Discovery of Novel Epigenetic Regulators of CD8+ T Cell Effector Function

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

Citation Tay, Rong En. 2019. Discovery of Novel Epigenetic Regulators of CD8+ T Cell Effector Function. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:42029527

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 Discovery of Novel Epigenetic Regulators of CD8+ T Cell Effector Function

A dissertation presented

Rong En Tay

The Department of Medical Sciences

in partial fulfilment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Immunology

Harvard University

Cambridge, Massachusetts

April 2019

Discovery of Novel Epigenetic Regulators of CD8+ T Cell Effector Function

Abstract

CD8+ cytotoxic T lymphocytes (CTLs) play a key role in acquired immunity by killing infected or cancerous cells. Upon antigen recognition and activation, the majority of naïve CD8+ T cells differentiate into potent but short-lived effector CTLs, while a small fraction generates long-lived memory cells that are poised to proliferate rapidly upon antigen re-encounter. While transcriptional control of CD8+ T cell differentiation and effector function has been extensively studied, little is known about epigenetic regulation of these processes.

Here we use two screening approaches to uncover novel epigenetic regulators of CD8+ T cell effector function. We first engineered a CRISPR-Cas9 genetic screening platform to reveal epigenetic regulators in tumour-infiltrating CD8+ T cells that could potentially be targeted to improve CD8+ anti- tumour function. This approach identified CARM1 as a strong candidate epigenetic regulator suppressing the accumulation of tumor-specific CD8+ T cells in the pre-clinical B16 melanoma tumour model.

We also used a functional pharmacological approach to screen for epigenetic regulators of CD8+

T cell effector function during in vitro T cell activation. We thus identified a novel role of the histone deacetylase HDAC3 as an epigenetic regulator of CD8+ T cell cytotoxicity and persistence. Hdac3-deficient

CD8+ T cells transiently acquired augmented cytotoxicity that was associated with early resistance to chronic LCMV infection but did not confer durable disease protection. Mechanistically, HDAC3 inhibited a gene programme of cytotoxicity-associated genes and terminal effector differentiation beginning early during CD8+ T cell activation, and required the transcription factors Runx3 and Blimp-1 as key downstream iii

mediators for its regulation of CD8+ T cell effector function. Targeting HDAC3 activity to epigenetically modulate the balance between a short-lived, potent cytotoxic effector state and a durable, long-lived response could potentially lead to the development of new approaches in adoptive T cell immunotherapy and in therapeutic vaccine development.

Table of Contents

Acknowledgements vi

Dedication vii

Chapter 1 – Introduction 1

Chapter 2 – Materials and Methods 7

Chapter 3 – Discovery of CARM1 as a suppressor of CD8+ T cell accumulation in 23 tumours using an in vivo CRISPR-Cas9 genetic screening platform

Chapter 4 – Discovery of HDAC3 as an inhibitor of CD8+ T cell cytotoxicity in an in 28 vitro functional pharmacologic screen

Chapter 5 – HDAC3 is required during T cell activation for persistence of antigen- 40 experienced CD8+ T cells

Chapter 6 – Effects of loss of HDAC3 activity in CD8+ T cells in settings of chronic 51 antigen burden

Chapter 7 – Transcriptional and epigenetic signatures of HDAC3 60

Chapter 8 – Uncovering potential mechanisms of HDAC3-mediated regulation of 71 CD8+ T cell effector function using genetic approaches

Chapter 9 – Summary and Discussion 87

Appendix 99

References 106

Acknowledgements

I reverently thank the LORD my God for sustaining me through this long and difficult journey. By His grace have I come thus far while keeping my honour and integrity intact.

I thank my PhD advisor, Kai Wucherpfennig, for his supervision and direction and for giving me the opportunity to work on an interesting project.

I am deeply and forever grateful to Hye-Jung Kim for her unwavering support of me, both scientifically and personally. Thank you for being a true friend and mentor to me and for believing in me.

I thank my collaborators Paloma Cejas, Henry Long, and Clifford Meyer (DFCI) for their invaluable help in performing the ChIP-seq analyses, and Peng Jiang (DFCI) for computational analysis of the in vivo screen.

I am very grateful to my dissertation advisory committee members Ulrich von Andrian (HMS), Myles Brown (DFCI), and Nicholas Haining (DFCI) for their helpful guidance and constructive feedback.

I thank Scott Hiebert (Vanderbilt), Christina Weng and David Fisher (MGH), Steven Elledge (HMS), and Ulrich von Andrian (HMS) for their kind gifts of tools and resources in support of my projects.

Special thanks to Olamide Olawoyin for her incredible hard work and dedication during our time working together. It was an honour and privilege to teach one so deserving as yourself.

I thank Sabrina Haag for her insightful comments and scientific input into my projects.

I thank my fellowship sponsors, the Agency for Science, Technology, and Research (A*STAR, Singapore), for their generous financial support over the course of my doctoral work.

I thank all other members of the Wucherpfennig lab for their collegiality. vi

Dedication

Dedicated to the LORD my God as a testimony of His faithfulness to me, to my loving family for their unwavering support and continual prayers for the safe return of their son and brother, to my dearest fiancée Kayla, for her patience and love during this period of separation, and to Kayla’s father Sam, for his faith that this day would come.

Give thanks to the LORD, for He is good.

His love endures forever.

- Psalm 136:1 (NIV)

vii

Chapter 1

Introduction

Upon antigen encounter during inflammation, naïve CD8+ T cells undergo phenotypic changes during activation to develop into effector cytotoxic T lymphocytes that mediate immunity via contact-dependent killing of infected or malignant transformed cells and by secreting effector cytokines such as IFN-γ and

TNF-α. By the peak of the CD8+ T cell effector response to acute antigen exposure, activated CD8+ T cells are already committed to one of two cell fates – a short-lived terminally-differentiated state with potent effector function1,2, or a long-lived memory precursor phenotype with high proliferative potential but weak effector function3,4. After antigen clearance, most of the terminally-differentiated effector cells die off, resulting in a quantitative and qualitative contraction of the CD8+ T cell response, whereas memory precursor cells persist and subsequently differentiate into memory cells, which are poised to mount a robust and rapid defence upon secondary antigen encounter. All these differentiation processes require co-ordination between the molecular pathways integrating extracellular signalling inputs with cell- intrinsic programming for proper acquisition of full effector CD8+ T cell function as well as for proper control and homeostasis of the CD8+ T cell immune response5.

The transcriptional regulation of the developmental processes following CD8+ T cell activation has been extensively described and studied. Advances in RNA-sequencing technology and analysis techniques have led to comprehensive profiling of the roles of such key transcription factors as T-bet, Eomes, Blimp-1, and

Bcl-6 in regulating the acquisition of effector functions of CD8+ T cells, as well as in regulating the commitment of activated CD8+ T cells to one of the two phenotypically-distinct cell fates following acute antigen encounter5,6. These studies have been indispensable to our current understanding of CD8+ T cell differentiation as a process governed by interactions of key driver molecules within an intricately complex transcriptional network7. Each transcription factor governs a distinct suite of effector functions or

2 phenotypes (e.g. T-bet is required to potentiate IFN-γ secretion and cytotoxicity, as well as effector phenotype acquisition8), and the co-ordination of several of these transcription factors across time thus constitutes a fundamental molecular control mechanism governing the acquisition of effector function and cell fate commitment.

Whereas control of CD8+ T cell function and differentiation is proximally mediated by transcription factors, it is the epigenetic landscape9 that provides the framework within which the gene regulatory network operates. Broadly speaking, epigenetic processes constrain the spectrum of potential phenotypes that a given cell can acquire by covalently modifying DNA bases or specific amino acid residues of chromatin- associated proteins such as histones. These modifications serve as beacons for ‘reader’ proteins that recognise and recruit specific transcription factors to the marked regions of chromatin. The patterns of epigenetic marks also generate regions of chromatin with differential conformational states and accessibility to transcriptional machinery (i.e. ‘open’ and ‘closed’ chromatin)10,11. Well-described examples of these processes include transcriptional silencing via cytosine methylation in regions rich with of CpG islands12 and the association of highly condensed, transcriptionally-silent heterochromatin with trimethylation of nearby histone 3 lysine 27 residues13. Epigenetic regulation thus governs the development of the functional specialisation of different cell types in multicellular organisms, and also serves to define terminal lineage commitment and loss of pluripotency as cells progress along a series of distinct developmental states14.

Epigenetic processes also play a central role in establishing and maintaining phenotypic differences between distinct lineages within the haematopoietic system, such as those between myeloid and lymphoid compartments15 or even between more closely-related subsets such as conventional and regulatory CD4+ T cells16. However, the contribution of epigenetic regulatory processes to the dynamic

3 reprogramming of the immune response within a particular lineage of cells, in the same manner that transcription factors regulate the effector phenotypes of immune cells, is less well understood. To address this gap in our knowledge, we focussed our efforts on discovering novel roles of epigenetic regulators during CD8+ T cell activation and differentiation, and to define specific suites of CD8+ T cell effector functions targeted by these regulators.

In contrast to the comprehensive literature regarding transcriptional regulation of CD8+ T cell activation, our current knowledge regarding epigenetic regulation of CD8+ T cell differentiation after activation is more limited and less systematic (recently reviewed here17). The majority of recent investigations working towards uncovering drivers of epigenetic regulation in CD8+ T cells and their mechanisms of action have taken either (1) a profiling approach characterising global patterns of epigenetic marks such as histone modifications or chromatin accessibility under different immunological conditions18-21, or have taken (2) a more classical genetics-type approach of characterising the phenotypes of mice genetically-deficient for a known epigenetic regulator using immunologically-relevant experimental models20,22. Two recent studies in particular have been highly informative for the field and are characteristic of each investigative approach. In the first study, the authors showed that histone 3 lysine 27 trimethylation mediated by the polycomb repressive complex (an epigenetic regulatory complex containing the histone methyltransferase EZH2) epigenetically silenced pro-memory genes in virus-specific CD8+ T cells in mice acutely infected with lymphocytic choriomeningitis virus. The authors also demonstrated that the memory development-associated transcription factor FOXO1 shielded these genes from epigenetic silencing, highlighting an interaction between epigenetic and transcriptional pathways governing CD8+ T cell fate commitment20. In the second study, the authors demonstrated that the histone methyltransferase SUV39H1 was responsible for trimethylation of histone 3 lysine 9 to silence stem-like memory phenotype-associated genes, thus regulating the balance between the development of memory

CD8+ cells that retained proliferative potential and terminally-differentiated effector CD8+ T cells22. Both of these studies highlight the complex interplay between transcriptional and epigenetic control of CD8+ T cell differentiation, and demonstrate a central role for histone covalent modifications in the epigenetic regulation of CD8+ T cell fate commitment.

While advances in epigenetic profiling technologies, particularly in low-input methods for ATAC- sequencing and ChIP-sequencing, have made whole-genome profiling approaches attractive, these experimental strategies are inherently limited in their ability to reveal unknown mechanisms of epigenetic regulation. This is because the readout signal of any given epigenetic mark is the net result of a many-to- one function with multiple input sources. Several epigenetic regulators (readers, erasers, and writers) all influence the levels of any given epigenetic modification23, such that the identities of these regulators are not easily deduced from the resultant traces of epigenetic marks alone. Moreover, there is a critical requirement for technically robust experimental systems in order to reproducibly generate functional phenotypes that can be reliably correlated with specific epigenetic marks with minimal biological noise and good resolution. Therefore, while most of the studies using profiling as their primary investigation approaches have generated large, richly detailed datasets, identifying the actual drivers of these epigenetic changes remains difficult due to the unfeasibility of systematically evaluating the contribution of every possible epigenetic regulator to the observed changes in epigenetic profiles.

On the other hand, reverse genetics approaches are limited by the availability of biological tools and model systems. Because the functional development of cells is also epigenetically regulated, careful experimental design and choice of biological tools are required to very specifically ablate or enhance the activity of any given epigenetic regulator gene. This is especially crucial when studying the activation of

CD8+ T cells, because their developmental processes utilise many of the same signalling mechanisms that

5 also govern the functional responses of mature cells in the periphery (e.g. TCR signalling pathways mediate both thymic development and activation of CD8+ T cells). Therefore, the utility of reverse genetics approaches is limited, particularly if non-specific conditional knockout models are used, or if temporal control of gene inactivation cannot be achieved. Furthermore, in such investigations, the functional phenotypes profiled are usually complex and broadly defined (e.g. memory formation). This further reduces the power of such experiments to link specific changes in phenotype to specific molecular perturbations. In cases of loss-of-function phenotypes associated with inactivation of epigenetic regulators, this inability to distinguish between causation due to aberrantly regulated epigenetic processes during development of CD8+ T cells, or causation due to epigenetic processes specifically in the context of CD8+ T cell activation, makes it difficult to establish direct epigenetic regulation of the functional phenotype.

In consideration of all these factors, we designed two different complementary approaches to uncover and characterise novel epigenetic regulators of CD8+ T cell effector function. We used both an unbiased reverse genetics approach in a preclinical tumour model system and a functional pharmacologic approach in an in vitro model of CD8+ T cell activation. To increase the likelihood of finding bona fide epigenetic regulators of CD8+ T cell functional phenotype, we specifically limited our search space to that of epigenetic regulators that affected the acquisition of effector functional by mature naïve CD8+ T cells, and also strove to define the function of candidate epigenetic regulators in terms of their effects on clearly- defined CD8+ T cell functions such as cytotoxicity, proliferation, and effector cytokine secretion. Using our two experimental approaches, we uncovered CARM1 and HDAC3 as novel epigenetic regulators of CD8+

T cells, and further showed that HDAC3 specifically regulates the cytotoxicity and persistence of the CD8+

T cell response following activation.

Chapter 2

Materials and Methods

Mice

All mice were maintained in specific pathogen-free (SPF) conditions and used in accordance to guidelines of the Dana-Farber Cancer Institute Institutional Animal Care and Use Committee (DFCI IACUC). C57BL/6

(strain 000664), CD45.1+ congenic (strain 002014), Rag1-/- (strain 002216), and OT-I+ (strain 003831) mice were purchased from The Jackson Laboratory. OT-I+ Cas9+ mice were generated by crossing OT-I+ mice with mice carrying a Rosa26-targeted knock-in of Streptococcus pyogenes Cas9 (Jackson strain 024858) for constitutive Cas9 expression; all Cas9+ mice in this study were homozygous for the Cas9 knock-in allele at the Rosa26 locus. Hdac3fl/fl mice were developed by and generously given by Scott Hiebert (Vanderbilt

University, TN)24. Hdac3fl/fl mice were crossed with E8I-Cre+ mice (Jackson strain 008766, kindly given by

Hye-Jung Kim) and OT-I+ mice to generate TCR-polyclonal and OT-I TCR-transgenic mice with a CD8 T cell- restricted deletion of Hdac3; the latter strain was further crossed with OT-I+ Cas9+ mice to generate OT-I+

Cas9+ E8I-Cre+;Hdac3fl/fl mice. Hdac3fl/fl mice were also crossed with UBC-Cre-ERT2 mice (Jackson strain

008085) and OT-I+ mice to generate OT-I TCR-transgenic mice with a tamoxifen-inducible deletion of

Hdac3. Hdac3-conditional knockout strains were maintained by crossing Cre+;Hdac3fl/fl mice with

Cre-;Hdac3fl/fl mice to ensure hemizygous inheritance of the Cre transgene and to generate Cre- Hdac3-

WT littermate controls within each generation of mice. All mice used as hosts in experiments were male mice 8 to 9 weeks of age. For adoptive transfer experiments, mice 6-12 weeks of age of both sexes were used as T cell donors.

Flow cytometry

Fluorochrome-conjugated antibodies against CD3ε (145-2C11), CD5 (53-7.3), CD8α (53-6.7), CD11b

(M1/70), CD11c (N418), CD25 (PC61), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), CD107a

(1D4B), CD127 (A7R34), H-2 (M1/42) (pan-MHC I), H-2Kb-SIINFEKL (25-D1.16), PD-1 (29F.1A12), TCRβ

(H57-597), Thy1.1 (OX-7), TIM-3 (RMT3-23), IFN-γ (XMG1.2), TNF-α (MP6-XT22), and T-bet (4B10) were purchased from Biolegend. Antibodies against CD4 (RM4-5) and Granzyme B (NGZB) were purchased from

Thermo Fisher Scientific. Antibodies against Runx3 (R3-5G4) were purchased from BD Biosciences. PE- conjugated Kb-SIINFEKL (OVA peptide recognised by OT-I TCR) and Db-KAVYNFATC (immunodominant

LCMV GP33 epitope) tetramers were purchased from MBL International Corporation. Antibodies against surface epitopes were used at 1:200 dilution, except α-CD44 and α-CD62L antibodies which were used at

1:100 dilution. Antibodies against intracellular epitopes were used at 1:50 dilution, except α-Runx3 antibodies which were used at 1:20 dilution.

All flow cytometry samples were first stained with Zombie UV fixable viability dye (Biolegend) for live/dead exclusion at 1:200 dilution, then treated with unconjugated α-CD16/32 antibodies (93, Biolegend) at 1:50 dilution to block non-specific binding to Fc receptors prior to staining. For cell sorting, live/dead staining was performed using 5 μg/ml DAPI (Biolegend) instead of Zombie UV dye. Where required, staining with tetramers was done at 1:100 dilutions after live/dead staining and prior to Fc blockade. For intracellular staining of cytokines and transcription factors, samples were fixed and permeabilised with the eBioscience™ FoxP3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific). Where indicated, samples were re-stimulated with PMA and Ionomycin using eBioscience™ Cell Stimulation Cocktail (plus protein transport inhibitors) (Thermo Fisher Scientific), or with 5 μg/ml of indicated peptide(s) plus eBioscience™ Protein Transport Inhibitor Cocktail (Thermo Fisher Scientific) for 3-4 hours prior to staining

9 for flow cytometry. Unstimulated samples (treated only with Protein Transport Inhibitor Cocktail) were acquired in parallel as gating controls. All flow cytometry data were acquired using a LSRFortessa X-20 with an attached HTS module (BD Biosciences) and all cell sorting was performed using a FACSAria IIIu (BD

Biosciences). Flow cytometry data were analysed using FlowJo software (FlowJo LLC).

Cell culture and cell lines

Cell culture media and supplements were purchased from Gibco (Thermo Fisher Scientific). All primary cell cultures and the MC38 cell line were grown in RPMI media supplemented with 10% (v/v) foetal bovine serum (FBS), 1x GlutaMAX™ (Gibco), 100 U penicillin-streptomycin, 1 mM sodium pyruvate, 20 mM HEPES, and 50 μM 2-Mercaptoethanol. All other cell lines were grown in DMEM media supplemented with 10%

(v/v) FBS, 1x GlutaMAX™, 100 U penicillin-streptomycin, and 10 mM HEPES. B16F10 melanoma and MC38 colon adenocarcinoma cell lines were obtained from the ATCC. OVA-expressing tumour cell lines were generated by lentiviral transduction of parental lines with a pHAGE expression vector bearing an N- terminally truncated variant of chicken ovalbumin that was sub-cloned from pcDNA3-deltaOVA (Addgene plasmid #6459525). zsGreen+ transductants were sorted to purity to establish the cell lines. B16-OVA and

MC38-OVA were validated by flow cytometry for zsGreen expression and to measure expression of the

OVA epitope SIINFEKL in complex with H2-Kb on the cell surfaces, as well as by functional evaluation of their ability to activate OT-I+ CD8+ T cells.

Preparation of cell suspensions from organs

For spleens and lymph nodes, organ samples were mechanically dissociated on 70 μm cell strainers with unsupplemented RPMI media. For liver and kidney samples, organs were minced with scalpels in RPMI

10 and further dissociated in GentleMACS C tubes on a GentleMACS dissociator (Miltenyi Biotec), and the resultant cell slurries were separated at 1200 x g on a 40%/70% discontinuous Percoll gradient (Sigma

Aldrich). B16 and MC38 tumours were dissected out with care to ensure that adjacent tumour-draining lymph nodes were not taken out together with the tumour samples. Tumours were minced with scalpels in RPMI, then further dissociated in GentleMACS C tubes on a GentleMACS dissociator (Miltenyi Biotec) using the ‘37C_m_TDK_1’ programme with an enzyme mix containing 1 mg/ml collagenase D (Sigma

Aldrich), 20 U/ml DNAse I (Sigma Aldrich), and 100 μg/ml hyaluronidase Type V (Sigma Aldrich).

Dissociated B16 tumour samples were further centrifuged twice at low-speeds (50 x g, 5 minutes), and the lymphocyte-enriched supernatants were collected and pooled together. Single-cell suspensions from spleens, livers, and tumours were treated with RBC lysis buffer (Biolegend) for one minute at room temperature after dissociation to remove erythrocytes.

CRISPR-Cas9-mediated gene knockout in CD8+ T cells by lentiviral transduction.

To efficiently transduce primary CD8+ T cells without prior TCR activation, we constructed the Thy1.1- marked lentiviral gRNA expression plasmid vector pLKO-gRNA-Thy1.1 by subcloning the gRNA expression cassette of lentiGuide-Puro (Addgene plasmid #5296326) into a pLKO.3G backbone (Addgene plasmid

#14748) and replacing the eGFP coding sequence with that of the surface marker Thy1.1. gRNA sequences targeting specific genes were picked from a list generated by the online sgRNA Design Tool (Broad

Institute, MA) and inserted into pLKO-gRNA-Thy1.1 on sticky ends generated by BsmBI digestion.

Lentivirus production in HEK 293 cells and CD8+ T cell transduction were performed as previously described27. Briefly, OT-I+ Cas9+ CD8+ T cells were cultured in 100 ng/ml IL-15 and 5 ng/ml IL-7 for 48 hours prior to spin-infection in retronectin-coated (Takara Bio) 24-well plates with concentrated supernatants lentivirus (MOI = 15). Spin-infection was done in RPMI media with 5 μg/ml protamine sulphate (Sigma

Aldrich) for 1.5-2 hours at 32°C. Cells were cultured for a further 72 hours post-transduction with 50 ng/ml

IL-15, 2.5 ng/ml IL-7, and 2 ng/ml IL-2 before magnetic enrichment for Thy1.1+ using an EasySep™ Mouse

CD90.1 Positive Selection Kit (StemCell Technologies) to ≥ 93% Thy1.1+ purity.

In vivo CRISPR-Cas9 screening for epigenetic regulators

For the pilot in vivo screens, we constructed 3 plasmid libraries of gRNA sequences (Supplementary Table

1) targeting 426 genes coding for epigenetic regulators in collaboration with Mamie Li, Qikai Xu, and

Steven Elledge (Harvard Medical School, Massachusetts). Each library contained 5 unique gRNA sequences targeting each of 142 candidate genes. In addition, 5 unique gRNA sequences targeting 7 genes previously shown to suppress CD8+ T cell accumulation in tumours as positive controls as well as 100 non-mouse genome-targeting gRNA sequences as negative controls. All gRNAs were cloned into the pLKO-gRNA-

Thy1.1 lentiviral expression vector. Concentrated lentiviral supernatants were prepared from transfected

HEK 293 cells as previously described27.

8-week old male C57BL/6 mice were injected subcutaneously with 2 x 105 B16F10 melanoma cells that had been passaged at least twice in vitro. 5 days later, tumour implantation was confirmed by physically checking the mice for palpable tumours before isolating CD8+ T cells from age-matched Cas9+ OT-I+ mice by negative magnetic selection using an EasySep™ Mouse CD8+ T Cell Isolation Kit (Stemcell Technologies).

Transduction with pooled lentiviral gRNA vectors and subsequent magnetic positive selection for expression of the lentiviral Thy1.1+ marker was performed as described in the previous section. Thy1.1+

OT-I+ CD8+ T cells were re-suspended in PBS and intravenously transferred into tumour-bearing hosts

(minimum tumour diameter of 7 mm) at 5 x 106 cells/mouse. A minimum of 10 tumour-bearing mice were injected per library of gRNAs screened. A small sample of the transferred cells (at least 1 x 106 cells) was

12 saved as the input fraction for quality control and later analysis, and a second sample was taken for flow cytometric measurement of Thy1.1+ purity.

10 days after CD8+ T cell transfer, recipient mice were randomly grouped into 2 groups with at least 5 mice each for sequential organ. Spleens and tumours were separately pooled across all mice within each group for preparation of single-cell suspensions as described previously. Total CD8+ T cells were isolated from splenic samples using the EasySep™ Mouse CD8+ T Cell Isolation Kit.

Total genomic DNA was prepared from live DAPI- CD8+ Thy1.1+ cells sorted from splenic and tumour organ suspensions and from the input samples using a Quick-DNA Microprep Plus Kit (Zymo Research). Deep- sequencing for gRNA sequences was performed by John Doench and the Genetic Perturbation Platform

(GPP) at the Broad Institute. Data from the screens of each of the 3 gRNA libraries were integrated and analysed using MAGeCK algorithms28 in collaboration with Peng Jiang and Shirley Liu (Dana-Farber Cancer

Institute, Massachusetts).

For the validation screen, a new gRNA library (Supplementary Table 2) was constructed in the pLKO-gRNA-

Thy1.1 plasmid vector in collaboration with the GPP (Broad Institute, Massachusetts) containing 6 gRNA sequences targeting each of 29 gene selected from the top hits of the pilot screens, as well as 6 gRNA sequences targeting Pdcd1 and Cblb as positive controls and a mix of (93 non-targeting + 93 intergenic region-targeting) negative gRNA sequences. Screening and data analysis were performed essentially as described for the pilot screens, except that we analysed a total of 12 tumour-bearing mice divided into 3 groups of 4 mice each for organ collection, cell sorting, and genomic DNA preparation.

In vitro activation of OT-I+ CD8+ T cells with OVA peptide-pulsed BMDCs

One week prior to T cell activation, bone marrow-derived dendritic cells (BMDCs) were generated by culturing bone marrow cells isolated from the femora and tibiae of C57BL/6-background mice for 6 days in the presence of 10 ng/ml GM-CSF (Biolegend) as previously described29. On day 6, immature BMDCs were activated with 100 ng/ml lipopolysaccharide (Sigma Aldrich) overnight. The next day, activated

BMDCs were γ-irradiated (3500 rad) and pulsed with 100 ng/ml SIINFEKL peptide at 37°C for at least an hour. OT-I+ CD8+ T cells were isolated from the spleens and peripheral lymph nodes (inguinal, brachial, and cervical) of OT-I TCR-transgenic mice by negative magnetic selection using an EasySep™ Mouse CD8+

T Cell Isolation Kit (Stemcell Technologies). BMDCs and OT-I+ CD8+ T cells were co-cultured in a 1:1 ratio at a concentration of 1 x 106 cells/ml in 96-well, 24-well, or 6-well tissue culture plates (at volumes of 0.1,

0.5, and 2 ml respectively). For in vitro HDAC3 inhibitor studies, RGFP966 (Selleck Chemicals) was prepared as a 10 mM stock in DMSO and diluted to the indicated concentrations directly into cell culture medium; control cell cultures were treated with an equivalent volume of DMSO vehicle.

In vitro functional pharmacological screen of epigenetic inhibitor drug library

96-well plates containing BMDC and OT-I+ CD8+ T cell co-cultures were set up as described in the preceding section. Immediately after OT-I+ CD8+ T cells were added to BMDCs, a custom small molecule library containing aliquots of 100 epigenetic inhibitor drugs in DMSO vehicle (Supplementary Table 3) was directly diluted into the 96-well plates using 96-well pin applicators (Thermo Fisher Scientific) to final concentrations of either 1 μM or 10 μM. The drug library was a generous gift from Christina Weng and

David Fisher (Massachusetts General Hospital, Massachusetts). Each 96-well plate also contained 10 control wells (DMSO vehicle only). Fresh media containing epigenetic inhibitor drugs or DMSO vehicle was

14 added on day 3 and 5 after activation. On day 7, cells were collected in 96-well plates and fixed for flow cytometric analysis of a panel of markers of effector CD8+ T cell function.

For each marker of CD8+ T cell effector phenotype, means and standard distributions of either the gMFIs

(for markers with unimodal staining patterns, e.g. CD25 and CD69), or of the percentage of positive cells

(for markers with bimodal distributions, e.g. Granzyme B and IFN-γ), were computed across the 10 control wells for each plate. For experimental wells, a marker was considered significantly perturbed if its gMFI was outside of 2 standard deviations of the mean gMFI of control wells (for unimodally-distributed parameters), or if the percentage of marker-positive cells was outside 1 standard deviation of the mean frequency of marker-positive cells in control wells (for bimodally-distributed parameters). The number of perturbed markers for each drug was enumerated and used as a score for ranking candidate drug hits.

Western blot analysis of gRNA knockout efficiency

After transduction with lentiviral gRNA expression vectors as previously described, Thy1.1+ cells were activated in vitro with 3 μg/ml plate-bound α-CD3ε (145-2C11, Biolegend) and 1μg/ml soluble α-CD28

(37.51, Biolegend). Cells were collected after 3 to 7 days post-activation for preparation of whole cell lysates in RIPA buffer with protease inhibitors (P8340, Sigma Aldrich) for subsequent immunoblot analysis.

Total cell lysates equivalent to 2-5 x 105 cells were loaded into each lane for SDS-PA gel electrophoresis and transfer to PVDF. Rabbit antibodies for Western blotting against HDAC3 (#2632, #85057), NCOR1

(#5948), Blimp-1 (#9115), Runx3 (#9647), and GAPDH (#2118) were purchased from Cell Signaling

Technology. Rabbit antibodies against NCOR2 (SMRT) (ab5802) were purchased from Abcam. Blot membranes were incubated with primary antibodies overnight at 4°C with gentle shaking at a dilution of

1:1000 in SuperBlock™ T20 (TBS) (Thermo Fisher Scientific), except α-GAPDH (1:5000) and α-NCOR1

15 antibodies (1:500). Blot membranes were then incubated with secondary goat α-rabbit HRP-conjugated antibodies (1:10000) (Cell Signaling Technology) for 1 hour at room temperature and then treated with

ECL reagent (Thermo Fisher Scientific). Protein bands were visualised on autoradiography film (Denville

Scientific) and the reduction of protein expression was quantified from scanned images using FIJI software

(NIH).

Cytotoxicity assays

For in vitro cytotoxicity assays, B16-OVA melanoma cells were treated with 10 ng/ml IFN-γ overnight and labelled with 51Cr the following day (1 hour, 37°C) and loaded into 96-well plates at 2 x 103 cells per well.

OT-I+ CD8+ T cells were collected from in vitro culture, resuspended at appropriate concentrations, and loaded on top of the B16-OVA target cells at effector-to-target ratios ranging from 10:1 to 1:1, with 4 replicates per condition. Plates were gently centrifuged to ensure cell-cell contact. 8 wells of B16-OVA targets without T cells were left untreated to measure spontaneous 51Cr release, and 8 wells of B16-OVA cells were fully lysed with 1% Triton-X in PBS to measure maximum 51Cr release. After 4 hours of co- culture, cells were pelleted, and the supernatant was collected for quantification of 51Cr radioactivity with a Microbeta2 microplate counter (Perkin Elmer). In vitro cytotoxicity was calculated as follows:

𝑆푎𝑚𝑝𝑙𝑒 𝑟𝑒𝑙𝑒푎𝑠𝑒 − 𝑆𝑝𝑜𝑛𝑡푎𝑛𝑒𝑜𝑢𝑠 𝑟𝑒𝑙𝑒푎𝑠𝑒 𝑆𝑝𝑒푐푖𝑓푖푐 𝑘푖𝑙𝑙푖𝑛𝑔 = 100% × 푀푎푥푖𝑚𝑢𝑚 𝑟𝑒𝑙𝑒푎𝑠𝑒 − 𝑆𝑝𝑜𝑛𝑡푎𝑛𝑒𝑜𝑢𝑠 𝑟𝑒𝑙𝑒푎𝑠𝑒

For in vivo cytotoxicity assays, target cells were prepared by depleting C57BL/6 splenocytes with biotin- conjugated α-CD3 (17A2, Biolegend) and α-NK1.1 (PK136, Biolegend) antibodies followed by negative magnetic selection with streptavidin beads (StemCell Technologies). Target cells were then loaded with

1 μM of either OVA SIINFEKL peptide (Sigma Aldrich) or the irrelevant control peptide gp100 KVPRNQDWL

(AnaSpec), then labelled with 1 μM CFSE (Biolegend) or eFluor670 (Thermo Fisher Scientific) respectively.

1 x 107 cells of a 1:1 mix of the two differentially-labelled target cells were intravenously injected into

OVA-immunised recipient mice that had previously been transferred with Hdac3-KO or -WT OT-I+ CD8+ T cells. Mice that had received OT-I+ CD8+ T cells without OVA immunisation were also injected with target cells and used as negative controls to measure basal levels of cytotoxicity. 9 hours post-target cell transfer, spleens from recipient mice were analysed by flow cytometry, and the numbers of CFSE+ OVA-labelled targets, eFluor670+ control peptide-labelled targets, and OVA peptide tetramer+ CD8+ T cells (OT-I TCR+

CD8+) were recorded. In vivo cytotoxicity was calculated as a ratio of depleted OVA-labelled targets relative to the numbers of OT-I+ CD8+ T cells as follows:

퐼𝑛 𝑣푖𝑣𝑜 𝑘푖𝑙𝑙푖𝑛𝑔 (𝑝𝑒𝑟 1000 𝑒𝑓𝑓𝑒푐𝑡𝑜𝑟𝑠)

#퐶𝐹𝑆𝐸 𝑒𝑣𝑒𝑛𝑡𝑠 푖𝑛 𝑛𝑒𝑔푎𝑡푖𝑣𝑒 푐𝑜𝑛𝑡𝑟𝑜𝑙 #𝑒𝐹𝑙𝑢𝑜𝑟 𝑒𝑣𝑒𝑛𝑡𝑠 푖𝑛 𝑠푎𝑚𝑝𝑙𝑒 × ( ) − #퐶𝐹𝑆𝐸 𝑒𝑣𝑒𝑛𝑡𝑠 푖𝑛 𝑠푎𝑚𝑝𝑙𝑒 #𝑒𝐹𝑙𝑢𝑜𝑟 𝑒𝑣𝑒𝑛𝑡𝑠 푖𝑛 𝑛𝑒𝑔푎𝑡푖𝑣𝑒 푐𝑜𝑛𝑡𝑟𝑜𝑙 = #푂𝑇퐼 𝑒𝑣𝑒𝑛𝑡𝑠 푖𝑛 𝑠푎𝑚𝑝𝑙𝑒 ( 1000 )

OVA immunisation studies

One day before immunisation, OT-I+ CD8+ T cells were prepared from pooled cell suspensions from spleens and peripheral lymph nodes as described earlier, and 5 x 105 OT-I+ CD8+ T cells in PBS were transferred into recipient mice by intravenous injection. Recipient mice were immunised by subcutaneous injection into the flanks with a dose of 10 μg ovalbumin (Imject™ Ovalbumin, Thermo Fisher Scientific) + 50 μg poly(I:C) (Sigma Aldrich) in PBS. For longitudinal studies, transferred OT-I+ populations were analysed by flow cytometry of retro-orbital bleeds. For terminal analysis of donor OT-I+ CD8+ T cell phenotype, total cell suspensions prepared from inguinal lymph nodes were re-stimulated with OVA SIINFEKL peptide

(Sigma Aldrich) for flow cytometry. For investigating lymph node egress and activation-induced cell death

17 during OVA immunisation, 0.5 mg/kg of FTY720 (Cayman Chemical) in a vehicle of 50% (v/v) ethanol in

PBS and/or 3 mg/kg of pan-caspase inhibitor peptide Ac-DEVD-CHO (Selleck Chemical) dissolved in PBS were injected intraperitoneally at indicated time points.

Lymphocytic Choriomeningitis Virus (LCMV) infections

The LCMV Armstrong strain expressing the ovalbumin peptide SIINFEKL epitope (LCMV-OVA) was a gift from Ulrich von Andrian and was generated by Juan de la Torre30. LCMV clone 13 was a gift from Hye-Jung

Kim. All viruses were propagated and titred in BHK cells as previously described31. For acute virus infection experiments tracking the responses of co-transferred Hdac3-KO and -WT virus-specific CD8+ T cells, TCR- polyclonal CD45.1+ congenic mice were adoptively transferred with 5 x 103 cells of a 1:1 mix of CD45.1-

CD45.2+ Hdac3-inducible KO (Cre/ERT2+;Hdac3fl/fl) and CD45.1+ CD45.2+ Hdac3-WT OT-I+ CD8+ T cells prepared from age- and sex-matched donors. Hdac3 deletion at indicated time points was induced by intraperitoneal injection of 2 mg tamoxifen (Sigma Aldrich) in corn oil daily for 3 consecutive days. 4 days post-adoptive transfer, host mice were infected with 2 x 105 plaque-forming units (PFU) of LCMV-OVA by intraperitoneal injection. Retro-orbital bleeds were analysed by flow cytometry for longitudinal studies of transferred OT-I+ populations. For terminal analysis of donor OT-I+ CD8+ T cell phenotype, total cell suspensions prepared from indicated organs were re-stimulated with OVA SIINFEKL peptide (Sigma

Aldrich) for flow cytometry.

For chronic LCMV infection experiments, TCR-polyclonal E8I-Cre+ Hdac3fl/fl mice with a CD8+ T cell- restricted deletion of Hdac3 or Cre- Hdac3fl/fl littermate controls of both sexes were infected with 2 x 106

PFU LCMV Clone 13 by intravenous injection. Weight loss was monitored every 2-3 days post-infection, and viraemia was monitored weekly by titration of peripheral blood serum on Vero cell monolayers and

18 counting of subsequent plaques as previously described31. For terminal analysis of virus-specific CD8+ T cell responses in viral reservoir organs, total cell suspensions prepared from indicated organs were re- stimulated with 5 μg/ml of LCMV GP33 and GP276 peptides (AnaSpec) for flow cytometric analysis.

Tumour growth studies

Mice were injected subcutaneously with 1.5 x 105 MC38 or MC38-OVA cells, or with 2 x 105 B16F10 cells into the right flank. Tumour growth was measured every 2 days beginning on day 5 with calipers, and tumour size was calculated as the product of the length (longest axis) by the breadth (axis perpendicular to the length). For HDAC3 inhibitor studies, 10 mg/kg RGFP966 (Selleck Chemicals) dissolved in a vehicle of 10% (v/v) DMSO + 30% (m/v) 2-hydroxypropyl-β-cyclodextrin (Cayman Chemical) in 100 mM sodium acetate pH 5.4 was administered by intraperitoneal injection every two days beginning on day 5. Where indicated, CD8+ T cells were depleted every 6 days beginning on day 5 by intraperitoneal injection of 100

μg of α-CD8β antibody (53-5.8, BioXCell). For adoptive T cell transfer experiments, MC38-OVA tumour- bearing mice received 1 x 105 Hdac3-KO or -WT OT-I+ CD8+ T cells by intravenous injection on day 5 post- tumour injection.

RNA-sequencing and analysis

OT-I+ CD8+ T cells were isolated and activated on irradiated SIINFEKL-pulsed BMDCs as described above and sorted to purity after 5 days. Total RNA was prepared using the RNAeasy Plus Mini Kit (Qiagen) according to the manufacturer’s instructions and submitted to the Dana-Farber Molecular Biology Core

Facility for library preparation and sequencing. A standard mRNA library preparation kit (RS-122-2101,

Illumina) was used for library preparation. Single-end 75 bp sequencing was done on an Illumina NextSeq

500. Statistics for differentially expressed genes were calculated by DESeq2 (version 3.5)32. Thresholds for determining significance were defined as log10(adjusted P-value) > 10 and |log2(fold change)| > 0.5.

Western blot analysis of histone acetyl-lysine residues

OT-I+ CD8+ T cells were isolated and activated on irradiated SIINFEKL-pulsed BMDCs in the presence of

3 μM RGFP966 or an equivalent volume of DMSO vehicle as described above and sorted to purity at different time points over a 7-day time course. Whole cell lysates were prepared in RIPA buffer with protease inhibitors (P8340, Sigma Aldrich) and 10 μM of the pan-HDAC inhibitor sodium butyrate (Sigma

Aldrich). Lysate volumes equivalent to 1-3 x 105 cells were loaded into each lane for gel electrophoresis and semi-quantitative Western blot analysis as previously described for the validation of gene knockout efficacy of gRNAs. Rabbit antibodies for Western blotting against HDAC3 (#2632), H3K9-ac (#9649),

H3K14-ac (#7627), H3K27-ac (#8173), Blimp-1 (#9115), and GAPDH (#2118) were purchased from Cell

Signaling Technology. Unconjugated mouse antibodies against T-bet (4B10) were purchased from

Biolegend, and HRP-conjugated horse anti-mouse secondary antibodies for detection of T-bet (1:5000 dilution) were purchased from Cell Signaling Technology.

Analysis of Histone 3 lysine 27-acetylation by ChIP-qPCR

OT-I+ CD8+ T cells were isolated and activated on irradiated SIINFEKL-pulsed BMDCs as described above and sorted to purity after 3 or 5 days. Sorted cells were fixed in 1% PFA (Sigma Aldrich) in PBS for 7 minutes, and the reaction was quenched with 125 mM glycine (Sigma Aldrich) in PBS for 5 minutes at room temperature, all with gentle rotation. Cells were then washed with ice-cold PBS in the presence of protease inhibitor cocktail (Sigma Aldrich P8340), 1 mM PMSF (Sigma Aldrich), and 10 mM sodium

20 butyrate (Sigma Aldrich). Dry pellets were stored at -80°C until all samples were collected. Total chromatin was prepared using a Zymo-SpinTM ChIP Kit (Zymo Research) according to the mechanical shearing protocol supplied by the manufacturer; chromatin was sheared by direct probe sonication using a Q125 sonicator with a 1/16 inch (1.6 mm) diameter probe tip (Qsonica), using 12 pulsed cycles of 15 seconds at

10% amplitude followed by 30 seconds rest at 4°C. Chromatin immunoprecipitation was performed with a 1:100 dilution of antibodies specific for H3K27ac (#8173) or an equivalent amount of normal rabbit IgG

(negative control) (#2729) (Cell Signaling Technology), using reagents from the Zymo-SpinTM ChIP Kit

(Zymo Research). After reverse-crosslinking and chromatin elution, samples were analysed by quantitative

PCR using PowerSYBR® Green reagents (Thermo Fisher Scientific) with validated primers targeting promoter regions (up to -1000 bp before TSS) of target gene loci. qPCR reactions were run on a C1000 thermocycler with a CFX96 optical reaction module (Bio-Rad) for 40 cycles of 15 seconds denaturation at

95°C followed by 60 seconds annealing/extension at 60°C.

Runx3 ChIP-sequencing and analysis

Hdac3-KO or -WT OT-I+ CD8+ T cells were isolated and activated on irradiated SIINFEKL-pulsed BMDCs as described above and sorted to purity after 5 days. Sorted cells were double-fixed with di(N-succinimidyl) glutarate (Sigma Aldrich) and 1% PFA (Sigma-Aldrich) according to a previously described protocol33. Cells were then washed with ice-cold PBS in the presence of 1x protease inhibitor cocktail (Sigma Aldrich

P8340), 1 mM PMSF (Sigma Aldrich), and 10 mM sodium butyrate (Sigma Aldrich) and dry pellets were stored at -80°C until all samples were collected. Cross-linked material was resuspended in sonication buffer (1% SDS, 50 mM Tris-HCl pH8, 10 mM EDTA) and sonicated in a Covaris E220 instrument. 40 μg of soluble chromatin was immunoprecipitated with 5 μg of rabbit anti-Runx3 antibody (ab11905) (Abcam).

Libraries were prepared with an Accel-NGS® 2S Plus DNA Library Kit (Swift Biosciences) following the

21 manufacturer's protocols. 36-bp paired end reads were sequenced on a NextSeq 2500 instrument

(Illumina).

The Runx3 regulatory potential34 score s of each gene g, representing its potential of being regulated by

Runx3, was calculated as follows:

where d0 is the half-decay distance, set to a value of 10 kb or of 100 kb as indicated. All k binding sites near the TSS of gene g (within the distance 15 * d0) are used in the calculation, and di is the distance between the centre of the ith peak and the TSS of g.

Chapter 3

Discovery of CARM1 as a suppressor of CD8+ T cell accumulation in tumours using an in vivo CRISPR-

Cas9 genetic screening platform

Discovery of CARM1 as a suppressor of CD8+ T cell accumulation in tumours using an in vivo CRISPR-

Cas9 genetic screening platform

To uncover novel epigenetic regulators of CD8+ T cell function, we first took a genetic approach by designing a CRISPR-Cas9-based screening platform to evaluate the phenotype of gene-knockout CD8+ T cells in an immunologically-relevant in vivo setting. Because our laboratory group had previously established a shRNA-based genome-wide screening platform for primary mouse CD8+ tumour-infiltrating lymphocytes (TILs) in a preclinical mouse model of melanoma27, we thus decided to leverage on our previous technical experience to develop a CRISPR-Cas9 screening system for CD8+ TILs.

The overall experimental design of the screening process is shown in Figure 3.1. Mice with a Rosa26- targeted knock-in of Cas9 were bred with OT-I TCR-transgenic mice to generate Cas9+ OT-I+ mice, which we then used as the source of CD8+ T cells specific for the ovalbumin model antigen expressed by our B16-

OVA melanoma tumour model. CD8+ T cells were transduced with a pool of lentiviral gRNA vectors to generate a pool of T cells deficient in genes coding for epigenetic regulators, and magnetically-purified

Thy1.1+ transductants (Figure 3.2a,b) were then transferred into mice with small established tumours.

10 days after cell transfer, OT-I+ CD8+ T cells were recovered from tumours and spleens (Figure 3.2c) for genomic DNA preparation and deep-sequencing of gRNA sequences. We then calculated the relative enrichment of each gRNA sequence within OT-I+ CD8+ TILs relative to the splenic OT-I+ CD8+ T cells as the readout for our screen, indicating the degree of accumulation of gene-inactivated, tumour antigen- specific CD8+ T cells within the tumour microenvironment.

Figure 3.1 | Experimental design of in vivo CRISPR-Cas9 screen for epigenetic regulators in CD8+ tumour-infiltrating lymphocytes. Cas9+ OT-I+ CD8+ T cells were lentivirally-transduced to express a pooled library of gRNA sequences targeting epigenetic regulators. Thy1.1+ transductants were intravenously transferred into C57BL/6 recipients with established B16-OVA tumours. 10 days later, Thy1.1+ OT-I+ CD8+ T cells were recovered from spleens and tumours for deep-sequencing to determine the relative enrichment of each gRNA species in tumours relative to spleens.

Figure 3.2 | Immunomagnetic positive selection of gRNA-transduced OT-I+ CD8+ T cells prior to transfer into B16-OVA tumour-bearing mice and subsequent recovery for deep-sequencing. a,b, Representative flow cytometry plots of Cas9+ OT-I+ CD8+ T cells before (a) and after magnetic positive selection (b), following transduction with pooled Thy1.1-marked lentiviral gRNA expression. Gated on live CD4- events (a) or live TCRβ+ CD8α+ CD4- events (b). c, Representative flow cytometry plots showing sort gates used to purify transferred Thy1.1+ gene-KO OT-I+ CD8+ T cells from indicated organs. Gated on live events in a broad forward/side scatter-based gate of lymphocyte-sized events.

We began our investigations by screening a set of 426 candidate genes with 5 gRNA sequences each, split across 3 separate pools (Supplementary Table 1) that were screened sequentially. These genes were selected based on GO term annotations for epigenetic activity, a loose definition that included canonical readers, writers, and erasers of epigenetic modifications as well as chromatin scaffolding proteins and transcriptional regulators with putative epigenetic functions. As positive controls, we added gRNA sequences targeting 7 genes that were previously shown to suppress the accumulation of CD8+ TILs

(Pdcd1, Ctla4, Cblb, Egr2, Smad2, Dgkz, and Ppp2r2d). We also included 100 gRNA sequences that were not predicted to bind to sites within the mouse genome as negative controls for data normalisation purposes. All gRNA sequences were cloned into the pLKO-gRNA-Thy1.1 gRNA expression plasmid, derived from the original Lentiguide backbone26, to obtain higher viral titres and increased transduction efficiency in primary CD8+ T cells.

We subsequently selected 29 of the top gene target hits for validation in a second round of screening

(Supplementary Table 2). Analysis of the results of both our pilot and validation screens revealed that inactivation of the Carm1 gene consistently resulted in an accumulation of tumour antigen-specific CD8+

T cells within B16-OVA tumours relative to the general circulation (Figure 3.3). The degree of accumulation of Carm1-deficient CD8+ T cells in tumours versus spleens was also consistently more than that of all positive controls tested, including CD8+ T cells in which expression of the co-inhibitory receptor PD-1 was genetically inactivated. Our results identified Carm1 as a potential inhibitor of tumour-specific CD8+ T cell accumulation within the tumour microenvironment for future follow-up work.

Figure 3.3 | Identification of Carm1 as a potential suppressor of CD8+ tumour-infiltrating lymphocyte accumulation in tumours in our in vivo CRISPR-Cas9 screen of epigenetic regulator genes. a,b, Summary of data from the three preliminary screens (a) and the subsequent validation screen (b). Experimental and positive control genes are represented as red and blue dots, respectively. For clarity, only genes with increased gRNA sequence reads in tumours relative to spleens are represented. Data from the preliminary screens are scored using the MAGeCK algorithm for normalisation across the three separate experiments used to screen the three pools.

Carm1 encodes Coactivator-Associated aRginine Methyltransferase 1 (also known as PRMT4), an enzyme that catalyses the methylation of the primary amino groups of arginine residues. CARM1 activity has been shown to drive transcriptional activation or repression of genes in a context- and arginine residue-specific manner35. While a role for CARM1/PRMT4 in peripheral T cells has yet to be described, there is evidence indicating that CARM1 is required for proper maturation of thymocytes, as Carm1-/- mouse embryos show a block in thymocyte development at the CD4+ CD8+ double-positive stage36,37. Further investigation of a potential role of CARM1 in T cells, or indeed in most primary somatic cells, has been hindered by the fact that a homozygous loss of Carm1 is embryonically lethal. However, a group at the MD Anderson Cancer

Centre (Texas) has recently developed mice bearing a floxed allele of Carm138, which will undoubtedly be extremely helpful in generating Carm1-conditional knockout mice for subsequent investigations into the potential role of CARM1 in regulating CD8+ T cell anti-tumour responses.

Chapter 4

Discovery of HDAC3 as an inhibitor of CD8+ T cell cytotoxicity in an in vitro functional

pharmacologic screen

Discovery of HDAC3 as an inhibitor of CD8+ T cell cytotoxicity in an in vitro functional pharmacologic screen

In parallel with the in vivo CRISPR-Cas9 genetic screen of epigenetic regulators in tumour-infiltrating CD8+

T cells, we also designed a functional pharmacological approach to identify potential epigenetic regulators of CD8+ T cell effector phenotype by screening a library of 100 small molecules (Supplementary Table 3) targeting diverse epigenetic pathways during in vitro T cell activation. We modelled CD8+ T cell activation by co-culturing ovalbumin-specific OT-I+ CD8+ T cells with irradiated Ova peptide-pulsed bone marrow- derived dendritic cells (BMDCs) in the presence of epigenetic inhibitor drugs for 7 days (Figure 4.1), and used high-throughput flow cytometry to identify candidate small molecule inhibitors that significantly perturbed the expression of markers of CD8+ T cell effector function such as activation-associated receptors (CD44, CD25), cytokine secretion (IFN-γ, TNF-α, IL-2), and cytotoxicity (Granzyme B, CD107a).

These findings are summarised in Table 1.

Figure 4.1 | Workflow of in vitro screening and validation of candidate small molecule epigenetic regulators of CD8+ T cell effector function. BMDCs were generated from total C57BL/6 bone marrow, pulsed with 100 nM Ova SIINFEKL peptide, and irradiated prior to co-culture with OT-I+ CD8+ T cells isolated by magnetic negative selection from OT-I transgenic mice. Small molecule inhibitors of epigenetic function or a DMSO vehicle control were added directly into culture media. After 7 days of activation, CD8+ T cell effector phenotype was assessed by multi-colour flow cytometry or by an in vitro 51Cr-release cytotoxicity assay against IFN-γ-treated B16-OVA target cells (in validation experiments).

Table 1 | Summary of significant perturbations of CD8+ T cell effector phenotype by small molecule epigenetic regulators in a model of in vitro CD8+ T cell activation.

Marker category Viability Surface markers of phenotypic state Cytotoxicity Effector cytokines

% CD62L+ % CD62L+ % Granzyme - + + + + + Drug % Live cells PD-1 gMFI CD127 gMFI CD25 gMFI CD44 Naïve CD44 TCM % CD107a B % IFN-γ % TNF-α

RGFP966

UNC0321 3-amino- benzamide

Valproic acid

Remodelin Phthalazinone pyrazole Delphinidin chloride

Tenovin-1

N-Oxalylglycine

2-PCPA

GSK-J5

PFI-3

CPTH2

Piceatannol

JGB1741

AGK2

Mirin

HNHA

RG-108

CBHA

M344

IOX1

GSK-J1

Nicotinamide Octyl-alpha- ketoglutarate

1-Naphthoic Acid

Increased expression of marker relative to controls No significant change of expression relative to controls Decreased expression of marker relative to controls Unable to confidently measure marker expression for technical reasons

HDAC3 inhibits CD8+ T cell cytotoxicity

Using this approach, we found that addition of the HDAC3-specific inhibitor RGFP96639 during in vitro CD8+

T cell activation increased the percentages of CD8+ T cells expressing the effector molecules Granzyme B,

IFN-γ, and TNF-α relative to vehicle-treated cells (Figure 4.2a). We further found that RGFP966-treated

CD8+ T cells exhibited potently increased cytotoxicity against ovalbumin-expressing B16F10 melanoma target cells (B16-OVA) compared to vehicle-treated cells as measured by an in vitro 51Cr release assay

(Figure 4.2b).

Figure 4.2 | Treatment of CD8+ T cells with the HDAC3-specific inhibitor RGFP966 during in vitro activation increases CD8+ T cell cytotoxicity. a,b, OT-I+ CD8+ T cells were treated with indicated concentrations of RGFP966 during 7 days of activation with irradiated Ova peptide-pulsed BMDCs. a, Flow cytometric measurement of expression of markers of CD8+ T cell effector function. b, In vitro cytotoxicity against ovalbumin-expressing B16 melanoma targets (B16-OVA) measured by 51Cr release assay after 4 hours of co-culture. a,b, Centre values, mean; error bars, s.d.; * P < 0.05, **** P < 0.0001; 2-way ANOVA. Data are representative of two independent experiments.

To validate that the increase in cytotoxicity was due to loss of HDAC3 activity, we transduced OT-I+ CD8+

T cells derived from Cas9+ transgenic mice with lentiviral gRNA expression vectors to inactivate Hdac3

(Figure 4.3a). Hdac3 gRNA-transduced CD8+ T cells showed increased cytotoxicity relative to CD8+ T cells transduced with a control LacZ gRNA after in vitro activation (Figure 4.3b), demonstrating that the increase in cytotoxicity was specific to the loss of HDAC3.

Figure 4.3 | Inactivation of Hdac3 increases CD8+ T cell cytotoxicity. a-c, OT-I+ CD8+ T cells derived from Cas9-transgenic mice were transduced with Thy1.1-marked lentiviral vectors bearing indicated gRNA sequences, and positively selected for Thy1.1 expression 3 days after transduction. a, Western blot analysis of HDAC3 levels in CD8+ T cells after 3 or 7 days post-infection (p.i.). M, mock-transduced; LZ, LacZ-targeting gRNA transduced (negative controls). Molecular weights in kDa are indicated. The red arrow indicates the gRNA sequence used for subsequent work. b, Thy1.1+ transductants were activated by co-culture with irradiated Ova peptide-pulsed BMDCs for 7 days, and in vitro cytotoxicity against B16- OVA targets was measured by 51Cr release assay after 4 hours of co-culture. c, Thy1.1+ transductants were activated by co-culture with irradiated Ova peptide-pulsed BMDCs for 7 days, and markers of CD8+ T cell effector phenotype were measured by flow cytometry. b,c, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, **** P < 0.0001; 2-way ANOVA. Data are representative of two independent experiments.

In addition, a higher proportion of Hdac3-inactivated CD8+ T cells expressed Granzyme B and the effector phenotype-associated transcription factor T-bet8 relative to control gRNA-transduced CD8+ T cells.

Moreover, addition of RGFP966 did not further increase the percentages of Granzyme B+ or T-bet+ Hdac3 gRNA-transduced cells (Figure 4.3c, left). In contrast, inactivation of Hdac3 in CD8+ T cells did not result in substantial changes in cytokine expression as was observed in RGFP966-treated CD8+ T cells, except for a small increase in IFN-γ expression (Figure 4.3d, right), suggesting that the increase in effector cytokine secretion as a result of RGFP966 treatment may have been due to indirect effects of the drug.

Generation and characterisation of CD8+ T cell-conditional Hdac3-KO mice

To investigate the potential role of HDAC3 in regulating CD8+ T cell cytotoxicity and effector function in vivo, we generated TCR-polyclonal and OT-I+ TCR-transgenic mice with a CD8+ T cell-restricted deletion of

Hdac3. Because HDAC3 is essential for proper thymopoiesis of T cells40-43, Hdac3 deletion was restricted to mature CD8+ T cells in the periphery by breeding mice bearing a floxed allele of Hdac3 to E8I-Cre+ mice.

E8I-Cre+ mice express Cre recombinase under the control of a regulatory element of Cd8a that is only active in the periphery44 but not within the thymus.

We verified that E8I-Cre+;Hdac3fl/fl mice (hereafter referred to as Hdac3-KO mice) only had Hdac3 inactivated in the CD8+ but not in the CD4+ compartment of mature T cells; similar results were seen in

OT-I TCR-transgenic E8I-Cre+;Hdac3fl/fl mice, where >95% of peripheral T cells are CD8+ (Figure 4.4a,b).

Interestingly, the degree of Hdac3 knockout in TCR-polyclonal E8I-Cre+;Hdac3fl/fl mice was strongest in

+ - + + + - + CD62L CD44 naïve CD8 T cells, followed by CD62L CD44 TCM cells and CD62L CD44 Teff/EM cells (Figure

4.4c). The major peripheral T cell compartments were intact in Hdac3-KO mice, albeit with slightly reduced total CD8+ T cell numbers relative to Hdac3-WT littermates (Figure 4.4d); CD8+ T cells from Hdac3-KO mice also had a distribution of naïve, TCM, and Teff/TEM phenotypes similar to Hdac3-WT mice (Figure 4.4e).

Finally, while CD8+ T cells from Hdac3-KO mice were not significantly different in Granzyme B or T-bet expression (Figure 4.4f), a smaller proportion produced IFN-γ or TNF-α in response to ex vivo pharmacological stimulation with PMA and ionomycin (Figure 4.4g).

Figure 4.4 | Characterisation of peripheral T cells in CD8-specific Hdac3-conditional KO mice. a,b, Evaluation of on-target knockout efficiency of Hdac3 in E8I-Cre+;Hdac3fl/fl mice. CD8+ and CD4+ T cells were sorted from in 6 week-old TCR-polyclonal E8I-Cre+ and Cre- Hdac3fl/fl littermates for assessment of knockout efficiency by PCR (a) and immunoblot (b) analysis, with parallel comparisons of magnetically- purified E8I-Cre+ and Cre- OT-I+ CD8+ T cells. 50 μg of purified genomic DNA was loaded as a template for PCR. Whole cell lysates prepared from 5 x 105 cells were loaded for immunoblotting. Molecular weights in bp or kDa are indicated on PCR gel and immunoblot images, respectively. (figure legend continues on next page)

Figure 4.4 (continued from previous page) c, Evaluation of Cre-mediated loss of Hdac3 in subpopulations of peripheral CD8+ T cells in E8I- + fl/fl + - + + - + Cre ;Hdac3 mice. Sort-purified total, CD62L CD44 naïve, CD62L CD44 TCM, and CD62L CD44 Teff/EM CD8+ T cells were analysed by PCR and immunoblot as in (a) and (b), respectively. d-g, Characterisation of peripheral T cell compartments in 10 week-old TCR-polyclonal E8I-Cre+ and Cre- Hdac3fl/fl littermates. Data are from one experiment with three mice per genotype. d, Measurement of sizes of splenic T cell + + - + + + compartments. CD4 Tconv, CD4 FoxP3 conventional T cells; CD4 Treg, CD4 FoxP3 regulatory T cells. + hi lo hi hi e-g, Characterisation of CD8 T cell phenotype. Naïve, CD62L CD44 cells; TCM, CD62L CD44 central lo hi memory cells; Teff/TEM, CD62L CD44 effector or effector memory cells. d-g, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, two-tailed Student’s t-test.

Importantly, we did not observe any substantial changes in thymic development in TCR-polyclonal Hdac3-

KO mice. Overall, there were no significant differences in the proportions of other thymocyte subpopulations between E8I-Cre+ and Cre- Hdac3fl/fl littermates, barring a slight decrease in the percentage of CD4+ CD8+ double-positive (DP) thymocytes and a slight increase in the percentage of CD4+ single-positive (SP) thymocytes in TCR-polyclonal Hdac3-KO mice (Figure 4.5a). Further examination of the phenotype of thymocyte subpopulations revealed that there was a slight decrease in the percentage of CD25lo CD44hi stage 1 CD4- CD8- double-negative thymocytes (Figure 4.5b), as well as a non-significant trend towards an increased proportion of CD5+ TCRβ+ cells in the DP thymocyte population (Figure 4.5d).

In addition, the proportion of CD4lo CD8lo ‘double-dull’ (DD) undergoing negative selection (CD69+ MHC Ilo cells) was unchanged (Figure 4.5c), as were the proportions of mature CD5+ thymocytes among the CD4+ and CD8+ SP compartments (Figure 4.5e,f).

Overall, our breeding strategy of generating E8I-Cre+;Hdac3fl/fl mice did indeed result in a peripheral CD8+

T cell-restricted knockout of Hdac3 without significantly affecting thymic development or ablating the peripheral CD8+ T cell compartment, in contrast to other Cre-based approaches used to generate T cell- restricted Hdac3 conditional-knockout mice40,41.

Figure 4.5 | Characterisation of thymopoiesis in CD8-specific Hdac3-conditional KO mice. a, Flow cytometric measurement of thymocyte developmental stages, gated on live CD11b- CD11c- cells in thymii from 10-week old male E8I-Cre+;Hdac3fl/fl mice (Hdac3-KO) or Hdac3fl/fl littermates (Hdac3-WT). DN, CD4- CD8- double-negative; DD, CD4lo CD8lo double-dull; DP, CD4+ CD8+ double-positive; CD4, CD4+ CD8- single- positive; CD8, CD4- CD8+ single-positive thymocytes. b-f, Phenotypic analysis of indicated thymocyte developmental stages. Representative flow cytometry plots and quantification of CD4- CD8- DN developmental stages (b), CD4lo CD8lo DD cells undergoing negative selection (c), CD4+ CD8+ DP cells undergoing positive selection (d), and CD5+ mature CD4 (e) and CD8 SP thymocytes (f). Data are representative of two independent experiments, each with three age- and sex-matched littermates per genotype. a-f, Centre values, mean; error bars, s.d.; * P < 0.05, two-tailed Student’s t-test.

Hdac3-KO CD8+ T cells exhibit increased cytotoxicity after activation in vitro and in vivo

We next evaluated the cytotoxicity of Hdac3-KO CD8+ T cells derived from E8I-Cre+;Hdac3fl/fl OT-I TCR- transgenic mice following in vitro activation with irradiated Ova peptide-pulsed BMDCs. Hdac3-KO OT-I+

CD8+ T cells were consistently more cytotoxic against B16-OVA melanoma target cells than Hdac3-WT

CD8+ T cells derived from Cre- littermates (Figure 4.6a), consistent with our previous results using Hdac3 gRNA-transduced CD8+ T cells.

Figure 4.6 | Hdac-KO CD8+ T cells exhibit increased cytotoxicity after activation in vitro and in vivo. a, OT-I+ CD8+ T cells from Hdac3-KO and -WT littermates were activated by co-culture with irradiated Ova peptide-pulsed BMDCs. In vitro cytotoxicity against B16-OVA targets was measured by 51Cr release assay after 4 hours of co-culture. Data are representative of two independent experiments. b, In vivo cytotoxicity of adoptively-transferred Hdac3-KO OT-I+ CD8+ T cells against Ova peptide-pulsed target cells. 5 x 105 Hdac3-KO or -WT OT-I+ CD8+ T cells were adoptively-transferred into wildtype mice, and recipient mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. 3 days later, 107 cells of a 1:1 mix of CFSE-labelled and Ova peptide-pulsed plus eFluor670-labelled and control irrelevant peptide-pulsed splenocytes were transferred into immunised mice. Spleens were harvested 9 hours later for flow cytometric quantification of relative depletion of Ova peptide-labelled versus control peptide-pulsed targets, normalised to numbers of Kb-SIINFEKL-tetramer+ OT-I+ CD8+ T cells. Data from one experiment with 7 host mice for each genotype of donor OT-I+ transferred. a,b, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, 2-way ANOVA (a) or two-tailed Student’s t-test (b).

We further validated this phenotype in an in vivo model of cytotoxicity by quantifying the depletion of

Ova peptide-pulsed target cells by activated Hdac3-KO and -WT OT-I+ CD8+ T cells. We first transferred naïve Hdac3-KO or -WT OT-I+ CD8+ T cells into C57BL/6 hosts and immunised the recipient mice with a low dose of ovalbumin (OVA) + poly(I:C). 3 days later, differentially-labelled Ova peptide- and irrelevant

37 control peptide-pulsed splenocyte targets were transferred into the immunised mice. 9 hours after injection of target cells, we measured the relative depletion of specific versus non-specific targets in the spleen by flow cytometry and found that Hdac3-KO OT-I+ CD8+ T cells killed higher numbers of Ova peptide-pulsed target cells on a per cell basis (Figure 4.6b). Together with our experiments using the

HDAC3-specific inhibitor RGFP966 and CRISPR-Cas9-mediated ablation of Hdac3 in CD8+ T cells, these data demonstrate that HDAC3 activity in CD8+ T cells suppresses CD8+ T cell cytotoxicity.

We also assessed cell-intrinsic changes in CD8+ T cell effector function caused by loss of HDAC3 in an in vivo setting by co-transferring congenically-distinct Hdac3-KO and -WT naïve OT-I+ CD8+ T cells into TCR- polyclonal hosts and evaluating the effector phenotypes of transferred OT-I+ CD8+ T cells within draining lymph nodes 4 days after immunisation with OVA + poly(I:C) (Figure 4.7).

Figure 4.7 | Experimental scheme of OT-I+ CD8+ co-transfer and immunisation. 5 x 105 cells of a 1:1 mix of congenically-distinct Hdac3-KO and -WT OT-I+ CD8+ T cells were transferred into CD45.1+ TCR- polyclonal recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. Inguinal lymph nodes draining the immunisation site were harvested after 4 days for analysis.

In agreement with our previous observations, an increased proportion of Hdac3-KO CD8+ T cells expressed

Granzyme B and T-bet compared to co-transferred Hdac3-WT cells (Figure 4.8a, left and centre; Figure

4.8b, top). Moreover, a reduced percentage of Hdac3-KO CD8+ T cells produced IFN-γ or TNF-α upon ex vivo re-stimulation with Ova peptide relative to co-transferred Hdac3-WT cells (Figure 4.8a, right; Figure

4.8b, bottom). Interestingly, across both genotypes, Granzyme B+ cells were almost exclusively cytokine- negative, and cytokine+ cells were also almost entirely Granzyme B-negative.

Figure 4.8 | Hdac-KO CD8+ T cells express increased levels of Granzyme B and T-bet after in vivo activation relative to Hdac3-WT cells. a,b, Congenically-distinct OT-I+ CD8+ T cells from Hdac3-KO and - WT mice were co-transferred into wildtype mice and activated with OVA + poly(I:C) immunisation as shown in Figure 4.7. Flow cytometric analysis (a) and quantification (b) of markers of effector phenotype of OT-I+ CD8+ T cells in inguinal lymph nodes draining the immunisation site harvested 4 days post- immunisation. Gated on live TCRβ+ CD8α+ CD4- events. Data are representative of two independent experiments with 5 recipient mice each. b, Centre values, mean; error bars, s.d.; ** P < 0.01, *** P < 0.001, **** P < 0.0001, two-tailed ratio-paired t-test.

In summary, our data show that HDAC3 activity suppresses the acquisition of CD8+ T cell cytotoxicity and the expression of T-bet in a cell-intrinsic manner following T cell activation. While inhibition of HDAC3 with RGFP966 during in vitro activation of CD8+ T cells resulted in increased secretion of the effector cytokines IFN-γ and TNF-α, this was not recapitulated in Hdac3-deficient CD8+ T cells activated in vitro or in vivo. We therefore focussed our subsequent investigations into the HDAC3-mediated regulation of CD8+

T cell effector phenotype on the functional aspect of cytotoxicity rather than on the secretion of effector cytokines.

Chapter 5

HDAC3 is required during T cell activation for persistence of antigen-experienced CD8+ T cells

Hdac3-KO CD8+ T cells fail to persist after activation in vivo

We were interested to determine whether the changes in CD8+ T cell effector phenotype as a result of

Hdac3-deficiency would affect the overall CD8+ T cell response to antigen challenge. We therefore first investigated whether loss of HDAC3 altered the dynamics of the CD8+ T cell functional response by tracking the relative frequencies of co-transferred Hdac3-KO and -WT OT-I+ CD8+ T cells over a time course following OVA + poly(I:C) immunisation (Figure 5.1a). In this setting, we found that Hdac3-KO OT-I+ CD8+

T cells did not persist in both peripheral blood and in the draining lymph nodes compared to co- transferred Hdac3-WT cells following contraction of the OT-I+ response between days 4 and 7. This lack of persistence could also not be rescued with a secondary boost of OVA + poly(I:C) (Figure 5.1b,c).

Figure 5.1 | Hdac-KO CD8+ T cells do not show long-term persistence after antigen priming. a-c, 5 x 105 cells of a 1:1 mix of congenically-distinct Hdac3-KO and -WT OT-I+ CD8+ T cells were transferred into CD45.1+ TCR-polyclonal recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. A secondary boost of ovalbumin with poly(I:C) was administered on day 14 post- priming. Dynamics of the Hdac3-KO and -WT OT-I+ CD8+ T cell responses were tracked longitudinally in peripheral blood (b) and in inguinal lymph nodes draining the immunisation site (c). Data are representative of two independent experiments with 5 mice per experiment (b) or 5 mice for each time point (c). b,c, Centre values, mean; error bars, s.e.m.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, two-way ANOVA.

Further analysis of the phenotypes of donor OT-I+ CD8+ T cells isolated from draining lymph nodes revealed that the proportions of cells expressing Granzyme B and T-bet were strongly reduced between days 4 and

7 after OVA immunisation, with this reduction being especially pronounced in Hdac3-KO CD8+ T cells

(Figure 5.2a,b and 5.2d, top). The frequencies of Granzyme B+ cells in Hdac3-KO CD8+ T cells were always higher than those of co-transferred Hdac3-WT cells, even after contraction of the OT-I+ CD8+ response on day 7. A higher percentage of Hdac3-KO CD8+ T cells than Hdac3-WT cells were also T-bet+ on day 4, consistent with previous data, but there was no significant difference by day 7. On the other hand, the fractions of IFN-γ+ and TNF-α+ cells across both genotypes did not change between days 4 and 7, although lower frequencies of cytokine+ cells following ex vivo Ova peptide re-stimulation were consistently observed in Hdac3-KO cells relative to co-transferred Hdac3-WT cells (Figure 5.2c and Figure 5.2d, bottom). Moreover, we again observed that the population of Granzyme B+ and TNF-α+ cells were almost mutually exclusive; this was also largely true between Granzyme B+ and IFN-γ+ cells.

The inability of Hdac3-KO CD8+ T cells to persist was not due to a general loss of survival fitness, because naïve non-autoreactive Hdac3-KO OT-I+ CD8+ T cells were able to proliferate and accumulate to the same extent as Hdac3-WT cells when allowed to undergo cytokine-driven homeostatic expansion in lymphopenic Rag1-/- hosts (Figure 5.3). This was an important point to establish because early work had shown that HDAC3 was required during S phase progression for proper mitosis of mouse embryonic fibroblasts45. Furthermore, inhibition of lymph node egress and/or activation-induced apoptosis (Figure

5.4a) did not rescue the failure of Hdac3-KO CD8+ T cells to accumulate in the draining lymph nodes in the context of the OT-I transfer and OVA immunisation model (Figure 5.4b), and also did not significantly alter the changes in phenotype associated with Hdac3-deficiency (Figure 5.4c). These findings suggested that the phenotypic changes observed in activated Hdac3-KO CD8+ T cells in vivo were likely to be a result of proximal effects of loss of HDAC3 activity, and not because of a loss of HDAC3-mediated regulation during

CD8+ T cell maturation and development.

Figure 5.2 | Granzyme B+ and T-bet+ responses of in vivo activated CD8+ T cells are short-lived relative to IFN-γ+ and TNF-α+ responses. a-d, 5 x 105 cells of a 1:1 mix of congenically-distinct Hdac3-KO and -WT OT-I+ CD8+ T cells were transferred into CD45.1+ TCR-polyclonal recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS, and donor OT-I+ CD8+ T cell responses were analysed in lymph nodes draining the immunisation site after 4 or 7 days. Representative flow cytometry plots (a-c) and quantification (d) of phenotypic changes in co-transferred Hdac3-KO and Hdac3-WT OT-I+ CD8+ T cells. Data are representative of two independent experiments with 5 mice for each time point analysed. Gated on live TCRβ+ CD8α+ CD4- events. d, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, two-way ANOVA.

Figure 5.3 | Hdac3-KO CD8+ T cells proliferate to a degree similar to Hdac3-WT cells in the context of cytokine-driven homeostatic proliferation in the absence of TCR activation. a-c, Hdac3-KO or -WT OT- I+ CD8+ T cells (non-autoreactive TCR) were labelled with CFSE and adoptively transferred into lymphopenic Rag1-/- hosts to undergo non-TCR driven, cytokine-mediated homeostatic proliferation. b, Representative flow cytometry plots and quantification of CFSE dilution in transferred OT-I+ CD8+ T cells 48 hours post-transfer. Gated on live TCRβ+ CD8α+ Kb-SIINFEKL-tetramer+ events. Data are representative of two independent experiments with 5 recipient mice for each genotype of OT-I+ CD8+ T cells transferred. c, Representative flow cytometry plots and quantification of homeostatic reconstitution in spleens of Rag1-/- recipient mice after 5 days. OT-I+ T cells were defined as live TCRβ+ CD8α+ Kb-SIINFEKL-tetramer+ events. Data are representative of two independent experiments with 5 recipient mice for each genotype of OT-I+ CD8+ T cells transferred. b,c, Centre values, mean; error bars, s.d.; data were analysed with two-tailed Student’s t-test.

Figure 5.4 | The reduced persistence of Hdac3-KO CD8+ T cells in draining lymph nodes after in vivo activation is not due to increased apoptosis or lymph node egress. a-c, CD45.1+ TCR-polyclonal mice were adoptively transferred with 1 x 106 cells of a 1:1 mix of congenically-distinct Hdac3-KO and -WT OT-I+ CD8+ T cells and immunised with ovalbumin adjuvanted with poly(I:C) in PBS. Recipient mice received S1PR downregulator FTY720 and/or pan-caspase inhibitor Ac-DEVD-CHO intraperitoneally at indicated time points after immunisation (n = 5, FTY720-treated groups; n = 6, FTY720-untreated groups). b, Flow cytometric quantification of changes in relative Hdac3-KO vs -WT frequencies normalised to pre- transfer ratios (left), and differences in absolute numbers (right) of co-transferred Hdac3-KO and -WT OT-I+ CD8+ T cells in inguinal lymph nodes 3 days after immunisation. c, Flow cytometric quantification of changes in relative ratios of Hdac3-KO vs -WT effector phenotypic markers in transferred OT-I+ CD8+ T cells between different treatment groups. Gated on live TCRβ+ CD8α+ CD4- events. b,c, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, two-tailed Student’s t-test (b), or two-tailed ratio-paired t-test (c).

HDAC3 is required specifically during T cell activation for persistence of antigen-experienced CD8+ T cells

In view of these observations, we hypothesised that the loss of persistence in Hdac3-KO CD8+ T cells was a consequence of a lack of HDAC3 activity specifically during T cell activation. To test this in a physiological setting, we investigated the effect of loss of Hdac3 beginning at different phases of the immune response in the context of a model of acute viral infection. We first co-transferred OT-I+ CD8+ T cells with a tamoxifen-inducible deletion of Hdac3 (Cre/ERT2+;Hdac3fl/fl mice, hereafter referred to as Hdac3-iKO mice) and congenically-distinct Hdac3-WT OT-I+ CD8+ T cells into wildtype recipients. Recipient mice were challenged by acute infection with lymphocytic choriomeningitis virus (Armstrong) engineered to express the SIINFEKL epitope of OVA recognised by the OT-I TCR (LCMV-OVA). Deletion of Hdac3 in Hdac3-iKO cells was induced by tamoxifen injection either before infection or on day 7 (Figure 5.5a), when the virus is typically cleared and the infection is resolved46. As a readout of the virus-specific CD8+ T cell response, we measured the relative frequencies of donor Hdac3-iKO and -WT OT-I+ CD8+ T cells in peripheral blood to track the persistence of donor OT-I+ CD8+ T cells responding to viral infection (Figure 5.5b-d).

In agreement with our previous observations using the experimental model of OT-I transfer followed by

OVA immunisation, genetic loss of Hdac3 prior to challenge with LCMV-OVA resulted in a lack of persistence of activated OT-I+ CD8+ T cells, although these cells were able to proliferate in response to viral infection (Figure 5.5b, Group 1). In this experimental group (Group 1), the numbers of Hdac3-iKO OT-

I+ CD8+ T cells relative to co-transferred Hdac3-WT cells were significantly reduced across the three major organ reservoirs of LCMV at 8 days after infection (Figure 5.5e). Additionally, in these mice, we also observed a higher proportion of Granzyme B+ Hdac3-iKO OT-I+ CD8+ T cells relative to co-transferred

Hdac3-WT cells, whereas the percentages of IFN-γ+ and TNF-α+ cells after ex vivo re-stimulation with Ova

Figure 5.5 | Inactivation of Hdac3 prior to infection but not after viral clearance results in reduced persistence of antigen-experienced CD8+ T cells during acute LCMV infection. a-f, 5 x 103 cells of a 1:1 mix of congenically-distinct Hdac3-inducible KO and -WT OT-I+ CD8+ T cells were transferred into CD45.1+ TCR-polyclonal recipients. Mice were infected intraperitoneally with 2 x 105 PFU of LCMV Armstrong expressing the OT-I-cognate ovalbumin epitope (LCMV-OVA), and Hdac3 deletion was induced at indicated time points by intraperitoneal injection of tamoxifen (2 mg per mouse) on three consecutive days. (figure legend continues on next page)

Figure 5.5 (continued from previous page) b-d, Donor OT-I+ CD8+ responses in peripheral blood were monitored longitudinally in indicated treatment groups (Group 1, n=4; Group 2, n=4; No deletion, n=3). Data are shown as frequencies of donor OT-I+ CD8+ T cells relative to total CD8+ T cells (b), as a ratio of Hdac3-KO OT-I+ frequencies normalised to the frequencies of Hdac3-WT OT-I+ cells within the same recipients for statistical analysis (c), or as representative flow cytometry plots (d). Gated on live TCRβ+ CD8α+ CD4- events. Data are representative of two independent experiments with 3-5 recipient mice per treatment group. e,f, Flow cytometric analysis of total numbers (e) and markers of effector phenotype (g) of co-transferred Hdac3-inducible- KO and Hdac3-WT OT-I+ CD8+ T cells at 8 days following LCMV-OVA infection from mice where tamoxifen was administered prior to infection. Gated on live TCRβ+ CD8α+ CD4- events. Data are representative of two independent experiments with 5 recipient mice each. b,c,e,f, Centre values, mean; error bars, s.e.m.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, two-way ANOVA (c) or two-tailed ratio-paired t- test (e,f). peptide were reduced (Figure 5.5f). These phenotypic changes are also consistent with our data from the

OT-I transfer and immunisation experimental model, indicating that the suppression of the Granzyme B+

CD8+ T cell cytotoxic response was not specific the OVA immunisation experimental system.

In contrast, tamoxifen-induced ablation of Hdac3 beginning on day 7 did not significantly alter the persistence of Hdac3-iKO OT-I+ CD8+ T cells relative to non-induced Hdac3-iKO cells (Figure 5.5b-d, Group

2). These data indicate that HDAC3 is required during initial activation of CD8+ T cells for the persistence of antigen-experienced cells, and also that HDAC3 is dispensable for the continued persistence of antigen- experienced CD8+ T cells after they have already become fully activated. Together with our previous results (Figure 5.3 and Figure 5.4), these data also support the idea that the reduced persistence of Hdac3-

KO CD8+ T cells was not merely due to a general loss of survival fitness, because the loss of Hdac3 after viral clearance did not adversely affect the persistence of antigen-experienced virus-specific CD8+ T cells.

HDAC3 functions early during activation to regulate acquisition of CD8+ T cell effector function

In view of the requirement for HDAC3 during T cell priming for persistence of antigen-experienced CD8+ T cells, and because the increased cytotoxicity and decreased cytokine response of Hdac3-deficient CD8+ T

48 cells were detectable by the peak of CD8+ T cell activation (~day 4 in our OVA immunisation model, ~day

7-8 in LCMV-OVA infection), we hypothesised that HDAC3 activity early during CD8+ T cell activation regulated subsequent acquisition of CD8+ T cell effector function.

We therefore investigated the kinetics of acquisition of CD8+ effector functions in Hdac3-KO and -WT CD8+

T cells by transferring CFSE-labelled OT-I+ CD8+ T cells into wildtype recipients and immunising the mice with OVA + poly(I:C). Within 48 hours post-activation, a larger proportion of Hdac3-KO CD8+ T cells already expressed Granzyme B and T-bet compared to Hdac3-WT CD8+ T cells, with the changes occurring within one cell division after activation (Figure 5.6a,b) and persisting at 96 hours post-activation (Figure 5.6d).

In contrast, there was no significant difference in the percentages of cytokine-producing cells between

Hdac3-KO and -WT CD8+ T cells after re-stimulation with Ova peptide within 48 hours following activation

(Figure 5.6c). Reduced percentages of IFN-γ+ or TNF-α+ in Hdac3-KO relative to Hdac3-WT CD8+ T cells were only observed by the 6th cell division post-activation at the 96-hour time point (Figure 5.6e). These data suggest that, during activation of naïve CD8+ T cells, HDAC3 suppresses the acquisition of cytotoxicity from an early time point and later supports effector cytokine production.

Taken together, our findings demonstrate that HDAC3 regulates the effector phenotype of CD8+ T cells early during T cell activation to dampen the cytotoxic (Granzyme B+ and T-bet+) response and to promote the durability of antigen-experienced CD8+ T cells. Moreover, the loss of persistence of CD8+ T cells as a result of Hdac3-deficiency was specific to TCR engagement and became permanent upon T cell activation.

These results suggest that HDAC3 may play a previously unknown role in regulating CD8+ T cell fate commitment early during priming.

Figure 5.6 | HDAC3 inhibits the CD8+ T cell cytotoxic response early during priming. a-e, 5 x 105 CFSE- labelled CD45.2+ Hdac3-KO or -WT OT-I+ CD8+ T cells were separately transferred into CD45.1+ TCR- polyclonal recipients. Mice were immunised with ovalbumin adjuvanted with poly(I:C) in PBS. Inguinal lymph nodes draining the immunisation site were collected 48 hours (a-c) or 96 hours (d,e) after immunisation for flow cytometric analysis to track the kinetics of acquisition of markers CD8+ T cell effector function with successive rounds of cell division. Flow cytometry plots are representative of two independent experiments with 5 recipient mice per genotype of donor OT-I+ CD8+ T cells transferred. a-e, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, **** P <0.0001, 2-way ANOVA.

Chapter 6

Effects of loss of HDAC3 activity in CD8+ T cells in settings of chronic antigen burden

Hdac3-KO mice show delayed disease progression in a model of chronic viral infection

We reasoned that any benefit to the immune response due to the increased cytotoxicity of Hdac3-KO

CD8+ T cells would be more readily observed in an experimental model in which antigen chronically persists without being completely cleared, rather than in the setting of an acute antigen challenge where antigen is quickly and robustly cleared. Furthermore, because the gain-of-cytotoxic function with Hdac3- deficient CD8+ T cells was observed very quickly following T cell activation, we hypothesised that Hdac3- deficient CD8+ T cells would boost the immune response to chronic antigen challenge, but that this boost would be transient because of the inability of the cells to persist.

To evaluate our hypothesis, we first compared disease progression in a mouse model of chronic viral infection in Hdac3-KO and Hdac3-WT mice. We infected E8I-Cre+;Hdac3fl/fl (Hdac3-KO) mice and

Cre-;Hdac3fl/fl littermates (Hdac3-WT) with a high dose of LCMV Clone 13 to induce chronic infection, and monitored disease progression and CD8+ T cell effector phenotypes in the infected mice. Consistent with our hypothesis, Hdac3-KO mice showed a moderate delay in initial disease progression as indicated by delayed weight loss and reduced viraemia on day 7, during the establishment stage of chronic viral infection47 (Figure. 6.1); however, this early delay did not confer protection from the disease, nor did it ameliorate disease severity at time points later than 7 days following infection.

Figure 6.1 | Hdac3-KO mice show delayed disease progression early during chronic infection with LCMV Clone 13. a,b, Hdac3-KO mice (n=8) or Hdac3-WT littermates (n=6) were infected with 2 x 106 PFU LCMV Clone 13 by intravenous injection. Weight loss (a) and viral load in peripheral blood serum (b) were monitored longitudinally over 5 weeks. Data are representative of 2 independent experiments with 6-8 mice per genotype each. a,b, Centre values, mean; error bars, s.e.m. (a) or s.d. (b); ** P < 0.01, **** P < 0.0001, 2-way ANOVA.

To evaluate the phenotype of CD8+ T cells responding to the infection, we analysed the CD8+ T cell response in the spleens, livers, and kidneys of infected mice at the early (day 8) and late (day 39) stages of infection by flow cytometry. On day 8, despite the early delay in disease progression in Hdac3-KO mice, we observed lower numbers of total and Db-GP33 tetramer+ virus-specific CD8+ T cells compared to WT littermates (Figure 6.2a, left and centre). Within the CD8+ T cell compartment, a higher proportion of CD8+

T cells in Hdac3-KO mice were Granzyme B+ (Figure 6.2b). CD4+ T cell numbers were similar between both groups except for a small increase in the kidneys of Hdac3-KO mice (Figure 6.2a, right). On day 39, numbers of total and virus-specific CD8+ T cells in Hdac3-KO mice were generally lower than those of

Hdac3-WT mice, although these differences were less pronounced than those observed on day 8; CD4+ T cell numbers were slightly increased in Hdac3-KO mice compared to Hdac3-WT mice (Figure 6.2c). We also continued to observe a higher proportion of Granzyme B+ CD8+ T cells in Hdac3-KO mice relative to

Hdac3-WT mice on day 39 (Figure 6.2d). In contrast, the percentages of IFN-γ+ and TNF-α+ CD8+ T cells were either unchanged (in spleen and kidneys) or lower (in liver) in Hdac3-KO mice relative to Hdac3-WT mice at both the day 8) and day 39 time points (Figure 6.2e,f).

Figure 6.2 | Comparison of Hdac3-KO and Hdac3-WT CD8+ T cell responses during chronic infection with LCMV Clone 13. a-f, Hdac3-KO mice or Hdac3-WT littermates were infected with 2 x 106 PFU LCMV Clone 13 by intravenous injection. Indicated organs were harvested at either early (day 8) or late (day 39) time points during the infection for analysis of the CD8+ T cell response by flow cytometry. a, Analysis of total T cells and LCMV-specific CD8+ T cells on day 8. b, Granzyme B+ CD8+ T cell responses on day 8. c, Analysis of total T cells and LCMV-specific CD8+ T cells on day 39. d, Granzyme B+ CD8+ T cell responses on day 39. e,f, Analysis of cytokine+ CD8+ T cell responses on day 8 (e) and day 39 (f). Data are pooled across two independent experiments each with 4 mice per genotype (a,b,e) or from one experiment with 5 mice per genotype (c,d,f). Functional responses in day 8 kidneys (†) were from one of the two independent experiments due to insufficient cell numbers recovered in the second repeat. a-f, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, two-tailed Student’s t-test.

These data suggest that CD8+ T cell-restricted loss of Hdac3 can confer increased resistance to chronic viral infection during the early establishment phase of infection, and are consistent with our previous

54 findings that HDAC3 inhibits the early development of the cytotoxic response and is required for persistence of activated CD8+ T cells.

HDAC3 inhibition delays the outgrowth of subcutaneous MC38 tumours in a CD8+ T cell-dependent manner

Because chronic viral infection and tumour growth both provide sources of persistent antigens and inflammation driving an immune response, we reasoned that the loss of HDAC3 activity in CD8+ T cells might also similarly delay the growth of implanted mouse tumours. Therefore, we challenged wildtype mice with subcutaneously implanted MC38 colon adenocarcinoma tumours (an ‘immunologically hot’ tumour that is known to generate a strong CD8+ T cell response48) and treated them with intraperitoneal injections of either the HDAC3 inhibitor RGFP966 or a vehicle control once tumours became palpable.

MC38 tumours in mice treated with RGFP966 showed a significant delay in outgrowth relative to mice in the control group (Figure 6.3a). As a result, RGFP966-treated mice survived longer than vehicle-treated mice, although the treatment did not cause tumour remission (Figure 6.3b). RGFP966-induced delay of tumour outgrowth required the presence of CD8+ T cells, as antibody-mediated depletion of CD8+ T cells abrogated the benefit of RGFP966 treatment (Figure 6.3a, dotted lines). To evaluate the effect of RGFP966 administration on CD8+ tumour-infiltrating lymphocytes (TILs), we analysed TILs in MC38 tumours on day

12 (after 4 doses of RGFP966 treatment), when tumour masses were comparable between RGFP966- and vehicle-treated groups (Figure 6.3c, left). At this time point, there were no significant differences between the numbers of CD8+ or CD4+ TILs per unit mass of tumour (Figure 6.3c, right). However, CD8+ TILs isolated from RGFP966-treated tumours showed a trend towards an increased frequency of Granzyme B+ cells, which translated into significantly higher numbers of Granzyme B+ CD8+ TILs per unit mass of tumour overall, relative to CD8+ TILs from vehicle-treated tumours (Figure 6.3d).

Figure 6.3 | Inhibition of HDAC3 with RGFP966 delays MC38 tumour growth and increases survival duration in a CD8+ T cell-dependent manner. a-d, C57BL/6 mice were subcutaneously injected with 1.5 x 105 MC38 colon adenocarcinoma cells, and treated with intraperitoneal injection of 10 mg/kg RGFP966 or an equivalent amount of DMSO vehicle every 2 days beginning on day 5 when tumours were palpable. CD8+ T cells were depleted with intraperitoneal injections of 100 μg α-CD8β antibody every 6 days beginning on day 5 in indicated treatment groups. Tumour burdens (a) and survival (b) of MC38- challenged mice. Data are representative of 2 independent experiments with 7-8 mice per treatment group each. c,d, Analysis of tumour-infiltrating lymphocytes (TILs) isolated from MC38 tumours 12 days after implantation. c, Tumour masses and TIL numbers per gram of tumour. d, Flow cytometric quantification of Granzyme B+ CD8+ TILs. Gated on live TCRβ+ CD8α+ CD4- events. Data are from one experiment with 5-6 mice per treatment group. a-d, Centre values, mean; error bars, s.e.m. (a) or s.d. (c,d). * P < 0.05, *** P < 0.001, **** P < 0.0001; 2-way ANOVA (a), log-rank Mantel-Cox test (b), or two- tailed Student’s t-test (c,d).

As a follow-up experiment using a genetic model of Hdac3 deficiency to verify that loss of HDAC3 activity caused the early delay in outgrowth of implanted MC38 tumours, we evaluated the tumour growth kinetics of subcutaneous MC38 and B16F10 tumours in TCR-polyclonal Hdac3-KO and -WT mice. However, in contrast to our previous findings using RGFP966 to inhibit HDAC3 activity, we did not observe significant differences between the tumour growth curves of Hdac3-KO and -WT mice in either tumour model (Figure

6.4). We theorised that this difference in the results from the two experimental systems was likely to have been caused by the permanent lack of persistence of activated Hdac3-KO CD8+ T cells, which might have negated any benefits of an early increase in cytotoxicity of the anti-tumour CD8+ T cell response in Hdac3-

KO mice. On the other hand, RGFP966-mediated inhibition of HDAC3 likely produced a beneficial response because the effects of the drug were only transient and therefore did not fully limit the durability of the augmented cytotoxic CD8+ T cell response.

Figure 6.4 | Hdac3-KO mice do not show significant delays in outgrowth of model subcutaneous tumours compared to Hdac3-WT mice. a, Hdac3-KO (n = 10) and Hdac3-WT littermates (n = 8) were subcutaneously injected with 1.5 x 105 MC38 colon adenocarcinoma cells. Data are representative of two independent experiments with at least 5 mice per genotype. b, Hdac3-KO (n = 11) and Hdac3-WT littermates (n = 11) were subcutaneously injected with 2 x 105 B16F10 melanoma cells. Data are representative of two independent experiments with at least 10 mice per genotype. a,b, Centre values, mean; error bars, s.e.m.; data were analysed by 2-way ANOVA.

We thus adopted a different genetic approach to evaluate whether these phenotypic changes were indeed due to loss of HDAC3 activity in tumour-specific CD8+ TILs. We first injected mice with MC38 tumours expressing ovalbumin (MC38-OVA), and then adoptively transferred congenically-marked Hdac3-

KO or -WT OT-I+ CD8+ T cells on day 5 when the tumours were palpable. Tumours were harvested 8 days after adoptive transfer for analysis of the tumour-specific donor OT-I+ CD8+ TILs (Figure 6.5a).

Figure 6.5 | Hdac3-KO tumour-infiltrating CD8+ T cells express more Granzyme B and are less persistent than Hdac3-WT CD8+ T cells. a-c, CD45.1+ congenic C57BL/6 mice were subcutaneously injected with 1.5 x 105 ovalbumin-expressing MC38 colon adenocarcinoma cells and adoptively transferred with 1 x 105 CD45.2+ OT-I+ CD8+ T cells from Hdac3-KO mice or Hdac3-WT littermates on day 5. Tumours were harvested for analysis 8 days after adoptive transfer of OT-I+ CD8+ T cells. b, Tumour masses and numbers of OT-I+ CD8+ TILs per gram of tumour. c, Flow cytometric analysis of Granzyme B response of OT-I+ CD8+ TILs. Gated on live TCRβ+ CD8α+ CD4- CD45.2+ events. Data are from one experiment with 8 recipient mice for each genotype of donor OT-I+ CD8+ T cells. b,c, Centre values, means; error bars, s.d.; * P < 0.05, **** P < 0.0001; two-tailed Student’s t-test.

At this time point, mice that received Hdac3-KO OT-I+ CD8+ T cells had slightly larger tumours on average compared to mice that received Hdac3-WT cells, although this difference was not significant (Figure 6.5b, left). There were also fewer OT-I+ CD8+ TILs present per unit mass of tumour in mice that received Hdac3-

KO cells relative to mice that received Hdac3-WT OT-I+ CD8+ T cells (Figure 6.5b, right). We also observed an increased proportion of Granzyme B+ OT-I+ CD8+ TILs in mice that received Hdac3-KO OT-I+ CD8+ T cells relative to mice that received Hdac3-WT cells, although this did not result in higher numbers of Granzyme

B+ cells per unit mass of tumour due to the reduced numbers of OT-I+ TILs and the larger overall tumour masses (Figure 6.5c). In addition, transfer of Hdac3-KO OT-I+ CD8+ T cells into MC38-OVA tumour-bearing hosts did not reduce the tumour burden in the long term relative to mice receiving Hdac3-WT T cells, likely because of the reduced persistence of Hdac3-KO tumour-infiltrating CD8+ T cells (data not shown). This is consistent with data from previous experiments in which we challenged TCR-polyclonal Hdac3-KO and -

WT mice with subcutaneous tumours.

Altogether, we show that inhibition of HDAC3 activity in a tumour model with a robust CD8+ T cell response can delay tumour outgrowth and improve survival duration in a CD8+ T cell-dependent manner, and that this improvement is also associated with a stronger Granzyme B+ CD8+ response.

Chapter 7

Transcriptional and epigenetic signatures of HDAC3

HDAC3 inhibits gene programmes associated with CD8+ T cell cytotoxic effector function

We next focussed on elucidating potential molecular mechanisms by which HDAC3 regulated the development of CD8+ T cell cytotoxic effector phenotype during T cell activation. Because HDAC3 functions as a histone deacetylase when in complex with nuclear co-repressors49,50, we hypothesised that

HDAC3-targeted deacetylation of acetylated histone tails during T cell activation could potentially play a role in the regulation of CD8+ T cell effector function. We therefore adopted a two-pronged approach to investigate this hypothesis by profiling changes in both the transcriptomic and epigenetic profiles of CD8+

T cells lacking HDAC3 activity during activation, and comparing these against those of wildtype CD8+ T cells activated in parallel. To facilitate the collection of large numbers of cells for these molecular analyses, and to minimise potential complications when working with in vivo experimental systems because of the lack of persistence of activated Hdac3-KO CD8+ T cells, we used the in vitro model of activating OT-I+ CD8+ T cells by co-culture with irradiated Ova peptide-pulsed BMDCs as the standard experimental system for our studies.

We first examined global transcriptional differences between sort-purified Hdac3-KO and -WT OT-I+ CD8+

T cells that were activated for 5 days in vitro (Figure 7.1a). Genes belonging to the cytotoxicity programme

(Gzma, Gzmb, Gzmc, Gzmd, Gzme, Prf1) showed the strongest differential expression and were more highly expressed in Hdac3-KO relative to Hdac3-WT CD8+ T cells, consistent with our previous data showing that HDAC3 suppresses CD8+ T cell cytotoxicity. In addition, genes encoding transcription factors that promote CD8+ T cell effector function and/or terminal effector differentiation (Tbx21, Id2, Prdm1)

Figure 7.1 | HDAC3 negatively regulates a cytotoxic effector-associated gene programme in CD8+ T cells. a-c, OT-I+ CD8+ T cells were co-cultured with irradiated Ova peptide-pulsed BMDCs in vitro for 5 days and sorted to purity for whole-genome RNA-sequencing. (figure legend continues on next page)

Figure 7.1 (continued from previous page) a,b, Analysis of whole genome RNA-sequencing of Hdac3-KO and -WT CD8+ T cells (a) or RGFP966-treated and vehicle-treated CD8+ T cells (b) 5 days after activation. Significance was determined using thresholds of log10(adjusted P-value) > 10 and |log2(fold change)| > 0.5. Data are from one experiment with three technical replicates per sample. c, Differentially-expressed genes common across Hdac3-KO OT-I+ CD8+ T cells and RGFP966-treated OT-I+ CD8+ T cells after 5 days of in vitro activation. were also more highly transcribed in Hdac3-KO than in Hdac3-WT CD8+ T cells. We also observed a concomitant decrease in transcription of other transcriptional regulators of CD8+ T cell differentiation such as Eomes, Fos, and Fosb. Finally, transcripts of a subset of chemokine receptor genes (Ccr5, Ccr2) and pro-inflammatory cytokine genes (Lta, Ccl5, Ccl4, Ccl3, and Tnf, but not Ifng or Il2) associated with T cell activation were also higher in Hdac3-KO compared to Hdac3-WT cells. We also performed a similar experiment using RGFP966 to ablate HDAC3 to generate a second set of differentially-expressed genes associated with loss of HDAC3 activity (Figure 7.1b).

Comparing both sets of differentially expressed genes from the two experiments, we found 20 genes that were differentially expressed in both Hdac3-KO CD8+ T cells and in HDAC3 inhibitor-treated cells (Figure

7.1c). Within this common set of 20 genes, we identified genes belonging to the 4 functional categories previously identified: cytotoxicity (Gzmb, Gzmc, Prf1), transcription factors associated with CD8+ T cell effector differentiation (Prdm1, Id2), as well as pro-inflammatory cytokines (Ccl5) and chemokine receptors (Ccr5, Ccr2). We further compared our data with data from CD8+ T cell populations surveyed by the Immunological Genome Project database (ImmGen)51 to glean further insights into the transcriptional signature associated with HDAC3 activity (Figure 7.2).

Figure 7.2 | Expression of HDAC3-regulated genes across activated CD8+ T cell populations. Gene expression profiles of the 20 differentially-expressed genes shown in Figure 6.1c across different CD8+ T cell populations as documented on the Immunological Genome Project online database. (figure legend continues on next page) 64

Figure 7.2 (continued from previous page) Thy, CD8+ CD4- thymocytes; Naïve, splenic CD62Lhi CD44lo αβ CD8+ T cells; TE and MP, splenic KLRG1+ CD127- and KLRG1- CD127+ P14 LCMV-specific TCR-transgenic CD8+ T cells on day 7 of acute LCMV infection, respectively; IEL, small intestinal intraepithelial P14+ CD8+ T cells on day 7 of acute LCMV + + + - + + infection; TCM and TEM, splenic CD62L CD44 and KLRG1 CD127 P14 CD8 T cells on day 180 after acute LCMV infection, respectively.

We found that the 17 genes with increased transcription in HDAC3 activity-deficient activated CD8+ T cells were generally upregulated in the activated CD8+ T cell populations surveyed by ImmGen and were expressed at lower levels in resting naïve CD8+ T cells. The majority of these genes (12 of 17) also had higher mRNA expression in terminally-differentiated effector (TE) cells than in memory progenitor (MP) cells in the context of acute LCMV infection (Figure 7.2, red and purple bars). In addition, the three genes with consistent downregulation of transcription in the absence of HDAC3 activity were most highly transcribed either in naïve (Ift80) or memory (Plxnc1, Bst1) CD8+ T cell populations. These findings served as a validation that the 20 genes identified in our RNA-seq approaches were indeed associated with CD8+

T cell effector function and activation. Finally, we also found that 6 out of the 17 genes upregulated in the absence of HDAC3 activity (Gzmc, Gzmb, B4galnt4, Dusp5, Prdm1, Dusp2) were also most highly expressed in intestinal epithelial CD8+ T cells (IELs) among all CD8+ T cell populations surveyed by the ImmGen project

(Figure 7.2, orange bars). This was particularly intriguing to us because the virus-specific CD8+ IEL population in the context of acute LCMV infection is a model of tissue-resident CD8+ T cell responses, and because Blimp-1 (encoded by Prdm1, the most highly upregulated transcription factor in HDAC3 activity- deficient CD8+ T cells) is a key transcriptional driver of tissue residency in CD8+ T cells52.

Overall, our RNA-seq data indicate that that HDAC3 negatively regulates gene programmes associated with CD8+ T cell cytotoxic effector function, including direct mediators of cytotoxicity (Gzmb, Gzmc, Prf1) and transcription factor genes that promote a terminally-differentiated effector phenotype in CD8+ T cells

(Prdm1, Id2). Our data also suggest the possibility that HDAC3 may play a role in regulating IEL tissue- resident CD8+ T cell responses.

Characterisation of histone deacetylation by HDAC3 in activated CD8+ T cells

Because HDAC3 has been shown to deacetylate a variety of histone acetyl-lysine residues53,54, and because the experimental systems and procedures to characterise epigenetic marks in CD8+ T cells were logistically and technically demanding, we first performed a pilot experiment to determine which histone acetyl- lysine residue to focus our investigations on. We activated OT-I+ CD8+ T cells by co-culturing them with irradiated Ova peptide-pulsed BMDCs in the presence of the HDAC3 inhibitor RGFP966 or an equivalent amount of DMSO vehicle, and profiled global levels of acetylation at the lysine residues 9, 14, and 27 of

Histone 3 (H3K9-ac, H3K14-ac, and H3K27-ac) as well as expression of HDAC3 by Western blot analysis

(Figure 7.3). We also profiled the protein expression levels of the transcription factors T-bet (encoded by

Tbx21) and Blimp-1 (encoded by Prdm1) as markers of CD8+ T cell activation and effector differentiation.

Our decision to use RGFP966 to pharmacologically inhibit HDAC3 activity rather than taking a genetic approach was made so as to avoid potential complications due to the lack of persistence consistently observed in Hdac3-KO CD8+ T cells upon activation.

Inhibition of HDAC3 activity increased the levels of H3K9-ac, H3K14-ac, and H3K27-ac in RGFP966-treated activated CD8+ T cells relative to control cells, especially during the period 3-7 days after initial activation

(Figure 7.3a, left and Figure 7.3b, top). This also coincided with the peak expression of the transcription factors T-bet and Blimp-1. Furthermore, we found that the levels of HDAC3 protein also increased with time over the time course of in vitro activation (Figure 7.3a, right, and Figure 7.3b, bottom). HDAC3 thus function as a bona fide histone deacetylase in CD8+ T cells.

Figure 7.3 | Dynamics of global levels of acetylation of Histone 3 lysine residues during CD8+ T cell activation. a,b, OT-I+ CD8+ T cells were co-cultured with irradiated Ova peptide-pulsed BMDCs in vitro in the presence of 3 μM RGFP966 or an equivalent volume of DMSO vehicle. At indicated time points, CD8+ T cells were sorted to purity for Western blot analysis (a) and quantification (b) of global levels of Histone 3 acetyl-lysine residues, HDAC3, T-bet, and Blimp-1 (right). Lamin B1 was used as a loading control. Whole cell lysates equivalent to 1 x 105 cells or 3 x 105 cells (H3K27-ac and HDAC3 blots only) were loaded in each lane.

While H3K9-ac and H3K27-ac levels in CD8+ T cells were both readily detectable by Western blot, H3K9-ac is associated with transcriptionally-active gene promoters, whereas H3K27-ac stably marks active enhancer regions of chromatin55. Both H3K9 and H3K27 can also be methylated, and both methylated states are highly associated with chromatin inactivation and transcriptional silencing55. Methylation and

67 acetylation of these residues are mutually exclusive as they both covalently modify the same primary amino group of lysine. Because we were interested in the possibility of bona fide epigenetic regulation of stable functional states of CD8+ T cells, and because previous studies had also focussed on profiling H3K27- ac and H3K27-trimethylation in activated CD8+ T cells in the context of acute viral infection, we decided to focus our subsequent work on profiling HDAC3-associated changes in H3K27-ac levels.

We examined the levels of H3K27-ac at the promoter regions of gene loci coding for CD8+ effector cytokines, cytotoxicity mediators, and transcription factors known to polarise CD8+ T cell differentiation.

We activated Hdac3-KO and -WT OT-I+ CD8+ T cells in vitro by co-culture with irradiated SIINFEKL-pulsed

BMDCs, sort-purified CD8+ T cells after 3 days for chromatin preparation, and performed chromatin immunoprecipitation using an antibody specific for H3K27-ac. Relative levels of H3K27-ac were then quantified by real-time quantitative PCR using primers targeting the genomic region up to 1000 bases upstream of the transcriptional start site of each gene locus (Figure 7.4a). We also used a similar approach to investigate the levels of H3K27-ac at the same set of gene promoter regions in RGFP966- and vehicle- treated OT-I+ CD8+ T cells on day 5 after in vitro activation (Figure 7.4b).

We observed higher levels of H3K27-ac at the promoters of cytotoxicity genes (Gzma, Gzmb, Fasl) in

Hdac3-KO relative to Hdac3-WT cells. Hdac3-KO CD8+ T cells also showed increased H3K27-ac at the

Prdm1 promoter, and at the promoters of the cytokine genes Tnf and Lta (Figure 7.4a). In RGFP966- treated cells, increased H3K27-ac was also observed at the promoters of Gzmb, Fasl, Prdm1, Tnf, and Lta relative to vehicle control-treated cells. In addition, higher levels of H3K27-ac were also observed at the

Ifng and Id2 promoters in activated RGFP966-treated CD8+ T cells (Figure 7.4b).

Figure 7.4 | HDAC3 targets a subset of Histone 3 acetyl-lysine 27 residues for deacetylation. a,b, OT-I+ CD8+ T cells were co-cultured with irradiated Ova peptide-pulsed BMDCs in vitro and sorted to purity for chromatin preparation and immunoprecipitation with an antibody specific to H3K27-ac. Relative levels of H3K27-ac were quantified by real-time quantitative PCR using primer probes targeting the promoter regions of indicated gene loci (-1000 bp to transcription start site). a, H3K27-ac at gene promoters in CD8+ T cells from Hdac3-KO mice or Hdac3-WT littermates on day 3. Data are representative of two experiments with three technical replicates per gene promoter. b, H3K27-ac at gene promoters in RGFP966- or vehicle-treated CD8+ T cells on day 5. Data are from one experiment with three technical replicates per gene promoter. a,b, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001; two-tailed Student’s t-test.

Our results indicate that HDAC3-mediated deacetylation of H3K27-ac at the promoters of cytotoxic effector genes such as Gzmb and at the promoter of Prdm1 was associated with negative regulation of transcription from these loci by HDAC3, suggesting that that HDAC3 may potentially suppress transcription of these genes in part by removing the activating H3K27-ac mark. However, because we have yet to observe a general systematic correlation between transcription and H3K27-ac, we are unable to definitively conclude that HDAC3-targeted deacetylation of H3K27-ac is a molecular mechanism by which it mediates the regulation of CD8+ T cell effector phenotype and function observed in our functional

69 studies. We are currently actively seeking to fill this gap in our knowledge by performing ChIP-sequencing on chromatin prepared from Hdac3-KO and Hdac3-WT CD8+ T cells after 3 days of in vitro activation using antibodies specific for H3K27-ac and using antibodies specific for HDAC3. We anticipate that the results of this experiment will enable us to track changes in H3K27-ac across chromatin regions that overlap with

HDAC3 binding sites, and will therefore provide novel insights into the molecular regulation of CD8+ T cell function by HDAC3.

Chapter 8

Uncovering potential mechanisms of HDAC3-mediated regulation of CD8+ T cell effector function

using genetic approaches

Uncovering potential mechanisms of HDAC3-mediated regulation of CD8+ T cell effector function using genetic approaches

Having determined in our functional studies that HDAC3 regulated both the cytotoxicity and persistence of CD8+ T cells after activation, and having uncovered molecular evidence supporting a role for HDAC3 in regulating CD8+ T cell cytotoxic effector phenotype, we next concentrated on understanding whether and how HDAC3 interacted with known transcriptional regulators of CD8+ T cell function and differentiation in order to integrate our discoveries with the wider body of literature on this topic. We probed for potential functional links between transcription factors and HDAC3 by investigating epistatic genetic interactions between Hdac3 and candidate transcription factor genes. To do so, we first bred OT-I+ Cas9+ double transgenic mice with E8I-Cre+;Hdac3fl/fl OT-I+ mice to generate Cas9+ Hdac3-KO OT-I+ mice. By transducing

OT-I+ CD8+ T cells isolated from these mice with lentiviral gRNA vectors, we generated transcription factor gene- and Hdac3-double knockout CD8+ T cells for subsequent functional phenotypic analyses to evaluate potential interactions between the gene of interest and Hdac3. Based on our findings from both the functional and molecular analyses of activated CD8+ T cells lacking HDAC3 activity, as well as our review of existing literature, we selected the transcription factors Blimp-1, T-bet, and Runx3 (encoded by Prdm1,

Tbx21, and Runx3, respectively) as the top three candidate transcription factors for our investigations.

Inactivation of Prdm1 (encoding Blimp-1) increases persistence of Hdac3-KO CD8+ T cells

In activated CD8+ T cells, Blimp-1 is known to drive differentiation into a terminal effector state, in which

CD8+ T cells possess high levels of effector function but lose their capacity for self-renewal and persistence56-59. Because our RNA-seq and ChIP-qPCR data suggested that Blimp-1 activity was increased in the absence of HDAC3 activity, we hypothesised that an aberrant increase in Blimp-1 activity might

72 account for the reduced accumulation of Hdac3-KO compared to Hdac3-WT CD8+ T cells after activation in vivo. To test this hypothesis, we first evaluated the efficacies of lentiviral vectors bearing Prdm1- targeting gRNA vectors in Cas9+ OT-I+ CD8+ T cells by confirming gene inactivation in Thy1.1+ transductants using Western blot analysis of Blimp-1 (Figure 8.1a).

Figure 8.1 | Genetic ablation of Blimp-1 (encoded by Prdm1) partially rescues the reduced persistence of Hdac3-KO CD8+ T cells after activation. a, Evaluation of knockout efficiency of Prdm1-targeting lentiviral gRNA expression vector in transduced Cas9+ OT-I+ CD8+ T cells. Whole cell lysates prepared from 2 x 105 magnetically-purified transductants were loaded per lane. LZ - LacZ targeting gRNA vector (negative control). Molecular weights in kDa are indicated on immunoblot images. The red arrow indicates the gRNA sequence used in subsequent experiments. b,c, Hdac3-WT Cas9+ OT-I+ CD8+ T cells were transduced with lentiviral vectors expressing Prdm1- or LacZ-targeting gRNA sequences, magnetically purified for Thy1.1+ expression, and adoptively transferred into Thy1.2+ C57BL/6 recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. Frequencies of donor OT-I+ CD8+ T cells relative to total CD8+ T cells in draining lymph nodes were quantified by flow cytometry after 7 days. Data are representative of two independent experiments with 5 mice per treatment group. c, Centre values, mean; error bars, s.d.; * P < 0.05, *** P < 0.001; 2-way ANOVA.

We next generated Prdm1- Hdac3-double KO OT-I+ CD8+ T cells by transduction of Cas9+ OT-I+ CD8+ T cells derived from Hdac3-KO mice or Hdac3-WT littermates with lentiviral vectors carrying our validated Prdm1 gRNA sequence or a control gRNA sequence (LacZ). Thy1.1+ transductants were subsequently transferred into congenic hosts that were then immunised with OVA + poly(I:C), and evaluated the persistence of each genotype of transferred OT-I+ CD8+ T cells in the draining lymph nodes after 7 days (Figure 8.2b), by which time Hdac3-KO OT-I+ CD8+ T cells were present at very low frequencies relative to co-transferred Hdac3-

WT cells (Figure 5.1, and Figure 5.2). We found that the reduction of Hdac3-KO relative to Hdac3-WT CD8+

T cells was partially rescued when Prdm1 was inactivated in Hdac3-KO cells (Figure 8.1c, Figure 8.2a).

These data support our hypothesis that increased Blimp-1 activity contributed to the loss of persistence observed in Hdac3-KO CD8+ T cells.

We also compared the effector phenotype of Prdm1- Hdac3-double KO CD8+ T cells to that of Hdac3-single

KO and Hdac3-WT cells in the OT-I transfer and OVA immunisation experimental system. We found that there was an increased frequency of T-bet+ cells in Prdm1- Hdac3-double KO CD8+ T cells on day 7 following immunisation compared to Hdac3-KO CD8+ T cells (Figure 8.2c). However, inactivation of Prdm1 did not otherwise result in significant differences in the percentages of Granzyme B+, IFN-γ+, or TNF-α+ cells in either Hdac3-KO or Hdac3-WT CD8+ T cells (Figure 8.2b,d,e). The significant increase in numbers of Prdm1- Hdac3-double KO CD8+ T cells resulted in an increased accumulation of Granzyme B+ and T-bet+

CD8+ T cells, both in comparison to Hdac3-KO and Hdac3-WT CD8+ T cells (Figure 8.2f, left). However, the total numbers of IFN-γ+ and TNF-α+ cells were not increased in Prdm1- Hdac3-double KO cells relative to

Hdac3-KO cells (Figure 8.2f, right).

Overall, our data indicate that Blimp-1 contributes to limiting the persistence of Hdac3-KO CD8+ T cells.

Furthermore, because inactivation of Prdm1 increased both the proportion and number of T-bet+ Hdac3-

KO CD8+ T cells, our data support a model in which HDAC3 may sustain the durability of the CD8+ T cell response by restraining expression of Blimp-1, which in turn limits the persistence of T-bet+ effector cells by driving them into a terminally-differentiated phenotype with reduced proliferative capacity.

Elucidating the biochemical aspects of this mechanism will be a critical next step towards defining a molecular pathway of CD8+ T cell phenotypic regulation from HDAC3 via Blimp-1.

Figure 8.2 | Phenotype of Prdm1-Hdac3-double KO CD8+ T cells following in vivo activation. a-f, Flow cytometric analysis of donor Hdac3-WT Thy1.1+ Cas9+ OT-I+ CD8+ T cells expressing indicated gRNA sequences in inguinal lymph nodes 7 days after adoptive transfer and in vivo activation as in Figure 7.1b. Data are representative of two independent experiments with 5 mice per treatment group. (figure legend continues on next page)

Figure 8.2 (continued from previous page) e,f, Quantification of phenotypic changes in transferred Cas9+ OT-I+ CD8+ T cells transduced with indicated gRNAs as total cell numbers (e) or as a proportion (f) of the donor OT-I+ CD8+ population in draining lymph nodes. e,f, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001; 2-way ANOVA.

T-bet is dispensable for the augmented cytotoxicity of Hdac3-deficient CD8+ T cells

Because loss of HDAC3 activity consistently increased the cytotoxicity of CD8+ T cells, we next assessed potential interactions between HDAC3 and known molecular drivers of CD8+ T cell cytotoxicity. Earlier studies had already flagged out a requirement for the transcription factors Runx3 and T-bet (encoded by

Tbx21) for the development of cytotoxicity in CD8+ T cells60. Furthermore, we had consistently observed an increase in T-bet expression in activated CD8+ T cells lacking HDAC3 activity in our functional studies of

Hdac3-KO and RGFP966-treated CD8+ T cells. We therefore evaluated the effect of T-bet or Runx3 ablation on the augmented cytotoxicity of Hdac3-KO CD8+ T cells using the tools of lentiviral gRNA transduction and the Cas9+ Hdac3-KO OT-I-transgenic mouse strain.

T-bet plays a well-characterised role in regulating effector phenotype development in CD8+ T cells and is required for the acquisition of many characteristic effector functions of CD8+ cytotoxic T cells such as contact-dependent killing and the secretion of IFN-γ8,59,61,62. We first generated and validated Tbx21- targeting gRNA sequences by confirming efficient reduction of T-bet protein by both Western blot (Figure

8.3a) and flow cytometric analysis (Figure 8.3b) of Cas9+ OT-I+ CD8+ T cells transduced with lentiviral gRNA vectors. Next, we activated Cas9+ Hdac3-KO OT-I+ CD8+ T cells transduced with Tbx21- or LacZ-targeting gRNAs by co-culture with irradiated Ova peptide-pulsed BMDCs for 7 days, and measured their cytotoxicity in a 4-hour 51Cr-release assay using IFN-γ pre-treated B16-OVA melanoma cells as targets. We found that inactivation of Tbx21 in Hdac3-KO CD8+ T cells did not significantly change their cytotoxicity

76 compared to Hdac3-KO CD8+ T cells (Figure 8.3c), which suggested that the augmentation of CD8+ T cell cytotoxicity as a result of Hdac3 loss was not dependent on T-bet.

Figure 8.3 | Inactivation of Tbx21 does not abrogate the increased cytotoxicity of Hdac3-KO CD8+ T cells. a,b, Validation of lentiviral gRNA vectors targeting Tbx21 in Cas9+ OT-I+ CD8+ T cells. a, Whole cell lysates from 2 x 105 magnetically-purified transductants were loaded per lane for immunoblot analysis. LZ, LacZ-targeting gRNA vector (negative control). Molecular weights (kDa) are indicated on immunoblot images. The red arrow indicates the gRNA sequence used in subsequent experiments. b, Flow cytometric measurement of T-bet expression in Cas9+ OT-I+ CD8+ T cells transduced with indicated gRNA sequences 3 days post-transduction. Gated on Thy1.1+ transductants. c, Hdac3-KO Cas9+ OT-I+ CD8+ T cells were transduced with indicated gRNA sequences, activated in vitro with irradiated Ova peptide-pulsed BMDCs for 7 days, and evaluated for cytotoxicity against ovalbumin-expressing B16 melanoma targets. Data are from 1 experiment with 4 replicates per condition. c, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, 2-way ANOVA.

However, given our consistent observation of increased frequencies of T-bet+ cells in HDAC3 activity- deficient CD8+ T cells (corroborated with increased transcription of Tbx21 in Hdac3-KO cells), it remains likely that T-bet plays a role in the regulation of other aspects of CD8+ T cell function and differentiation by HDAC3. Understanding what these aspects are, as well as the underlying mechanism of interaction between HDAC3 and T-bet, could potentially reveal an interesting pathway in the cross-talk between epigenetic and transcriptional regulators of CD8+ T cell phenotype.

Runx3 is required for regulation of CD8+ T cell cytotoxicity by HDAC3

Runx3 is a transcription factor that plays multiple roles in the development of CD8+ cytotoxic T cells, both during lineage commitment during thymic development63,64, and during the activation and maturation of

77 cytotoxic CD8+ T cells60,65,66. Runx3 also co-operates with the effector function-associated transcription factor T-bet (encoded by Tbx21) during activation for full development of CD8+ T cell cytotoxicity60.

Therefore, in parallel with our work on Tbx21 in Hdac3-KO CD8+ T cells, we also engineered Runx3- targeting lentiviral gRNA vectors and validated them in Cas9+ OT-I+ CD8+ T cells to develop tools for investigating genetic interactions between Hdac3 and Runx3 in the regulation of CD8+ T cell cytotoxicity.

We first evaluated the efficacy of a Runx3 gRNA by Western blot (Figure 8.4a), and followed up by evaluating the in vitro cytotoxicity of Runx3- Hdac3-double KO CD8+ T cells against B16-OVA melanoma targets using an approach similar to the one we used to evaluate Tbx21- Hdac3-double KO cells. We found that inactivation of Runx3 severely abrogated the enhanced cytotoxic phenotype in Hdac3-KO CD8+ T cells

(Figure 8.4b), indicating that Runx3 was required for the regulation of CD8+ T cell cytotoxicity by HDAC3.

Figure 8.4 | Runx3 is required for the increased cytotoxicity of Hdac3-KO CD8+ T cells. a, Validation of lentiviral gRNA vectors targeting Runx3 in Cas9+ OT-I+ CD8+ T cells. Whole cell lysates from 2 x 105 magnetically-purified transductants were loaded per lane for immunoblot analysis. LZ, LacZ-targeting gRNA vector (negative control). Molecular weights (kDa) are indicated on immunoblot images. b, Hdac3-KO Cas9+ OT-I+ CD8+ T cells were transduced with indicated gRNA sequences, activated in vitro with irradiated Ova peptide-pulsed BMDCs, and evaluated for cytotoxicity against ovalbumin-expressing B16 melanoma targets. Data are from 1 experiment with 4 replicates per condition. b, Centre values, mean; error bars, s.d.; *** P < 0.001, **** P < 0.0001, 2-way ANOVA.

To further investigate whether other aspects of the phenotype of Hdac3-KO CD8+ T cells were associated with Runx3, we compared the functional phenotype of Runx3- Hdac3-double KO CD8+ T cells and Hdac3- single KO cells using the OT-I+ transfer and OVA immunisation model (Figure 8.5a). Inactivation of Runx3

78 in Hdac3-KO CD8+ T cells resulted in a significant reduction in the proportion of Granzyme B+ CD8+ T cells, similar to the level of Hdac3-WT cells, and also resulted in a slight decrease in T-bet+ cells (Figure 8.5b,c, and Figure 8.5e, left). Runx3 inactivation also resulted in a reduced cytokine response in CD8+ T cells following ex vivo restimulation with Ova peptide, independent of the presence of HDAC3 (Figure 8.5d, and Figure 8.5e, right).

We also evaluated whether Runx3 contributed to the decline in persistence observed in activated Hdac3-

KO CD8+ T cells by analysing the relative frequencies of Cas9+ OT-I+ CD8+ T cells transduced with Runx3- or

LacZ-targeting gRNAs on day 7. Genetic loss of Runx3 did not significantly rescue the decline in the persistence of activated Hdac3-KO CD8+ T cells. We also observed no significant effect of Runx3 ablation on the persistence of Hdac3-WT cells (Figure 8.6), although the reduction in frequency of Runx3- Hdac3- double KO CD8+ T cells relative to Hdac3-WT cells was less significant compared to the reduction in frequency of Hdac3-single KO CD8+ T cells.

Overall, these functional data indicated that expression of Runx3 was required for the regulation of CD8+

T cell cytotoxicity by HDAC3. Our data also suggested that the regulation of T-bet expression by HDAC3 might be linked to Runx3 as well, as inactivation of Runx3 in Hdac3-KO CD8+ T cells reduced the frequencies of T-bet+ cells and also further reduced the T-bet-associated IFN-γ response relative to Hdac3- single KO cells.

Figure 8.5 | Phenotypic analysis of Runx3-Hdac3-double KO CD8+ T cells following in vivo activation. a-e, Hdac3-KO or -WT Cas9+ OT-I+ CD8+ T cells were transduced with lentiviral vectors expressing Runx3- or LacZ-targeting gRNA sequences, magnetically purified for Thy1.1+ expression, and adoptively transferred into Thy1.2+ C57BL/6 recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. Data are representative of two independent experiments each with 5 mice per treatment group. e, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001; 2-way ANOVA.

Figure 8.6 | Reduced persistence of Hdac3-KO CD8+ T cells is independent of Runx3. a,b, Hdac3-KO or -WT Cas9+ OT-I+ CD8+ T cells were transduced with lentiviral vectors expressing Runx3- or LacZ-targeting gRNA sequences, magnetically purified for Thy1.1+ expression, and adoptively transferred into Thy1.2+ C57BL/6 recipients. Mice were immunised subcutaneously with ovalbumin adjuvanted with poly(I:C) in PBS. After 7 days, the frequencies of donor Thy1.1+ OT-I+ CD8+ T cells were analysed by flow cytometry (a) and quantified relative to the total T cell compartment (b) in the draining lymph nodes. Data are from one experiment with 3 recipient mice for each Hdac3-WT OT-I+ group and 5 recipient mice for Hdac3-KO OT-I+ group. b, Centre values, mean; error bars, s.d.; * P < 0.05, ** P < 0.01; 2-way ANOVA.

To investigate whether Runx3 played a role in the transcriptional regulation of the HDAC3-associated gene signature (Figure 7.1 and Figure 7.2), we performed ChIP-sequencing with an antibody specific to Runx3 on chromatin prepared from sort-purified Hdac3-KO and -WT OT-I+ CD8+ T cells activated by co-culture with irradiated Ova peptide-pulsed BMDCs. We found that Runx3 was indeed bound to chromatin near the loci of genes that showed increased transcription in Hdac3-KO CD8+ T cells after activation (Figure

8.7). While we observed increased Runx3 binding at regulatory elements near the loci of a subset of genes encoding mediators of cytotoxicity (Gzmb, Prf1, Fasl) (Figure 8.7, top row), this was not consistently observed across the other cytotoxicity-associated gene loci (Gzma, Gzmc, Gzmd) that showed increased transcription in activated Hdac3-KO CD8+ T cells (Figure 8.7, second row). In addition, we did not observe increased binding of Runx3 in Hdac3-KO CD8+ T cells near the genes coding for pro-inflammatory cytokines that had increased transcriptional activity in Hdac3-KO CD8+ T cells (Figure 8.7, third row).

Figure 8.7 | Runx3 binds to chromatin near the loci of genes with increased transcription in activated CD8+ T cells. OT-I+ CD8+ T cells from Hdac3-KO mice or Hdac3-WT littermates were co-cultured with irradiated Ova peptide-pulsed BMDCs and sorted to purity for chromatin preparation at indicated time points. ChIP-sequencing was performed using an antibody specific to Runx3. (figure legend continues on next page)

Figure 8.7 (continued from previous page) Sequencing reads were aligned to the mm10 build of the Mus musculus genome and visualised with the IGV Genome Browser (Broad Institute, Massachusetts). Grey columns indicate peaks with differential binding of Runx3 between Hdac3-KO and Hdac3-WT cells.

There was also no increase in Runx3 binding near the loci of Prdm1, Tbx21, or Id2 in Hdac3-KO CD8+ T cells relative to Hdac3-WT CD8+ T cells across the time points surveyed (Figure 8.7, fourth row). Interestingly, however, we found increased Runx3 binding near the Ccr2 and Ccr5 genomic loci in Hdac3-KO CD8+ T cells after activation (Figure 8.7, bottom row).

To systematically search for potential correlations of Runx3 binding with transcriptional activity in an unbiased manner, we performed deeper bioinformatics analysis on our Runx3 ChIP-seq and previous RNA- seq data to evaluate the regulatory potential34 of Runx3 binding across all transcribed genes before and after in vitro activation of CD8+ T cells. These analyses revealed that increased Runx3 binding in the genomic vicinity of a particular gene locus was associated with increased transcriptional activity only in the set of genes whose transcripts were upregulated in Hdac3-KO CD8+ T cells (Figure 8.8, red plots).

Figure 8.8 | Runx3 binding is associated with increased transcription of the set of genes upregulated in Hdac3-KO CD8+ T cells. Box-and-whisker plots showing association of Runx3 binding (regulatory potential) with genes differentially-expressed between Hdac3-KO and -WT OT-I+ CD8+ T cells, grouped by direction of change. Datasets for analysis were drawn from the Runx3 ChIP-sequencing data in Figure 8.7 and the RNA-sequencing data in Figure 7.1. Empty circles indicate individual outlier genes.

In contrast, we observed very little difference between the Runx3 regulatory potential of genes that were not differentially transcribed and that of genes whose transcription was downregulated in Hdac3-KO cells

(Figure 8.8, grey and blue plots).

Taken together, our molecular analyses suggest that Runx3 plays a role either downstream of, or in parallel with HDAC3 in regulating the gene programmes associated with HDAC3 activity. Moreover, in the context of its interaction with HDAC3, Runx3 likely functions as a positive regulator of transcription of the subset of genes with increased transcriptional activity in the absence of HDAC3 activity. In light of recent discoveries describing a role of Runx3 in regulating the multiple aspects of CD8+ T cell function, including memory formation66 and tissue residency67, further investigation into the molecular interaction between

HDAC3 and Runx3 could be highly informative for understanding the epigenetic regulation of these diverse aspects of CD8+ T cell phenotype following activation.

Inactivation of the HDAC3 binding partners NCOR1 or NCOR2 does not increase CD8+ T cell cytotoxicity

Previous work from other groups has demonstrated that HDAC3 binds the nuclear co-repressors NCOR1 or NCOR2 (also known as SMRT) to form complexes with histone deacetylase activity49,50. In vitro, binding of HDAC3 by either nuclear co-repressor activates HDAC3 deacetylase activity, while recombinant HDAC3 only possesses minimal deacetylase activity49. We therefore investigated whether the loss of the HDAC3 binding partners NCOR1 or NCOR2 would recapitulate the augmented cytotoxicity consistently observed in CD8+ T cells lacking HDAC3 activity, as an indication of whether the biological activities of HDAC3/NCOR complexes played a role in regulating the cytotoxicity of activated CD8+ T cells.

We first generated and validated Ncor1- and Ncor2-targeting lentiviral gRNA vectors by using Western blot analysis to assess the reduction in NCOR1 and NCOR2 protein levels in transduced Cas9+ OT-I+ CD8+ T cells (Figure 8.9a,b). We then measured the cytotoxicity of purified Cas9+ OT-I+ CD8+ T cells transduced with Ncor1- or Ncor2-targeting gRNAs against B16-OVA cells in a 4-hour 51Cr-release assay, and compared these results to the cytotoxicity of isogenic CD8+ T cells transduced with LacZ- or Hdac3-targeting gRNAs as negative and positive controls, respectively. Neither Ncor1- nor Ncor2-deficient CD8+ T cells showed significantly increased cytotoxicity relative to LacZ gRNA-transduced cells, whereas Hdac3-deficient CD8+

T cells showed strongly increased cytotoxicity over LacZ gRNA-transduced CD8+ T cells (Figure 8.9c).

Figure 8.9 | Inactivation of Ncor1 or Ncor2 does not increase CD8+ T cell cytotoxicity. a-c, OT-I+ CD8+ T cells derived from Cas9-transgenic mice were transduced with Thy1.1-marked lentiviral vectors bearing indicated gRNA sequences, and positively selected for Thy1.1 expression 3 days after transduction. a,b, Western blot analysis of NCOR1 (a) or NCOR2 (b) levels in CD8+ T cells 3 days post-infection (p.i.). M, mock-transduced; LZ, LacZ-targeting gRNA transduced (negative controls). Molecular weights in kDa are indicated. The red arrows indicate the gRNA sequences used for subsequent work. c, Thy1.1+ transductants were activated by co-culture with irradiated Ova peptide-pulsed BMDCs for 7 days, and in vitro cytotoxicity against B16-OVA targets was measured by 51Cr release assay after 4 hours of co-culture. c, Centre values, mean; error bars, s.d.; ** P < 0.01, **** P < 0.0001; comparison of Hdac3 gRNA- transduced cells with LacZ gRNA-transduced cells by 2-way ANOVA. No significant differences (P > 0.05) were observed when comparing Ncor1 gRNA- or Ncor2 gRNA-transduced cells to LacZ gRNA-transduced cells. Data are representative of two independent experiments.

These results indicate that the augmentation of CD8+ T cell cytotoxicity in the absence of HDAC3 activity was not due to the loss of biological activity of HDAC3/NCOR complexes. Our data suggested the very intriguing possibility that HDAC3 may regulate CD8+ T cell cytotoxicity by deacetylating acetylated histone residues, or even other protein substrates, in complex with previously unknown binding partners. Further biochemical work beyond the scope of this study will be required to identify the target substrates of

HDAC3 deacetylase activity as well as putative binding partners for HDAC3 in the context of CD8+ T cell activation.

Chapter 9

Summary and Discussion

Experimental design considerations for discovering epigenetic regulators of CD8+ T cell function and differentiation

We used two different approaches to discover novel epigenetic regulators of CD8+ T cell function – by using a CRISPR-Cas9 platform to screen epigenetic regulator genes for perturbations to the accumulation of CD8+ tumour-specific cells, and also by functionally screening small molecule inhibitors of epigenetic function for perturbations to CD8+ T cell effector function after activation. In retrospect, both of these experimental systems had advantages and limitations in overcoming the challenges associated with uncovering drivers of epigenetic regulation in CD8+ T cells.

The primary difficulty associated with finding and characterising epigenetic regulators in CD8+ T cells, or indeed any haematopoietic cell type, is the complex developmental and differentiation pathways that the cells go through prior to becoming mature functional cells. CD8+ T cells proceed through multiple developmental stages, first in the bone marrow and later in the thymus, before finally passing into general circulation as fully-fledged mature naïve CD8+ T cells. Naïve CD8+ T cells continually re-circulate until they encounter their cognate antigen presented in the context of inflammatory signals such as CD28 co- stimulation and cytokine-mediated signalling. Once activated, CD8+ T cells rapidly acquire their cytotoxic effector functions, and can further differentiate into a number of distinct cell fates, including terminal effectors, memory cells, and a dysfunctional state of exhaustion46,47,68. All of these developmental processes are largely irreversible, with advancement through successive stages being associated with a progressive reduction in the differentiation potential of the cell and a concomitant increase in commitment to a particular lineage.

Because the phenotypic changes marking these distinct functional states of CD8+ T cell development are epigenetically inherited, as are most developmental processes in biology14,15, any given epigenetic regulator is likely to play multiple roles at different time points within the series of stages of development of CD8+ T cells. For example, HDAC3 has been shown to play multiple critical roles in the thymic development of CD8+ T cells by downregulating the expression of the transcription factor RORγt at the

CD4+ CD8+ double-positive (DP) stage42, and subsequently in regulating the lineage commitment of DP thymocytes to the CD4+ and CD8+ lineages43. Therefore, any experimental approach aimed at uncovering and characterising mechanisms of epigenetic regulation must either account for effects arising from earlier stages of CD8+ T cell development or must be sufficiently well-targeted (e.g. by judicious choice of conditional gene-knockouts) to investigating epigenetic regulation at a specific stage of CD8+ T cell development.

Both of our experimental approaches were developed to uncover the involvement of epigenetic regulator genes at specific stages in the differentiation of naïve CD8+ T cells into antigen-experienced cells, while minimising the likelihood that any functional perturbations observed were due to the disruption of epigenetic regulation at an earlier stage of CD8+ T cell development. In both our genetic and pharmacologic screens, we used peripheral post-thymic CD8+ T cells as our starting material, thus circumventing any potential complications associated with perturbing epigenetic regulation during earlier stages of thymic development. Furthermore, whereas our CRISPR-Cas9 screening approach did not allow us to specifically distinguish when or how a particular epigenetic regulator affected the phenotype of tumour-infiltrating CD8+ T cells (we consciously traded a high degree of resolution of phenotypic changes for the ability to conduct high-throughput in vivo screening in a technically-challenging model), we were able to dissect this more carefully in our in vitro functional screening approach with small molecule inhibitors.

This difference between the nature of readouts between our two screening approaches highlights the second major consideration in designing experiments to investigate epigenetic regulators of CD8+ T cell phenotype – experimental readouts of phenotypic changes need to be clearly defined because these complex changes are likely to be the result of changes in multiple molecular pathways. Because CD8+ T cells acquire a diverse set of biological functions after activation, having the capability to identify which specific aspects of effector function are influenced by a given epigenetic regulator at the screening stage will significantly accelerate the process of discovery by facilitating the design of more specific follow-up experiments. This would in turn allow faster and more accurate elucidation of the underlying molecular mechanisms linking specific epigenetic modifications to the observed functional phenotype.

This principle was reflected in the different rates at which we were able to make progress with our two different screening approaches. Our in vivo CRISPR-Cas9 screen was intentionally designed to uncover potential epigenetic regulators of tumour-infiltrating CD8+ T cells in a preclinical model that simulated the tumour-specific restraints on the anti-tumour CD8+ T cell response observed in patients, with the hope that any candidate hits would also have a high likelihood of being relevant in clinical settings. However, because of the technical challenges required to establish and validate the screening platform, and because of the limited numbers of cells that could be recovered from an already heterogenous immune response, we were compelled to simplify the readout in order to be able to screen candidate genes in a high- throughput manner. As a result, while we are confident that CARM1 plays a role in suppressing the intratumoral accumulation of tumour-specific CD8+ T cells, at least within the B16 melanoma model, we have little information on which specific aspects of effector CD8+ T cell function it regulates. Indeed, the phenotype of increased accumulation of Carm1-deficient CD8+ T cells could have been a result of changes in multiple aspects of CD8+ T cell function such as proliferation, cell death, or trafficking patterns. With the increasing sophistication and cost-effectiveness of single-cell RNA-sequencing, future in vivo genetic

90 screens could potentially be coupled with single-cell sequencing and genomic barcoding platforms to match each genetic perturbation to its associated transcriptional profile.

In contrast, because the setup for our in vitro pharmacologic screen was simpler compared to the highly- engineered in vivo screening platform, this allowed us to perform in vitro screening experiments more quickly and also enabled us to recover higher numbers of CD8+ T cells for phenotypic analysis. We were therefore able to profile a wider array of CD8+ T cell functional markers and define a functional profile associated with each epigenetic perturbation. Together, these advantages allowed us to identify HDAC3 and to profile its regulation of CD8+ T cell effector function in the same amount of time it took us to identify and validate CARM1 as a robust hit from our in vivo CRISPR-Cas9 screen. We were therefore able to more quickly focus our investigations on the mechanism of HDAC3-mediated regulation of CD8+ effector phenotype and to explore possible interactions between HDAC3 and known transcriptional regulators of

CD8+ T cell function and differentiation.

Finally, designing experiments to characterise a phenotype across multiple parameters would also allow for more careful distinctions to be made when investigating loss-of-function phenotypes of CD8+ T cells lacking a given epigenetic regulator. Because epigenetic regulation plays a critical role in the development of a functional CD8+ T cell, any loss-of-function phenotype, such as a loss of memory recall capacity or of cytokine secretion, could in principle merely be the result of aberrant CD8+ T cell development.

Characterising complex phenotypes by profiling multiple errors would allow investigators to better resolve such proximal and distal effects of perturbing epigenetic regulation.

Overall, in consideration of the critical role of epigenetic regulation in the development of CD8+ T cells prior to their becoming mature naïve T cells in the periphery, we argue that a bottom-up approach to first

91 more fully understand the impact of a given epigenetic regulator on a wider set of CD8+ T cell effector functions within a less complex experimental setting is less prone to error than a top-down approach profiling multi-causal phenotypes in a complex experimental system.

A potential role for HDAC3 in the epigenetic regulation of CD8+ T cell fate following activation

Our studies showed that HDAC3 functions early during T cell activation to inhibit the development of a short-lived, strongly cytotoxic state in CD8+ T cells. Loss of HDAC3 activity during activation resulted in a gain-of-function phenotype with increased CD8+ T cell cytotoxicity, in contrast to the loss-of-function phenotypes described in recent studies of epigenetic regulators in CD8+ T cells20,22. Because increased cytotoxicity was only observed when CD8+ T cells lacked Hdac3 during initial antigen encounter, but not when CD8+ T cells lost HDAC3 activity after development of the peak effector response, we argue that

HDAC3 epigenetically regulates CD8+ T cell cytotoxicity specifically during the time window of T cell activation. This is further corroborated by our observations that increased proportions of Granzyme B+ and T-bet+ Hdac3-KO CD8+ T cells relative to Hdac3-WT cells could be observed by the first cell division after activation in our OT-I transfer and OVA immunisation model system. Additionally, because we did not observe any defects in cell survival or proliferation in the absence of TCR activation, we interpret the lack of persistence of Hdac3-deficient cytotoxic CD8+ T cells as a reflection of commitment to a short-lived effector phenotype after T cell activation, and not as a mere consequence of a loss of survival fitness.

Indeed, the role of HDAC3 in mediating acquisition of the cytotoxic effector phenotype in CD8+ T cells as well as the role of HDAC3 in controlling stability of this phenotype both fit classical hallmarks of epigenetic control over phenotype development and lineage commitment9,69.

While we found that HDAC3 activity negatively regulates the transcription of gene programmes associated with CD8+ T cell cytotoxic effector function, the relationship between the transcriptional signature of

HDAC3 and its molecular function in CD8+ T cells remains unclear. Although we observed significant increases in levels of activation-associated H3K27 acetylation at the promoters of a subset of genes whose transcription was upregulated in the absence of HDAC3 activity (e.g. Gzmb, Prdm1), this was not the case across all genes with increased transcription in HDAC3 activity-deficient CD8+ T cells. Therefore, we cannot definitively conclude that HDAC3-mediated deacetylation of H3K27-ac directly drives the transcriptional changes observed in CD8+ T cells lacking HDAC3 activity, although it may well be that HDAC3 is recruited to specific chromatin loci by an unidentified binding partner to deacetylate only a subset of acetylated histone lysine residues. Nonetheless, because inhibition of HDAC3 enzymatic activity with the highly- specific small molecule RGFP96639 was sufficient to observe an augmented cytotoxic phenotype in CD8+ T cells, HDAC3-mediated deacetylation plays a role in inhibiting CD8+ T cell cytotoxicity and terminal effector differentiation, possibly by targeting other acetyl-histone marks (e.g. H3K9-ac) or other transcription factors such as NF-κB, as reported in other biological contexts70,71.

Our results showing that Ncor1- or Ncor2-deficient CD8+ T cells did not display the same increase in cytotoxicity following in vitro activation suggests that the HDAC3 may not primarily regulate CD8+ T cell effector phenotype as part of a HDAC3/NCOR complex. Because HDAC3 lacks a DNA-binding domain and therefore requires a binding partner (e.g. the nuclear-corepressors, or the transcription factor Yin Yang 1

(YY1)72,73) for chromatin binding, it is likely that HDAC3 associates with other transcription factors or even other epigenetic regulators (e.g. the histone acetyltransferase p30074) to form a complex of regulator molecules mediating both transcriptional and epigenetic regulation of its target loci. Further biochemical analyses, possibly taking a co-immunoprecipitation and mass spectrometry approach, would be needed to elucidate the identity of binding partners in this putative complex.

While we have not identified the exact molecular pathways that transduce extracellular or cell-intrinsic input signals to trigger the activation and/or recruitment of HDAC3-containing chromatin regulatory complexes, our data suggest that signals resulting from TCR and/or co-stimulatory receptor activation are likely to be the proximal signalling inputs triggering HDAC3 activity. This is because the HDAC3-associated functional phenotype of CD8+ T cells is specific to TCR activation but not to cytokine-driven proliferation, because HDAC3 protein expression increases after CD8+ T cell activation, and because we were able to observe the phenotypic changes associated with loss of Hdac3 as early as one cell division following T cell activation. Furthermore, HDAC3 is critically required for the survival and maturation of double-positive thymocytes receiving signals for positive selection that are transduced from TCR engagement40-42.

Although all these pieces of circumstantial evidence point strongly towards a TCR activation-induced role of HDAC3 in CD8+ T cells, directly proving this interaction is likely going to be technically demanding, both due to the relatively short time window in which potential interactions might be glimpsed, as well as the difficulty of resolving a physical and/or biochemical interaction between a membrane-bound protein complex and a nuclear-localised protein.

Nonetheless, one interesting lead that has come out of earlier work investigating HDAC3 involvement during neuronal development is that HDAC3 is phosphorylated by the β isoform of the serine/threonine kinase glycogen synthase kinase-3 (GSK-3β). GSK-3 (both α and β isoforms) is well-established as a key intermediate in the canonical Wnt signalling pathway in many cell types75, including T cells. In resting naïve

T cells, GSK-3 is constitutively active, but becomes transiently inactivated and sequestered upon TCR and

CD28 engagement76,77. Because phosphorylation of HDAC3 at its Ser424 residue has been shown to markedly increase its deacetylase activity78, we suggest that the transient inactivation of GSK-3 by T cell activation via engagement of both TCR and co-stimulatory molecule pathways may provide a brief window in which the deacetylase activity of HDAC3 is temporarily suppressed to allow histone acetylation and

94 thus transcriptional activation of key activation and effector function-associated genes in CD8+ T cells. This model is consistent with our observation that histone 3 lysine acetylation levels generally increase very rapidly upon activation, but decline sharply after the peak of activation.

With regards to the interactions of HDAC3 with other known regulators of CD8+ T cell effector phenotype, we demonstrated that the increased cytotoxic phenotype of CD8+ T cells following T cell activation required the presence of Runx3, and that increased expression of Blimp-1 in Hdac3-KO CD8+ T cells contributed to their lack of persistence following activation. Runx3 is a critical transcription factor that is known to determine commitment to the CD8+ T cell lineage during thymopoiesis63,64, mediate activation and maturation of cytotoxic CD8+ T cell and NK lineages60,65,66, and to co-operate with the effector function-associated transcription factor T-bet (encoded by Tbx21) during activation for full development of CD8+ T cell cytotoxicity60. Runx3 has also recently been shown to be required for memory CTL development during acute viral infection, and was also found to be a pioneer transcription factor binding to cis-regulatory elements beginning early during CD8+ T cell activation66. Furthermore, HDAC3 was also recently reported to bind to a super-enhancer regulatory region of the Runx3 locus and was associated with suppression of Runx3 transcription for stabilising CD4+ lineage over CD8+ lineage commitment in developing thymocytes43. In view of all these data from our group and others, we are intrigued by the possibility that HDAC3 and Runx3 might both be recruited to cytotoxic gene loci and/or possibly interact to regulate gene expression early during CD8+ T cell activation.

We were somewhat surprised that T-bet was not required for the increased cytotoxic phenotype of

Hdac3-KO CD8+ T cells. However, given its known role in driving effector differentiation of CD8+ T cells1,8, and because we have consistently observed an increase in expression of T-bet in HDAC3 activity-deficient

CD8+ T cells, it is likely that T-bet plays a role in the development of the HDAC3 activity-associated

95 functional phenotype of CD8+ T cells. Specifically, T-bet may indirectly potentiate acquisition of cytotoxicity early during activation by driving effector differentiation of activated CD8+ T cells, but not be directly required for cytotoxicity in the same manner that Runx3 is required. Indeed, the same study that described a requirement for T-bet in regulating CD8+ T cell cytotoxicity also reported that it was early induction of T-bet during T cell activation that set the stage for acquisition of cytotoxicity60. We further hypothesise that the increased early induction of T-bet expression in Hdac3-deficient CD8+ T cells after activation may polarise the CD8+ T cell response towards increased effector activity, possibly at the expense of memory development. This would be a very interesting hypothesis to test, as it could potentially uncover a direct link between epigenetic modifications (mediated by HDAC3) and cell fate commitment (mediated by T-bet). However, the lack of persistence of Hdac3-KO CD8+ T cells have made experimental evaluation of this hypothesis a very difficult endeavour with our current toolkit.

The genetic interaction between Hdac3 and Prdm1 (Blimp-1) is also particularly interesting. At the molecular level, Blimp-1 expression is induced by IL-2 signalling (but not exclusively), whereupon it represses transcription of both Il2 and its activator Fos, functioning as a negative feedback loop on IL-2 activity58,79. Blimp-1 is also required for proper acquisition of cytotoxicity and trafficking of activated CD8+

T cells in the context of acute viral infection56,57,80, both of which contribute to the development of the terminal effector phenotype of CD8+ T cells. Moreover, in response to cytokine signalling, Blimp-1 induces epigenetic changes at the genetic locus coding for the high-affinity IL-2 receptor α-subunit (CD25) in activated CD8+ T cells via its association with the histone methyltransferase G9a (Ehmt2) or with the histone deacetylase Hdac280. Because Blimp-1 regulates such diverse aspects of CD8+ T cell effector function, we speculate that HDAC3 may have a context-dependent role in epigenetically regulating commitment to a particular functional phenotype after CD8+ T cell activation.

Overall, we propose a model in which HDAC3 epigenetically regulates a network of genes including Runx3 and Blimp-1 in CD8+ T cells to suppress cytotoxicity and to promote the persistence of activated T cells

(Figure 9.1). We theorise that the deacetylation of target protein substrates by HDAC3 activity beginning early during CD8+ T cell activation thus directly regulates the nature of the CD8+ T cell response to antigen, and further suggest that HDAC3-mediated protein deacetylation may potentiate the development of bona fide CD8+ T cell memory by de-committing CD8+ T cells from a terminally-differentiated effector cell fate driven by the transcription factors T-bet and Blimp-1.

Figure 9.1 | Proposed model of HDAC3-mediated regulation of CD8+ T cell immunity. HDAC3 may function as an epigenetic negative feedback loop by antagonising development and acquisition of a terminally-differentiated cytotoxic effector phenotype, thus potentiating formation of durable memory from long-lived antigen-experienced cells.

Potential therapeutic applications of modulating HDAC3 activity in CD8+ T cells

To understand the physiological impact of loss of HDAC3 in CD8+ T cells on the overall immune response to antigen challenge, we investigated the impact of genetic or pharmacological ablation of HDAC3 activity on the CD8+ T cells in a variety of experimental models including both acute and chronic antigen challenge settings. We did not observe additional benefits to the CD8+ T cell immune response in the context of

97 acute antigen challenge, instead finding that the reduced persistence of Hdac3-KO CD8+ T cells could potentially compromise the efficacy of the immune response to secondary encounter of the same antigen.

However, loss of HDAC3 activity did provide a transient benefit to CD8+ T cell-mediated immunity in experimental models of chronic viral infection and tumour challenge.

These preliminary data suggest that modulation of the CD8+ T cell response using small molecule inhibitors targeting HDAC3 could in principle provide an additional epigenetic dimension of control in therapeutic settings, either to temporarily boost cytotoxic CD8+ T cell immunity to achieve better clinical outcomes or to limit the persistence of an effector CD8+ response in order to reduce immune-related adverse effects.

These applications may be relevant in the settings of therapeutic vaccines or in cell transfer-based therapies, respectively. While further testing and work will be certainly needed prior to any form of translational application of HDAC3-mediated immunoregulation, particularly with regards to fully elucidating the molecular mechanism of HDAC3-mediated regulation of CD8+ effector function, HDAC3 may prove to be an interesting and useful candidate target for future immunotherapy against chronic inflammatory conditions.

Appendix

Supplementary Table 1 | List of genes in pilot CRISPR-Cas9 screen of CD8+ TILs.

Epigenetic Regulator Pool 1 Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Nono 53610 Suz12 52615 Aebp2 11569 Kdm3b 277250 Nap1L1 53605 Tdg 21665 Baz1A 217578 Kdm4a 230674 Paxip1 55982 Trim28 21849 Ctnnb1 12387 L3Mbtl3 237339 Phf1 21652 Uhrf1 18140 Kdm2b 30841 Lbr 98386 Phf13 230936 Wdr5 140858 Gatad1 67210 Limd1 29806 Phf19 74016 Whsc1 107823 Hr 15460 Max 17187 Phf21A 192285 Paf1 54624 Ints12 71793 Mina 67014 Phf5A 68479 Atad2 70472 Uhrf2 109113 Kmt2a 214162 Phf6 70998 Atf4 11911 Kdm1a 99982 Kmt2e 69188 Prdm1 12142 Blm 12144 Asxl2 75302 Mta1 116870 Prmt1 15469 Bmi1 12151 Atrx 22589 Mta3 116871 Prmt5 27374 Brip1 237911 Baz1B 22385 Kat7 217127 Prmt7 214572 Cbx1 12412 Baz2A 116848 Kat6a 244349 Ptbp1 19205 Cbx3 12417 Bptf 207165 Ncor1 20185 Rad54L 19366 Cbx5 12419 Brd2 14312 Nipbl 71175 Rbbp4 19646 Chaf1A 27221 0610010K14Rik 104457 Nsd1 18193 Rbbp7 245688 Chaf1B 110749 Cbx7 52609 Pbrm1 66923 Rif1 51869 Dek 110052 Chd1 12648 Phf12 268448 Setd6 66083 Dnmt1 13433 Chd2 244059 Phf20L1 239510 Setd8 67956 Dpy30 66310 Chd3 216848 Phf3 213109 Smarca5 93762 Eed 13626 Chd4 107932 Rnf20 109331 Smarcad1 13990 Ehmt2 110147 Chd7 320790 Rnf40 233900 Smarcc1 20588 Eif4B 75705 Chd8 67772 Setd2 235626 Smarcc2 68094 Ezh2 14056 Ctbp1 13016 Setd5 72895 Smarce1 57376 Gadd45B 17873 Ctcf 13018 Sin3A 20466 Smchd1 74355 H2Afz 51788 Dpf2 19708 Sirt2 64383 Smndc1 76479 Hat1 107435 Ep300 328572 Sirt7 209011 Smyd2 226830 Hcfc1 15161 Supt20 56790 Smarca4 20586 Smyd5 232187 Hdac1 433759 Kdm2a 225876 Sp110 109032 Snd1 56463 Hdac2 15182 Hdac7 56233 Srcap 100043597 Sp100 20684 Hdac3 15183 Ikbkb 16150 Stat1 20846 Sp140 434484 Hells 15201 Kdm5a 214899 Tcf20 21411 Ssrp1 20833 Jun 16476 Jarid2 16468 Ubtf 21429 Supt16 114741 Kdm6B 216850 Kdm7a 338523 Tox 252838 Supt3 109115 Mbd2 17191 Kdm3a 104263 Hmgb2 97165 Suv39H1 20937 Mbd3 17192

100

Supplementary Table 1 (continued)

Epigenetic Regulator Pool 2 Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Wdr82 77305 Kdm4c 76804 Kdm4b 193796 Alkbh2 231642 Whsc1L1 234135 Kat2A 14534 Jmjd6 107817 Bcor 71458 Yy1 22632 Lcor 212391 Jmjd8 72106 Brd3 67382 Satb1 20230 Mga 29808 Kat5 81601 Brpf1 78783 Mllt3 70122 Kmt2c 231051 L3Mbtl2 214669 Cdyl2 75796 Mllt6 246198 Pagr1a 67278 Lrwd1 71735 Chd6 71389 Mtf2 17765 Recql 19691 Men1 17283 Chd9 109151 Ncoa2 17978 Rsf1 233532 Mier1 71148 Clock 12753 Ncoa3 17979 Shprh 268281 Morf4L1 21761 Cxxc1 74322 Pcmt1 18537 Smarca2 67155 Rbbp5 213464 Egln2 112406 Jade 76901 Suv420H1 225888 Rnf2 19821 Ehmt1 77683 Phip 83946 Tada1 27878 Setd1A 233904 Ercc6 319955 Prmt3 71974 Mllt10 17354 Setd1B 208043 G2E3 217558 Kmt2b 75410 Msl3 17692 Setd3 52690 Gtf3C4 269252 Zgpat 229007 Kat8 67773 Sin3B 20467 Hdac10 170787 Zmynd11 66505 Phf2 18676 Sirt1 93759 Hspbap1 66667 Arid4A 238247 Prdm4 72843 Smarcd1 83797 Ing1 26356 Arid5B 71371 Adnp 11538 Taf1 270627 Ing4 28019 Ash1L 192195 Alkbh3 69113 Msh6 17688 Jade3 382207 Ash2L 23808 Arid4B 94246 Mta2 23942 Kat2B 18519 Atad2B 320817 Brd4 57261 Ncor2 20602 Kdm1B 218214 Bap1 104416 Brd8 78656 Padi2 18600 Mbd1 17190 Brd1 223770 Brd9 105246 Pcgf6 71041 Mbd4 17193 Brd7 26992 Brwd1 93871 Phf10 72057 Mbtd1 103537 Chd1L 68058 Emsy 233545 Phf14 75725 Mecp2 17257 Crebbp 12914 Carm1 59035 Jade1 269424 Mlh1 17350 Dnmt3A 13435 Cbx4 12418 Phf20 228829 Kmt2d 381022 Dot1L 208266 Egln1 112405 Pogz 229584 Psip1 101739 Elp3 74195 Epc2 227867 Prdm2 110593 Pygo2 68911 Ep400 75560 Hdac4 208727 Trim33 94093 Rai1 19377 Epc1 13831 Hira 15260 Setdb1 84505 Rnf8 58230 Ezh1 14055 Ing3 71777 Pcgf5 76073 Sertad2 58172 Ing5 66262 Iws1 73473 Tet3 194388 Sirt6 50721 Ino80 68142 Kdm5c 20591 Tp53Bp1 27223 Hmgn5 50887 Jade2 76901 Kdm5d 20592 Zmynd8 228880 Mxd1 17119 Kdm5b 75605 Jmjd1C 108829

101

Supplementary Table 1 (continued)

Epigenetic Regulator Pool 3 Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Smyd3 69726 Prdm8 77630 Jmjd4 194952 Sfmbt2 353282 Supt7L 72195 Recql4 79456 Kdm8 77035 Sirt3 64384 Suv420H2 232811 Scml2 107815 L3Mbtl1 241764 Sirt4 75387 Taf3 209361 Smarcd3 66993 L3Mbtl4 320858 Sirt5 68346 Tbp 21374 Smyd1 12180 Lrif1 321000 Smarca1 93761 Ncoa1 17977 Suv39H2 64707 Mbd5 109241 Smyd4 319822 Pcgf1 69837 Vhl 22346 Meis2 17536 Tdrd3 219249 Phc3 241915 2410016O06Rik 71952 Mettl8 228019 Tet1 52463 Phf8 320595 9930021J03Rik 240613 Mgmt 17314 Tet2 214133 Prdm15 114604 Actl6B 83766 Recql5 170472 Trim24 21848 Setd7 73251 Asxl3 211961 Ring1 19763 Trim66 330627 Pcgf3 69587 Baz2B 407823 Rybp 56353 Uty 22290 Phc1 13619 Bcl9 77578 Tada3 101206 Yaf2 67057 Prmt6 99890 Bcorl1 320376 Phc2 54383 Cbx6 494448 Usp22 216825 Brdt 114642 Prdm10 382066 Hmgxb4 70823 Kdm6a 22289 Brpf3 268936 Prdm11 100042784 Jmjd7 433466 Wdr11 207425 Brwd3 382236 Setd4 224440 Mphosph8 75339 Wiz 22404 Tyw5 68736 Setdb2 239122 Prdm12 381359 Aire 11634 Cbx2 12416 Smarcal1 54380 Hist1h2ac 319164 Bard1 12021 Cbx8 30951 Mst1 15235 Hist1h2bg 319181 Cdh1 12550 Sgf29 75565 Mum1 68114 Hist1h4i 319158 Cecr2 330409 Cdyl 12593 Kat6b 54169 Hist1h2bc 68024 Ctbp2 13017 Chd5 269610 Nap1L2 17954 Hist1h1c 50708 Dpf3 70127 Coil 12812 Pkd2 18764 Sap30l 50724 Fancm 104806 Cpa4 71791 Phf7 71838 Cpeb2 231207 Gadd45A 13197 Dmap1 66233 Prdm13 230025 Utp20 70683 Hdac11 232232 Dnmt3B 13436 Prdm14 383491 Crebbp 12914 Hdac9 79221 Dnmt3L 54427 Prdm6 225518 Nap1l4 17955 Rad54B 623474 Dpf1 29861 Prdm9 213389 Nop2 110109 Morc4 75746 Elp4 77766 Prmt2 15468 Elk4 13714 Nap1L3 54561 Ercc6L2 76251 Prmt8 381813 Prmt3 71974 Orc1 18392 Hdac5 15184 Sap30 60406 Pa2g4 18813 Padi4 18602 Hdac6 15185 Scmh1 29871 Snrnp70 20637 Pcgf2 22658 Hdac8 70315 Setmar 74729 Ikzf2 22779 Prdm16 70673 Ing2 69260 Sfmbt1 54650 Hmgn3 94353 Prdm5 70779 Kdm4d 244694

102

Supplementary Table 1 (continued)

Positive Controls Gene symbol Gene ID Ppp2r2d 52432 Smad2 17126 Dgkz 104418 Egr2 13654 Cblb 208650 Pdcd1 18566 Ctla4 12477

103

Supplementary Table 2 | List of genes in validation CRISPR-Cas9 screen of CD8+ TILs.

Preliminary MAGeCK Gene Gene ID Screen Rank Score Symbol 1 1.855 Carm1 59035 2 1.490 Setd5 72895 Positive 1.261 Cblb 208650 Control 3 1.161 Arid 4b 94246 4 1.153 Satb1 20230 Positive 1.074 Pdcd1 18566 Control 5 1.024 Kdm5d 20592 6 0.765 Ints12 71793 7 0.749 Cpa4 71791 8 0.732 Ncoa1 17977 9 0.677 Pbrm1 66923 10 0.568 Hist1h1c 50708 11 0.567 Smyd4 319822 12 0.537 Cbx8 30951 13 0.529 Ehmt2 110147 14 0.515 Prdm5 70779 15 0.513 Hdac4 208727 16 0.509 Brwd3 382236 17 0.506 L3Mbtl1 241764 18 0.482 Actl6B 83766 19 0.435 Arid5B 71371 20 0.419 Uhrf2 109113 23 0.361 Sirt6 50721 24 0.344 Elk4 13714 25 0.321 Phf2 18676 27 0.283 Kdm6B 216850 30 0.183 Hdac9 79221 31 0.179 Jarid2 16468 33 0.155 Sirt1 93759 36 0.134 Mgmt 17314 40 0.097 L3Mbtl3 237339

104

T cell effector

Topoisomerase I

NAT10

HDAC3

EZH2 histoneEZH2 methyltransferase

DOT1L methyltransferase

HDAC6

HDAC1

HIF-PH

COX-1

SMARCA2, SMARCA4, PB1 SMARCA4, SMARCA2,

EZH2 histoneEZH2 methyltransferase

CBP and p300 HATs

SIRT1 and SIRT2

EZH2 histoneEZH2 methyltransferase

L3MBTL3

methyltransferases

PRMT proteinPRMT arginine

WDR5

(JMJD2A, 2C, 2E) (JMJD2A,

a-ketoglutarate enzymes dependent

HDACs

methyltransferases

SET domain-containingSET

EZH2 histoneEZH2 methyltransferase

HDACs

Known Target(s)/Mechanism

hecin

Hydroxycamptot

(S)-10-

Remodelin

RGFP966

EPZ-6438

EPZ-5676

(Rocilinostat)

ACY-1215

RSC-133

CAY10398

Trichostatin A

DMOG

2,4-DPD

Resveratrol

trans-

PFI-3

UNC 1999UNC

I-CBP112

SIRT1/2 inhibitor

GSK 343

UNC1215

AMI-1

WDR5-0103

N-oxalylglycine

Pyroxamide

Sinefungin

Deazaneplanocin

Chidamide

Drug

G9a histone methyltransferase

Multiple metabolic pathways

DNA methyltransferasesDNA

DNA methyltranferasesDNA

HDACs

HDAC1

HDACs

Class I HDACs

MRN (Mre11-Rad50-Nbs1)MRN complex

SIRT1 and SIRT2

HDACs

HDAC1 andHDAC1 HDAC6

HDAC1 andHDAC1 HDAC3

SIRT2

methyltransferase

SUV39H1 histone

HDAC6

SIRT2

p300 HAT

G9a histone methyltransferase

HDACs

SIRT1, SIRT2, SIRT3

SIRT1 and SIRT2

Class I HDACs, GSK3

Gcn5 HAT

Known Target(s)/Mechanism

UNC0224

steine

adenosylhomocy

S-

5-azacytidine

2',3',5'-triacetyl-

RG-108

HNHA

MS-275

(S)-HDAC-42

KD 5170

106

diphenylamide

Pimelic

Mirin

Salermide

Oxamflatin

M 344M

CBHA

Splitomicin

Chaetocin

CAY10603

AGK2

Anacardic acid

BIX01294

Sodium butyrate

Tenovin-6

Tenovin-1

Valproic acid

CPTH2

Drug

JMJD3

(KDM2A, PHF8,(KDM2A, KDM7A)

JMJC histoneJMJC demethylases

Prolyl hydroxylase stimulator

MGMT MGMT

Gadd45a

Ribonucleotide reductases and

Menin

2-OG oxygenases, JMJD2A

HDACs

PARP1

Control for (+)-JQ1

BRD4

DNA methyltransferasesDNA

BRD2, BRD4

Calmodulin

HD2, HD-1B, HD-1A

NO, EGFR, p300/CBP EGFR, NO, HATs

DNA methyltransferasesDNA

pCAF HAT pCAF

Known Target(s)/Mechanism

GSK-J5

GSK-J4

GSK-J2

GSK-J1

Daminozide

ketoglutarate

Octyl-alpha-

Lemoguatrib

Gemcitabine

MI-nc

MI-2

IOX1

phenylbutyrate

Sodium 4-

acid

1-Naphthoic

BSI-201

(-)-JQ1

(+)-JQ1

Decitabine

5-Azacytidine

PFI-1

Nicotinamide

Suramin

ITF2357

chloride

Delphinidin

Zebularine

CAY10669 Drug

List of small molecules in functional pharmacological screen of epigenetic regulators of CD8

inhibitor

ary Table 3

Rho pathway activation of SRF

Quinone reductase-1 agonist

Aurora kinase A

G9a histone methyltransferase

SIRT1

PAD4, PAD1, PAD3 PAD1, PAD4,

PAD4

SAH hydrolaseSAH

G9a histone methyltransferase

HDAC3

HDAC1 andHDAC1 HDAC3

HDACs

H3R17 methylation

p300 HAT

SIRT1 and SIRT2

HDAC1 andHDAC1 HDAC6

HDAC8

HDACs

PARP PARP

LSD1 LSD1 inhibitor

Class I, II HDACs

SIRT1 inhibitor

SIRT1 activator, TNF-

kinases, NF-

Tyrosine and serine/threonine

HDACs Known Target(s)/Mechanism

Supplement phenotype.

CCG-100602

Isoliquirtigenin

pyrazole

Phthalazinone

UNC0638

JGB1741

salt)

(trifluoroacetate

F-amidine

Cl-amidine

(-)-Neplanocin A

UNC0321

Apicidin

ic acid

Suberohydroxam

Scriptaid

Ellagic acid

C646

Sirtinol

4-iodo-SAHA

PCI 34051

SB SB 939

benzamide

3-amino

2-PCPA

SAHA

Ex-527

CAY10591

Piceatannol

CAY10433 Drug

105

References

1 Joshi, N. S. et al. Inflammation directs memory precursor and short-lived effector CD8(+) T cell fates via the graded expression of T-bet transcription factor. Immunity 27, 281-295, doi:10.1016/j.immuni.2007.07.010 (2007).

2 Sarkar, S. et al. Functional and genomic profiling of effector CD8 T cell subsets with distinct memory fates. The Journal of experimental medicine 205, 625, doi:10.1084/jem.20071641 (2008).

3 Schluns, K. S., Kieper, W. C., Jameson, S. C. & Lefrançois, L. Interleukin-7 mediates the homeostasis of naïve and memory CD8 T cells in vivo. Nature immunology 1, 426, doi:10.1038/80868 (2000).

4 Kaech, S. M. et al. Selective expression of the interleukin 7 receptor identifies effector CD8 T cells that give rise to long-lived memory cells. Nature immunology 4, 1191-1198, doi:10.1038/ni1009 (2003).

5 Kaech, S. M. & Cui, W. Transcriptional control of effector and memory CD8+ T cell differentiation. Nature reviews. Immunology 12, 749-761, doi:10.1038/nri3307 (2012).

6 Chang, J. T., Wherry, E. J. & Goldrath, A. W. Molecular regulation of effector and memory T cell differentiation. Nature immunology 15, 1104, doi:10.1038/ni.3031 (2014).

7 Best, J. A. et al. Transcriptional insights into the CD8(+) T cell response to infection and memory T cell formation. Nature immunology 14, 404-412, doi:10.1038/ni.2536 (2013).

8 Lazarevic, V., Glimcher, L. H. & Lord, G. M. T-bet: a bridge between innate and adaptive immunity. Nature Reviews Immunology 13, 777, doi:10.1038/nri3536 (2013).

9 Goldberg, A. D., Allis, C. D. & Bernstein, E. Epigenetics: a landscape takes shape. Cell 128, 635-638, doi:10.1016/j.cell.2007.02.006 (2007).

10 Patel, D. J. A Structural Perspective on Readout of Epigenetic Histone and DNA Methylation Marks. Cold Spring Harbor perspectives in biology. 8, a018754, doi:10.1101/cshperspect.a018754 (2016).

11 Allshire, R. C. & Madhani, H. D. Ten principles of heterochromatin formation and function. Nature reviews. Molecular cell biology 19, 229-244, doi:10.1038/nrm.2017.119 (2018).

106

12 Zhang, Y., Li, E. & Zhang, Y. DNA Methylation in Mammals. Cold Spring Harbor perspectives in biology. 6, a019133-a019133, doi:10.1101/cshperspect.a019133 (2014).

13 Paro, R., Grossniklaus, U. & Paro, R. Transcriptional Silencing by Polycomb-Group Proteins. Cold Spring Harbor perspectives in biology. 6, a019331-a019331, doi:10.1101/cshperspect.a019331 (2014).

14 Surani, M. A., Reik, W. & Surani, M. A. Germline and Pluripotent Stem Cells. Cold Spring Harbor perspectives in biology. 7, a019422, doi:10.1101/cshperspect.a019422 (2015).

15 Cedar, H. & Bergman, Y. Epigenetics of haematopoietic cell development. Nature Reviews Immunology 11, 478, doi:10.1038/nri2991 (2011).

16 Morikawa, H. et al. Differential roles of epigenetic changes and Foxp3 expression in regulatory T cell-specific transcriptional regulation. Proceedings of the National Academy of Sciences 111, 5289, doi:10.1073/pnas.1312717110 (2014).

17 Henning, A. N., Roychoudhuri, R. & Restifo, N. P. Epigenetic control of CD8(+) T cell differentiation. Nature reviews. Immunology 18, 340-356, doi:10.1038/nri.2017.146 (2018).

18 Alice E. Denton, B. E. R., Peter C. Doherty, Sudha Rao, and Stephen J. Turner. Differentiation- dependent functional and epigenetic landscapes for cytokine genes in virus-specific CD8+ T cells. Proceedings of the National Academy of Sciences of the United States of America 108, 15306- 15311 (2011).

19 Russ, B. E. et al. Distinct epigenetic signatures delineate transcriptional programs during virus- specific CD8(+) T cell differentiation. Immunity 41, 853-865, doi:10.1016/j.immuni.2014.11.001 (2014).

20 Gray, S. M., Amezquita, R. A., Guan, T., Kleinstein, S. H. & Kaech, S. M. Polycomb Repressive Complex 2-Mediated Chromatin Repression Guides Effector CD8(+) T Cell Terminal Differentiation and Loss of Multipotency. Immunity 46, 596-608, doi:10.1016/j.immuni.2017.03.012 (2017).

21 Philip, M. et al. Chromatin states define tumour-specific T cell dysfunction and reprogramming. Nature 545, 452-456, doi:10.1038/nature22367 (2017).

22 Pace, L. et al. The epigenetic control of stemness in CD8+ T cell fate commitment. Science 359, 177, doi:10.1126/science.aah6499 (2018).

107

23 Zhang, T., Cooper, S. & Brockdorff, N. The interplay of histone modifications - writers that read. EMBO reports 16, 1467-1481, doi:10.15252/embr.201540945 (2015).

24 Knutson, S. K. et al. Liver-specific deletion of histone deacetylase 3 disrupts metabolic transcriptional networks. The EMBO journal. 27, 1017-1028, doi:10.1038/emboj.2008.51 (2008).

25 Diebold, S. S., Cotten, M., Koch, N. & Zenke, M. MHC class II presentation of endogenously expressed antigens by transfected dendritic cells. Gene Therapy 8, 487, doi:10.1038/sj.gt.3301433 (2001).

26 Sanjana, N. E., Shalem, O. & Zhang, F. Improved vectors and genome-wide libraries for CRISPR screening. Nature methods 11, 783-784, doi:10.1038/nmeth.3047 (2014).

27 Zhou, P. et al. In vivo discovery of immunotherapy targets in the tumour microenvironment. Nature 506, 52, doi:10.1038/nature12988 (2014).

28 Li, W. et al. MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome biology 15, 554-554, doi:10.1186/s13059-014-0554-4 (2014).

29 Madaan, A., Verma, R., Singh, A. T., Jain, S. K. & Jaggi, M. A stepwise procedure for isolation of murine bone marrow and generation of dendritic cells. Journal of Biological Methods; Vol 1, No 1 (2014).

30 Gerlach, C. et al. The Chemokine Receptor CX3CR1 Defines Three Antigen-Experienced CD8 T Cell Subsets with Distinct Roles in Immune Surveillance and Homeostasis. Immunity 45, 1270-1284, doi:10.1016/j.immuni.2016.10.018 (2016).

31 Welsh, R. M. & Seedhom, M. O. Lymphocytic choriomeningitis virus (LCMV): propagation, quantitation, and storage. Current protocols in microbiology Chapter 15, Unit-15A.11, doi:10.1002/9780471729259.mc15a01s8 (2008).

32 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2. Genome Biology 15, 550, doi:10.1186/s13059-014-0550-8 (2014).

33 Tian, B., Yang, J. & Brasier, A. R. Two-step cross-linking for analysis of protein-chromatin interactions. Methods in molecular biology (Clifton, N.J.) 809, 105-120, doi:10.1007/978-1-61779- 376-9_7 (2012).

108

34 Wang, S. et al. Modeling cis-regulation with a compendium of genome-wide histone H3K27ac profiles. Genome research 26, 1417-1429, doi:10.1101/gr.201574.115 (2016).

35 Shishkova, E. et al. Global mapping of CARM1 substrates defines enzyme specificity and substrate recognition. Nature communications 8, 15571, doi:10.1038/ncomms15571 (2017).

36 Kim, J. et al. Loss of CARM1 results in hypomethylation of thymocyte cyclic AMP-regulated phosphoprotein and deregulated early T cell development. J Biol Chem 279, 25339-25344, doi:10.1074/jbc.M402544200 (2004).

37 Li, J. et al. Coactivator-associated arginine methyltransferase 1 regulates fetal hematopoiesis and thymocyte development. Journal of immunology (Baltimore, Md. : 1950) 190, 597-604, doi:10.4049/jimmunol.1102513 (2013).

38 Bao, J. et al. The arginine methyltransferase CARM1 represses p300•ACT•CREMτ activity and is required for spermiogenesis. Nucleic acids research 46, 4327-4343, doi:10.1093/nar/gky240 (2018).

39 Xu, C. et al. Chemical Probes Identify a Role for Histone Deacetylase 3 in Friedreich's Ataxia Gene Silencing. Chemistry & Biology 16, 980-989, doi:https://doi.org/10.1016/j.chembiol.2009.07.010 (2009).

40 Hsu, F. C. et al. Histone Deacetylase 3 Is Required for T Cell Maturation. Journal of immunology 195, 1578-1590, doi:10.4049/jimmunol.1500435 (2015).

41 Stengel, K. R. et al. Histone Deacetylase 3 Is Required for Efficient T Cell Development. Molecular and cellular biology 35, 3854-3865, doi:10.1128/MCB.00706-15 (2015).

42 Philips, R. L. et al. HDAC3 Is Required for the Downregulation of RORgammat during Thymocyte Positive Selection. Journal of immunology 197, 541-554, doi:10.4049/jimmunol.1502529 (2016).

43 Lee, J.-H. et al. HDAC3 restrains CD8-lineage genes to maintain a bi-potential state in CD4+CD8+ thymocytes for CD4-lineage commitment. ELife. 8, doi:10.7554/eLife.43821 (2019).

44 Hammad Hassan, S. S., Mari Tenno, Aglaja Kopf, Nicole Boucheron, Andrea C. Carpenter, Takeshi Egawa, Ichiro Taniuchi, and Wilfried Ellmeier. Cd8 enhancer E8I and Runx factors regulate CD8α expression in activated CD8+ T cells. Proceedings of the National Academy of Sciences of the United States of America 108, 18330-18335 (2011).

109

45 Bhaskara, S. et al. Deletion of Histone Deacetylase 3 Reveals Critical Roles in S Phase Progression and DNA Damage Control. Molecular Cell 30, 61-72, doi:https://doi.org/10.1016/j.molcel.2008.02.030 (2008).

46 Wherry, E. J., Blattman, J. N., Murali-Krishna, K., van der Most, R. & Ahmed, R. Viral persistence alters CD8 T-cell immunodominance and tissue distribution and results in distinct stages of functional impairment. Journal of virology 77, 4911-4927, doi:10.1128/JVI.77.8.4911-4927.2003 (2003).

47 Zajac, A. J. et al. Viral immune evasion due to persistence of activated T cells without effector function. The Journal of experimental medicine 188, 2205-2213 (1998).

48 Efremova, M. et al. Targeting immune checkpoints potentiates immunoediting and changes the dynamics of tumor evolution. Nature communications 9, 32-32, doi:10.1038/s41467-017-02424- 0 (2018).

49 Guenther, M. G., Barak, O. & Lazar, M. A. The SMRT and N-CoR Corepressors Are Activating Cofactors for Histone Deacetylase 3. Molecular and cellular biology 21, 6091, doi:10.1128/MCB.21.18.6091-6101.2001 (2001).

50 Wen, Y.-D. et al. The histone deacetylase-3 complex contains nuclear receptor corepressors. Proceedings of the National Academy of Sciences 97, 7202, doi:10.1073/pnas.97.13.7202 (2000).

51 Heng, T. S. P. et al. The Immunological Genome Project: networks of gene expression in immune cells. Nature immunology 9, 1091, doi:10.1038/ni1008-1091 (2008).

52 Mackay, L. K. et al. Hobit and Blimp1 instruct a universal transcriptional program of tissue residency in lymphocytes. Science 352, 459, doi:10.1126/science.aad2035 (2016).

53 Seto, E. & Yoshida, M. Erasers of histone acetylation: the histone deacetylase enzymes. Cold Spring Harbor perspectives in biology 6, a018713-a018713, doi:10.1101/cshperspect.a018713 (2014).

54 Vermeulen, M. et al. In Vitro Targeting Reveals Intrinsic Histone Tail Specificity of the Sin3/Histone Deacetylase and N-CoR/SMRT Corepressor Complexes. Molecular and cellular biology 24, 2364- 2372, doi:10.1128/mcb.24.6.2364-2372.2004 (2004).

55 Wang, Z. et al. Combinatorial patterns of histone acetylations and methylations in the human genome. Nature genetics 40, 897-903, doi:10.1038/ng.154 (2008).

110

56 Rutishauser, R. L. et al. Transcriptional repressor Blimp-1 promotes CD8(+) T cell terminal differentiation and represses the acquisition of central memory T cell properties. Immunity 31, 296-308, doi:10.1016/j.immuni.2009.05.014 (2009).

57 Kallies, A., Xin, A., Belz, G. T. & Nutt, S. L. Blimp-1 transcription factor is required for the differentiation of effector CD8(+) T cells and memory responses. Immunity 31, 283-295, doi:10.1016/j.immuni.2009.06.021 (2009).

58 Martins, G. A., Cimmino, L., Liao, J., Magnusdottir, E. & Calame, K. Blimp-1 directly represses Il2 and the Il2 activator Fos, attenuating T cell proliferation and survival. The Journal of experimental medicine 205, 1959-1965, doi:10.1084/jem.20080526 (2008).

59 Xin, A. et al. A molecular threshold for effector CD8(+) T cell differentiation controlled by transcription factors Blimp-1 and T-bet. Nature immunology 17, 422-432, doi:10.1038/ni.3410 (2016).

60 Cruz-Guilloty, F. et al. Runx3 and T-box proteins cooperate to establish the transcriptional program of effector CTLs. The Journal of experimental medicine 206, 51-59, doi:10.1084/jem.20081242 (2009).

61 Lewis, M. D., Miller, S. A., Miazgowicz, M. M., Beima, K. M. & Weinmann, A. S. T-bet's ability to regulate individual target genes requires the conserved T-box domain to recruit histone methyltransferase activity and a separate family member-specific transactivation domain. Molecular and cellular biology 27, 8510-8521, doi:10.1128/MCB.01615-07 (2007).

62 Takemoto, N., Intlekofer, A. M., Northrup, J. T., Wherry, E. J. & Reiner, S. L. Cutting Edge: IL-12 Inversely Regulates T-bet and Eomesodermin Expression during Pathogen-Induced CD8+ T Cell Differentiation. The Journal of Immunology 177, 7515-7519, doi:10.4049/jimmunol.177.11.7515 (2006).

63 Taniuchi, I. et al. Differential Requirements for Runx Proteins in CD4 Repression and Epigenetic Silencing during T Lymphocyte Development. Cell 111, 621-633, doi:https://doi.org/10.1016/S0092-8674(02)01111-X (2002).

64 Woolf, E. et al. Runx3 and Runx1 are required for CD8 T cell development during thymopoiesis. Proceedings of the National Academy of Sciences of the United States of America. 100, 7731-7736, doi:10.1073/pnas.1232420100 (2003).

65 Lotem, J. et al. Runx3-mediated Transcriptional Program in Cytotoxic Lymphocytes. PLOS ONE 8, e80467, doi:10.1371/journal.pone.0080467 (2013).

111

66 Wang, D. et al. The Transcription Factor Runx3 Establishes Chromatin Accessibility of cis- Regulatory Landscapes that Drive Memory Cytotoxic T Lymphocyte Formation. Immunity 48, 659- 674.e656, doi:https://doi.org/10.1016/j.immuni.2018.03.028 (2018).

67 Milner, J. J. et al. Runx3 programs CD8+ T cell residency in non-lymphoid tissues and tumours. Nature 552, 253, doi:10.1038/nature24993 (2017).

68 Doering, T. A. et al. Network analysis reveals centrally connected genes and pathways involved in CD8+ T cell exhaustion versus memory. Immunity 37, 1130-1144, doi:10.1016/j.immuni.2012.08.021 (2012).

69 Deans, C. & Maggert, K. A. What do you mean, "epigenetic"? Genetics 199, 887-896, doi:10.1534/genetics.114.173492 (2015).

70 Chen, L.-f., Fischle, W., Verdin, E. & Greene, W. C. Duration of Nuclear NF-κB Action Regulated by Reversible Acetylation. Science 293, 1653, doi:10.1126/science.1062374 (2001).

71 Xiao, L. et al. Histone Deacetylase 3 Interacts with and Deacetylates Myocyte Enhancer Factor 2. Molecular and cellular biology : MCB. 27, 1280-1295, doi:10.1128/MCB.00882-06 (2007).

72 Yang, W.-M., Yao, Y.-L., Sun, J.-M., Davie, J. R. & Seto, E. Isolation and Characterization of cDNAs Corresponding to an Additional Member of the Human Histone Deacetylase Gene Family*. The Journal of Biological Chemistry 272, 28001-28007 (1997).

73 Sankar, N. et al. p300 provides a corepressor function by cooperating with YY1 and HDAC3 to repress c-Myc. Oncogene 27, 5717, doi:10.1038/onc.2008.181 (2008).

74 Zhang, L. et al. Hdac3 Interaction with p300 Histone Acetyltransferase Regulates the Oligodendrocyte and Astrocyte Lineage Fate Switch. Developmental cell 36, 316-330, doi:10.1016/j.devcel.2016.01.002 (2016).

75 Wu, D. & Pan, W. GSK3: a multifaceted kinase in Wnt signaling. Trends in biochemical sciences 35, 161-168, doi:10.1016/j.tibs.2009.10.002 (2010).

76 Ohteki, T. et al. Negative regulation of T cell proliferation and interleukin 2 production by the serine threonine kinase GSK-3. The Journal of experimental medicine 192, 99-104 (2000).

112

77 Wood, J. E., Schneider, H. & Rudd, C. E. TcR and TcR-CD28 engagement of protein kinase B (PKB/AKT) and glycogen synthase kinase-3 (GSK-3) operates independently of guanine nucleotide exchange factor VAV-1. J Biol Chem 281, 32385-32394, doi:10.1074/jbc.M604878200 (2006).

78 Zhang, X. et al. Histone deacetylase 3 (HDAC3) activity is regulated by interaction with protein serine/threonine phosphatase 4. Genes & development 19, 827-839, doi:10.1101/gad.1286005 (2005).

79 Kalia, V. et al. Prolonged interleukin-2Ralpha expression on virus-specific CD8+ T cells favors terminal-effector differentiation in vivo. Immunity 32, 91-103, doi:10.1016/j.immuni.2009.11.010 (2010).

80 Shin, H. M. et al. Epigenetic modifications induced by Blimp-1 Regulate CD8(+) T cell memory progression during acute virus infection. Immunity 39, 661-675, doi:10.1016/j.immuni.2013.08.032 (2013).

113