Epigenetic Regulation of T cell Adaptive Immunity

Summary

The conceptualization of adaptive immunity, founded on the observation of immunological memory, has served as the basis for modern vaccination and immunotherapy approaches. This fundamental concept has allowed immunologists to explore mechanisms that enable humoral and cellular lymphocytes to tailor immune response functions to a wide array of environmental insults and remain poised for future pathogenic encounters. Until recently, for T cells it has remained unclear how memory differentiation acquires and sustains a gene expression program that grants a cell with a capacity for a heightened recall response. Recent investigations into this critical question have identified epigenetic programs as a causal molecular mechanism governing T cell subset specification and immunological memory. Here we outline the studies that have illustrated this concept and posit on how insights into T cell adaptive immunity can be applied to improve upon existing immunotherapies.

Introduction

Upon recognizing an aberrant or infected host cell, antigen-specific T cells clonally expand, traffic to the site of initial antigen encounter, and kill the target cells. Following antigen-control, the expanded population of T cells undergo contraction resulting in elimination of short-lived effector cells and leaving a subset of cells that survive and develop into long-lived functional memory T cells. Memory differentiation grants T cells the ability to undergo antigen-independent homeostatic self-renewal driven by gamma-chain cytokines and a heightened recall response upon subsequent encounters with their cognate antigen. However, this process of memory differentiation is highly sensitive to the duration of antigen-exposure and antigen load. In the setting of acute antigen exposure, briefly outlined above, T cells can indeed form long-lived immunological memory that is capable of swiftly eliminating threats upon antigen re-encounter (Figure 1) (1). However, if the source of antigen persists, T cells can undergo an alternative adaptive path whereby they progressively suppress their effector functions to limit the immunopathology associated with a sustained cytolytic T cell response (2). These broad examples of divergent T cell fates within a genetically fixed background have raised several important questions that the field has worked to unravel: Once a T cell has a fixed T cell receptor (TCR), how then does the cell adapt in real-time to the constantly evolving antigen load and subsequently rapidly recall this response days, months, or even years, later? What are the regulatory mechanisms that control the cell fate decision that delineates the development of a T cell into either short-lived effectors or memory precursor effector cells? Once a T cell has differentiated and acquired a subset-specific gene expression program, can this T cell fate be reversed? From exploration of these questions, a critical role for epigenetic mechanisms has been established in regulating many of these processes and provided a molecular mechanism to explain how gene expression programs are propagated from parental memory T cell to its progeny, even in the absence of antigen.

Figure 1. Memory T cell differentiation in response to acute and chronic diseases.

During acute infection, naive T cells clonally expand and differentiate into short lived (terminal) effector cells (SLECs/TEs, KLRG1+CD127−) or memory precursors (MPs, KLRG1−CD127+) initially defined based on their expression of KLRG1 and IL7R and survival in adoptive transfer studies. Following antigen clearance, T cells develop into long-lived memory T cells that are capable of antigen-independent self-renewal for years. In chronic diseases where antigens and inflammatory stimuli persist, T cells retain expression of TF Tox that enables their survival as they progress through a developmental path that results in suppression of effector functions. This differential program promotes the cell’s survival and limits immunopathology. At the beginning of the “exhaustion” lineage exists a population of self-renewing [ICB responsive] Tex progenitors (Tox+ Tcf-1+Tim-3−) that give rise to terminally exhausted T cells (Tox+Tcf-1− Tim-3+). Teff, Tmem and Tex are not only distinguishable in function but also in transcriptional and chromatin accessibility programs. A generalized chromatin accessibility state is shown next each T cell subset.

Epigenetic regulation of T cell differentiation was first described over twenty years ago, with studies utilizing DNA methylation sensitive or insensitive restriction enzymes to profile the epigenetic states of loci in T helper (Th) cell subsets (3–5). With the advent of high-throughput sequencing combined with DNA manipulation methods, whole genome epigenetic profiling has made it possible to detail the expansive development-associated gene expression programs that become reinforced during the various stages of T cell differentiation in response to acute and chronic sources of antigen. From these studies, we have gained genome-wide snapshots of T cell epigenetic landscapes, as well as identified the enzymes directly involved in chromatin remodeling, allowing the field to gain a deeper understanding of how T cells protect against a wide range of diseases.

Not only have these insights led to a more fundamental understanding of T cell biology, but they have also enabled the development of novel diagnostics and therapeutic approaches to treat and cure diseases that were once universally fatal. A prime example of how the specificity and durability of T cell immune responses can be exploited is highlighted by the recent and major advances in cancer immunotherapy (6). Among T cell-based therapies, the two major categories are therapeutics that currently rely on antibody blocking of T cell inhibitory receptors to their cognate ligand, and adoptive T cell therapy of unedited or T cells engineered to express chimeric-antigen receptors (CAR) (7). While these approaches have elicited a substantial improvement in the way various cancers are treated, many studies have demonstrated that the capacity of a T cell to elicit a sustained immune response depends on mechanisms regulating cell fate commitment.

In this review, we will detail the work describing the mechanisms involved in and governing epigenetic regulation of T cell differentiation. We will also discuss how our contemporary understanding of epigenetic regulation can be exploited to enhance current and future T cell-based therapeutics.

1. Epigenetic regulation of CD8⁺ T cell development and differentiation

The chromatin landscape of a cell is controlled by various histone modifications and DNA methylation that specify accessibility of transcription factors (TFs) to gene regulatory elements. Histone proteins allow the vast genome to be condensed into nucleosomes. ‘Overpackaging’ of DNA renders the genome inaccessible and silenced, referred to as heterochromatin while ‘under-packed’ DNA results in accessible euchromatic regions (8). The state of chromatin accessibility is commonly dependent on modifications of lysine residues in the histone tail, although other amino acid residues and the histone core can also be modified. Initial gene expression studies focused on transcriptional start sites (TSSs) where accessibility is associated with transcriptional potential, however attention has shifted to chromatin modifications which are ubiquitous throughout the genome, and collectively work to control transcription (9).

One of the best documented histone modifications that results in chromatin remodeling is acetylation of histone. This covalent modification disfavors nucleosome interactions allowing for increased access to DNA. Histone acetylation is mediated by histone acetyltransferases (HATs) and removed by histone deacetylases (HDACs). Slightly more intricate, histone methylation can be associated with a chromatin repressive (H3K9, H3K27) or permissive (H3K4, H3K36) state dependent on the lysine residue location, and vary in methylation valence (R-me1, R-me2, R-me3). There is a myriad of histone modifications that can occur at different amino acid residues on each histone protein (H2, H3, H4) including ADP ribosylation, deamination, phosphorylation, ubiquitylation, and SUMOylation (8,10). The distinct combination of modifications, referred to as the histone code, provides a sophisticated network of signals that collectively work to regulate gene expression in a context dependent manner.

DNA methylation encompasses another layer of epigenetic regulation that can prevent the activity of DNA binding factors on gene transcription. Cytosine methylation at CpG dinucleotides, denoted as 5-Methylcytosine (5mC), is the most common form of DNA methylation and is catalyzed by the family of DNA methyltransferases (DNMTs) (11). Owing to its palindromic nature, the CpG methylation status within a cell can be propagated during cell division by a DNA methyltransferase (DNMT1) that has an affinity for hemi-methylated DNA found on newly synthesized DNA daughter strands. Other members of the DNMT family, DNMT3a and DNMT3b, are responsible for de novo DNA methylation. Counter balancing DNMT activity, DNA demethylation is achieved by the family of ten-eleven translocation (TET) enzymes where DNA undergoes oxidation steps of 5mC and its intermediates resulting in passive (cell division mediated) and/or active (enzyme catalyzed) demethylation (12). While these individual epigenetic modifications are frequently described separately, mechanisms involving DNA methylation and histone modifications often intersect and work in concert to reinforce chromatin accessibility. For instance, DNMT3a prefers the unmethylated H3K4 and exhibits reduced affinity when methylated (H3K4me3) which also signals chromatin accessibility and is associated with active gene expression (11,13). The above briefly described chromatin modifications are critical in regulating cell and tissue specialization, and have been linked to the commitment of lymphoid and myeloid cell fate (14). Building upon these molecular and cellular concepts, a role for epigenetics in adaptive T cell responses has been recently explored.

Over the past few decades, a large body of work has begun to detail the epigenetic modifications coupled to T cell thymic development and the differentiation process that occurs following a T cell’s encounter with its cognate antigen in the periphery. As lymphoid progenitors egress from the bone marrow and migrate to the thymus, signaling within the thymus results in T cell lineage commitment and leads to the generation of CD4⁻CD8⁻ (double negative, DN) T cells which subsequently give rise to either TCRαβ⁺ or TCRγδ⁺ T cells. DN αβ⁺ T cells progress through several developmental stages to attain a naive stage of differentiation in which they possess a complete, rearranged TCR specific for a given antigen and bearing an optimal affinity for MHC (major histocompatibility complex), and express CD4 or CD8 co-receptors (reviewed in (15)). It has become clear that transcriptional and epigenetic regulatory programs govern each of these developmental steps. In early reports, Sawada and colleagues discovered an enhancer 13kb upstream of the CD4 transcriptional initiation site and an intronic regulatory region in mice which collectively seemed to mediate, in part, co-receptor expression in DN through single-positive (SP) T cell differentiation (16–18). Further work by the Littman laboratory demonstrated that this silencer was necessary for CD4 silencing only at distinct stages of T cell development. Deletion of this silencer prior to lineage commitment resulted in the de-repression of CD4 throughout T cell differentiation, but CD4 silencing was maintained when deleted post-CD8⁺ commitment (19). The investigators further demonstrated that silencing of CD4 pre- and post-CD8⁺ commitment depended on Runx1 and Runx3, respectively (20). Collectively, these events were broadly demonstrated to be controlled by DNA methylation given that disruption of maintained methylation (Dnmt1) at this region resulted in improper regulation of lineage commitment (21). Along similar lines, Lee et al. used a Lck-Cre; Dnmt1^2lox model to delete Dnmt1 in mice during the DN stage and noted that loss of maintenance methylation resulted in a significant reduction of double-positive (DP) and SP TCRαβ⁺ T cells in secondary lymphoid organs. Conversely, a unique population of CD8⁺ TCR γδ⁺ T cells were found to be enriched in the thymus, spleen, and lymph nodes of these mice in response to Dnmt1 loss (22). Work by Kwesi-Maliepaard et al. reported that Lck-Cre mediated deletion of the H3K79 methyltransferase, disruptor of telomeric silencing 1-like (DOT1L), in early T cell development resulted in a global loss of thymic cellularity with reductions in the double-positive (DP) cells, and in CD4⁺ and CD8⁺ SP compartments (23). These alterations were also found to result in an increase in CD8⁺ T cells with a memory-like (CD44⁺CD62L⁺) phenotype (23). Collectively, these findings served to establish epigenetic mechanisms as a critical regulatory process in the cell fate decisions that occur during T cell thymic development. However, once a T cell has undergone CD4 or CD8 lineage commitment, the way in which these epigenetic mechanisms affect the adaptive response remained to be determined.

Once naive CD8⁺ T cells encounter their cognate antigen on an antigen presenting cell (APC), a cascade of transcriptional and epigenetic reprogramming events transpire that are coupled to clonal expansion of the cells and endows them with effector functions and specific tissue homing properties. The initial step of the process happens within minutes to hours of antigen exposure and is accompanied by chromatin modifications. Notably, Zhou et al. found that within 10 minutes of TCR engagement, BRG1 (encoded by SMARCA4), a member of the SWI/SNF subfamily of ATP-dependent chromatin modifiers, was necessary for β-actin and BAF53 association with the BAF complex to stably associate with chromatin (24). In addition to these pioneering chromatin regulatory events, a multitude of epigenetic changes occur during the transition from naive to effector T cells. These changes include effector genes switching “on” with concurrent naïve-associated genes such as lymph node homing molecules switching “off”. Much of the early work exploring epigenetic mechanisms involved in T cell differentiation focused on changes in DNA methylation that occurred during the effector stage of the immune response, with a specific focus on the Ifng locus. Fukunaga et al. showed early on that in certain T cell lines the level of IFNG expression was inversely correlated to the extent of hypomethylation of the gene (25). Melvin et al. examined the degree of methylation of the IFNG loci in a spectrum of primary T cells isolated from humans (thymic, neonatal, adult, CD4⁺, CD8⁺) and concluded that a proportion of hypomethylation in the 5’ flank and first intron of the gene was coupled to the capacity of these cells to express the cytokine (4). In two separate reports, Fitzpatrick et al. demonstrated that demethylation of Ifng and Il3 loci, particularly in the promoter region, was linked to increased mRNA expression of these cytokines, and that these states were heritable in cells that were derived from activated (CD44^high) CD8⁺ T cells (26). Further, they went on to show that the differentiation-associated epigenetic state of the Ifng promoter was maintained within CD44^high CD8⁺ T cells, regardless of whether the cells clonally expanded, divided, or stimuli was withdrawn (27) suggesting that these programs, once established, are faithfully propagated to their progeny.

The above-described studies provided important insights into effector differentiation and highlighted questions surrounding the relationships between parental and expanded T cells. To further explore the relationship between effector and memory T cells several labs turned to using model systems which enabled tracking of in vivo antigen-specific T cell responses. Combining these models with loci-specific and genome-wide epigenetic profiling approaches has shed light on the developmental path for effector and memory differentiation. Using whole-genome DNA methylation analysis, Scharer et al. performed an initial characterization of the changes in DNA methylation that occur during effector differentiation of lymphocytic choriomeningitis virus (LCMV)-specific T cells. DNA methylomes of naive antigen-specific T cells and those isolated at the peak of the effector response (8 days post infection [dpi]) from mice infected with the acute strain of LCMV were strikingly distinct from one another (28). These analyses revealed over 350,000 differentially methylated regions (DMRs) in 8dpi effector T cells relative to their naive counterparts. Importantly, DMRs were present in around ~50% of RefSeq promoters. Many of the genes were identified as differentially expressed between naive and effector T cells; in particular, several genes required for effector function were demethylated in the expanded population of T cells including Gzmb, Zbtb32, Ifng, Klrg1, Klrc1, Ctla4, Casp1, and Batf, among others, while other genes associated with a stem-like naive state were methylated in effectors such as Cxcr2, Tcf7, Itgae, Ccr7, Ccr9, Lef1, and Cxxc5 (28). Further, the histone modifications H3K4me3 and H3K27me3 were observed to correlate with DNA methylation states of effector versus naive T cells (28). Similar to what was seen during T cell thymic development, Dnmt1 was found to be required for T cell expansion in response to acute LCMV infection. Using a Granzyme B-Cre mediated deletion of Dnmt1, Chappell et al. demonstrated that despite similar activation phenotypes, there was a ≥84% reduction in antigen-specific T cells lacking Dnmt1 at the height of the effector response (8dpi) versus littermate controls—an effect determined to be due to delayed cellular proliferation (29). Taken together, such studies illustrate a critical role for epigenetics in the initial phase of the effector response. Given that optimal stimulation of a naive T cell (with a fixed TCR) results in the generation of a pool of effector cells with differing cellular fates and multipotent capacities (i.e., short-lived effector cells (SLECs/terminal effectors [TE]) and memory precursor (MP) cells) (Figure 1), a major question for the field was whether epigenetic mechanisms control these bifurcative decision(s), and do these epigenetic programs impact on memory T cell functions?

Such open question(s) prompted our group and others to explore epigenetic mechanisms and modifications coupled to T cell memory subset specification. Work from our group on the dynamic role of de novo DNA methylation during T cell differentiation focused on genes that have an “on-off-on” pattern of expression over the course of naive to effector to memory development where we characterized de novo programs acquired in the known effector subsets (TE vs MP). We initially focused on the methylation status of CD62L (encoded by Sell, and also referred to as L-selectin) in wild-type antigen-specific MP and TE cells isolated at the peak of the effector response in mice acutely infected with LCMV (8dpi). This analysis revealed that both subsets equally methylated the Sell promoter region with associated low levels of mRNA expression levels, while this region was demethylated with high levels of mRNA expression in memory T cells isolated at 37dpi. Thus, MP cells acquired a transcriptionally repressive DNA methylation state at the Sell promoter early during T cell activation, and this repressive state was relieved in a subset of memory T cells (30). Whole genome bisulfite sequencing (WGBS) of these populations (TE and MP) at early (4.5 dpi) and peak (8 dpi) timepoints demonstrated significant remodeling as compared to naive T cells, as many of the DMRs comprised naive-associated genes such as Sell (CD62L), Ccr7 and Tcf7 as well as effector-associated genes such as Prf1, Gzmb, and Ifng, and in which a majority of these methylating events occurred relatively early within the response (30). Importantly, these data suggested that MP cells transitioned through an effector state prior to differentiating into memory T cells (Figure 1). We next proceeded to ask what would happen if the promoter could not be methylated. To do this, we utilized a Granzyme B-Cre system in which de novo DNA methyltransferase 3a (Dnmt3a) activity would be lost once the T cells became fully activated and upregulated granzyme B expression, and again examined genes targeted by Dnmt3a by WGBS. WGBS of WT and cKO cells revealed that ~2000 regions between WT naive and effector cells acquired de novo methylation during this differentiation stage, with ~1000 of these regions identified as Dnmt3a targets (Sell, Ccr7, Tcf7, Lef1, Il6st). Importantly, we found that maintenance DNA methylation remained fully intact in the cKO cells (30). Strikingly, while there was no difference in viral clearance nor magnitude of the response (as these Dnmt3a cKO cells did acquire similar effector programming as the WT cells had and mounted the correct antiviral response), we found that CD62L re-expression occurred much earlier in Dnmt3a cKO T cells as compared to their WT controls. This result translated to enrichment of these cells in secondary lymphoid organs and non-lymphoid tissues (30). In a similar study, using Dnmt3a conditional knockout under the control of CD4-Cre or Lck-Cre, Ladle et al. reported that antigen-specific T cells lacking Dnmt3a resulted in more MP cells (CD127⁺KLRG1⁻), rather than TE cells (CD127⁻KLRG1⁺), at the peak of the response as compared to controls in three different acute infection models (VacOva, PR8 influenza, LCMV). Using genome wide methyl-CpG binding domain sequencing (MBD-seq) to determine Dnmt3a-dependent de novo DNA methylation, 8,600 regions were identified with ~13% of these (1,134 regions) located within the promoter regions of genes, with the most notable gene identified as Tcf7 (encoding TCF1) (31). Given the importance of TCF-1 in the maintenance of stemness in memory T cells, it is likely that, in early effector cells (CD127⁻KLRG1⁻) Dnmt3a catalyzes de novo DNA methylation in certain cells at the Tcf7 locus contributing to the terminal state of the cell. Taken together, these results demonstrate that early DNMT3a activity restricts the multipotent capacity of effector cells and regulates their conversion into distinct T cell subsets.

Once a cell acquires new epigenetic modifications, proteins often generally described as epigenetic “readers” bind to the modified genome and implement the specific program. In an effort to better understand how DNA methylation is interpreted in T cell function, methyl-binding proteins have been examined in the context of effector and memory differentiation. Using methyl-CpG binding domain protein 2 (MBD2)-deficient animals, Kersh reported that MBD2^−/− effector T cells derived from an acute LCMV infection had an altered cytokine profile in which they had a selective reduction in IFNG and were generally more apoptotic than their wild-type counterparts but were overall capable of mediating viral clearance. More dramatically, MBD2^−/− T cells during the effector phase demonstrated a reduction, as well as a general delay in the generation of antigen-specific MP populations which translated to a reduced memory CD8⁺ T cell pool and failure of secondary protection during reinfection (32).

To further explore the broader impact of epigenetic regulation in effector differentiation, other groups have sought to define additional mechanisms which control the development of naive T cells into effectors. Denton et al. assessed the epigenetic landscape of T cells responding using an acute influenza murine infection model that contains OVA (A/HKx31-OVA) and enabled the investigators to track epigenetic changes associated with differentiation of OVA-specific TCR transgenic T cells (OT-I cells). Once activated a significant reduction of histone 3 (H3) protein and H3K27 trimethylation (H3K27me3) marks, with a concurrent increase in H3K9 acetylation (H3K9ac) and H3K4 trimethylation (H3K4me3) at the proximal Ifng promoter was observed in effector OT-I cells relative to naive T cells. Such permissible marks (H3K9ac and H3K4me3) located at the proximal promoter region of Ifng was further shown to be coupled to an increase in RNA pol II docking at this specific region (thus promoting their expression) (33).

As alluded to earlier, under normal conditions, once antigen is cleared from the host, the majority of T cells undergo a contraction through the process of apoptosis, while effector cells with memory potential continue their differentiation into bona fide memory cells to provide long-term protection (Figure 1). These MP effector cells have the capacity to differentiate into several different types of memory cells, such as stem-cell memory T cells (T_SCM), central memory T cells (T_CM), resident memory T cells (T_RM), and effector memory T cells (T_EM), in which each of these subsets vary in their multipotency, homing capacity, and cytotoxic function. Notably, T_RM localize and reside in tissues acting as a first line of defense against reinfection. Recent efforts to explore T_RM cell development have highlighted the utility of coupling epigenetic profiling approaches with cellular functions to assess the differentiation potential of memory T cell subsets. Epigenetic profiling of small intestine intraepithelial lymphocyte T_RM cells post LCMV infection (> 3 months) revealed that T_RM cells grouped closer with T_CM and T_EM cells than naive or TEs (at different time points) as determined by their DNA methylation profile and using a CD8⁺ T cell methylation-based plasticity index, or multipotency index (MPI, discussed in more detail below) (34). These results provided support for an ‘outside-in’ model where in response to reinfection, epigenetically plastic T_RM cells expand at their designated peripheral site and differentiate into protective memory subsets that can re-enter circulation. Despite having distinct functional differences, a shared aspect among these cells was that T_CM and T_RM CD8⁺ T cells have an extraordinary capacity to remain relatively quiescent until they re-encounter their cognate antigen and subsequently rapidly proliferate and re-express effector programs. Given that memory T cells undergo homeostatic proliferation in a largely antigen-independent fashion (35), significant work has gone into understanding how memory T cells are able to heritably “remember” the programs that were established during the effector stage of the immune response.

Thus far, we have discussed the distinct ways that DNA methylation can impact the differentiation and maintenance of programs in T cells. Not surprising, there has been reports demonstrating that DNA methylation also plays a role in memory CD8⁺ T cells. Kersh et al. assessed the degree of DNA methylation and whether this impacted the differential capacity of naive versus memory CD8⁺ T cells to express IFNG. Using acute LCMV infection, methylation levels of the Ifng promoter were assessed in antigen-specific T cells at the naive, effector, and memory stage. Expectedly, the Ifng promoter demethylated in the transition from naive to effector, however, the promoter underwent partial re-methylation in memory cells, but to a lesser degree than what was observed in naive cells. Importantly, once re-challenged, stimulated memory cells rapidly demethylated the Ifng promoter at a much faster rate compared to naive cells (36).

Characterization of DNA methylation throughout the various stages of memory differentiation have illustrated the critical association between demethylation of effector loci in the poised potential of resting memory T cells. To further explore the impact of this process on memory differentiation, studies have focused on the DNA demethylation regulator Tet2. Specifically, Carty et al. showed that loss of Tet2, a member of the family of methylcytosine dioxygenases, not only preferentially generated MP cells during acute LCMV infection with enhanced cytotoxicity relative to WT T cells, but also that Tet2 loss resulted in an increase in central memory-like CD8⁺ T cells in the spleens of mice at 45dpi. Upon secondary challenge with Listeria monocytogenes expressing LCMV epitope, GP33, Tet2-deficient memory T cells were able to mediate bacterial clearance better than their WT counterparts (37). Taken together, these results suggest that while memory T cells are capable of re-eliciting effector programs at a faster rate than naive T cells, modification of enzymes associated with DNA methylation during the effector response can alter these programs to modify the capacity of memory cells to respond during secondary responses. In essence, epigenetic manipulation can enhance effector propagation of MPs.

To get a more global view of the chromatin remodeling that occur during acute infection, Scharer et al. performed ATAC-seq on LCMV-specific T cells showing that memory CD8⁺ T cells acquire a unique chromatin accessibility signature relative to other populations generated during the course of the infection. Specifically, integrating the chromatin accessibility data with gene expression data demonstrated that memory-associated differentially accessible regions (DARs) were mostly reprogrammed immediately during naive to effector differentiation, with only a small subset reprogrammed later during effector to memory transition (38). Memory CD8⁺ T cells showed open promoter accessibility in genes that were actively upregulated, importantly however, was that most of the downregulated genes maintained an open promoter which could be rapidly induced upon stimulation. Indeed, upon ex vivo stimulation, antigen-specific memory cells expressed Ifng, Zbtb32, Itga1, and Eomes, much faster compared to ex vivo stimulated naive antigen-specific T cells (38).

Taken together, the studies described above underscore the idea that memory T cells transition through an initial effector stage of differentiation and acquire epigenetic modifications that allow them to maintain open, “poised” DNA methylation and chromatin states that can be heritably maintained during homeostasis. Broadly, the studies collectively explain how epigenetic programs serve as a mechanism to enable adaptative immunity in a genetically fixed environment.

2. Epigenetic regulation of T cell exhaustion

The formation of functional T cell memory occurs during an acute infection where antigen is successfully cleared. However, in certain cases in which antigen is not cleared, chronic T cell stimulation results in a transcriptional and epigenetic reprogramming of the T cell resulting in a progressive suppression of effector potential (39) (Figure 2). This epigenetic repression of the T cell is functionally demarcated by impaired proliferative capacity, diminished polyfunctional cytotoxic potential, and the sustained expression of inhibitory receptors, collectively referred to as “T cell exhaustion” (39). Biologically, the process of T cell exhaustion is hypothesized to mitigate host immunopathology during chronic viral/tumor exposure. However, acquisition of exhaustion-associated epigenetic programs also presents a challenge for T cell based immunotherapy (39).

Figure 2. Assessment of T cell multipotency based on WGBS DNA methylation datasets.

T cell differentiation results in remodeling of the epigenetic landscape whereby distinct epigenetic changes are coupled to T cell subset specification and in the cell’s developmental potential. The epigenetic modification DNA methylation serves to reinforce changes in the chromatin landscape. DNA methylation profiling has revealed that naive T cell loci associated with self-renewal and homing to secondary lymphoid organs are demethylated while effector associated loci are fully methylated. Once T cells encounter their cognate antigen, LN homing genes become methylated and effector-associated loci are demethylated. After the source of antigen is cleared, naive-associated genes are demethylated in long-lived memory cells which is coupled to cell’s multipotent developmental capacity. Importantly, effector-associated loci stay unmethylated, poising them for an effective secondary response. In terminal exhausted T cells, key stem and effector associated loci remain highly methylated enforcing the dysfunctional cellular state. Using these known epigenetic states, an epigenetic based multipotency score has been generated based on a set of 245 CpG sites obtained by a machine learning analysis of the whole-genome methylation datasets from classically defined T cell subsets. This MPI in turn can predict the developmental hierarchy of CD8 T cells. Bound within a range of 0 and 1, multipotency is denoted from least (0) to most (1). The MPI is being used as a biomarker for predicting the developmental plasticity of CD8 T cells in clinical settings.

T cell exhaustion was initially defined in LCMV, in which antigen-specific T cells were found to lack the capacity to perform their effector functions against the chronic forms of LCMV, namely the Clone-13 and Docile strains (40,41). This cellular state was subsequently identified in humans suffering from chronic viral infections such as HIV (42–44). Since then, a plethora of work has gone into investigating the various cell-intrinsic and extrinsic mechanisms by which T cell exhaustion occurs, with a significant emphasis on the study of surface inhibitory receptors that repress effector transcription during chronic infection and cancer. Following the discovery of the inhibitory receptor, programmed cell death protein 1 (PD-1) (45) and its expression on activated T and B cells (46), it was realized that modulation of this receptor could result in improved therapeutic T cell responses against tumors in mice (47). Importantly, it was determined by Barber et al. that exhausted T cells actively upregulated PD-1, and suppression of PD-1 signaling by monoclonal antibody blockade resulted in “rescue” of these T cells to impart them with an improved capacity to proliferate and perform their effector functions in mice infected with chronic LCMV (48). Application of these results were extended to primate and human systems prompting further investigation into the mechanisms enabling a reinvigorated T cell response (49–53). In pursuit of a deeper understanding of T cell responsiveness to PD-1 blockade, Blackburn et al. observed that there were actually two distinct subsets within the pool of exhausted T cells – one which was capable of responding to PD-1 blockade (later termed the ‘progenitor exhausted’ pool), and another which was more terminally exhausted and unable to fully respond to this “reinvigoration” (54). It was later worked out that these two populations had differential PD-1 expression (PD1^int versus PD1^high) that depended on the transcription factors, T-bet and Eomes (55) (Figure 1). Importantly, both populations were necessary for controlling chronic infection as loss of either population resulted in viral escape (55). Building on these results, the major question became: Why, if both populations express PD-1, is there a dichotomous cellular response to PD-1 blockade? Work by our group thus examined the DNA methylation state of both the murine and human PD-1 locus. During acute infection of LCMV and during yellow fever (YF) vaccination, the Pdcd1 locus became demethylated in effector cells following TCR signaling but was remethylated following the differentiation of these cells into functional memory (56). Conversely, in chronic viral infection such as those found in chronic LCMV, Epstein-Barr virus (EBV) and Cytomegalovirus (CMV) infection, exhausted T cells demethylated the Pdcd1 locus and it remained demethylated even in light of decreasing viral titers (56). Moreover, this observation was further confirmed by examining the PDCD1 locus DNA methylation status in donor-matched naive and HIV-specific CD8⁺ T cells isolated from antiretroviral therapy (ART)-treated subjects or from elite controllers (ECs), who naturally control HIV-infection. We documented that, similar to cells derived from EBV and CMV infections, HIV-specific CD8⁺ T cells demonstrated an unmethylated PDCD1 promoter while naive T cells retained a methylated region, regardless of later viral control, (57). Collaborative work by our group went on to establish that this demethylation of the Pdcd1 promoter was established early on during the effector phase in a chronic antigenic environment, and that this demethylated promoter region was maintained in T cells that survived to form memory in acutely infected mice which ultimately translated to impaired responses upon secondary challenge (58). Taken together, these data suggested that exhausted T cells reached a terminal fate that impacts the responsiveness of PD-1 blockade in terminally exhausted T cells.

Of note, exhausted T cells in the absence of antigen were also found to heritably maintain this exhaustion phenotype and the cells were re-expanded in response to an acute LCMV infection (59), suggesting that regardless of resting environment, the differentiation process to T cell exhaustion resulted in programming that could not be reversed. The question then became what underlying mechanisms were driving the heritable nature of this exhaustion programming? Could this programming potentially be reversed or blocked so that exhausted T cells could remain responsive to PD-1 blockade? Given that we had previously demonstrated that loss of Dnmt3a at the effector stage could lead to enhanced memory conversion during acute infection, we wondered whether de novo DNA methylation programs could also reinforce the differentiation of exhausted cells. Using a Dnmt3a conditional knockout murine mode (cKO), we were able to show that blocking de novo DNA methylation programs resulted in reduced CD8⁺ T cell contraction (following peak expansion), sustained demethylation in effector-associated loci resulting in cell-intrinsic maintenance of IFN-gamma and IL-2 potential production (>60dpi), and preservation of the overall TCR repertoire after PD-1 blockade (60). WGBS comparisons between WT and cKO T cells isolated from 8dpi (effector stage) and at 35dpi (exhaustion stage) demonstrated that Dnmt3a targeted >900 regions from the naive to effector stage and a striking 1200 regions from the effector to exhaustion stage, thereby determining that in this differentiation pathway from effector to exhaustion, many of these methylating events are due to de novo DNA methylation (60). Closer assessment of an effector associated DMR in the Ccr7 locus revealed that de novo methylation can occur both during the effector stage and gain additional methylation during exhaustion. Many of the regions targeted were those associated with TCR-signaling, and IL-7/IL-2 signaling, with specific targeting of the loci of Ifng, Myc, Tcf7, Ccr7, Tbx21, and Eomes. Importantly, WT exhausted T cells treated with PD-1 blockade did not undergo erasure of de novo DNA methylation programs at any of the above loci suggesting that once established, exhaustion-associated programming is stable, and that PD-1 treatment does not erase these methylation programs (60). Further, when we examined Dnmt3a cKO and WT antigen-specific T cells in chronic infection, blocking of de novo DNA methylation programs (Dnmt3a cKO) resulted in improved retention of these cells during the exhaustion stage with an enhanced proliferation, heightened cytokine recall potential, and a more stem-like state (higher Tcf1 expression) (60). Thus, these data demonstrate that Dnmt3a plays a key role in regulating the development of T cell exhaustion. Of note, blocking de novo DNA methylation by modulating Dnmt3a coupled to PD-1 blockade does result in significant effects on the subsequent functionality of T cells (60), which we will discuss later in this review.

In recent years, there has been an explosion of work demonstrating the various transcriptional and epigenetic regulatory proteins involved in the development of exhaustion as well as in the maintenance of this exhausted progenitor population, recently summarized by Lugli et al. (in (61)). Here, we will discuss some of the recent investigations of a key TF associated with the development of T cell exhaustion.

Through a series of experiments that involved analysis of gene expression profiles, TOX has been identified as a potential mediator of T cell exhaustion. Notably it was observed during chronic LCMV infection that TOX is highly expressed in exhausted T cells whereas this was not observed in T cells during acute infection (62–64). Importantly, in both chronic LCMV and in tumor models, modulation of TOX resulted in a reduction in or failure to upregulate certain inhibitory receptors expression on antigen-specific T cells (62–64). In line with this, we found that the Tox locus remains methylated in acute infection, however, in chronic infection the Tox locus becomes unmethylated and this unmethylated state was retained even after 8 weeks post-infection (62). Taken together, TOX regulation is critical in the formation and survival of exhausted CD8⁺ T cells (62,64), however, TOX activity is likely context dependent as it has also been demonstrated to be involved in other processes (63).

3. Epigenetic programs can define CD8⁺ T cell differentiation states

While much of the field’s knowledge of T cell subset specification comes from characterization of cell surface marker expression, anatomical location, and effector functions, several lines of evidence suggest that T cell subsets can be demarcated based on epigenetic modifications (Figure 2). Recent efforts by our lab to establish epigenetically-defined correlates of T cell differentiation status focused on establishing genome-wide DNA methylation profiles of well-defined naive and memory T cell subsets isolated from humans (T_naive, CCR7⁺CD45RO⁻CD95⁻; T_SCM, CCR7⁺CD45RO⁻CD95⁺; T_CM, CCR7⁺CD45RO⁺, and T_EM, CCR7⁻CD45RO⁺) (65,66). Our initial characterization of human T cell subsets led to the observation that the vast majority of differentiation-associated changes in DNA methylation were located within or near genes, specifically promoter, intronic or distal regions (65). To assess the stability of these programs, we examined the DNA methylation profile of polyclonal memory T cells isolated from patients after haploidentical donor hematopoietic cell transplantation (HSCT) (65). Similar to the methylation profiles of the T_EM cells prior to infusion, the IFNG and PRF1 loci of T_EM cells isolated one month from HSCT-patients after lymphocyte infusion remained demethylated. Importantly, the infused memory T cells had undergone significant antigen-independent proliferation suggesting that the effector-associated epigenetic modifications were propagated during homeostatic proliferation. Complementing our studies using adoptively transferred memory T cells in transplant patients, we also examined the DNA methylation and open chromatin states of antigen-specific T cells longitudinally generated in response to yellow fever vaccination of healthy adults. Importantly, we observed that YFV-specific memory CD8⁺ T cells retained demethylation of the granzyme B and perforin loci for ~10 years (endpoint of study) (67). Thus, while the expression of these molecules was downregulated in the YFV-specific memory T cells, the cells appeared to be epigenetically-poised to recall these effector functions. Similarly, Crompton et al. performed global mapping of H3 lysine 4 and H3 lysine 27 trimethylation maps in (T_naive, CD62L⁺CD44⁻SCA-1⁻; T_SCM, CD62L⁺CD44⁻SCA-1⁺; T_CM, CD62L⁺CD44⁺, and T_EM, CD62L⁻CD44⁺) and found that chromatin remodeling occurs in a progressive nature (T_naive -> T_SCM -> T_CM -> T_EM) (68). Specifically, the T_SCM subset was enriched for permissive H3K4me3 and absent for H3K27me3 marks in memory associated genes (Tcf1, Lef1, Foxo1, Klf2) with this H3K4me3 and H3K27me3 pattern reversing in T_CM and T_EM subsets, thus showing that these subsets can be further delineated using DNA methylation and histone modifications. Collectively, these studies indicated that human memory CD8⁺ T cell subset-specification is coupled to acquisition and retention of epigenetic programs that facilitate a poised effector response.

Building on these initial observations, we have worked toward identifying epigenetic programs that can serve as predictive biomarkers of specific T cell differentiation states. Using a machine learning approach, inspired by work from Malta et al. (69), epigenetic signatures spanning the developmental bounds of mouse and human T cell differentiation were used to identify CpG sites that served as a “DNA methylation-based T cell ‘multipotency index’ (MPI)”. The MPI was validated using CD8⁺ T cell subsets isolated from healthy human donors and now can be used to assess the relative multipotent potential of a T cell of unknown developmental status based on its DNA methylation state (70) (Figure 2). As a proof of concept, we determined that beta-cell specific CD8⁺ T cells isolated from type 1 diabetic individuals contained a hybrid epigenetic state of naive and effector-like programs, with an MPI higher than T_CM and T_EM cells, but lower than naive T cells (70). Thus, we were able to apply this new bioinformatic tool to document the hybrid stem-like/effector nature of T cells associated with autoimmune disease.

Building upon the concept that epigenetic states can be used to predict T cell developmental potential, we incorporated our MPI analysis into the assessment of murine CD8 T cell memory fate commitments. In a recent study by the Masopust laboratory, we performed whole-genome DNA methylation analyses of LCMV-specific memory CD8 T cell subsets, including resident memory T cells (T_RM) isolated from the small intestine. Quite surprisingly, our MPI analysis demonstrated similarities between T_RM cells and T_CM cells, suggesting a far more plastic nature of T_RM cells than previously documented (34). Consistent with these findings, parallel adoptive transfer and transcriptional studies demonstrated that T_RM cells have the capacity to differentiate into other memory subsets and re-circulate upon secondary stimulation, while heritably maintaining a proclivity to home back to their tissue of origin (34). These data contrasted with the prevailing idea that T_RM cells were terminally differentiated owing to similarities in their effector function relative to terminal effectors. Collectively, these recent human and murine studies underscore the concept that epigenetic programs serve as a mechanism for the maintenance, plasticity, and enforcement of memory T cell subset specification, and can be used to predict the differentiation status of a T cell of unknown origin and development potential.

4. Epigenetic regulation of effector and memory CD4⁺ T cell differentiation

While this review has primarily focused on epigenetic modifications involved in CD8⁺ T cell effector and memory differentiation, the concept of epigenetic reinforcement of T cell subset specification is largely borne out of studies focused on CD4⁺ T helper (Th) cell specification. CD4⁺ T cells have been studied through the paradigm of T helper (Th) cell subset specification. Upon activation, conventional naive CD4⁺ T cells differentiate into specialized subsets in a cytokine dependent manner and provide support to the immune system in the form of cytokine signaling and co-stimulation. These subsets include Th1, Th2, Th17, and T_FH (follicular helper) cells and are initially described by expression of program-determining TFs, such as T-bet, GATA3, ROR-gamma-t, and Bcl-6, respectively (71). The study of Th subset specification has been covered extensively by others, so we will focus on the memory formation of CD4⁺ T cells and their parallels with CD8⁺ T cells.

Naive T cells, not yet dedicated to a particular subset, have multiple regulatory mechanisms that prevent premature cytokine expression which include epigenetic mechanisms. These mechanisms include enrichment of repressive H3K27me3 marks at loci such as Ifng and Il4 (72,73). In addition, the naive T cell Il4 and Ifng loci maintain DNA methylation not present in cytokine-positive cells (4,5,74). A similar repressive H3K27me3 pattern is observed among differentiated-associated transcriptions factors including Tbx21 (Th1), Gata3 (Th2), and Rorc (Th17) (72), which serve to restrain differentiation programs silenced in the absence of antigenic stimulus. This clearly shows several epigenetic components responsible for the silencing of effector programs in naive T cells which in turn require epigenetic reprogramming upon activation for expression.

Upon CD4⁺ T cell activation (TCR, co-stimulation and cytokine), chromatin modifications, specifically histone acetylation, are acquired at loci responsible for differentiation and effector function. Particularly, the histone acetyltransferase enzyme p300 is found to be enriched in super enhancer sites of cytokines and cytokine receptors, in a subset-specific manner, including Ifng, Il10, Il17a, Il17f, Il2ra, Il7r, although it is most highly enriched in the T cell differentiation regulator Bach2 (73,75,76). Histone acetylation is attributed to characteristic STAT4/T-bet signaling for the Ifng locus in Th1 cells, and STAT6/GATA3 signaling for the Il4 locus in Th2 cells. T-bet mediated reduction of histone acetylation is observed at the Il4 locus and GATA3-mediated reduction at the Ifng locus (73). Additionally, it has recently been reported that the Th17 master transcriptional regulator ROR-gamma-t itself is necessary to induce chromatin accessibility of IL-17A, IL-17F, and IL-23R through the SWI/SNF chromatin remodeling complex, as expression of these proteins are downregulated with the loss of the SRG3 subunit in CD4⁺ T cells (CD4^CreSRG3^fl/fl) (77).

While the study of CD4⁺ T cell memory has been hindered due to their differentiation into multiple subsets as well as a substantially limited degree of cellular expansion compared to their CD8⁺ T cell counterparts (78), recent observations have provided evidence that Th subset specification can be sustained in CD4 memory T cells. Particularly, the lineage defining T_FH markers Bcl6 and CXCR5 have been observed in CD4⁺ T_SCM and T_CM populations (79) revealing overlap of markers in the T_FH lineage and CD4⁺ T cell memory.

Using the acute model of LCMV, work in understanding the longevity of T_FH responses by Hale et al. explored the concept of distinct memory formation through the assessment of epigenetic changes that occurred from the naive to memory stage of the immune response from adoptively transferred antigen-specific memory CD4⁺ Th1 and T_FH cells from SMARTA mice. Notably we reported that Th1 cells demethylate the granzyme B locus and maintain this demethylated state into memory. Interestingly, the same region in T_FH cells remains methylated (80). Despite Il21 and Ifng expression ascribed to T_FH and Th1, respectively, these loci, along with that of Pdcd1, were all demethylated in both CD4⁺ subsets following activation, and they remained demethylated into the memory stage (80). Not only were these epigenetic programs preserved long after the cells had been activated (up to 101 days), these cells were poised to recall the distinct properties of their T helper specification. Regarding reactivation, when SMARTA memory T_FH cells were transferred into B cell deficient mice and followed by LCMV infection, these T_FH cells were able to maintain their T_FH phenotype (35–45% by day 10) while the frequency of CXCR5⁺ effector T_FH-like cells were diminished (<10% by day 10) in the transferred naive T_FH cell group (80). Taken together, these results show that memory T_FH cells can facilitate GC formation as they persist, remain committed to their lineage, and even maintain a distinct epigenetic program.

Around the same time, Hashimoto et al. studied the DNA methylation changes in antigen specific memory CD4⁺ T cells compared to naïve CD4⁺ T cells from OVA specific mice and found that a majority of DMRs were located outside promoter regions, were enriched in cytokine-associated genes, but did not always correlate with gene expression (81). Their findings are in line with other studies on DMR localization, methylation status of Th-specific cytokine loci, and demethylated poised profiles in memory (65,82), demonstrating the importance of epigenetic programs in antigen specific CD4⁺ T cell memory formation.

As we have discussed above, epigenetic regulation is critical to the maintenance of naive, effector and memory programs in CD4⁺ T cells. In naive CD4⁺ T cells, there is enrichment of repressive histone marks (H3K27me3) and DNA methylation at effector-associated loci. Following activation, CD4⁺ T cells undergo dramatic changes in their epigenetic landscape including histone acetylation and permissive histone marks (H3K4me3). Despite different CD4⁺ T cell lineage subsets (Th1 and T_FH) involved and gene expression overlap (IL-21), epigenetic profiling enabled the identification of distinct memory formation of both LCMV specific T_FH and Th1 cells. These studies show the utility of epigenetic profiling in fine-tuning our understanding of the diversity of CD4⁺ T cell differentiation.

5. Impact of epigenetics on T cell therapeutics

Given the exquisite specificity and capacity for T cells to acquire memory from a previously encountered cognate antigen, such understanding is now being widely exploited in various immunotherapeutic strategies. Although not exhaustive, such strategies include monoclonal antibody blockade of inhibitory receptors with their ligands, thus blocking their interaction and mitigating such inhibitory signals (referred to as ‘immune checkpoint blockade’ [ICB]) (83). Additional strategies involve genetically modifying a patient’s own T cells with an engineered receptor called a chimeric-antigen receptor (CAR), which allows for the T cells to recognize new antigens (namely, proteins expressed on cancer cells) and receive co-stimulatory signals directly from the CAR (84). Unfortunately, not all patients respond to ICB therapy and CAR T cell approaches. Given this, it is imperative to identify cell intrinsic and extrinsic mechanisms which limit the T cell response to improve upon existing T cell immunotherapy approaches (7). Below we will discuss the recent work from our lab and others which underscore the critical role of T cell epigenetics and explore potential strategies to improve immunotherapies.

In PD-1 checkpoint blockade a subset of T cells have the capacity to become rejuvenated, but prior to our studies it was questioned whether PD-1 blockade could reverse epigenetic programs that causally reinforce T cell exhaustion. Ghoneim et al. observed that despite treatment with anti-PD-L1 antibody, exhausted T cells had little to no changes in their methylation profiles at effector associated loci (60). Thus (as mentioned above), we explored the consequences of de novo DNA methylation on PD-1 blockade responsiveness and demonstrated that Dnmt3a deficient CD8⁺ T cells lacked the heritable-exhaustion methylation programs found in WT T cells, and, importantly, did not only have superior effector function, but were more responsive to PD-1 blockade relative to their WT counterparts (60). Overall, these data provide evidence that de novo methylation is a critical mediator of the acquisition of exhaustion-associated programs. Unfortunately, these data also demonstrate that despite responding to PD-1 blockade, such T cells will not gain durable differences in exhaustion-associated DNA methylation programs acquired during either the effector or post-effector time-points. When comparing the DMRs of the PD-1 blockade treated versus untreated controls, less than 2% of the DMRs (84 out of 5964) were Dnmt3a-mediated programs, and, compared to functional memory, less than 1% of treated-group DMRs were found in functional memory subsets. Essentially, this showed that PD-1 blockade had no significant effect on DNA methylation programs. In line with the idea of stable epigenetic programs, none of the DMRs found were part of the exhaustion-associated DNA methylation program such as Ifng, Myc, Tcf7, and Tbx21 loci (60). Similarly, another study demonstrated that HIV-specific CD8⁺ T cells isolated from ECs and ART-patients were found to maintain a stable, demethylated PD1-locus even after years of sustained viral control (57), thus, supporting the concept that once acquired, these exhaustion-associated epigenetic programs are maintained long-term, with or without antigen control. In other words, disruption of de novo methylation may be necessary to disrupt irreversible terminal exhausted programs.

In addition to the Dnmt3a knockout model, we demonstrated that cells pre-treated with the demethylation agent decitabine (DAC) can synergize with PD-1 blockade to modulate stable exhaustion epigenetic programs (60). These data support the rationale for a combinational therapeutic approach which includes either targeted erasure of DNA methylation marks and/or blockade of de novo DNA methylation with ICB therapy. Other studies have also investigated the possibility of exploiting manipulation of epigenetic mechanisms in combination of existing therapies. Zhang et al. treated T_EX cells with histone deacetylase inhibitors prior to adoptive transfer into mice. Importantly, these experiments found that these previously exhausted CD8⁺ T cells did go on to form protective memory (85). Fraitta and colleagues had success with a CD19 CAR T cell in a patient with chronic lymphocytic leukemia where it was identified that TET2 deficiency led to an altered epigenetic state, and thus altered T cell fate, that translated into anti-tumor activity followed by complete remission (86). Taken together, these studies highlight that modulating epigenetic mechanisms (histone acetylation, de novo methylation, and demethylation) has significant therapeutic potential to improve treatment in patients which do not respond to ICB treatment; further, these strategies may also improve upon the existing limitations of CAR T cell therapy, namely, persistence and sustained function. Similar to what was observed in TET-deficient CAR T cells, in which effectiveness was discovered after-the-fact, we have developed tools to assess cellular therapeutic potential of T cell-based therapies in advance of the costly generation and infusion into patients. As mentioned above, using our novel MPI, we now have the capacity to predict whether a patient’s T cells will most likely lead to a responsive versus nonresponsive outcome using epigenetic correlates of response (Figure 2).

As the field continues to gain new insights into T cell exhaustion, such studies have highlighted the further demarcation of CD8⁺ T cell lineage development during chronic versus acute antigen exposure. Here, through the work of many laboratories, we have learned that stem-like PD-1⁺Tcf1⁺Tim-3⁻ cells give rise to the ICB-responsive, transitory PD-1⁺Tim-3⁺CXC3R1⁺ cells before differentiating into the ICB-refractory, PD-1⁺Tim-3⁺CD101⁺ exhausted population (87–89) (Figure 2). Alternatively, a more recent study showed that Tcf1⁺ Tim-3⁻ and Tcf1⁻ Tim-3⁺ T cell populations develop early in both acute and chronic viral infections with the Tcf1⁺ population giving rise to the pool of exhausted T cells (90). Jadhav and colleagues examined the chromatin accessibility of these PD-1⁺CXCR5⁺Tim-3− stem-like cells against the PD-1⁺CXCR5⁺Tim-3⁺ exhausted T cell populations. They found that the stem-like population were enriched for TF motifs from the HMG, RHD, and TBX families, while the exhausted population were enriched for ETS and Runx motifs (91), although based on their exhaustion markers were unable to differentiate between the transitory-effector and terminally exhausted bifurcation as defined by CD101 expression. They noted that the stem-like population did have a unique epigenetic profile which differed from memory and effector populations, but there was no evidence of a bivalency in their exhausted PD-1⁺CXCR5⁺Tim-3⁺ population to distinguish presence of the more recently identified transitory population from the terminally exhausted subset. Further analysis of chromatin histone modifications and DNA methylation are needed to uncover the intrinsic functional programing, or lack thereof, within these defined CD8⁺ T cell subsets. We have discussed here the stable epigenetic landscape that renders T_EX cells unable to re-unleash functional programs or allow reinvigoration by ICB, as well as some studies showing the efficacy of modulating epigenetic mechanisms to re-establish functional abilities in T cells.

Conclusion

From the collective work described above it has become clear that histone modifications and DNA methylation play a critical role in enabling T cells to acquire long-lived gene expression programs that allow a cellular adaptive immune response to form immunological memory. Notably, new insights into T cell differentiation gained from a focus on epigenetic mechanisms have helped clarify the lineage relationship between effector and memory T cells, reconciling the previously thought-to-be “contradictory” multipotent and effector T cell properties. With the incorporation of genome-wide epigenetic profiling approaches to address open questions surrounding the developmental relationship between naïve, effector and memory T cells, a surprising breadth of epigenetic reprogramming events has been observed to occur during memory T cell development. This collective body of work has now cemented the importance of epigenetic regulation in T cell differentiation and identified specific cellular fates that are coupled to distinct epigenetic programs.

In addition to providing new insights into long-standing questions surrounding the fundamental mechanisms of memory T cell development, elucidation of specific epigenetic programs and the enzymes that catalyze such programs has begun to enable novel methods for determining response to T cell based therapy and improve clinical outcomes by engineering specific T cell fates. Efforts to use epigenetic profiles to predict T cell differentiation states are now being embraced as a way to further assess correlates of clinical outcome in therapeutic settings that utilize both endogenous and adoptively transferred T cells. Our development of the DNA methylation-based MPI serves as an example of how mechanistic insights into epigenetic programs that limit T cell developmental potential can be translated into biomarkers of clinical outcome. Furthermore, approaches that attempt to select for or induce specific epigenetically-reinforced T cell differentiation states are now being explored as a way to improve upon the efficacy of T cell-based immunotherapies. While great strides have been made in identifying, and therapeutically exploiting, the molecular mechanisms regulating T cell differentiation, it is likely that the advances described in this review represent just the beginning of what promises to be an exciting stage in clinical and basic immunological research.