduction domains mediate enhanced survival of T cells and increased antileukemic efficacy in vivo. Mol Ther. 2009;17(8):1453-1464.

7. Kawalekar OU, O'Connor RS, Fraietta JA, et al. Distinct Signaling of Coreceptors Regulates Specific Metabolism Pathways and Impacts Memory Development in CAR T Cells. Immunity. 2016;44(2):380-390.

8. Lafferty KJ, Cunningham AJ. A new analysis of allogeneic interactions. Aust J Exp Biol Med Sci. 1975;53(1):27-42.

9. Quill H, Schwartz RH. Stimulation of normal inducer T cell clones with antigen presented by purified Ia molecules in planar lipid membranes: specific induction of a long-lived state of proliferative nonresponsiveness. J Immunol. 1987;138(11):3704-3712.

10. Mueller DL, Jenkins MK, Schwartz RH. An accessory cell-derived costimulatory signal acts independently of protein kinase C activation to allow T cell 154

proliferation and prevent the induction of unresponsiveness. J Immunol. 1989;142(8):2617-2628.

11. Bluestone JA, St Clair EW, Turka LA. CTLA4Ig: bridging the basic immunology with clinical application. Immunity. 2006;24(3):233-238.

12. Sharpe AH. Mechanisms of costimulation. Immunol Rev. 2009;229(1):5-11.

13. Sharpe AH, Abbas AK. T-cell costimulation--biology, therapeutic potential, and challenges. N Engl J Med. 2006;355(10):973-975.

14. Esensten JH, Helou YA, Chopra G, Weiss A, Bluestone JA. CD28 Costimulation: From Mechanism to Therapy. Immunity. 2016;44(5):973-988.

15. Greenwald RJ, Freeman GJ, Sharpe AH. The B7 family revisited. Annu Rev Immunol. 2005;23:515-548.

16. Boomer JS, Green JM. An enigmatic tail of CD28 signaling. Cold Spring Harb Perspect Biol. 2010;2(8):a002436.

17. Fraser JD, Irving BA, Crabtree GR, Weiss A. Regulation of interleukin-2 gene enhancer activity by the T cell accessory molecule CD28. Science. 1991;251(4991):313-316.

18. Janardhan SV, Praveen K, Marks R, Gajewski TF. Evidence implicating the Ras pathway in multiple CD28 costimulatory functions in CD4+ T cells. PLoS One. 2011;6(9):e24931.

19. Andres PG, Howland KC, Nirula A, et al. Distinct regions in the CD28 cytoplasmic domain are required for T helper type 2 differentiation. Nat Immunol. 2004;5(4):435-442.

20. Boise LH, Minn AJ, Noel PJ, et al. CD28 costimulation can promote T cell survival by enhancing the expression of Bcl-xL. Immunity. 1995. 3: 87-98. J Immunol. 2010;185(7):3788-3799.

155

21. Frauwirth KA, Riley JL, Harris MH, et al. The CD28 signaling pathway regulates glucose metabolism. Immunity. 2002;16(6):769-777.

22. de Kouchkovsky D, Esensten JH, Rosenthal WL, Morar MM, Bluestone JA, Jeker LT. microRNA-17-92 regulates IL-10 production by regulatory T cells and control of experimental autoimmune encephalomyelitis. J Immunol. 2013;191(4):1594- 1605.

23. Salomon B, Lenschow DJ, Rhee L, et al. B7/CD28 costimulation is essential for the homeostasis of the CD4+CD25+ immunoregulatory T cells that control autoimmune diabetes. Immunity. 2000;12(4):431-440.

24. Krummel MF, Allison JP. CD28 and CTLA-4 have opposing effects on the response of T cells to stimulation. J Exp Med. 1995;182(2):459-465.

25. Schildberg FA, Klein SR, Freeman GJ, Sharpe AH. Coinhibitory Pathways in the B7-CD28 Ligand-Receptor Family. Immunity. 2016;44(5):955-972.

26. Sanchez-Paulete AR, Labiano S, Rodriguez-Ruiz ME, et al. Deciphering CD137 (4-1BB) signaling in T-cell costimulation for translation into successful cancer immunotherapy. Eur J Immunol. 2016;46(3):513-522.

27. Pollok KE, Kim YJ, Zhou Z, et al. Inducible T cell antigen 4-1BB. Analysis of expression and function. J Immunol. 1993;150(3):771-781.

28. Broll K, Richter G, Pauly S, Hofstaedter F, Schwarz H. CD137 expression in tumor vessel walls. High correlation with malignant tumors. Am J Clin Pathol. 2001;115(4):543-549.

29. Melero I, Johnston JV, Shufford WW, Mittler RS, Chen L. NK1.1 cells express 4- 1BB (CDw137) costimulatory molecule and are required for tumor immunity elicited by anti-4-1BB monoclonal antibodies. Cell Immunol. 1998;190(2):167- 172.

30. Wilcox RA, Chapoval AI, Gorski KS, et al. Cutting edge: Expression of functional CD137 receptor by dendritic cells. J Immunol. 2002;168(9):4262-4267.

156

31. Croft M, Duan W, Choi H, Eun SY, Madireddi S, Mehta A. TNF superfamily in inflammatory disease: translating basic insights. Trends Immunol. 2012;33(3):144-152.

32. Watts TH. TNF/TNFR family members in costimulation of T cell responses. Annu Rev Immunol. 2005;23:23-68.

33. Wang C, Lin GH, McPherson AJ, Watts TH. Immune regulation by 4-1BB and 4- 1BBL: complexities and challenges. Immunol Rev. 2009;229(1):192-215.

34. Chen L, Flies DB. Molecular mechanisms of T cell co-stimulation and coinhibition. Nat Rev Immunol. 2013;13(4):227-242.

35. Zhu Y, Zhu G, Luo L, Flies AS, Chen L. CD137 stimulation delivers an antigen- independent growth signal for T lymphocytes with memory phenotype. Blood. 2007;109(11):4882-4889.

36. Kim YJ, Kim SH, Mantel P, Kwon BS. Human 4-1BB regulates CD28 costimulation to promote Th1 cell responses. Eur J Immunol. 1998;28(3):881-890.

37. Maus MV, Thomas AK, Leonard DG, et al. Ex vivo expansion of polyclonal and antigen-specific cytotoxic T lymphocytes by artificial APCs expressing ligands for the T-cell receptor, CD28 and 4-1BB. Nat Biotechnol. 2002;20(2):143-148.

38. Shuford WW, Klussman K, Tritchler DD, et al. 4-1BB costimulatory signals preferentially induce CD8+ T cell proliferation and lead to the amplification in vivo of cytotoxic T cell responses. J Exp Med. 1997;186(1):47-55.

39. Yonezawa A, Dutt S, Chester C, Kim J, Kohrt HE. Boosting Cancer Immunotherapy with Anti-CD137 Antibody Therapy. Clin Cancer Res. 2015;21(14):3113-3120.

40. Kim YJ, Mantel PL, June CH, Kim SH, Kwon BS. 4-1BB costimulation promotes human T cell adhesion to fibronectin. Cell Immunol. 1999;192(1):13-23.

157

41. Melero I, Shuford WW, Newby SA, et al. Monoclonal antibodies against the 4- 1BB T-cell activation molecule eradicate established tumors. Nat Med. 1997;3(6):682-685.

42. Chester C, Sanmamed MF, Wang J, Melero I. Immunotherapy targeting 4-1BB: mechanistic rationale, clinical results, and future strategies. Blood. 2018;131(1):49-57.

43. Houot R, Goldstein MJ, Kohrt HE, et al. Therapeutic effect of CD137 immunomodulation in lymphoma and its enhancement by Treg depletion. Blood. 2009;114(16):3431-3438.

44. Smith SE, Hoelzinger DB, Dominguez AL, Van Snick J, Lustgarten J. Signals through 4-1BB inhibit T regulatory cells by blocking IL-9 production enhancing antitumor responses. Cancer Immunol Immunother. 2011;60(12):1775-1787.

45. Zhang P, Gao F, Wang Q, et al. Agonistic anti-4-1BB antibody promotes the expansion of natural regulatory T cells while maintaining Foxp3 expression. Scand J Immunol. 2007;66(4):435-440.

46. van der Stegen SJ, Hamieh M, Sadelain M. The pharmacology of second- generation chimeric antigen receptors. Nat Rev Drug Discov. 2015;14(7):499- 509.

47. Choi YS, Kageyama R, Eto D, et al. ICOS receptor instructs T follicular helper cell versus effector cell differentiation via induction of the transcriptional repressor Bcl6. Immunity. 2011;34(6):932-946.

48. Paulos CM, Carpenito C, Plesa G, et al. The inducible costimulator (ICOS) is critical for the development of human T(H)17 cells. Sci Transl Med. 2010;2(55):55ra78.

49. Guedan S, Chen X, Madar A, et al. ICOS-based chimeric antigen receptors program bipolar TH17/TH1 cells. Blood. 2014;124(7):1070-1080.

50. Guedan S, Posey AD, Jr., Shaw C, et al. Enhancing CAR T cell persistence through ICOS and 4-1BB costimulation. JCI Insight. 2018;3(1).

158

51. Nocentini G, Riccardi C. GITR: a modulator of immune response and inflammation. Adv Exp Med Biol. 2009;647:156-173.

52. Pascutti MF, Geerman S, Slot E, et al. Enhanced CD8 T cell responses through GITR-mediated costimulation resolve chronic viral infection. PLoS Pathog. 2015;11(3):e1004675.

53. Riccardi C, Ronchetti S, Nocentini G. Glucocorticoid-induced tumor necrosis factor receptor as a therapeutic target for immunotherapy. Expert Opin Ther Targets. 2018.

54. Rogers PR, Song J, Gramaglia I, Killeen N, Croft M. OX40 promotes Bcl-xL and Bcl-2 expression and is essential for long-term survival of CD4 T cells. Immunity. 2001;15(3):445-455.

55. Schaer DA, Hirschhorn-Cymerman D, Wolchok JD. Targeting tumor-necrosis factor receptor pathways for tumor immunotherapy. J Immunother Cancer. 2014;2:7.

56. Xiao X, Balasubramanian S, Liu W, et al. OX40 signaling favors the induction of T(H)9 cells and airway inflammation. Nat Immunol. 2012;13(10):981-990.

57. Hombach AA, Abken H. Costimulation by chimeric antigen receptors revisited the T cell antitumor response benefits from combined CD28-OX40 signalling. Int J Cancer. 2011;129(12):2935-2944.

58. Zhang X, Xiao X, Lan P, et al. OX40 Costimulation Inhibits Foxp3 Expression and Treg Induction via BATF3-Dependent and Independent Mechanisms. Cell Rep. 2018;24(3):607-618.

59. Kumar P, Marinelarena A, Raghunathan D, et al. Critical role of OX40 signaling in the TCR-independent phase of human and murine thymic Treg generation. Cell Mol Immunol. 2018.

60. Croft M. The role of TNF superfamily members in T-cell function and diseases. Nat Rev Immunol. 2009;9(4):271-285.

159

61. Carr JM, Carrasco MJ, Thaventhiran JE, et al. CD27 mediates interleukin-2- independent clonal expansion of the CD8+ T cell without effector differentiation. Proc Natl Acad Sci U S A. 2006;103(51):19454-19459.

62. Kawamura T, Ogawa Y, Shimozato O, et al. CD70 is selectively expressed on Th1 but not on Th2 cells and is required for Th1-type immune responses. J Invest Dermatol. 2011;131(6):1252-1261.

63. Song DG, Ye Q, Poussin M, Harms GM, Figini M, Powell DJ, Jr. CD27 costimulation augments the survival and antitumor activity of redirected human T cells in vivo. Blood. 2012;119(3):696-706.

64. Maus MV, Fraietta JA, Levine BL, Kalos M, Zhao Y, June CH. Adoptive immunotherapy for cancer or viruses. Annu Rev Immunol. 2014;32:189-225.

65. June CH, Riddell SR, Schumacher TN. Adoptive cellular therapy: a race to the finish line. Sci Transl Med. 2015;7(280):280ps287.

66. Wang X, Riviere I. Clinical manufacturing of CAR T cells: foundation of a promising therapy. Mol Ther Oncolytics. 2016;3:16015.

67. Kebriaei P, Izsvak Z, Narayanavari SA, Singh H, Ivics Z. Gene Therapy with the Sleeping Beauty Transposon System. Trends Genet. 2017;33(11):852-870.

68. Gomes-Silva D, Mukherjee M, Srinivasan M, et al. Tonic 4-1BB Costimulation in Chimeric Antigen Receptors Impedes T Cell Survival and Is Vector-Dependent. Cell Rep. 2017;21(1):17-26.

69. Quintarelli C, Orlando D, Boffa I, et al. Choice of costimulatory domains and of cytokines determines CAR T-cell activity in neuroblastoma. Oncoimmunology. 2018;7(6):e1433518.

70. Hoffmann JM, Schubert ML, Wang L, et al. Differences in Expansion Potential of Naive Chimeric Antigen Receptor T Cells from Healthy Donors and Untreated Chronic Lymphocytic Leukemia Patients. Front Immunol. 2017;8:1956.

160

71. Geldres C, Savoldo B, Dotti G. Chimeric antigen receptor-redirected T cells return to the bench. Semin Immunol. 2016;28(1):3-9.

72. Courtney AH, Lo WL, Weiss A. TCR Signaling: Mechanisms of Initiation and Propagation. Trends Biochem Sci. 2018;43(2):108-123.

73. Dotti G, Gottschalk S, Savoldo B, Brenner MK. Design and development of therapies using chimeric antigen receptor-expressing T cells. Immunol Rev. 2014;257(1):107-126.

74. Jamnani FR, Rahbarizadeh F, Shokrgozar MA, et al. T cells expressing VHH- directed oligoclonal chimeric HER2 antigen receptors: towards tumor-directed oligoclonal T cell therapy. Biochim Biophys Acta. 2014;1840(1):378-386.

75. Hammill JA, VanSeggelen H, Helsen CW, et al. Designed ankyrin repeat proteins are effective targeting elements for chimeric antigen receptors. J Immunother Cancer. 2015;3:55.

76. Kulemzin SV, Gorchakov AA, Chikaev AN, et al. VEGFR2-specific FnCAR effectively redirects the cytotoxic activity of T cells and YT NK cells. Oncotarget. 2018;9(10):9021-9029.

77. Spear P, Barber A, Rynda-Apple A, Sentman CL. NKG2D CAR T-cell therapy inhibits the growth of NKG2D ligand heterogeneous tumors. Immunol Cell Biol. 2013;91(6):435-440.

78. Lee L, Draper B, Chaplin N, et al. An APRIL-based chimeric antigen receptor for dual targeting of BCMA and TACI in multiple myeloma. Blood. 2018;131(7):746- 758.

79. Chmielewski M, Hombach A, Heuser C, Adams GP, Abken H. T cell activation by antibody-like immunoreceptors: increase in affinity of the single-chain fragment domain above threshold does not increase T cell activation against antigen- positive target cells but decreases selectivity. J Immunol. 2004;173(12):7647- 7653.

161

80. Walker AJ, Majzner RG, Zhang L, et al. Tumor Antigen and Receptor Densities Regulate Efficacy of a Chimeric Antigen Receptor Targeting Anaplastic Lymphoma Kinase. Mol Ther. 2017;25(9):2189-2201.

81. Hudecek M, Lupo-Stanghellini MT, Kosasih PL, et al. Receptor affinity and extracellular domain modifications affect tumor recognition by ROR1-specific chimeric antigen receptor T cells. Clin Cancer Res. 2013;19(12):3153-3164.

82. Lamers CH, Willemsen R, van Elzakker P, et al. Immune responses to transgene and retroviral vector in patients treated with ex vivo-engineered T cells. Blood. 2011;117(1):72-82.

83. Jonnalagadda M, Mardiros A, Urak R, et al. Chimeric antigen receptors with mutated IgG4 Fc spacer avoid fc receptor binding and improve T cell persistence and antitumor efficacy. Mol Ther. 2015;23(4):757-768.

84. Watanabe N, Bajgain P, Sukumaran S, et al. Fine-tuning the CAR spacer improves T-cell potency. Oncoimmunology. 2016;5(12):e1253656.

85. Fitzer-Attas CJ, Schindler DG, Waks T, Eshhar Z. Harnessing Syk family tyrosine kinases as signaling domains for chimeric single chain of the variable domain receptors: optimal design for T cell activation. J Immunol. 1998;160(1):145-154.

86. Shirasu N, Kuroki M. Functional design of chimeric T-cell antigen receptors for adoptive immunotherapy of cancer: architecture and outcomes. Anticancer Res. 2012;32(6):2377-2383.

87. Guest RD, Hawkins RE, Kirillova N, et al. The role of extracellular spacer regions in the optimal design of chimeric immune receptors: evaluation of four different scFvs and antigens. J Immunother. 2005;28(3):203-211.

88. Frigault MJ, Lee J, Basil MC, et al. Identification of chimeric antigen receptors that mediate constitutive or inducible proliferation of T cells. Cancer Immunol Res. 2015;3(4):356-367.

89. Hudecek M, Sommermeyer D, Kosasih PL, et al. The nonsignaling extracellular spacer domain of chimeric antigen receptors is decisive for in vivo antitumor activity. Cancer Immunol Res. 2015;3(2):125-135. 162

90. Imai C, Mihara K, Andreansky M, et al. Chimeric receptors with 4-1BB signaling capacity provoke potent cytotoxicity against acute lymphoblastic leukemia. Leukemia. 2004;18(4):676-684.

91. Carpenito C, Milone MC, Hassan R, et al. Control of large, established tumor xenografts with genetically retargeted human T cells containing CD28 and CD137 domains. Proceedings of the National Academy of Sciences of the United States of America. 2009;106(9):3360-3365.

92. Zhong XS, Matsushita M, Plotkin J, Riviere I, Sadelain M. Chimeric antigen receptors combining 4-1BB and CD28 signaling domains augment PI3kinase/AKT/Bcl-XL activation and CD8+ T cell-mediated tumor eradication. Mol Ther. 2010;18(2):413-420.

93. Irving BA, Weiss A. The cytoplasmic domain of the T cell receptor zeta chain is sufficient to couple to receptor-associated signal transduction pathways. Cell. 1991;64(5):891-901.

94. Srivastava S, Riddell SR. Engineering CAR-T cells: Design concepts. Trends Immunol. 2015;36(8):494-502.

95. van der Merwe PA, Dushek O. Mechanisms for T cell receptor triggering. Nat Rev Immunol. 2011;11(1):47-55.

96. Kuhns MS, Davis MM, Garcia KC. Deconstructing the form and function of the TCR/CD3 complex. Immunity. 2006;24(2):133-139.

97. Monks CR, Freiberg BA, Kupfer H, Sciaky N, Kupfer A. Three-dimensional segregation of supramolecular activation clusters in T cells. Nature. 1998;395(6697):82-86.

98. Cantrell DA. T-cell antigen receptor signal transduction. Immunology. 2002;105(4):369-374.

99. Bridgeman JS, Ladell K, Sheard VE, et al. CD3zeta-based chimeric antigen receptors mediate T cell activation via cis- and trans-signalling mechanisms: implications for optimization of receptor structure for adoptive cell therapy. Clin Exp Immunol. 2014;175(2):258-267. 163

100. Davenport AJ, Jenkins MR, Cross RS, et al. CAR-T Cells Inflict Sequential Killing of Multiple Tumor Target Cells. Cancer Immunol Res. 2015;3(5):483-494.

101. Schroten C, Kraaij R, Veldhoven JL, et al. T cell activation upon exposure to patient-derived tumor tissue: a functional assay to select patients for adoptive T cell therapy. J Immunol Methods. 2010;359(1-2):11-20.

102. Davenport AJ, Cross RS, Watson KA, et al. Chimeric antigen receptor T cells form nonclassical and potent immune synapses driving rapid cytotoxicity. Proc Natl Acad Sci U S A. 2018;115(9):E2068-E2076.

103. Harris DT, Kranz DM. Adoptive T Cell Therapies: A Comparison of T Cell Receptors and Chimeric Antigen Receptors. Trends Pharmacol Sci. 2016;37(3):220-230.

104. Huang J, Brameshuber M, Zeng X, et al. A single peptide-major histocompatibility complex ligand triggers digital cytokine secretion in CD4(+) T cells. Immunity. 2013;39(5):846-857.

105. Valitutti S, Muller S, Cella M, Padovan E, Lanzavecchia A. Serial triggering of many T-cell receptors by a few peptide-MHC complexes. Nature. 1995;375(6527):148-151.

106. Maher J, Brentjens RJ, Gunset G, Riviere I, Sadelain M. Human T-lymphocyte cytotoxicity and proliferation directed by a single chimeric TCRzeta /CD28 receptor. Nat Biotechnol. 2002;20(1):70-75.

107. Savoldo B, Ramos CA, Liu E, et al. CD28 costimulation improves expansion and persistence of chimeric antigen receptor-modified T cells in lymphoma patients. J Clin Invest. 2011;121(5):1822-1826.

108. Maude SL, Frey N, Shaw PA, et al. Chimeric antigen receptor T cells for sustained remissions in leukemia. N Engl J Med. 2014;371(16):1507-1517.

109. Davila ML, Riviere I, Wang X, et al. Efficacy and toxicity management of 19-28z CAR T cell therapy in B cell acute lymphoblastic leukemia. Sci Transl Med. 2014;6(224):224ra225.

164

110. Neelapu SS, Locke FL, Bartlett NL, et al. Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B-Cell Lymphoma. The New England journal of medicine. 2017;377(26):2531-2544.

111. Schuster SJ, Svoboda J, Chong EA, et al. Chimeric Antigen Receptor T Cells in Refractory B-Cell Lymphomas. N Engl J Med. 2017;377(26):2545-2554.

112. Porter DL, Hwang WT, Frey NV, et al. Chimeric antigen receptor T cells persist and induce sustained remissions in relapsed refractory chronic lymphocytic leukemia. Sci Transl Med. 2015;7(303):303ra139.

113. Lee DW, Kochenderfer JN, Stetler-Stevenson M, et al. T cells expressing CD19 chimeric antigen receptors for acute lymphoblastic leukaemia in children and young adults: a phase 1 dose-escalation trial. Lancet. 2015;385(9967):517-528.

114. Brentjens RJ, Davila ML, Riviere I, et al. CD19-targeted T cells rapidly induce molecular remissions in adults with chemotherapy-refractory acute lymphoblastic leukemia. Sci Transl Med. 2013;5(177):177ra138.

115. Zhang H, Snyder KM, Suhoski MM, et al. 4-1BB is superior to CD28 costimulation for generating CD8+ cytotoxic lymphocytes for adoptive immunotherapy. J Immunol. 2007;179(7):4910-4918.

116. Mamonkin M, Mukherjee M, Srinivasan M, et al. Reversible Transgene Expression Reduces Fratricide and Permits 4-1BB Costimulation of CAR T Cells Directed to T-cell Malignancies. Cancer Immunol Res. 2018;6(1):47-58.

117. Long AH, Haso WM, Shern JF, et al. 4-1BB costimulation ameliorates T cell exhaustion induced by tonic signaling of chimeric antigen receptors. Nature medicine. 2015;21(6):581-590.

118. Wei X, Ni S, Correll PH. Uncoupling ligand-dependent and -independent mechanisms for mitogen-activated protein kinase activation by the murine Ron receptor tyrosine kinase. J Biol Chem. 2005;280(42):35098-35107.

119. Eyquem J, Mansilla-Soto J, Giavridis T, et al. Targeting a CAR to the TRAC locus with CRISPR/Cas9 enhances tumour rejection. Nature. 2017;543(7643):113-117. 165

120. Brentjens RJ, Riviere I, Park JH, et al. Safety and persistence of adoptively transferred autologous CD19-targeted T cells in patients with relapsed or chemotherapy refractory B-cell leukemias. Blood. 2011;118(18):4817-4828.

121. Grupp SA, Kalos M, Barrett D, et al. Chimeric antigen receptor-modified T cells for acute lymphoid leukemia. N Engl J Med. 2013;368(16):1509-1518.

122. Kochenderfer JN, Dudley ME, Feldman SA, et al. B-cell depletion and remissions of malignancy along with cytokine-associated toxicity in a clinical trial of anti- CD19 chimeric-antigen-receptor-transduced T cells. Blood. 2012;119(12):2709- 2720.

123. Porter DL, Kalos M, Zheng Z, Levine B, June C. Chimeric Antigen Receptor Therapy for B-cell Malignancies. J Cancer. 2011;2:331-332.

124. Porter DL, Levine BL, Kalos M, Bagg A, June CH. Chimeric antigen receptor- modified T cells in chronic lymphoid leukemia. N Engl J Med. 2011;365(8):725- 733.

125. Kalos M, Levine BL, Porter DL, et al. T cells with chimeric antigen receptors have potent antitumor effects and can establish memory in patients with advanced leukemia. Sci Transl Med. 2011;3(95):95ra73.

126. Enblad G, Karlsson H, Gammelgard G, et al. A PHASE I/IIa TRIAL USING CD19-TARGETED THIRD GENERATION CAR T CELLS FOR LYMPHOMA AND LEUKEMIA. Clin Cancer Res. 2018.

127. Raje NS, Berdeja JG, Lin Y, et al. bb2121 anti-BCMA CAR T-cell therapy in patients with relapsed/refractory multiple myeloma: Updated results from a multicenter phase I study. ASCO anual meeting. 2018.

128. Schneider D, Xiong Y, Wu D, et al. A tandem CD19/CD20 CAR lentiviral vector drives on-target and off-target antigen modulation in leukemia cell lines. J Immunother Cancer. 2017;5:42.

129. Sotillo E, Barrett DM, Black KL, et al. Convergence of Acquired Mutations and Alternative Splicing of CD19 Enables Resistance to CART-19 Immunotherapy. Cancer Discov. 2015;5(12):1282-1295. 166

130. Maus MV, June CH. Making Better Chimeric Antigen Receptors for Adoptive T- cell Therapy. Clin Cancer Res. 2016;22(8):1875-1884.

131. O'Rourke DM, Nasrallah MP, Desai A, et al. A single dose of peripherally infused EGFRvIII-directed CAR T cells mediates antigen loss and induces adaptive resistance in patients with recurrent glioblastoma. Sci Transl Med. 2017;9(399).

132. Scarfo I, Maus MV. Current approaches to increase CAR T cell potency in solid tumors: targeting the tumor microenvironment. J Immunother Cancer. 2017;5:28.

133. Lee DW, Gardner R, Porter DL, et al. Current concepts in the diagnosis and management of cytokine release syndrome. Blood. 2014;124(2):188-195.

134. Giavridis T, van der Stegen SJC, Eyquem J, Hamieh M, Piersigilli A, Sadelain M. CAR T cell-induced cytokine release syndrome is mediated by macrophages and abated by IL-1 blockade. Nat Med. 2018;24(6):731-738.

135. Norelli M, Camisa B, Barbiera G, et al. Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syndrome and neurotoxicity due to CAR T cells. Nat Med. 2018;24(6):739-748.

136. Sakaguchi S. Regulatory T cells: history and perspective. Methods Mol Biol. 2011;707:3-17.

137. Golubovskaya V, Wu L. Different Subsets of T Cells, Memory, Effector Functions, and CAR-T Immunotherapy. Cancers (Basel). 2016;8(3).

138. Weninger W, Crowley MA, Manjunath N, von Andrian UH. Migratory properties of naive, effector, and memory CD8(+) T cells. J Exp Med. 2001;194(7):953-966.

139. Sarkar S, Kalia V, Haining WN, Konieczny BT, Subramaniam S, Ahmed R. Functional and genomic profiling of effector CD8 T cell subsets with distinct memory fates. J Exp Med. 2008;205(3):625-640.

140. Klebanoff CA, Gattinoni L, Restifo NP. CD8+ T-cell memory in tumor immunology and immunotherapy. Immunol Rev. 2006;211:214-224.

167

141. Vaccari M, Trindade CJ, Venzon D, Zanetti M, Franchini G. Vaccine-induced CD8+ central memory T cells in protection from simian AIDS. J Immunol. 2005;175(6):3502-3507.

142. Castiglioni P, Gerloni M, Zanetti M. Genetically programmed B lymphocytes are highly efficient in inducing anti-virus protective immunity mediated by central memory CD8 T cells. Vaccine. 2004;23(5):699-708.

143. Wherry EJ, Teichgraber V, Becker TC, et al. Lineage relationship and protective immunity of memory CD8 T cell subsets. Nat Immunol. 2003;4(3):225-234.

144. Gattinoni L, Klebanoff CA, Restifo NP. Paths to stemness: building the ultimate antitumour T cell. Nat Rev Cancer. 2012;12(10):671-684.

145. Tumes DJ, Papadopoulos M, Endo Y, Onodera A, Hirahara K, Nakayama T. Epigenetic regulation of T-helper cell differentiation, memory, and plasticity in allergic asthma. Immunol Rev. 2017;278(1):8-19.

146. Bluestone JA, Mackay CR, O'Shea JJ, Stockinger B. The functional plasticity of T cell subsets. Nat Rev Immunol. 2009;9(11):811-816.

147. Terhune J, Berk E, Czerniecki BJ. Dendritic Cell-Induced Th1 and Th17 Cell Differentiation for Cancer Therapy. Vaccines (Basel). 2013;1(4):527-549.

148. Patel DD, Kuchroo VK. Th17 Cell Pathway in Human Immunity: Lessons from Genetics and Therapeutic Interventions. Immunity. 2015;43(6):1040-1051.

149. Sakaguchi S, Yamaguchi T, Nomura T, Ono M. Regulatory T cells and immune tolerance. Cell. 2008;133(5):775-787.

150. Sakaguchi S, Sakaguchi N, Asano M, Itoh M, Toda M. Immunologic self- tolerance maintained by activated T cells expressing IL-2 receptor alpha-chains (CD25). Breakdown of a single mechanism of self-tolerance causes various autoimmune diseases. J Immunol. 1995;155(3):1151-1164.

151. Fontenot JD, Gavin MA, Rudensky AY. Foxp3 programs the development and function of CD4+CD25+ regulatory T cells. Nat Immunol. 2003;4(4):330-336.

168

152. Bennett CL, Christie J, Ramsdell F, et al. The immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome (IPEX) is caused by mutations of FOXP3. Nat Genet. 2001;27(1):20-21.

153. Yamada A, Arakaki R, Saito M, Tsunematsu T, Kudo Y, Ishimaru N. Role of regulatory T cell in the pathogenesis of inflammatory bowel disease. World J Gastroenterol. 2016;22(7):2195-2205.

154. Noval Rivas M, Chatila TA. Regulatory T cells in allergic diseases. J Allergy Clin Immunol. 2016;138(3):639-652.

155. Di Ianni M, Falzetti F, Carotti A, et al. Tregs prevent GVHD and promote immune reconstitution in HLA-haploidentical transplantation. Blood. 2011;117(14):3921- 3928.

156. Gliwinski M, Iwaszkiewicz-Grzes D, Trzonkowski P. Cell-Based Therapies with T Regulatory Cells. BioDrugs. 2017;31(4):335-347.

157. Kretschmer K, Apostolou I, Jaeckel E, Khazaie K, von Boehmer H. Making regulatory T cells with defined antigen specificity: role in autoimmunity and cancer. Immunol Rev. 2006;212:163-169.

158. Kawahata K, Misaki Y, Yamauchi M, et al. Generation of CD4(+)CD25(+) regulatory T cells from autoreactive T cells simultaneously with their negative selection in the thymus and from nonautoreactive T cells by endogenous TCR expression. J Immunol. 2002;168(9):4399-4405.

159. van Nieuwenhuijze A, Liston A. The Molecular Control of Regulatory T Cell Induction. Prog Mol Biol Transl Sci. 2015;136:69-97.

160. Belkaid Y, Piccirillo CA, Mendez S, Shevach EM, Sacks DL. CD4+CD25+ regulatory T cells control Leishmania major persistence and immunity. Nature. 2002;420(6915):502-507.

161. Samy ET, Parker LA, Sharp CP, Tung KS. Continuous control of autoimmune disease by antigen-dependent polyclonal CD4+CD25+ regulatory T cells in the regional lymph node. J Exp Med. 2005;202(6):771-781.

169

162. Dieckmann D, Plottner H, Berchtold S, Berger T, Schuler G. Ex vivo isolation and characterization of CD4(+)CD25(+) T cells with regulatory properties from human blood. J Exp Med. 2001;193(11):1303-1310.

163. Takahashi T, Kuniyasu Y, Toda M, et al. Immunologic self-tolerance maintained by CD25+CD4+ naturally anergic and suppressive T cells: induction of autoimmune disease by breaking their anergic/suppressive state. Int Immunol. 1998;10(12):1969-1980.

164. Schmidt A, Oberle N, Krammer PH. Molecular mechanisms of treg-mediated T cell suppression. Front Immunol. 2012;3:51.

165. Vignali DA. Mechanisms of T(reg) Suppression: Still a Long Way to Go. Front Immunol. 2012;3:191.

166. Vignali DA, Collison LW, Workman CJ. How regulatory T cells work. Nat Rev Immunol. 2008;8(7):523-532.

167. Green EA, Gorelik L, McGregor CM, Tran EH, Flavell RA. CD4+CD25+ T regulatory cells control anti-islet CD8+ T cells through TGF-beta-TGF-beta receptor interactions in type 1 diabetes. Proc Natl Acad Sci U S A. 2003;100(19):10878-10883.

168. Nakamura K, Kitani A, Strober W. Cell contact-dependent immunosuppression by CD4(+)CD25(+) regulatory T cells is mediated by cell surface-bound transforming growth factor beta. J Exp Med. 2001;194(5):629-644.

169. Collison LW, Workman CJ, Kuo TT, et al. The inhibitory cytokine IL-35 contributes to regulatory T-cell function. Nature. 2007;450(7169):566-569.

170. Fontenot JD, Rasmussen JP, Gavin MA, Rudensky AY. A function for interleukin 2 in Foxp3-expressing regulatory T cells. Nat Immunol. 2005;6(11):1142-1151.

171. Burchill MA, Yang J, Vang KB, Farrar MA. Interleukin-2 receptor signaling in regulatory T cell development and homeostasis. Immunol Lett. 2007;114(1):1-8.

170

172. Gasteiger G, Kastenmuller W. Foxp3+ Regulatory T-cells and IL-2: The Moirai of T-cell Fates? Front Immunol. 2012;3:179.

173. Thornton AM, Donovan EE, Piccirillo CA, Shevach EM. Cutting edge: IL-2 is critically required for the in vitro activation of CD4+CD25+ T cell suppressor function. J Immunol. 2004;172(11):6519-6523.

174. Deaglio S, Dwyer KM, Gao W, et al. Adenosine generation catalyzed by CD39 and CD73 expressed on regulatory T cells mediates immune suppression. J Exp Med. 2007;204(6):1257-1265.

175. Tang Q, Adams JY, Tooley AJ, et al. Visualizing regulatory T cell control of autoimmune responses in nonobese diabetic mice. Nat Immunol. 2006;7(1):83- 92.

176. Tadokoro CE, Shakhar G, Shen S, et al. Regulatory T cells inhibit stable contacts between CD4+ T cells and dendritic cells in vivo. J Exp Med. 2006;203(3):505- 511.

177. Qureshi OS, Zheng Y, Nakamura K, et al. Trans-endocytosis of CD80 and CD86: a molecular basis for the cell-extrinsic function of CTLA-4. Science. 2011;332(6029):600-603.

178. Huang CT, Workman CJ, Flies D, et al. Role of LAG-3 in regulatory T cells. Immunity. 2004;21(4):503-513.

179. Cao X, Cai SF, Fehniger TA, et al. Granzyme B and perforin are important for regulatory T cell-mediated suppression of tumor clearance. Immunity. 2007;27(4):635-646.

180. Velaga S, Ukena SN, Dringenberg U, et al. Granzyme A Is Required for Regulatory T-Cell Mediated Prevention of Gastrointestinal Graft-versus-Host Disease. PLoS One. 2015;10(4):e0124927.

181. Choi BD, Gedeon PC, Herndon JE, 2nd, et al. Human regulatory T cells kill tumor cells through granzyme-dependent cytotoxicity upon retargeting with a bispecific antibody. Cancer Immunol Res. 2013;1(3):163.

171

182. Sula Karreci E, Eskandari SK, Dotiwala F, et al. Human regulatory T cells undergo self-inflicted damage via granzyme pathways upon activation. JCI Insight. 2017;2(21).

183. Ohkura N, Kitagawa Y, Sakaguchi S. Development and maintenance of regulatory T cells. Immunity. 2013;38(3):414-423.

184. Barbi J, Pardoll D, Pan F. Treg functional stability and its responsiveness to the microenvironment. Immunol Rev. 2014;259(1):115-139.

185. Kitagawa Y, Sakaguchi S. Molecular control of regulatory T cell development and function. Curr Opin Immunol. 2017;49:64-70.

186. Rubtsov YP, Niec RE, Josefowicz S, et al. Stability of the regulatory T cell lineage in vivo. Science. 2010;329(5999):1667-1671.

187. Hori S, Nomura T, Sakaguchi S. Control of regulatory T cell development by the transcription factor Foxp3. Science. 2003;299(5609):1057-1061.

188. Sugimoto N, Oida T, Hirota K, et al. Foxp3-dependent and -independent molecules specific for CD25+CD4+ natural regulatory T cells revealed by DNA microarray analysis. Int Immunol. 2006;18(8):1197-1209.

189. Akhmetzyanova I, Zelinskyy G, Littwitz-Salomon E, et al. CD137 Agonist Therapy Can Reprogram Regulatory T Cells into Cytotoxic CD4+ T Cells with Antitumor Activity. J Immunol. 2016;196(1):484-492.

190. Brunstein CG, Miller JS, Cao Q, et al. Infusion of ex vivo expanded T regulatory cells in adults transplanted with umbilical cord blood: safety profile and detection kinetics. Blood. 2011;117(3):1061-1070.

191. Zwang NA, Leventhal JR. Cell Therapy in Kidney Transplantation: Focus on Regulatory T Cells. J Am Soc Nephrol. 2017;28(7):1960-1972.

192. Kalil AC, Sandkovsky U, Florescu DF. Severe infections in critically ill solid organ transplant recipients. Clin Microbiol Infect. 2018.

172

193. Manuel O, Husain S, Kumar D, et al. Assessment of cytomegalovirus-specific cell-mediated immunity for the prediction of cytomegalovirus disease in high-risk solid-organ transplant recipients: a multicenter cohort study. Clin Infect Dis. 2013;56(6):817-824.

194. Issa F, Wood KJ. CD4+ regulatory T cells in solid organ transplantation. Curr Opin Organ Transplant. 2010;15(6):757-764.

195. Rezvani K, Mielke S, Ahmadzadeh M, et al. High donor FOXP3-positive regulatory T-cell (Treg) content is associated with a low risk of GVHD following HLA-matched allogeneic SCT. Blood. 2006;108(4):1291-1297.

196. Edinger M, Hoffmann P, Ermann J, et al. CD4+CD25+ regulatory T cells preserve graft-versus-tumor activity while inhibiting graft-versus-host disease after bone marrow transplantation. Nat Med. 2003;9(9):1144-1150.

197. Nadig SN, Wieckiewicz J, Wu DC, et al. In vivo prevention of transplant arteriosclerosis by ex vivo-expanded human regulatory T cells. Nat Med. 2010;16(7):809-813.

198. Kohm AP, Carpentier PA, Anger HA, Miller SD. Cutting edge: CD4+CD25+ regulatory T cells suppress antigen-specific autoreactive immune responses and central nervous system inflammation during active experimental autoimmune encephalomyelitis. J Immunol. 2002;169(9):4712-4716.

199. Safinia N, Grageda N, Scotta C, et al. Cell Therapy in Organ Transplantation: Our Experience on the Clinical Translation of Regulatory T Cells. Front Immunol. 2018;9:354.

200. Liu W, Putnam AL, Xu-Yu Z, et al. CD127 expression inversely correlates with FoxP3 and suppressive function of human CD4+ T reg cells. J Exp Med. 2006;203(7):1701-1711.

201. Hoffmann P, Eder R, Boeld TJ, et al. Only the CD45RA+ subpopulation of CD4+CD25high T cells gives rise to homogeneous regulatory T-cell lines upon in vitro expansion. Blood. 2006;108(13):4260-4267.

173

202. Chen W, Jin W, Hardegen N, et al. Conversion of peripheral CD4+CD25- naive T cells to CD4+CD25+ regulatory T cells by TGF-beta induction of transcription factor Foxp3. J Exp Med. 2003;198(12):1875-1886.

203. Rossetti M, Spreafico R, Saidin S, et al. Ex vivo-expanded but not in vitro- induced human regulatory T cells are candidates for cell therapy in autoimmune diseases thanks to stable demethylation of the FOXP3 regulatory T cell-specific demethylated region. J Immunol. 2015;194(1):113-124.

204. Wright GP, Notley CA, Xue SA, et al. Adoptive therapy with redirected primary regulatory T cells results in antigen-specific suppression of arthritis. Proc Natl Acad Sci U S A. 2009;106(45):19078-19083.

205. Tang Q, Vincenti F. Transplant trials with Tregs: perils and promises. J Clin Invest. 2017;127(7):2505-2512.

206. Hoeppli RE, MacDonald KG, Levings MK, Cook L. How antigen specificity directs regulatory T-cell function: self, foreign and engineered specificity. HLA. 2016;88(1-2):3-13.

207. Stephens LA, Malpass KH, Anderton SM. Curing CNS autoimmune disease with myelin-reactive Foxp3+ Treg. Eur J Immunol. 2009;39(4):1108-1117.

208. Kim YC, Zhang AH, Su Y, et al. Engineered antigen-specific human regulatory T cells: immunosuppression of FVIII-specific T- and B-cell responses. Blood. 2015;125(7):1107-1115.

209. Elinav E, Waks T, Eshhar Z. Redirection of regulatory T cells with predetermined specificity for the treatment of experimental colitis in mice. Gastroenterology. 2008;134(7):2014-2024.

210. Fransson M, Piras E, Burman J, et al. CAR/FoxP3-engineered T regulatory cells target the CNS and suppress EAE upon intranasal delivery. J Neuroinflammation. 2012;9:112.

211. Blat D, Zigmond E, Alteber Z, Waks T, Eshhar Z. Suppression of murine colitis and its associated cancer by carcinoembryonic antigen-specific regulatory T cells. Mol Ther. 2014;22(5):1018-1028. 174

212. Skuljec J, Chmielewski M, Happle C, et al. Chimeric Antigen Receptor- Redirected Regulatory T Cells Suppress Experimental Allergic Airway Inflammation, a Model of Asthma. Front Immunol. 2017;8:1125.

213. Boardman DA, Philippeos C, Fruhwirth GO, et al. Expression of a Chimeric Antigen Receptor Specific for Donor HLA Class I Enhances the Potency of Human Regulatory T Cells in Preventing Human Skin Transplant Rejection. Am J Transplant. 2017;17(4):931-943.

214. Yoon J, Schmidt A, Zhang AH, Konigs C, Kim YC, Scott DW. FVIII-specific human chimeric antigen receptor T-regulatory cells suppress T- and B-cell responses to FVIII. Blood. 2017;129(2):238-245.

215. Grupp SA, Kalos M, Barrett D, et al. Chimeric antigen receptor-modified T cells for acute lymphoid leukemia. The New England journal of medicine. 2013;368(16):1509-1518.

216. Kochenderfer JN, Dudley ME, Kassim SH, et al. Chemotherapy-refractory diffuse large B-cell lymphoma and indolent B-cell malignancies can be effectively treated with autologous T cells expressing an anti-CD19 chimeric antigen receptor. J Clin Oncol. 2015;33(6):540-549.

217. Lander ES. The new genomics: global views of biology. Science. 1996;274(5287):536-539.

218. Yang Y, Kohler ME, Chien CD, et al. TCR engagement negatively affects CD8 but not CD4 CAR T cell expansion and leukemic clearance. Sci Transl Med. 2017;9(417).

219. Cholia RP, Nayyar H, Kumar R, Mantha AK. Understanding the Multifaceted Role of Ectonucleotide Pyrophosphatase/Phosphodiesterase 2 (ENPP2) and its Altered Behaviour in Human Diseases. Curr Mol Med. 2015;15(10):932-943.

220. Drenkard D, Becke FM, Langstein J, et al. CD137 is expressed on blood vessel walls at sites of inflammation and enhances monocyte migratory activity. FASEB J. 2007;21(2):456-463.

175

221. Aijo T, Edelman SM, Lonnberg T, et al. An integrative computational systems biology approach identifies differentially regulated dynamic transcriptome signatures which drive the initiation of human T helper cell differentiation. BMC Genomics. 2012;13:572.

222. Lund R, Aittokallio T, Nevalainen O, Lahesmaa R. Identification of novel genes regulated by IL-12, IL-4, or TGF-beta during the early polarization of CD4+ lymphocytes. J Immunol. 2003;171(10):5328-5336.

223. Pegram HJ, Lee JC, Hayman EG, et al. Tumor-targeted T cells modified to secrete IL-12 eradicate systemic tumors without need for prior conditioning. Blood. 2012;119(18):4133-4141.

224. Tian Y, Cox MA, Kahan SM, Ingram JT, Bakshi RK, Zajac AJ. A Context- Dependent Role for IL-21 in Modulating the Differentiation, Distribution, and Abundance of Effector and Memory CD8 T Cell Subsets. J Immunol. 2016;196(5):2153-2166.

225. Shalek AK, Satija R, Adiconis X, et al. Single-cell transcriptomics reveals bimodality in expression and splicing in immune cells. Nature. 2013;498(7453):236-240.

226. Picelli S, Faridani OR, Bjorklund AK, Winberg G, Sagasser S, Sandberg R. Full- length RNA-seq from single cells using Smart-seq2. Nat Protoc. 2014;9(1):171- 181.

227. Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 2009;10(3):R25.

228. Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics. 2011;12:323.

229. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47.

176

230. Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000;25(1):25-29.

231. The Gene Ontology C. Expansion of the Gene Ontology knowledgebase and resources. Nucleic Acids Res. 2017;45(D1):D331-D338.

232. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545-15550.

233. Liberzon A, Subramanian A, Pinchback R, Thorvaldsdottir H, Tamayo P, Mesirov JP. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011;27(12):1739-1740.

234. Dawson NAJ, Vent-Schmidt J, Levings MK. Engineered Tolerance: Tailoring Development, Function, and Antigen-Specificity of Regulatory T Cells. Front Immunol. 2017;8:1460.

235. Zhang R, Borges CM, Fan MY, Harris JE, Turka LA. Requirement for CD28 in Effector Regulatory T Cell Differentiation, CCR6 Induction, and Skin Homing. J Immunol. 2015;195(9):4154-4161.

236. Zheng G, Wang B, Chen A. The 4-1BB costimulation augments the proliferation of CD4+CD25+ regulatory T cells. J Immunol. 2004;173(4):2428-2434.

237. Yang Y, Lin T, Jacoby E, et al. CD4 CAR T Cells Mediate CD8-like Cytotoxic Anti-Leukemic Effects Resulting in Leukemic Clearance and Are Less Susceptible to Attenuation By Endogenous TCR Activation Than CD8 CAR T Cells. Blood. 2015;126(23):100-100.

238. Kmieciak M, Gowda M, Graham L, et al. Human T cells express CD25 and Foxp3 upon activation and exhibit effector/memory phenotypes without any regulatory/suppressor function. J Transl Med. 2009;7:89.

239. Thornton AM, Shevach EM. CD4+CD25+ immunoregulatory T cells suppress polyclonal T cell activation in vitro by inhibiting interleukin 2 production. J Exp Med. 1998;188(2):287-296.

177

240. Maynard CL, Harrington LE, Janowski KM, et al. Regulatory T cells expressing interleukin 10 develop from Foxp3+ and Foxp3- precursor cells in the absence of interleukin 10. Nat Immunol. 2007;8(9):931-941.

241. Kehrl JH, Wakefield LM, Roberts AB, et al. Production of transforming growth factor beta by human T lymphocytes and its potential role in the regulation of T cell growth. J Exp Med. 1986;163(5):1037-1050.

242. Tran DQ, Andersson J, Hardwick D, Bebris L, Illei GG, Shevach EM. Selective expression of latency-associated peptide (LAP) and IL-1 receptor type I/II (CD121a/CD121b) on activated human FOXP3+ regulatory T cells allows for their purification from expansion cultures. Blood. 2009;113(21):5125-5133.

243. Nowak A, Lock D, Bacher P, et al. CD137+CD154- Expression As a Regulatory T Cell (Treg)-Specific Activation Signature for Identification and Sorting of Stable Human Tregs from In Vitro Expansion Cultures. Front Immunol. 2018;9:199.

244. Hagness M, Henjum K, Landskron J, et al. Kinetics and activation requirements of contact-dependent immune suppression by human regulatory T cells. J Immunol. 2012;188(11):5459-5466.

245. Gondek DC, Lu LF, Quezada SA, Sakaguchi S, Noelle RJ. Cutting edge: contact- mediated suppression by CD4+CD25+ regulatory cells involves a granzyme B- dependent, perforin-independent mechanism. J Immunol. 2005;174(4):1783- 1786.

246. Sommermeyer D, Hill T, Shamah SM, et al. Fully human CD19-specific chimeric antigen receptors for T-cell therapy. Leukemia. 2017;31(10):2191-2199.

247. Johnson LA, Scholler J, Ohkuri T, et al. Rational development and characterization of humanized anti-EGFR variant III chimeric antigen receptor T cells for glioblastoma. Sci Transl Med. 2015;7(275):275ra222.

248. Choi BK, Bae JS, Choi EM, et al. 4-1BB-dependent inhibition of immunosuppression by activated CD4+CD25+ T cells. J Leukoc Biol. 2004;75(5):785-791.

178

249. Hippen KL, Harker-Murray P, Porter SB, et al. Umbilical cord blood regulatory T- cell expansion and functional effects of tumor necrosis factor receptor family members OX40 and 4-1BB expressed on artificial antigen-presenting cells. Blood. 2008;112(7):2847-2857.

250. Grossman WJ, Verbsky JW, Barchet W, Colonna M, Atkinson JP, Ley TJ. Human T regulatory cells can use the perforin pathway to cause autologous target cell death. Immunity. 2004;21(4):589-601.

251. Efimova OV, Kelley TW. Induction of granzyme B expression in T-cell receptor/CD28-stimulated human regulatory T cells is suppressed by inhibitors of the PI3K-mTOR pathway. BMC Immunol. 2009;10:59.

252. Sun M, Fink PJ. A new class of reverse signaling costimulators belongs to the TNF family. J Immunol. 2007;179(7):4307-4312.

253. Ju S, Ju S, Ge Y, et al. A novel approach to induce human DCs from monocytes by triggering 4-1BBL reverse signaling. Int Immunol. 2009;21(10):1135-1144.

254. Srivastava RM, Trivedi S, Concha-Benavente F, et al. CD137 Stimulation Enhances Cetuximab-Induced Natural Killer: Dendritic Cell Priming of Antitumor T-Cell Immunity in Patients with Head and Neck Cancer. Clin Cancer Res. 2017;23(3):707-716.

255. Novy P, Huang X, Leonard WJ, Yang Y. Intrinsic IL-21 signaling is critical for CD8 T cell survival and memory formation in response to vaccinia viral infection. J Immunol. 2011;186(5):2729-2738.

256. Ramsburg EA, Publicover JM, Coppock D, Rose JK. Requirement for CD4 T cell help in maintenance of memory CD8 T cell responses is epitope dependent. J Immunol. 2007;178(10):6350-6358.

257. Wang JC, Livingstone AM. Cutting edge: CD4+ T cell help can be essential for primary CD8+ T cell responses in vivo. J Immunol. 2003;171(12):6339-6343.

258. Thomas AK, Maus MV, Shalaby WS, June CH, Riley JL. A cell-based artificial antigen-presenting cell coated with anti-CD3 and CD28 antibodies enables rapid

179

expansion and long-term growth of CD4 T lymphocytes. Clin Immunol. 2002;105(3):259-272.

259. Disis ML, Bernhard H, Jaffee EM. Use of tumour-responsive T cells as cancer treatment. Lancet. 2009;373(9664):673-683.

260. White L, Krishnan S, Strbo N, et al. Differential effects of IL-21 and IL-15 on perforin expression, lysosomal degranulation, and proliferation in CD8 T cells of patients with human immunodeficiency virus-1 (HIV). Blood. 2007;109(9):3873- 3880.

261. Myers DR, Zikherman J, Roose JP. Tonic Signals: Why Do Lymphocytes Bother? Trends Immunol. 2017;38(11):844-857.

262. Brocker T. Survival of mature CD4 T lymphocytes is dependent on major histocompatibility complex class II-expressing dendritic cells. J Exp Med. 1997;186(8):1223-1232.

263. Martin B, Becourt C, Bienvenu B, Lucas B. Self-recognition is crucial for maintaining the peripheral CD4+ T-cell pool in a nonlymphopenic environment. Blood. 2006;108(1):270-277.

264. Allen F, Bobanga ID, Rauhe P, et al. CCL3 augments tumor rejection and enhances CD8(+) T cell infiltration through NK and CD103(+) dendritic cell recruitment via IFNgamma. Oncoimmunology. 2018;7(3):e1393598.

265. Dons EM, Raimondi G, Cooper DK, Thomson AW. Induced regulatory T cells: mechanisms of conversion and suppressive potential. Hum Immunol. 2012;73(4):328-334.

266. Williams JM, Duckworth CA, Burkitt MD, Watson AJ, Campbell BJ, Pritchard DM. Epithelial cell shedding and barrier function: a matter of life and death at the small intestinal villus tip. Vet Pathol. 2015;52(3):445-455.

267. Huebner EA, Strittmatter SM. Axon regeneration in the peripheral and central nervous systems. Results Probl Cell Differ. 2009;48:339-351.

180

268. So T, Lee SW, Croft M. Immune regulation and control of regulatory T cells by OX40 and 4-1BB. Cytokine Growth Factor Rev. 2008;19(3-4):253-262.

269. Ramello MC, Fang B, Koomen JM, Haura EB, Abate-Daga D. Novel Phosphorylation Sites in the CD28 Costimulatory Domain Shape CAR-T Cell Function. American Society of Gene & Cell Terapy Abstract. 2018.

270. Fujii H, Luo ZJ, Kim HJ, et al. Humanized Chronic Graft-versus-Host Disease in NOD-SCID il2rgamma-/- (NSG) Mice with G-CSF-Mobilized Peripheral Blood Mononuclear Cells following Cyclophosphamide and Total Body Irradiation. PLoS One. 2015;10(7):e0133216.

271. Roybal KT, Rupp LJ, Morsut L, et al. Precision Tumor Recognition by T Cells With Combinatorial Antigen-Sensing Circuits. Cell. 2016;164(4):770-779.

181