Cellular and molecular waypoints along the path of T cell exhaustion

Abstract

Thirty years of foundational research investigating molecular and cellular mechanisms promoting T cell exhaustion are now enabling rational design of T cell-based therapies for the treatment of chronic infections and cancer. Once described as a static cell fate, it is now well-appreciated that the developmental path towards exhaustion is comprised of a heterogeneous pool of cells with varying degrees of effector potential that ultimately converge on a terminally differentiated state. Recent description of the developmental stages along the differentiation trajectory of T cell exhaustion has provided insight into past immunotherapeutic success and future opportunities. Here, we discuss the hallmarks of distinct developmental stages occurring along the path to T cell dysfunction, and the impact of these discrete CD8⁺ T cell fates on cancer immunotherapy.

One-sentence summary:

Defining discrete cellular subsets of T cell exhaustion is enabling new approaches for improving T cell-based immunotherapies.

Introduction

In scenarios of optimal priming and antigen clearance, CD8⁺ T cells can establish immunity that provides a host with life-long protection against repeated pathogenic challenge. Development of immunological memory is mediated in part by the CD8⁺ T cells undergoing a developmental adaptation that is tailored to the duration and tropism of the pathogen. Upon rechallenge, memory T cells have a heightened capacity to recall effector functions relative to their original naïve progenitor. This remarkable immunological feature has served as the basis for many current vaccines and T cell-based therapies designed to control tumors and protect against disease relapse (1). While these observations provide a clear rationale for employing CD8⁺ T cells in the context of immunotherapy, there remains tremendous work to establish long-term immunological protection for most cancer patients. One clear and significant barrier limiting the durability of T cell responses in therapeutic settings is the development of T cell exhaustion(2). Defining the cellular and molecular steps that govern the generation of functional memory and exhausted T cells is needed to better understand and overcome this hurdle.

In order to replicate the process of memory T cell differentiation for therapeutic purposes, research has focused on defining the steps necessary for a naïve CD8⁺ T cell to transition into a potent effector and subsequently into a long-lived memory T cell. Substantial effort has gone into detailing the steps involved in priming a naïve CD8⁺ T cell by antigen-presenting cells (APCs), which then results in clonal expansion and provides the host with a numerically large quantity of cytolytic effector cells. Notably, a small subset of these effector cells retains the potential to further differentiate into long-lived quiescent memory T cells that have acquired effector gene expression programs and remain poised to mount a robust recall response (3, 4) (Figure 1). However, in settings of sustained antigen exposure, the effector response is diverted away from functional memory T cell differentiation to an alternative exhaustion developmental trajectory. Generation of ‘exhausted’ T cells was initially described in the lymphocytic choriomeningitis virus (LCMV) murine model system due to the ability to directly compare antigen-specific T cells generated in response to acute versus chronic infection(5–7). These initial observations revealed that chronically stimulated T cells did not simply die, but rather they became impaired in their ability to proliferate and express effector cytokines. Further investigation of the features defining exhaustion demonstrated that limited proliferation was coupled to the expression of inhibitory receptors (IRs) that restrain TCR signaling (8–10), which facilitated the development of immune checkpoint blockade (ICB) therapy that enhances endogenous T cell responses using a blocking antibody that disrupts the IR-ligand interaction(11). Notably, only a subset of chronically stimulated T cells responds to ICB, which has motivated recent efforts to further dissect the heterogeneity that exists among the pool of exhausted T cells at transcriptomic, phenotypic and epigenetic levels(12–14). Recent insights into the progression of T cell exhaustion are guiding three general modalities of T cell immunotherapy design to improve the potency and persistence of immunological protection: i) preventing T cell exhaustion; ii) delaying or diverting T cell terminal exhaustion prior to fate commitment; iii) reversing exhaustion after fate commitment (Figure 1). In this review, we will summarize recent advances in understanding the cellular and molecular mechanisms modulating distinct developmental stages of T cell exhaustion in chronic viral infection and cancer. We then will discuss the therapeutic obstacles and opportunities for targeting T cell exhaustion in cancer.

Figure 1. A progressive model depicting the stages of CD8⁺ T exhaustion.

CD8⁺ T cell differentiation post-activation is a process whereby T cells acquire specialized functions that are adapted to the duration of antigen (Ag) exposure through the transitioning of a continuum of cellular states. The first major bifurcation in CD8⁺ T cell developmental path is to become a terminal effector, a committed cell fate with little differentiation potential (1), versus becoming a precursor T cell that retains the developmental plasticity to further differentiate into long-live memory T cell (Tmem) (during acute Ag exposure) or exhausted progenitor T cell (Tpex) (during chronic Ag exposure) (2). The adaptation of CD8⁺ T cells to chronic stimulation starts from Tpex by stably expressing Tox and PD-1 promoting cell survival while enforcing progeny T cells with dampened proliferative and killing capacity. In contrast, the progenies of Tmem have heightened effector potential. The fates of Teff-like cells derived from Tpex at later chronic infection stages are determined by the presence of CD4⁺ T cells. In the absence of CD4⁺ T cells, Tpex-derived Teff-like subset (Tim3⁺CD101⁻) is a transitory state between Tpex and terminal Tex (3). However, CD4⁺ T cells can divert effector-like (CX3CR1⁺) T cells to a terminally differentiated state with superior cytotoxicity (4). Both Teff-like subsets serve as the source of cytotoxicity of the exhausted T cell pool. This overview of CD8⁺ T cell differentiation provides a rationale for therapeutically targeting T cell exhaustion in chronic infection and cancer, among which major directions are diverting T cells away from exhaustion lineage into more potent effectors, preserving Tpex and/or transitory effectors to delay T cell terminal exhaustion, and reprogramming exhausted T cells after fate commitment.

Defining the developmental path to T cell exhaustion

When exposed to chronic antigens, CD8⁺ T cells gradually reduce their effector response to balance immunological protection and immunopathology. This progressive repression of T cell function results in a spectrum of T cell developmental potential ranging from the ability to mount a therapeutic response to being fully indolent. To achieve optimal efficacy of T cell-based immunotherapies, it is critical to define the point at which CD8⁺ T cells commit to the exhausted cell lineage. Initial insight into the progressive changes T cells undergo during the process of exhaustion came from longitudinal analysis of endogenous and adoptively transferred T cells using the murine model of clone 13 LCMV chronic viral infection. LCMV-specific CD8⁺ T cells displayed a progressive suppression and hierarchical loss of effector polyfunctionality characterized by initial loss of IL-2 expression, followed by TNFa, and ultimately IFNg (15). Moreover, virus-specific CD8⁺ T cells isolated from mice 8 days post-infection retained greater memory potential compared to the CD8⁺ T cells isolated at later stages (16). Despite prolonged rest in uninfected mice prior to secondary challenge, CD8⁺ T cells isolated after the peak of the effector response to chronic LCMV infection retained phenotypic and functional hallmarks of T cell exhaustion (17). This progressive loss of developmental plasticity among chronically stimulated T cells was also observed in an inducible murine liver tumor model where tumor-specific CD8⁺ T cells from pre and early malignant lesions were able to restore effector function and acquire a memory phenotype after transfer into antigen-free hosts. In contrast, T cells isolated after 30 days of tumor implantation failed to reacquire effector potential and retained high expression of multiple IRs regardless of PD-1 blockade(18). Collectively, these pioneering studies indicate that the delineation between a functional memory versus an exhausted state occurs early in the immune response and is reinforced through continued TCR stimulation and exposure to immunosuppressive tissue environment.

Upon demonstration that exhaustion is a discrete developmental pathway, further work focused on how non-proliferating, terminally differentiated T cells can persist for years after initial antigen encounter. Subsequent studies have established that the pool of exhausted T cells is sustained by a stem-like progenitor that can self-renew but also give rise to terminally differentiated T cells during chronic antigen exposure. Conceptualization of this progenitor/daughter cell relationship emerged from studies examining the role of Tbet and Eomes during T cell effector differentiation in chronically infected mice (19). Subsequent efforts to better resolve the developmental relationship among exhausted T cell subsets revealed Tcf1 and CXCR5 as the hallmarks of this stem-like progenitor. Notably, the presence of such Tcf1⁺ progenitors is responsible for the proliferative burst in response to ICB therapy and positively correlated with therapeutic outcomes in a variety of cancers(12–14, 20–22). Recent advances in single cell sequencing have enabled researchers to simultaneously track the transcriptome, epigenome, and TCR repertoire of chronically stimulated T cells (23). Such efforts have comprehensively mapped T cell adaptations to chronic antigens as well as enabled better prediction of a T cell’s developmental potential in a specific therapeutic setting. It is now clear that during murine chronic LCMV infection, the developmental bifurcation between the path toward progenitor and effector T cells can be detected as early as day 5 after infection. While this initial branch point is conceptually similar to the bifurcation of effector and memory precursors present in acutely resolved infections, these exhausted progenitor T cells (Tpex: PD-1⁺Tox⁺Tcf1⁺Tim3⁻), as well as their progeny including terminally exhausted T cells (termTex), retain higher expression of PD-1 and the transcription factor Tox (22, 24, 25). Heterogeneity exists within the pool of Tpex; relative to the CD62L⁻ subset, Tpex expressing CD62L were shown to be the bona fide exhausted progenitors responsible for the proliferative burst upon PD-1 blockade during LCMV infection(26). Distinct progenitor subpopulations were also reported in a murine lung carcinoma model where Ly108⁺Tcf1⁺CD8⁺ T cells retained a greater proliferative capacity and ability to infiltrate the tumor relative to the Ly108⁻Tcf1⁺ population(27). Similarly, the Ly108⁺CD69⁻ T cell subset appears to be more proliferative and has a greater propensity for entering the blood compared to the Ly108⁺CD69⁺ quiescent population in chronically infected mice(28).

While Tpex were initially identified as being enriched for cells committed to a dysfunctional state, it is now apparent that Tpex are more multipotent than previously believed. Evidence of their intrinsic developmental potential came from studies assessing the various effector T cell subsets that are derived from Tpex. A recent study examining CD8⁺ T cell responses to chronic LCMV infection in the absence of CD4⁺ T cells described the development of a transitory Teff-like subset (Tim3⁺CD101⁻) from Tpex that ultimately progressed to a terminally exhausted state (Figure 1) (29). A similar linear differentiation model (Tpex→Teff-like→termTex) was proposed in a study examining high-grade serous ovarian cancer(30). However, in the presence of CD4⁺ T cell help, Tpex-derived effector-like CX3CR1⁺ T cells were driven into a distinct lineage from termTex distinguished by lower PD-1 expression yet higher Teff gene expression including Tbet and Zeb2 (31). Despite the acquisition of a heightened effector program, this Teff-like CD8⁺ T subset retained a partial exhaustion signature denoted by Tox expression. Another two independent studies documented a similar effector subset enriched for a natural killer signature including selective expression of killer lectin-like receptors (KLRs) (32, 33). This Tex-KLR signature was also enriched in a CXCR3⁺ CD45RA⁺ effector memory (Temra) CD8⁺ T cell population in a variety of human cancers(33). The developmental plasticity of Tpex under appropriate signaling conditions was further demonstrated by recent preclinical studies showing enhanced cytotoxicity of Tpex progeny in the therapeutic response of anti-PD-L1 combined with IL-2 agonist (34, 35). These proof-of-principle studies also serve as a rationale for developing new combination therapies that leverage our understanding of the receptors and cellular partners regulating the effector potential of exhausted T cell subsets.

The well-defined developmental path of exhaustion discussed above was largely borne out of preclinical murine models where the progression of T cell exhaustion starts from a naïve state. However, the developmental starting point of human tumor-reactive T cells may not always be naïve, and importantly, autologous T cells used to generate products for adoptive cellular therapy (ACT) are largely derived from memory T cells(36). This biological disconnect between preclinical and clinical settings has prompted investigators to ask to what extent memory T cells contribute to the pool of exhausted T cells and whether a ‘memory-origin’ model of T cell exhaustion might be more relevant (37, 38). Parallel to human ACT studies, investigations of human tumors have identified a variety of functional memory phenotypes, including central-memory (Tcm) and effector-memory (Tem) among endogenous T cells, further raising the question of whether these cells are tumor-reactive and if they undergo exhaustion(39, 40). TCR sequencing of such tumor-infiltrating T cells combined with transcriptional profiling has identified shared TCRs among Tcm, Tem, and Tex, suggesting a trajectory of Tcm (IL7R^hi)→Tem (Gzmk⁺)→Tex as a common path to terminal exhaustion in human cancer(41). Additional ‘memory-origin’ studies of exhaustion have come from investigating intratumoral tissue-resident (Trm) memory T cells. It is worth noting that human tumor-infiltrating Trm are typically defined by expression of cell surface markers such as CD103 and CD69, rather than specific markers associated with developmental stage, indicating Trm can encompass both progenitor and effector-like exhausted T cells. While virus-specific and tumor-irrelevant Trm were observed in human cancers, T cells co-expressing tissue retention and exhaustion phenotypic markers can be tumor-reactive and predictive of therapeutic outcomes(42, 43). Consistent with these observations, a differentiation trajectory of Tcm→Trm (ZNF683⁺CXCR6⁺)→Tex has been inferred as another path to exhaustion in various types of cancer(41). Virally induced tumors represent another human setting in which the development of virus-specific tumor-infiltrating T cells is of interest. HPV-specific stem-like T cells were indeed detected in human-papilloma virus-associated head and neck cancers and were suggested to undergo exhaustion(44). However, whether these cells were derived from the primary viral infection remains to be determined. Further work is needed to develop a better working model to understand ‘memory-origin’ exhaustion.

Anatomical partitioning and migration of exhausted T cell subsets

A key observation from the studies described above highlighting the developmental path toward terminal T cell exhaustion is that the surrogate markers used to define many of exhausted T cell subsets are tissue homing or retention molecules, suggesting that developmental transitions occur at discrete anatomical locations. Anatomical partitioning of the T cell adaptive response generally begins with a naïve T cell encountering its cognate antigen through interaction with APCs that capture antigens from peripheral tissue and bring them to specific regions of lymphoid organs. Following this priming event, T cells downregulate lymphoid-homing molecules and upregulate tissue-infiltrating receptors to enable trafficking to the source of the antigen where they undergo further differentiation. Early investigation into the anatomical partitioning of functional and dysfunctional antigen-specific T cells found that tumor-infiltrating antigen-specific T cells were largely unable to mount a recall response when stimulated ex vivo relative to tumor-specific T cells isolated from lymphoid tissues (45). The developmental relationship between functional cells in lymphoid tissue and dysfunctional cells at the source of the antigen became clear upon observation that the progenitor subset was enriched in lymphoid organs whereas terminally exhausted T cells reside in both lymphoid and peripheral tissue during chronic LCMV infection (12, 13). Further insight into the anatomical distribution of different subsets came from assessment of the origin of tumor-infiltrating T cells after ICB therapy (46–48). These studies suggest that T cells responding to PD-1 blockade in cancer patients can be derived from lymphoid tissues and circulation, which has contributed to the theory that Tpex residing in lymphoid tissue serve as a self-renewing population that can traffic to and subsequently undergo terminal differentiation in tumors thus replenishing the intratumoral pool of exhausted T cells. A direct assessment of this hypothesis revealed that blocking the migration of lymphoid-residing T cells into circulation substantially limited the number of Tpex in tumors (21, 22, 27, 46, 49). This replenishment model was further strengthened by a recent report that observed T cells trafficking through tumor-associated high endothelial venules (TA-HEVs) using intravital microscopy (50). Such TA-HEV formation increased the abundance of progenitor CD8⁺ T cells in tumors ultimately enhancing the response to combination ICB therapy. Notably, a recent study investigating CD8⁺ T cell responses in human head and neck squamous cell carcinoma patients found that Tpex were enriched in regional lymph nodes relative to tumors and were able to mount a response to anti-PD-L1(51). However, this ICB responsiveness of Tpex was disrupted in metastatic lymph nodes. From these studies, it appears that a protective niche distinct from the tumor microenvironment is needed to preserve Tpex to sustain an anti-tumor response.

After tissue or tumor infiltration, T cells are faced with critical decisions involving residency and further differentiation. While both effector-like and terminally exhausted T cells acquire tissue-residency programs that positively correlate with anti-tumor response(52), it is less clear whether Tpex become resident or may egress from the tumor after infiltration. A recent study tracking movement of in vivo labeled tumor-infiltrating T cells demonstrated that PD-1⁺Tcf1⁺CD8⁺ T cells can recirculate between the tumor periphery and tumor-draining LNs(53). This ability to transition between tumor and lymph node was likely attributed to higher expression of lymphoid-homing molecules on Tpex. In line with this notion, tumor-infiltrating T cells were also found to exit from tumors through tumor-associated lymphatic vessels via CXCR4/CXCL12, thus limiting the quantity of tumor-reactive T cells(54). Another recent study investigating the differentiation trajectory of stem-like tumor-infiltrating lymphocytes (TILs) in human ovarian cancer reported that tissue residency was a hallmark of TIL tumor reactivity and only the presence of tumor-resident stem-like progenitor T cells positively correlated with disease outcome(30). In the chronic LCMV infection murine model, the Ly108⁺CD69⁻ Tpex subset was shown to retain the capacity to recirculate (28). However, results from a parabiosis experiment performed at the late stage of chronic LCMV infection demonstrated that PD-1⁺Tcf1⁺ cells remain resident in lymphoid organs rather than recirculate (55). One possibility is that a quiescent resident subpopulation within the heterogenous Tpex is enriched as disease progresses, whereas a migrating subset contracts due to sustained inhibitory signals occurring later during chronic disease. In addition to the residency decision after tissue infiltration, CD8⁺ T cells undergo adaptations specific to each tissue during development of exhaustion (33, 56). Distinct distribution patterns among exhausted T cell subsets have been reported during chronic LCMV infection. In a recent study, lung, spleen and blood were found to be the major sites for CX3CR1⁺ effector-like subsets, whereas terminally exhausted subsets (CX3CR1⁻CXCR6⁺) were mostly found in the bone marrow and liver(56). Whether such differences are predetermined prior to tissue infiltration or develop within tissues remains unclear.

In addition to their migration between lymphoid and non-lymphoid tissues, exhausted CD8+ T cell subsets are positioned in specific niches within tissues. Notably, CXCR5⁺ Tpex are maintained in the white pulp of the spleen and are redirected to the red pulp in a CXCR3/CXCL10-dependent manner for terminal differentiation (57). Parallel to their discrete positioning within lymphoid tissues, Tpex were found to be enriched among tertiary lymphoid structures or T cell-APC zones in several human solid tumors, and the establishment of this immune niche correlated with improved clinical outcomes(58–60). More than just being positioned proximally to Tpex, cDC1 have also been shown in murine chronic LCMV and lung tumor models to be essential for preserving Tpex in the spleen and tumor-draining lymph nodes respectively(27, 61). Furthermore, in a murine melanoma model, CCR7⁺DCs were densely populated in the perivascular niche of the tumor stroma where they engaged with Teff-like cells via CXCR6/CXCL16, promoting the survival of effector-like CD8⁺ T cells(62). Rather than remaining inert within tumors, exhausted T cells have been reported to actively recruit monocytes into the tumor microenvironment via CCL3/4/5 secretion, which subsequently promotes monocyte differentiation into tumor-associated macrophages (TAM) that can prime T cells and further promote exhaustion (63). Collectively, these results detail the interactions between exhausted T cell subsets and other immune cells along the developmental path to terminal exhaustion (Figure 2).

Figure 2. Anatomical partitioning of exhausted T cell subsets.

In chronic viral infection, Tpex are maintained in the white pulp of spleen by cDC1 and redirected through CXCR3 and CXCL9/10 signaling to the red pulp where they undergo further differentiation into transitory effectors and ultimately terminal exhausted T cells. The distribution of distinct exhausted subsets also varies across tissues. CX3CR1⁺ effector-like subsets are mostly found in the lung, spleen and blood while terminal exhausted subsets (CX3CR1⁻CXCR6⁺) are enriched in the liver, where the phenotypes of antigen-specific T cells are the most homogeneous. In the setting of tumors, Tpex are generated and maintained in tumor-draining lymph nodes, which serve as a reservoir that sustains intratumoral Tpex. A critical entry gate to extravasate into tumors is via tumor-associated high endothelial venules (TA-HEV). Once arriving at tumor sites, CD8⁺ T cells interact with a variety of antigen-presenting cells and tumor-associated macrophages (TAM). Specifically, APC-T cell zones can be found in tumors and are critical for the maintenance of tumor-infiltrating Tpex. CCR7⁺DCs are indispensable for the survival of CXCR6⁺ effectors in tumors. Exhausted T cells can secret chemokines such as CCL3/4/5 to actively recruit monocytes and shape their differentiation into suppressive TAM, promoting the terminal exhaustion of CD8⁺ T cells. CXCR4⁺ Tpex can exit tumor via tumor-associated lymphatic vessels (TA-LV) and traffic to tumor-draining lymph nodes.

A diminishing return from sustained TCR stimulation

The spatial and temporal stages that demarcate the generation of exhausted CD8⁺ T cells incorporate signals from TCR, co-stimulation and cytokines through a well-orchestrated series of cellular encounters. How each of these signals contributes to the developmental transition between progenitor and terminally exhausted T cells provides critical insight into the lineage relationship among T cell subsets and the downstream signals governing their transition between stages. Among these signals, there is uniform acknowledgement that TCR signaling is critical for the development of exhaustion. An early study investigating the mechanisms of exhausted T cell maintenance during chronic infection reported that exhausted T cells failed to persist when adoptively transferred into naïve hosts or those infected with an epitope-mutant virus(64). This antigen-dependent maintenance of exhausted T cells was later refined by examining the requirement of TCR engagement at distinct stages of exhausted T cell differentiation. Specifically, adoptive transfer studies showed that Tpex can survive for long periods of time in the absence of antigen(17). Furthermore, re-exposure to cognate antigen is necessary and sufficient to initiate Tpex differentiation into terminally exhausted T cells (22). While these studies have shown Tpex maintenance is independent of antigen, the impact of chronic TCR engagement on Tpex self-renewal capacity remains to be fully elucidated.

Transcriptional profiling of phenotypically defined exhaustion subsets isolated from murine viral and tumor model systems revealed that the Teff-like and terminally exhausted T cells were enriched for a TCR signaling signature, whereas this signature was less apparent in Tpex, raising questions about the strength of TCR signaling distinct subsets experience (25, 49). A recent investigation examining TCR signal strength using a Nr4a-GFP reporter in P14 transgenic T cells shed new light on this question. While the P14 CD8⁺ T cells initially underwent rapid induction of TCR signaling, inhibition of the TCR signal subsequently occurred following PD-1 ligation, which could be successfully mitigated by PD-1 blocking antibody treatment (65). Consistent with this observation, TCR signaling was also increased following anti-PD-1 treatment in a murine MC38 tumor model(66). Similar to these preclinical murine systems, correlative studies from melanoma patients treated with anti-PD-1 therapy identified a strong TCR gene signature positively correlated with response (66). These results suggest that while a T cell’s cognate antigen can be present throughout the course of chronic infection or cancer, the magnitude of TCR signaling experienced by the pool of exhausted T cells varies. Assessment of TCR signal strength occurring among individual subsets has shown that during chronic LCMV infection Tpex experienced substantially stronger TCR signaling compared to terminal exhausted T cells(26). These observations, along with the recently defined role of cDC1 for preserving Tpex in an MHC-I-dependent manner(61), imply a potential role of TCR stimulation in Tpex maintenance during chronic antigen exposure.

The strength of TCR stimulation is determined by the binding avidities between TCR and cognate peptide-MHC. The contribution of TCR-peptide-MHC affinities on the kinetics of T cell activation and differentiation has been extensively studied under acute antigen challenge but remains to be investigated in chronic disease settings. In murine lung carcinoma and fibrosarcoma models where affinity-altered TCRs and altered peptide ligands were employed to tune TCR-pMHC affinities respectively, TCRs with a lower affinity promoted a stem-like phenotype (PD-1⁺Tcf1⁺) while higher affinity TCRs promoted terminal exhaustion in tumors(67, 68). However, it is important to note that the cells with a stem-like phenotype described in these studies were unable to generate Teff-like cells or mount a proliferative burst in response to PD-1 blockade, raising the question of whether T cells activated by a low-affinity interaction develop into bona fide Tpex. High antigen levels have been shown to promote an exhaustion phenotype in CD8⁺ T cells during chronic infection(69); however, further refinements to this generalization have been suggested by multiple studies. For example, PD-1 blockade in settings of suboptimal TCR stimulation were found to promote T cell terminal exhaustion(70). In line with this, a low neoantigen burden has been coupled to a heightened development of terminal exhaustion in tumor-specific CD8⁺ T cells, and higher viral titers during chronic LCMV infection promoted the formation of Tpex versus Teff cells (24, 26, 71). Two additional studies suggest that lower affinity TCR-pMHC interaction denoted by lower Gp33 tetramer fluorescent intensity skews CD8⁺ T cells towards a KLRG⁺ Teff-like phenotype while higher affinity TCR clones are biased toward a terminal exhaustion trajectory but also enriched in Tpex(72, 73). Future investigations into the role of TCR signal strength at discrete developmental stages of T cell exhaustion, along with the anatomical sites where these differentiation stages occur, can provide a better understanding of its impacts on exhausted T cell lineage specification and maintenance.

Developmental fine tuning via co-regulatory receptors

Important to note, many of the above studies interrogating the dynamics of TCR signaling and cellular trafficking relied on clonal TCR transgenics. Using engineered T cells with a single TCR specificity suggests that the observed population dynamics cannot be accounted for by heterogeneity among the TCR repertoire and highlights the importance of co-stimulation and cytokine signaling. A broad network of costimulatory and IRs has now been well-described that either enhance or blunt T cell activation. This combination of activating and/or repressing signals depends on the ligands available on professional APCs and non-conventional antigen presenters such as tumor or stroma cells. While the role of co-stimulatory receptors in T cell priming has been intensively studied, their function in subset specification during the developmental of exhaustion is not fully understood. Costimulatory ligand-receptor pairs encompass a large variety of members including the well-characterized CD28 and tumor necrosis factor receptor (TNFR) superfamilies. Emerging evidence describing an enrichment of CD28 and ICOS among CXCR5⁺Tcf1⁺ CD8⁺ T cells has prompted further investigation into the role that co-stimulatory receptors play in the transition between distinct exhaustion states and the impact these receptors have on CD8⁺ T cell responses to PD-1 blockade(74). Specifically, specific deletion of CD28 in CD8⁺ T cells substantially compromised their proliferation after PD-1 blockade. Consistent with these findings, lung cancer patients treated with PD-1 therapy had a dominant population of circulating CD28⁺ T cells (74, 75). Based on the observation that ICB-responsive populations arise from a less differentiated state, CD28 could be critical in promoting the terminal differentiation of Tpex to generate Teff-like cells and/or amplify Teff-like subsets, which has been recently demonstrated in tumor settings(76).

Further insight into the regulatory effects of co-stimulatory receptors on CD8⁺ T cell exhaustion came from a phenotypical and functional assessment of TILs after combinatory ICB and co-stimulatory agonist treatment. In murine ovarian and breast cancer models, the combination of inhibiting PD-1 and activating the co-stimulatory receptor GITR enhanced the effector potential of exhausted T cells and promoted their progenitor phenotype in a CD266-dependent manner(77). Additionally, OX40 plays a critical role in antigen-specific CD8⁺ T cell survival through the Bcl2/Bcl-xl axis in both chronic LCMV infection and tumors (78, 79). Results from a phase Ib clinical trial in which the anti-OX40 agonist antibody was used to treat head and neck squamous cell carcinoma patients showed an increase in CD103⁺CD39⁺CD8⁺ T cells in tumor after treatment(80). These results suggest that OX40 may play a role similar to CD28 in enhancing the differentiation of Tpex into terminal Tex. While the co-stimulatory receptors described above are mostly enriched in Tpex, 4-1BB was found to be upregulated in terminally differentiating PD-1^hi CD39⁺ CD8⁺ TILs(81, 82). Nonetheless, 4-1BB⁺ TILs retain greater proliferative potential than 4-1BB⁻ TILs and enrichment of 4-1BB related gene signature was associated with HCC patient survival(81). Indeed, 4-1BB agonistic antibodies enhanced the performance of CD8⁺ TILs synergistically with anti-PD-1(82). It is noteworthy that these co-stimulatory receptors are not restricted to CD8⁺ T cells and many of them also play a critical role in Treg activation(83). Thus, it remains to be fully elucidated whether the effects observed from various broadly acting antibody treatments are CD8⁺ T cell-intrinsic and their specific roles in the generation of Tpex.

While persistent TCR stimulation is sufficient to induce functional exhaustion in vitro, generation of exhausted T cells in vivo involves additional signaling from co-inhibitory pathways that can contribute to the generation of distinct phenotypes. Sustained expression of multiple IRs is one of the primary adaptive strategies that serve to counterbalance TCR and co-stimulatory signals and prevent activation-induced cell death during chronic antigen exposure(84). However, acutely activated T cells also upregulate IRs transiently prior to memory development, and IR expression in chronic diseases can be decoupled from T cell dysfunction(85, 86). For example, CD8⁺ T cells deficient in the intracellular zinc chaperones metallothioneins exhibit superior effector function and tumor control while maintaining a high level of IRs(85). Among these IRs, the role of PD-1 in the development of T cell exhaustion has been well characterized. As discussed above, PD-1-mediated inhibition of T cell effector potential was initially described in chronically infected mice that were treated with PD-L1 blocking antibodies, while subsequent studies genetically disrupted the gene encoding PD-1 in CD8⁺ T cells. This resulted in enhanced clonal expansion during the effector stage of chronic antigen response, suggesting a role for PD-1 in limiting early terminal effector T cell development while also skewing T cells toward the exhaustion lineage(87, 88). Consistent with the concept that PD-1 signaling regulates the generation of Tpex, the quantity of exhausted PD-1-deficient T cell numbers declined much faster than wildtype T cells over time despite their increased expansion during the effector stage of the immune response(88, 89). The greater proliferative and effector response of Tpex subset following PD-1 blockade indicates that PD-1 signaling prevents the terminal differentiation of Tpex(26, 90). Notably, other IRs including CD83, CD160, CD200r1, CD200r2 and Lag3 are also enriched among Tpex relative to termTex, implying a yet-to-be-defined role for these receptors in regulating Tpex development and maintenance.

Cytokine-mediated specification of exhausted T cell subsets

In addition to the signal input from antigen and co-regulatory receptors, CD8⁺ T cells respond to the environment through para- and autocrine cytokines that either amplify or dampen a response. Originally described as a T cell growth factor, Il-2 has now become well appreciated as a cytokine utilized by many immune cells to support their survival. IL-2 is an FDA-approved immunotherapy for treating certain types of cancers due to its superb ability to expand T cells in combination with TCR stimulation(91). Predominantly secreted by activated CD4 T cells, IL-2 signals through IL-2R consisting of IL-2Ra (CD25), IL-2Rb (CD122) and IL-2Rr (CD132). Together, these three subunits associate to form the high-affinity version of IL-2R. Early during acute infection, CD25^hi activated T cells are prone to differentiating into terminal effectors while the CD25^lo subset develops into memory T cells(92). Building on these observations, early studies in a chronic disease setting indicated that while low-dose IL-2 treatment partially reversed the exhaustion phenotype of CD8⁺ T cells, there was minimal effect on virus clearance. However, combined with anti-PD-1, IL-2 synergistically enhanced T cell effector potential and consequently improved virus control(93). Closer inspection of the exhausted T cell subsets underlying such synergistic effect revealed that Tpex express high levels of CD25 and are believed to be the primary IL-2-responsive subset(34). A similar effect was observed in tumor models using optimized IL-2R agonists in combination with PD-1 inhibition(35, 94). These results suggest the co-dependence of TCR and IL-2 signaling in driving the differentiation and expansion of Tpex-derived Teff-like T cells to their full potential during chronic antigen exposure. Consistent with this, prolonged IL-2 signaling alone has been shown to promote T cell exhaustion in tumors, and an IL-2 agonist lowering STAT5 activation enhanced the stemness of chimeric antigen receptor T (CAR-T) cells and improved tumor control(95, 96). Sharing the IL-2Rb and IL-2Rr chains, IL-15R also has a high affinity form that contains IL-15Ra. DCs, a major source of IL-15, can trans-present IL-15 via IL-15Ra to promote the survival and homeostatic proliferation of memory T cells(97, 98). Given that IL-15 signaling supports the generation and maintenance of memory T cells, IL-15 is now a prominent target to enhance CAR-T cell stemness(99, 100). Notably, delivery of IL-15 linked to anti-PD-1 enhanced anti-tumor efficacy while limiting systemic toxicity in murine system(101). An expansion of the Teff-like cells following this PD-1-cis-IL-15 treatment indicates a potentially similar mechanism as employed by a PD-1-cis-IL-2 agonist.

Counterbalancing the effects of pro-inflammatory cytokines described above, anti-inflammatory cytokines also play a role in the development of T cell exhaustion, which has become more appreciated. Early studies in the chronic LCMV infection model demonstrated that IL-10 suppressed the anti-viral T cell response and prolonged infection(102, 103). Such immunosuppression by IL-10 was also observed in tumor settings(104, 105). However, a protective role of IL-10 has been proposed from studies demonstrating that IL-10(R) deficiency was negatively associated with tumor control while IL-10 supplement enhanced the anti-tumor response(106). Furthermore, we recently reported that IL-10 signaling was critical for the increased functionality of Dnmt3a-KO CAR T cells(107). These results suggest that IL-10 cannot be exclusively described as an inhibitory cytokine, and further investigation into subset-specific regulatory roles it plays during T cell exhaustion is warranted. A recent study in a murine chronic lymphocytic leukemia model demonstrated that IL-10R signaling can contribute to the preservation of Tpex(108). Additionally, a half-life extended IL-10 agonist antibody enhanced the expansion and effector potential of terminally exhausted T cells by promoting oxidative phosphorylation in the B16-F10 tumor model (109).

Previously described as an immunosuppressive cytokine during chronic infection and tumor development(110, 111), it is now appreciated that the impact TGFb has on T cell differentiation is stage-dependent. Recent studies revisiting the role of TGFb signaling in newly defined exhausted subsets identified a critical function in preserving Tpex through lymphoid tissue retention and as a regulator of mTOR activity(112–115). It has been observed that TGFb’s inhibitory features promote Tpex formation during priming while compromising the effector potential of Teff-like cells that derive from Tpex(114, 115). Similarly, type 1 interferon (IFN-I) has been reported to play an opposing role in distinct stages of exhaustion during chronic viral infection. In studies where IFN-I was administered early during infection, T cell expansion was enhanced, ultimately contributing to improved viral control(116–118). However, treatment with IFN-I at a late stage of infection promoted CD8⁺ T cell terminal exhaustion that has been recently shown to be mediated by IFN-induced transcription factor (Irf2) (119). Further investigations into the CD8⁺ T cell-intrinsic role of IFN-I signaling demonstrated that the formation of Tpex is suppressed by IFNAR1-mediated signaling(120). Together, these findings highlight the importance of inhibitory signaling through IRs and cytokines in the maintenance of exhausted progenitors and providing long-term protection against chronic antigens.

Transcriptional and epigenetic rewiring of exhaustion programs

Stable adaptation by a CD8⁺ T cell to the multitude of signals described above occurs in the context of a genetically fixed background. This heritable maintenance of the T cell’s newly acquired phenotype and function prompted investigations into whether intrinsic changes to gene regulation underly the long-lived nature of adaptive T cell responses. Transcriptional profiling of T cells in settings of acute and chronic antigen exposure documented that naïve, effector, memory and exhausted T cells express vastly distinct gene expression signatures that are coupled to the specific function of each T cell subset(121, 122). Specifically, relative to Teff and memory T cells, exhausted T cells are enriched for IR transcripts and have lower levels of transcripts associated with TCR signaling such as Lck. Additionally, genes involved in effector function and cytokine signaling, migration, and metabolic pathways were altered at a transcriptional level in exhausted T cells. The initial description of the striking difference in gene expression profiles delineating functional and exhausted T cells motivated a shift in research focus toward defining transcriptional regulators involved in the lineage specification and maintenance of exhaustion. Here, we summarize the role of TFs that play a critical role in the development of exhausted T cells (Table 1) and briefly highlight several recent discoveries that have provided insight into the transcriptional changes that occur during transitions between exhausted T cell subsets.

Table 1.

Transcriptional regulation of CD8⁺ T cell differentiation

TF	T cell subset regulated by TF (Acute)	T cell subset regulated by TF (Chronic)	Model (Chronic)	Mechanism (Chronic)	References (Chronic)
Bach2	↑Memory	↑Tpex	LCMV cl-13	↓Terminal differentiation, possibly via Runx3/Prdm1	(129)
Batf	↑Effector	↑Tpex to Teff-like; ↓Exhaustion (CAR-T)	LCMV cl-13 murine tumor	↑Enhancer accessibility of effector-related regions (Tbx21, Klf2); ↑NFAT-AP1 activity	(140, 144)
Bcl6	↑Memory	↑Tpex	LCMV cl-13	↓Blimp1	(120)
Bhlhe40	↑Trm	↑Effector	Murine tumor	↑Metabolic and epigenetic fitness	(173)
Blimp	↑Effector	↑Effector/Tex	LCMV cl-13, human tumor	Expression level-dependent	(174, 175)
c-Jun	↑Effector	↑Effector	Tumor xenograft	↓AP-1–IRF complex	(172)
c-Myb	↑Memory	↑Tpex	LCMV cl-13, murine tumor	↑Tcf1, CD62L ↓Zeb2	(26, 89, 176)
c-Myc	↑Cell division ↑Effector	↑Exhaustion	Murine and human tumor	Metabolic reprogramming	(148, 177)
E2A	↑Memory	↑Tpex	LCMV cl-13	↑Cxcr5	(13, 178)
Egr2	CD8+ intrinsic effect not clear	Enriched in Tpex; Tex maintenance	LCMV cl-13	↑IR expression	(127)
Ets1	↑Effector	Locus accessible in Tex	LCMV cl-13	Unknown	(139)
Fli1	↓Effector	↓Effector	LCMV cl-13, murine tumor	↓Chromatin accessibility at ETS-RUNX sites	(128)
Foxo1	↑Memory	↑Tpex	LCMV cl-13	↓T-bet	(179, 180)
Foxp1	Unknown	↓Proliferation and effector	In vitro, murine and human tumor	Smad2/3-mediated c-Myc repression; TGFb-induced c-Jun repression	(181)
Hif1a	↑Effector	↑Effector	LCMV cl-13, murine tumor	↑Glycolytic metabolism	(182–184)
Id2	↑Effector	↑Tex	LCMV cl-13	↓E2A	(13, 185–188)
Id3	↑Memory	↑Tpex; ↑Exhaustion (CAR-T)	LCMV cl-13 /docile, in vitro, Murine tumor	↑Cell survival
Ikzf2	Enriched in naïve and memory	Enriched in Tpex; ↓Effector (CAR-T)	LCMV cl-13, in vitro	Unknown	(125, 189, 190)
Irf4	↑Effector	↑Tex	LCMV cl-13, murine tumor	↓Tcf1	(172, 191, 192)
NFAT	Nfat1-↑Effector Nfat2-↑Memory	↑Exhaustion	LCMV cl-13, HIV	NFAT-Nr4a-Tcf1; nuclear translocation; bind PD-1/Tim3	(193, 194)
Nr4a	↑Early activation	↑Exhaustion	LCMV cl-13, murine and human tumor	NFAT-Nr4a/AP1	(195, 196)
Runx1	Redundant to Runx3	↓Effector	LCMV cl-13	Unknown	(128)
Runx2	↑Memory	Enriched in Tex	LCMV cl-13, human tumor	Unknown	(139, 197)
Runx3	↑Effector ↑Trm	↑Effector	LCMV cl-13, murine tumor	↓Tcf1, Bcl6 and CXCR5	(128, 198)
Satb1	↓Effector program in naive	↓Exhaustion	In vitro, murine and human tumor	↓Pd1	(199)
Smad4	↑Activation and ↑effector	↑Activation and effector	Murine tumor	Myc-mediated proliferation; TCR-induced nuclear translocation	(200, 201)
Tcf1	↑Memory	↑Tpex	LCMV cl-13, HIV, murine and human tumor	↑Wnt/Catenin, ↑Bcl6; chromatin remodeling	(12, 14, 89, 120)
Tox	Cytokine-induced expression in Tmem	↑Survival ↑Exhaustion	LCMV cl-13, murine and human tumor, human HBV	↑IR expression; epigenetic reprogramming	(86, 123–125, 202)
Tbet	↑Teff	↑Teff-like	LCMV cl-13, human melanoma, HIV, AML	Tbet/Eomes ratio; subcellular location	(203, 204) (19, 205)
Eomes	↑Memory	↑Tpex and termTex
Xbp1	↑Effector	↑Exhaustion	LCMV cl-13, human tumor	↑IR expression; ↓mitochondrial activity	(206, 207)
Zeb2	↑Effector	Enriched in Teff-like	LCMV cl-13	Unknown	(31)

↑ denotes positive regulation; ↓ denotes negative regulation; Teff: effector T cell, Tex: exhausted T cell, Tmem: memory T cell, Tpex, exhausted progenitor T cell, Trm: tissue-resident memory T cell, CAR: chimeric antigen receptor, IR: inhibitory receptor, LCMV: lymphocytic choriomeningitis virus,

The characterization of Tox and its function during CD8⁺ T cell differentiation in settings of acute and chronic stimulation documented that Tox is highly expressed throughout the exhausted T cell lineage relative to effector and memory T cell subsets(86, 123–125). Ectopic expression of Tox in T cells reinforced the expression of IRs and supported T cell survival, shown by a failure of exhausted T cells to persist when Tox function was disrupted. Similarly, Egr2 was also found to be upregulated in exhausted T cells responding to chronic infection or tumors, and specifically enriched in the Tpex subset. These studies found that Egr2 was critical for maintaining exhausted T cells through a transcriptional mechanism that directly enhanced IR expression (126, 127). Notably, in many cases, TFs promoting a specific cell fate also restrict alternative lineage potential. For example, Tcf1 reinforces Tpex formation but also limits the initial effector T cell differentiation(89). Similarly, deletion of Fli1 promotes effector differentiation while partially compromising the development of the progenitor subset(128). This fate choice of Teff versus Tpex has provided a therapeutic entry point for approaches that target this developmental bifurcation to generate a robust initial effector response at the expense of preserving Tpex capable of sustaining an effector response. Furthermore, transcriptional mechanisms controlling the transition of Tpex into Teff cells highlight the delicate balance between the maintenance of Tpex versus the replenishment of effectors. While elevated expression of Bach2 increased the quantity of Tpex, it also blocked their terminal differentiation by preventing transition to an effector state, which likely explained the observed lack of viral control in mice(24, 129). It is now well-appreciated that a deeper understanding of the transcriptional regulators controlling the transition between Tpex, Teff and Tex can enable future T cell therapies tailored to the disease context requiring either a ‘fast and furious’ effector response or a ‘slow but steady’ maintenance strategy.

The highly dynamic transcriptional reprogramming that ensues during the formation of memory and exhausted T cells necessitates a mechanism to stabilize the newly acquired cellular programs and functions. To account for the transcriptional heritability among T cell subsets, epigenetic rewiring events have been explored. Prior to studies examining T cell exhaustion, work investigating T helper cell fate decisions demonstrated an association between chromatin accessibility of the loci for genes that were traditionally used to define the helper subsets (130). Extension of this concept to CD8⁺ T cells provided additional support for defining T cell exhaustion as a committed cell fate. As a model to explore epigenetic stability, DNA methylation profiles of Pdcd1 promotor were examined in both murine and human settings of acute and chronic viral infection. These studies were the first to show that while CD8⁺ T cells maintain epigenetic plasticity during acute infection, these programs become reinforced during chronic infections(131, 132).

The stability of such programming events across the broader genomic landscape was subsequently analyzed in murine tumor and chronic viral settings (133–136). ATAC-seq profiling demonstrated that T cell exhaustion was associated with thousands of regions that were uniquely accessible, including IRs, Il-10 pathway mediators, and negative regulators of cytokine production and NF-kb activity. These differentially accessible regions in exhausted T cells exhibited features of enhancers that were closely associated with subset-specific transcriptional activity and function. Accessible TF motifs acquired in exhausted T cells include those for BATF, Eomes, Fos, Nr4a, IRF and NFAT, whereas TF motifs for Tcf1, Lef1 and Foxo1 were enriched in functional memory T cells, suggesting that the distinct transcriptional programming of functional versus exhausted T cells are epigenetically imprinted. These exhaustion-associated chromatin accessibility profiles appear to remain in a fixed state since very few changed in response to PD-1 blockade, a finding that was further supported by the correlation between the chromatin states and cellular plasticity of tumor-specific CD8⁺ T cells observed in a murine tumor setting. Several recent studies have further highlighted this epigenetic “scarring” of a previous exhaustion state whereby the Tox, Batf and Hif1a loci remain accessible even after the removal of antigen for months (25, 137, 138).

Extending the concept to exhausted T cell subsets, Tpex, Teff-like and Tex are distinctly rewired in their epigenetic programs (22, 24, 32, 33, 139–141). Specifically, open chromatin regions enriched in termTex involve regulatory pathways regulating cytotoxicity, apoptosis and cell cycle, while Tpex have increased chromatin accessibility at gene regions regulating IL-2, IL-7, Wnt and NF-kb signaling. When compared to Tpex and termTex, the TF motifs that are more accessible in Teff-like cells include those for RUNX2, T-bet and KLF3. The TFs critical for subset specification are also differentially imprinted at an epigenetic level among exhaustion subsets. For example, Id3, Tcf7 and Satb1 loci are more accessible in Tpex while Id2, Hif1a and Nfkb2 remain inaccessible relative to Tex. In addition to being epigenetically controlled for their expression and accessibility of their binding sites, TFs also instruct cell fate commitment by specifying epigenetic remodeling events during the progression of exhaustion. Several studies support this concept, showing that TFs such as Tcf1, Tox, and Bhlhe40 can remodel the epigenetic landscape of T cells in response to chronic antigen exposure(142, 143). Recently, BATF was demonstrated to not only act at the transcriptional level but also modulate the enhancer accessibility of Teff-like associated genes (e.g., Cx3cr1, Gzmb, and KLRs) and TF targets (ETS, RUNT and IRF) during Tpex to Teff-like transition(140, 144). Together, these findings highlight the interplay between transcriptional and epigenetic regulation of T cell differentiation.

The chromatin remodeling events that occur during the development of exhaustion are a result of multiple covalent modifications to DNA and histones. Genome-wide changes in H3K4me3 and H3K27me3 at promoter regions and H3K27ac at enhancer regions have been observed during the lineage specification of exhausted T cell subsets(125, 140). Specifically, Tpex signature genes Tcf7 and Id3 acquire a permissive pattern of histone modifications (H3K4me3^Hi H3K27me3^Low; H3K27ac^Hi H3K27me3^Low), which are later replaced with suppressive marks (H3K4me3^LowH3K27me3^Hi; H3K27ac^Low H3K27me3^Hi) during the transition of Tpex into termTex. In contrast, Teff-like-related genes such as Cx3cr1, S1pr5, Klrg1, and Zeb2 acquire transcriptionally permissive marks (H3K4me3^Hi H3K27me3^Low) during the transition of Tpex into Teff-like cells. However, genes such as Fasl, Gzmb, and Ccl5 remain in a poised state in Tpex based on histone methylation patterns (H3K4me3^Hi and H3K27me3^Hi). Notably, it was reported that a significant proportion of genes with permissive histone marks do not correlate with transcript expression in termTex TILs, while anti-PD-1 or anti-4-1BB treatment can upregulate the transcription of these anticorrelated genes in a murine tumor model(145). Similar to altered histone modifications, broad changes also occur in DNA methylation that accompany the functional decline of T cells as they transition from the effector to exhaustion stage of an immune response(146). The unmethylated Ifng locus in effector T cells became hypermethylated during prolonged antigen exposure. Additionally, the Tcf7 locus was fully methylated at the late stage of chronic infection, supporting the hypothesis that progressive epigenetic modifications may reinforce restriction on the developmental plasticity of CD8⁺ T cells as they undergo exhaustion.

More than just correlates of the various transcriptional states along the developmental trajectory, epigenetic modifications have now become well-appreciated as determinants of exhausted T cell fate. This causal relationship was supported by studies demonstrating that DNA-methyltransferase 3A (Dnmt3A)-KO CD8⁺ T cells retain greater proliferative capacity and effector potential compared to wildtype cells(146). Whole-genome DNA methylation analysis showed that Dnmt3A downstream targets included genes associated with homing (Sell and Ccr7) and stemness (Lef1 and Tcf7) that were hypermethylated in wildtype exhausted T cells, consistent with Dnmt3A-KO T cells exhibiting properties associated with less differentiated T cells. In addition to preserving effector function during chronic viral infection, Dnmt3a deletion also preserved the T cell’s ability to mount a proliferative response during ICB. Together with ATACseq results from exhaustion studies, these data imply that epigenetic programs reinforce the transition between the ICB-responsive Tpex population and the ICB-nonresponsive terminal population. Other mediators of DNA methylation have also been found to limit T cell function. Specifically, tet methylcytosine dioxygenase 2 (Tet2) was inadvertently disrupted in CD19 CAR T cells administered to a CLL patient who subsequently underwent complete remission. It was concluded that Tet2 disruption was responsible for the durable anti-tumor T cell response and epigenetic preservation of CAR T cell function(147). These findings collectively demonstrate the importance of DNA methylation in reinforcing T cell exhaustion and illustrate the therapeutic potential of epigenetic reprogramming to limit T cell dysfunction, which has motivated further investigations into the role of other epigenetic regulators on CD8⁺ T cell differentiation in chronic disease settings (Figure 3).

Figure 3. Translating epigenetic programming into cancer immunotherapies.

During the terminal differentiation of naïve or stem-like progenitor into effectors, CD8⁺ T cells acquire de novo epigenetic programs where stemness-associated loci are repressed through DNA methylation and chromatin remodeling. Similarly, effector-associated loci are suppressed during the development of exhaustion. Blocking the acquisition of the epigenetic programs limiting T cell stemness and effector potential is a major therapeutic opportunity. In adoptive cell therapy approaches, autologous or allogenic T cells can be genetically engineered or modified using chemicals that disrupt the catalytic function of epigenetic enzymes. Examples of ‘epigenetic blockade’ of exhaustion include deletion of Dnmt3a or Tet2 to maintain progenitor T cells, which results in heightened tumor control. LSD1 deletion in LCMV-specific CD8⁺ T cells also promotes enrichment of Tpex and a better response to anti-PD-1. Inhibiting Suv39h1 activity or its deletion enhances effector capacity and tumor killing. cBAF components, Arid1a and Smarcd2 also contribute to exhausted CD8⁺ T cell differentiation and could represent therapeutic targets.

Recently, CRISPR screens identified Arid1a, a component of the chromatin remodeling cBAF complex, as a critical player in memory T cell formation and T cell exhaustion(148, 149). Notably, Arid1a-KO T cells were refractory to the acquisition of Tex-associated chromatin accessibility. In addition to cBAF, several other chromatin modifiers have been recently examined in the context of memory T cell differentiation and exhaustion. Brd4, which recognizes acetylated lysine residues on histones, has been shown to promote the terminal differentiation of CD8⁺ T cells in murine tumors(150). Additional studies exploring the role of histone modifications in regulating CD8⁺ T cell function have further established these epigenetic mechanisms as critical determinants of the cell fate steps along the path to exhaustion. Inhibition of histone deacetylase (HDAC) activity using small molecules impaired tumor growth, which was coupled to enhanced proliferative and functional capacity of CD8⁺ T cells(151). Ex vivo inhibition of HDAC activity in adoptively transferred CD8⁺ T cells increased IFNg expression during chronic LCMV infection(152). Furthermore, tumor-infiltrating CTLs in the colon tumor microenvironment have been reported to upregulate H3K9me3-specific histone methyltransferase SUV39H, which repressed Gzmb, Prf1, Faslg, and Ifng expression in CD8⁺ T cells. Suppression of these effector-associated molecules was hypothesized to enable cancer immune escape (153). Accordingly, SUV39H-KO CD8⁺ T cells prevented H3K9me3 of the effector loci allowing for a better response to PD-1 blockade in B16 tumor model (154). Deletion of enhancer of Zeste 2 Polycomb Repressive Complex 2 (EZH2) in Pmel CD8⁺ T cells was incapable of inhibiting tumor growth due to reduced IFNg production in late staged tumors compared to wildtype cells, implying the persistence of CD8⁺ T cells may be controlled by EZH2 (155). Additionally, the histone demethylase LSD1 acts to enforce an epigenetic program in progenitor exhausted CD8⁺ T cells that antagonizes Tcf1-mediated progenitor maintenance and to promote terminal differentiation(156). Taken together, histone and DNA epigenetic modifications reinforce the developmental transition between Tex subsets, and insight into these mechanisms can be leveraged to improve the efficacy of various immunotherapy strategies (Figure 3).

Translating subsets into therapy

As new insights into the exhaustion differentiation trajectory become incorporated into approaches to enhance and extend T cell-targeted therapy, an equally important question that has implications for T cell persistence and function remains: how do we optimize tumor-specificity without compromising the durability of the T cell effector response? Typically, tumor-specific T cells were identified through culturing TILs with autologous tumor cells or lysates, with the goal of isolating an expanded population of T cells. In cases where tumor-associated antigens (TAA) and neoantigens have been identified, tumor-reactive T cells can be selected by specific peptide stimulation or HLA-multimer-based cell sorting. Yet, the low affinity and density of tumor antigens presented on MHC, as well as the compromised capacity of exhausted tumor-reactive T cells to respond to peptide stimulation, often limit the sensitivity of these approaches. As an alternative to in vitro expansion approaches, other investigators have relied on enrichment of a T cell exhaustion signature as a surrogate for identifying tumor-specific clonal expansion(157, 158). While many TILs infiltrate the TME in a bystander manner, studies have identified human tumor-reactive T cells by the expression of CD39(159, 160). Recently, a deeper investigation into the tumor-specificity of TILs in human metastatic cancers demonstrated that a majority of neoantigen-specific T cells acquired exhausted T cell programs including enrichment of Tox, CXCL13, CXCR6 and IRs, whereas virus-specific TCR clones were largely found in Tem and Trm populations(161). However, strategies relying on a restricted exhaustion phenotype may inadvertently exclude progenitor T cells subset due to their lower expression of exhaustion markers in certain human tumors, which may ultimately impact the durability of ACT response. A recent report showed Cxcl13 can identify both tumor-reactive progenitor and terminal exhausted T cell subsets(162). It remains to be determined how the balance of progenitor and effector-like populations can be achieved through the selection of specific T cell clones derived from TILs.

Combining the insights into T cell exhaustion that were borne out of ICB studies with lessons learned from tumor-reactivity analyses, these concepts are now being incorporated into the development of T cell-based therapies with engineered specificities. CAR engineering endows polyclonal T cells with a new antigen specificity that targets tumors in an MHC-unrestricted manner. Characterization of CAR T cells documented that these engineered cells indeed acquire hallmarks of exhaustion in ex vivo culturing settings (163). For example, during ex vivo expansion of human HA-28z CAR T cells, genome-wide changes in chromatin accessibility and three-dimensional chromosome conformation occurred, which preceded expression of exhaustion-related genes (164). Extending these observations to an in vivo setting, longitudinal epigenetic profiling of CD19 CAR T cells isolated from treated B-ALL patients further confirmed that CD19 CAR T cells progressively enrich for causal regulators of T cell exhaustion(165, 166). Specifically, DNA methylation programs that resemble the recently described human Tpex subset and that also directly limit anti-tumor responses in preclinical settings, were acquired within the first several weeks of clinical response. Importantly, these cellular and molecular hallmarks of exhaustion are associated with an inability of the CAR T cells to mount a recall response after the re-emergence of antigen-expressing cells.

Different approaches that leverage our understanding of cellular and molecular mechanisms regulating the development of exhaustion are being pursued to enhance CAR T cell anti-tumor responses by preserving cellular plasticity. For example, including MyD88 and CD40 costimulatory endo-domains maintained CAR T cells in a less differentiated state which enhanced their proliferative capacity and antitumor activity (167). Other groups have attempted to limit CAR T cell terminal differentiation through calibration of the chimeric receptor’s immunoreceptor tyrosine-based activation motifs (ITAMs), thereby preventing overactivation and preserving functional potential(168). Along similar lines, strategies have focused on preventing the generation of CAR T cells residing in a stem-like state. Using a wide range of human CAR T cell model systems, genetic disruption of Dnmt3a in CAR T cells prior to their chronic antigen exposure prevented the acquisition of bona fide exhaustion DNA methylation programs(107). More importantly, the CAR T cells retained a capacity to proliferate and mount a sustained anti-tumor response in difficult-to-resolve tumor or rechallenge settings. Similarly, treating CAR T cells with the hypomethylating agent decitabine resulted in an improvement in anti-tumor and proliferative function coupled to lower exhaustion-associated gene expression (169).

Given that epigenetic reprogramming approaches have the potential to sustain T cells in chronic antigen environments, how these approaches impact mechanisms controlling cellular proliferation is an area of active investigation. A recent study found that disruption of Tet2 resulted in BATF3-induced clonal expansion that was independent of antigen(170). Despite susceptibility to genome instability, Tet2-KO CAR T cells did not exhibit hallmarks of leukemic transformation (171). Additionally, Tet2 disruption in CAR T cells in a treated patient showed no evidence of substantial outgrowth over a four-year period(147). While current approaches to enhance T cell anti-tumor function do not appear to be drivers of malignancy, the use of such strategies to treat refractory disease should proceed with caution. Modulating transcriptional factor activity also represents a viable approach to boosting CAR T cell function. CAR T cells engineered to overexpress c-Jun exhibited superior expansion and effector potential along with less terminal differentiation, which resulted in better tumor control in multiple preclinical tumor models(172). Together, these studies demonstrate that a better understanding of the developmental stages underlying T cell exhaustion and the molecular mechanisms controlling this differentiation process can identify cellular and molecular barriers impeding the translation of T cell-based therapies for treating cancer and chronic viral infections.

Concluding Remarks

Recent advances in understanding the lineage specification of exhausted T cells have provided a more granular view of how CD8⁺ T cells adapt to chronic exposure to viral and tumor antigens. Translating these insights into therapies holds great promise for advancing efforts to leverage the immune system to provide durable control of cancer and chronic infections. As the discrete stages of T cell exhaustion are becoming clearer, mechanism-driven approaches to improve T cell immunotherapies now include preventing, delaying, or reversing the developmental transitions toward a terminal state (Figure 1). Given that a T cell’s current differentiation state determines its future developmental trajectory, therapeutic approaches should embrace the tremendous heterogeneity within each patient’s total pool of T cells. In order to advance T cell-based immunotherapy research and design, several outstanding questions remain to be addressed, including: 1) Which epigenetic modifications are critical for preserving the chromatin landscape of exhausted CD8⁺ T cells? By defining the crosstalk between DNA methylation and histone modifications, we are likely to identify new entry points for intervening in the exhaustion trajectory. 2) With single cell technologies more readily available, which molecular programs should we use to define T cell developmental states and how will this inform therapeutic interventions? 3) Are there discrete cellular and molecular regulatory mechanisms that can be used to further delineate the process of T cell exhaustion from senescence, and what implications can these provide for the design of immunotherapies for young and aged individuals? Addressing these questions may help usher in the next generation of T cell-based therapies and expand our understanding of T cell dysfunction in chronic diseases.