Rewriting history: Epigenetic reprogramming of CD8⁺ T cell differentiation to enhance immunotherapy

Abstract

The full potential of T cell-based immunotherapies remains limited by a variety of T cell extrinsic and intrinsic immunosuppressive mechanisms that can become imprinted to stably reduce the antitumor ability of T cells. Here we discuss recent insights into memory CD8⁺ T cell differentiation and exhaustion and the association of these differentiation states with clinical outcomes during immune checkpoint blockade and CAR T cell therapeutic modalities. We consider the barriers limiting the efficacy of immunotherapy with a focus on epigenetic regulation impeding efficacy of adoptively transferred T cells and other approaches that augment T cell responses such as immune checkpoint blockade. Further, we outline conceptual and technical breakthroughs that can be applied to existing therapeutic approaches and to the development of novel cutting-edge strategies.

The past and present of T cell immunotherapy

Immunotherapies are being embraced as an integral component of cancer treatment. Instead of following protocols using a standardized chemotherapeutic approach, we are moving toward individualized treatment modalities which include using a patient’s own immune system to fight against disease. Despite immunotherapy options being on the cutting edge of present day cancer treatments, the concept of using the patient’s immune system to combat disease is far from novel (Box 1). Early efforts to better define the cellular partitioning of immunological properties were spearheaded by Drs. Max Cooper and Jacques Miller, who made the fundamental discovery that the adaptive immune system is composed of two types of lymphocytes, B and T cells [1]. T cells were subsequently identified as possessing the killing capacity to combat malignant cells which facilitated modern therapeutic approaches involving the rational design of T cells to control tumor growth.

Box 1: The Dawn of Immunotherapy.

Foundational immunotherapy approaches began as early as the late 1800s with Dr. William Coley who noted that inducing a fever in patients could lead to tumor regression. After reviewing medical records, Coley learned about a patient who had an inoperable sarcoma of the neck that resolved after he developed the streptococcal infection erysipelas. The hypothesis that the infection activated the immune system to treat cancer led Coley to inject his first patient with streptococcal organisms in 1891. Shrinkage of the patient’s malignant tumor prompted him to treat other sarcoma patients and ultimately led to the development of Coley’s Toxin, a combination of heat-killed streptococcal organisms with Serratia marcescens [90, 91]. Despite criticism due to treatment inconsistencies and variations in outcome, Coley’s pioneering work laid the foundation for harnessing the potential of the immune system to treat tumors.

Exemplary immunotherapeutic options to augment cancer-specific T cell responses include immune checkpoint blockade (ICB) and adoptive cellular therapies such as chimeric antigen receptor (CAR) T-cell therapy. ICB works by disrupting ligation of T cell inhibitory receptors to reverse T cell suppression and restore immunological function [2–7]. The first checkpoint inhibitor, ipilimumab (cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade), was FDA approved in 2011 for treatment of melanoma [8] and was followed shortly thereafter by the approval of programed cell death receptor 1 (PD-1) inhibitors such as pembrolizumab and nivolumab in 2014 [9, 10]. Patients with advanced melanoma, who historically faced a dismal prognosis, can now be offered these cutting-edge treatment options. However even checkpoint blockade offers suboptimal outcomes in these patients, given that the overall survival (OS) has been 52% at five years post-treatment with the combination of nivolumab and ipilimumab [11]. Another milestone for immunotherapy occurred in 2017 with tisagenlecleucel, the first FDA approved CAR T-cell therapy for the treatment of B-cell acute lymphoblastic leukemia (ALL) in children and young adults with relapsed or refractory disease, a patient population previously facing limited options and a grim prognosis. With the use of tisagenlecleucel, the rate of event-free survival was 50% and OS was 76% at 12 months post treatment, indicating that while CAR T-cell therapy is promising, further improvements are necessary to increase patient survival [12]. Failure to achieve a sustained response in a proportion of the patients treated with immunotherapeutic approaches raises questions about the underlying cellular and molecular mechanisms that lead to the discrepant treatment outcomes. Consequently, there has been a specific focus in defining the cell intrinsic and extrinsic factors that limit therapeutic T cell responsiveness.

While there have been great successes, the current limitations that we now face with T cell-based immunotherapies have served as a catalyst for the field to better understand the developmental drivers of T cell effector and memory properties, as well as how these programs are imprinted during T cell differentiation. Several research groups have begun to further identify the cellular subsets that contribute to the clinical response and define the cell fate-determining mechanisms that reinforce anti-tumor properties [13, 14]. While current T cell therapeutic approaches largely reflect a ‘one-size fits all’ strategy, future approaches may be able to exploit recent insights into T cell subset specification and cell-fate defining mechanisms to rationally tailor the T cell effector response, and meet the needs of individual diseases. Here, we describe how experimental exploration of the mechanisms governing mammalian CD8⁺ T cell effector and memory differentiation have been applied to the exploitation of various characteristics such as effector function, proliferation, tissue homing, and survival; we hypothesize how these approaches can be implemented in the rational augmentation of current T cell-based therapeutic approaches.

Erosion of T cell developmental plasticity during chronic stimulation

During an acute infection or vaccination, naïve antigen-specific CD8⁺ T cells undergo clonal expansion and differentiate into effector cells that are either terminally differentiated (terminal effectors) or cells that retain the potential (memory precursor effector cells) to further develop into long-lived memory T cells (functional memory) [15–17]. The branching point in this developmental process likely arises from a multipotent progenitor cell that is able to directly or indirectly give rise to the cell types depicted in Figure 1. While the majority (~95%) of effector CD8⁺ T cells are destined to die, the remaining ~5% survive the process of cellular contraction giving rise to the pool of memory CD8⁺ T cells (Box 2) [18].

Figure 1. Antigen stimulation is essential for establishing the progressive epigenetic states of T cell memory differentiation and exhaustion.

A naïve CD8⁺ T cell exposed to its cognate antigen undergoes differentiation which can result in several distinct T cell fates. In the context of acute stimulation, the naïve CD8⁺ T cell proliferates and clonally expands. Multipotent progenitor cells are able to give rise to any of the subsequent cell types. The vast majority of cells generated are terminal effectors (TE), which die fulfilling their duty of antigen clearance. After antigen clearance, a small subset of cells, described as memory precursor effector cells (MPEC – derived from a multipotent progenitor), survive the contraction stage of the immune response and develop into subsets of functional memory (FM) T cells [44]. These functional memory T cells remain poised to rapidly recall effector functions upon antigen reencounter. In the case of chronic stimulation, persistent antigen exposure results in an exhausted (Exh.) fate with T cells unable to illicit an effector response after antigen re-challenge [21]. The question mark indicates a need for further resolution of the branching point among the developmentally plastic cells that give rise to exhausted versus functional memory T cells. The general epigenetic state of effector and stem genes are represented at each of these stages of differentiation. White circles represent an unmethylated CpG and black circles represent a methylated CpG.

Box 2: Mammalian Memory T cell subsets.

Broadly, the memory T cell compartment is comprised of two main populations which can be defined by the expression of the lymph node homing molecules CD62L and CCR7. Central memory T cells (Tcm) exhibit a CD62L⁺ CCR7⁺ phenotype and access lymphoid tissue. Conversely, effector memory T cells (Tem) are CD62L⁻ CCR7⁻, reside in circulation, home to tissue, and are more cytolytic [92, 93]. recently, stem cell memory T cells (Tscm) were defined in humans based on cell surface expression of CD95 and CD122 [14, 94]. Human Tscm are similar to naïve CD8⁺ T cells in that they are CCR7⁺ and CD45RO⁻ but possess a heightened ability for self-renewal and can give rise to Tcm and Tem. Consistent with their differentiation hierarchy, the less differentiated Tscm exhibit a more sustained antitumor response than Tcm which are in turn more effective than the more terminally differentiated Tem [14].

The development of T cells with memory potential occurs in the setting of acute antigen exposure; however, sustained TCR signaling and inflammation occur during many chronic viral infections and cancers in humans and mice, limiting the differentiation potential of these cells [19, 20]. Persistent antigen exposure drives a decline in the T cell effector and memory potential to a terminal nonfunctional state of differentiation, often referred to as T cell ‘exhaustion’ [21]. The progressive development of CD8⁺ T cell exhaustion is now thought to occur linearly (Figure 1) [22]. Specifically, a subset of developmentally plastic T cells continue to undergo changes in their differentiation program which includes a gradual suppression of their capacity to give rise to other T cell subsets, and of their ability to rapidly recall effector functions [23, 24]. Not all T cells undergo similar levels of exhaustion with the less terminally differentiated cells remaining more functional in settings of chronic antigen exposure and serving as progenitors for terminal populations [24]. These states of T cell differentiation are often generalized as a pool, but are actually heterogeneous in terms of homing capacity, recall response, and self-renewal [21, 25, 26]. Reinforcement of these cell-fate processes suggests that an epigenetic mechanism is responsible for T cell exhaustion. In addition to epigenetic programs, specific transcription factors have been identified as essential regulators of the exhaustion program. For instance, the transcription factors TOX and NR4A have been shown to induce exhaustion programs in mouse and human model systems of chronic viral and tumor antigen settings [27–30]. Understanding the key components in establishing T cell exhaustion provides an opportunity for stably modifying cellular functions toward the generation of T cells with desired properties that may be ideal for cellular therapy.

Epigenetic maintenance of cell fates

Epigenetic modifications can be utilized to reinforce cell fate decisions of specific CD8⁺ T cell subsets by providing a mechanism to heritably propagate acquired gene expression programs in a dividing population of cells. Our group and others recently demonstrated a causal relationship between epigenetic programming and the maintenance of effector and memory-associated functions during T cell homeostasis to sustain immunity [31–34]. During the development of long-lived memory CD8⁺ T cells, activated naïve antigen-specific CD8⁺ T cells transition through the effector stage of differentiation enabling a subset of cells to acquire effector-associated programs prior to their continued development into memory CD8⁺ T cells in humans and mice [33, 35] (Figure 1). The transient exposure to effector-promoting signals imparts memory T cells with long-lived effector-associated epigenetic programs that endow them with a heightened ability to recall effector functions while retaining the naïve-like capacity to develop into other memory and effector cell types [32, 33, 36, 37]. The blend of naïve and effector properties among memory CD8⁺ T cells is reflected by their DNA methylation profiles being similar to both naïve and effector T cells, (Figure 1) further supporting the epigenetic characterization of cell fates [31, 33, 38].

In addition to establishing memory programs in the context of antigen clearance, epigenetic modifications contribute to the cell-intrinsic mechanism for heritable maintenance of exhaustion gene expression programs in antigen-specific CD8⁺ T cells from individuals with chronic viral infections and cancer [34, 39–47]. Previous studies using the chronic lymphocytic choriomeningitis virus (LCMV; Clone 13) model system have shown that once antigen-specific CD8⁺ T cells acquire exhaustion-associated gene expression and epigenetic programs, that can persist even after reduction of antigen load [34, 46]. DNA methyltransferase 3a (Dnmt3a) has been identified as the de novo methyltransferase responsible for implementing exhaustion-associated programs in CD8⁺ T cells in the setting of chronic antigen exposure [34]. Indeed, deletion of Dnmt3a (granzyme B cre × Dnmt3a^flox/flox) inhibits acquisition of de novo DNA methylation programs in murine antigen-specific CD8⁺ T cells and prevents T cell exhaustion [34]. Thus, inhibiting DNA methylation programs that are causal in establishing exhaustion is one strategy for improving T-cell based immunotherapies. However, successful application of this technique requires an understanding of the epigenetic changes that occur during memory T cell differentiation and exhaustion, in addition to a well-defined epigenetic signature to track during the manipulation of these cells.

Epigenetic definition for memory T cell differentiation

Historically, functional and phenotypic features of effector and memory T cells have been used to position CD8⁺ T cells along a differentiation hierarchy; however these ever-expanding features have begun to convolute our traditional collection of memory T cell hallmarks. Recently, several groups have utilized murine models of acute infection to examine epigenetic programs associated with effector and memory CD8⁺ T cell differentiation. Notably, these kinetic analyses have documented sequential modifications in DNA methylation during the progression from naïve, to memory precursor effector cells (MPEC), to memory CD8⁺ T cells [33]. Further, memory CD8⁺ T cell differentiation has been associated with extensive T cell subset-specific DNA methylation remodeling which imparts the cells with the capacity to rapidly re-elicit the effector response in the event that the cells re-encounter their cognate antigen [33, 48]. Collectively, insights gained from the above described epigenetic programming events involved in mouse and human memory CD8⁺ T cell differentiation have made it apparent that epigenetic modifications may be used as a universal definition for describing the differentiation status of a T cell. Toward this end, we have recently worked to establish an epigenetic atlas of both murine and human CD8⁺ T cells. Specifically, using naïve and exhausted CD8⁺ T cells as a bounds, we identified ~250 CpG sites that are predictive of the cell’s developmental potential and can therefore be used to delineate the differentiation status of CD8⁺ T cells [49]. Further, the collection of epigenetic modifications has been normalized based on the methylation of key CpG sites. Generation and application of this novel “T cell multipotency index” across species further supports the idea that DNA methylation serves as a conserved mechanism in defining memory T cell differentiation [49, 50] (Key Figure, Figure 2A). As we move toward the development of therapeutic strategies to instill T cells with desired properties tailored to treat specific diseases, our ability to also characterize the differentiation status of these T cells is crucial. Incorporation of epigenetic programs into a modified definition of T cell effector and memory differentiation will enable a richer assessment of engineered T cells and their therapeutic outcome. Below we further describe how insights into the processes of memory subset specification and T cell exhaustion have provided new insights into immune check point blockade (ICB) and the rational design of the next generation of CAR-T cell therapies.

Key Figure, Figure 2. Mechanisms of T cell subset specification and immunotherapy design.

A. Schematic depiction of how the DNA methylation-based T cell multipotency index can inform on clinical response to ICB and CAR T cell therapy. We speculate that determining the differentiation status of CD8⁺ T cells by using the multipotency index can identify a ‘therapeutic window’ of stem-like CD8⁺ T cells which are most likely to induce a clinical response in both ICB and CAR T cell therapy. FM, functional memory; TE, terminal effector; Exh, exhausted.

B. Diagram of epigenetic events regulating T cell responses during immunotherapy. DNA methylation programming that occurs during CD8⁺ T cell differentiation involves DNMT3A and TET2. We hypothesize that augmentation of T cells to enhance response to immunotherapy can occur by inhibiting DNMT3A-mediated de novo methylation of stem-associated genes and TET2-mediated demethylation of effector-associated genes.

C. Proposed approaches for using insights into epigenetic regulation of T cell differentiation to guide the design of T cell immunotherapy. Genetic engineering approaches should focus on preservation of CD8⁺ T cells in a stem-like state as characterized by a high MPI. We anticipate strategies for maintaining this stem-like state include inhibition of DNMT3A and TET2-mediated epigenetic programs.

Improving current approaches to immune checkpoint blockade

Acquisition of an exhaustion differentiation program in T cells occurs progressively during the prolonged exposure to antigen and is coincident with upregulation of inhibitory receptors including CTLA-4, PD-1, and T cell immunoglobulin and mucin domain-containing protein 3 (Tim-3). The development of T cell exhaustion specifically reduces the proliferative capacity of T cells and restricts their ability to recall the expression of effector cytokines and cytolytic molecules [19, 20]. While suppression of effector responses is necessary to mitigate overactivation of the immune system and prevent autoimmunity, it also leaves the host with a low quantity of antigen-specific CD8⁺ T cells that are unable to kill tumors, and is a major impediment for successful T cell-based immunotherapies [21, 26]. ICB acts to circumvent these immunological checkpoints and allow T cells to maintain effector functions [51]. As the original checkpoint inhibitor, the CTLA-4 antibody ipilimumab provided the conceptual foundation for additional checkpoint inhibitors. Given that CTLA-4 inhibition has been well described elsewhere [52, 53], our discussion focuses primarily on mechanisms regulating responses to PD-1 blockade.

Early anecdotal observations suggested that T cells with greater and sustained TCR signaling resulted in impaired T cell responses to ICB. These observations, as well as the need to better define prognostic indicators of clinical outcome have guided efforts to identify the cellular subsets that contribute to the clinical response during ICB (Key Figure, Figure 2A). The progressive development of T cell exhaustion results in a heterogeneous pool of antigen-specific T cells that can be parsed into cellular subsets roughly based on their surface expression of inhibitory receptors [21]. In the setting of chronic viral antigen exposure (e.g. LCMV), CD8⁺ T cell differentiation progresses from a PD-1⁺TCF1⁺ stem-like state, to a PD-1⁺Tim3⁺ TCF1⁻ CX3CR1⁺ transitory state, to a PD-1⁺ Tim3⁺ CD101⁺ exhausted state. The antigen-specific transitory CD101⁻ Tim3⁺ CD8⁺ T cells proliferate and control virus after PD-1 pathway blockade, indicating these are the cells responsive to checkpoint therapy [54]. Adding to the present limitation that only a small subset of cells are responsive to ICB, several studies demonstrated that CD8⁺ T cell exhaustion programs were preserved in the absence of antigen; this in turn suggested that the inability of fully exhausted T cells to respond to ICB therapy is a cell-intrinsic property that can be heritably maintained [42, 43, 46, 55, 56].

The finding that T cells with terminal-exhaustion epigenetic programs do not respond to ICB has raised further questions regarding which T cell subset exerts antitumor effects during ICB. Recent studies have shown that ICB recruits T cells with limited expression of inhibitory receptors into the anti-tumor response and these cells are predominantly derived from a pool of CD8⁺ T cells that have a “progenitor”-like gene expression profile [57]. Conversely, PD-1^hi tumor infiltrating lymphocytes (TILs) acquire exhaustion-associated DNA-methylation programs [57]. Downregulation of the TCF1 protein, (encoded by TCF7), in the memory-precursor-like T cell subset was coupled to a failure to respond to immunotherapy, while the less terminally-differentiated TCF1⁺ memory-precursor-like T cell subset correlated with a stronger response to ICB in melanoma patients [57]. Furthermore, single cell transcriptional profiling of immune cells isolated from the tumor of melanoma patients demonstrated that the presence of TCF1 in CD8⁺ T cells could help predict the clinical response to ICB [58]. Given that TCF1 plays a key role in preserving stem cell-like memory cells, these less differentiated cells could be presumed to give rise to the transitory cells that contribute to tumor control [59].

The observation that ICB can only modulate T cells that are not terminally exhausted, may explain in part why PD-1 blockade has limited therapeutic efficacy in some individuals and cancer types, while it is highly effective in others [56]. Although the presence of less terminally differentiated T cell subsets has been associated with improved clinical outcomes, formalized prediction tools are not available to determine which patients will respond to ICB. In this regard, determining the multipotency of freshly isolated TILs using our developed index might be useful to predict upfront ICB responses (Key Figure, Figure 2A). Not only do chronically stimulated T cells undergo irreversible exhaustion-associated epigenetic remodeling, but reinforcement of these programs prior to administration of ICB limits the therapeutic potential of this approach [34, 60] (Key Figure, Figure 2B). Given that de novo DNA methylation programming promotes T cell exhaustion, strategies to inhibit these methylation programs may be used to facilitate T cell responsiveness to ICB. Previous work demonstrated that Dnmt3a conditional knockout (cKO) mice (granzyme B cre × Dnmt3a^flox/flox) treated with anti-PD-L1 antibody exhibited an increase in the frequency and quantity of virus-specific CD8⁺ T cells compared to wildtype virus-specific CD8⁺ T cells during a chronic LCMV infection [56]. Similar strategies could be potentially implemented for human tumor-specific CD8⁺ T cells by using small molecule inhibitors to target DNMT3A in tumor-specific T cells prior to an autologous infusion. These possibilities certainly merit further investigation.

Developing a long-lived CAR-T cell therapy

Unlike ICB which relies on modulating the endogenous tumor-specific T cell response, adoptive cellular therapy with genetically-modified T cells utilizes a non-tumor specific functional T cell population that is reappropriated for anti-tumor purposes. For instance, chimeric antigen receptor (CAR) T cell therapy involves engineering T cells with specific receptors that target tumor associated antigens ex vivo before these cells are infused into patients. CARs consist of an antigen binding domain, in most cases derived from a single chain variable fragment (scFv) of a monoclonal antibody, a hinge, a transmembrane domain, and an intracellular signaling domain [61]. Commonly employed signaling domains include the CD3z chain for T cell activation and costimulatory domains derived from CD28 or 4–1BB [61]. Thus far, CAR-T cell therapy has been successful for B cell lineage malignancies including acute lymphoblastic leukemia (ALL), lymphoma, and multiple myeloma [62–64]. Two CD19-CAR-T cell products, tisagenlecleucel and axicabtagene ciloleucel, received FDA approval in 2017 for the treatment of CD19⁺ B-cell ALL or lymphoma. While the initial response for both CD19-CAR T cell products was high, many patients relapsed within the first year post CAR-T cell infusion [65, 66]. The etiology of relapse is most likely multifactorial and includes limited T-cell persistence and/or the emergence of CD19 antigen loss variants [65, 66]. In addition, initial peak expansion of infused CAR-T cells has been correlated to clinical response, and also to long-term CAR-T cell persistence [67].

In contrast to the success for certain hematological malignancies, CAR-T cell therapies for solid tumors and brain tumors have had limited efficacy in early phase clinical studies despite promising results in preclinical models [68]. Many factors have likely contributed to the disappointing clinical results including limited T cell expansion and persistence, heterogenous antigen expression, limited ability of CAR-T cells to home and penetrate solid tumors, and the hostile tumor microenvironment [69]. Numerous studies are underway to improve the effector function of CAR-T cells and are reviewed in detail elsewhere [69–71]. We now focus on the efforts to maintain less differentiated T cells during ex vivo CAR T-cell production (Box 3), as well as after CAR-T cell infusion and exposure to tumor cells.

Box 3: Brief overview of ex vivo CAR T cell production.

The majority of clinical studies have used selected or unselected bulk T cells from peripheral blood mononuclear cells (PBMCs) as a starting cell source. These T cells are then activated with anti-CD3/CD28 antibodies, genetically modified to express the CAR, and expanded with IL2 for 7 to 14 days [95]. While this allows for the reliable generation of clinical CAR T cell products, T cells differentiate into terminal effectors during this culture process. Investigators have shown that preselecting naïve T cells and Tcm subsets prior to T-cell activation results in improved effector function of CAR T cells [96]. However, peripheral T cells of cancer patients, who are heavily pretreated, have a predominant Tem phenotype [96]. Other efforts have focused on replacing IL2 for T cell expansion. For example, the use of IL7/IL15 is thought to preserve a Tscm-like phenotype (CD8⁺CD45RA⁺CCR7⁺CD95⁺) among CAR T cells during the expansion process[97]. Lastly, halting CD8⁺ CAR T cell differentiation by targeting the Wnt-β-catenin pathway with small molecules such as glycogen synthase-3β (GSK-3β) inhibitors remains an attractive approach [13, 59]. While GSK-3β inhibitors arrest T cell differentiation, they also inhibit T cell proliferation and effector function necessitating alternative approaches to generate a T cell with sustained antitumor abilities [59].

Prevention of CAR-T cell exhaustion

Recently, the mechanisms contributing to exhaustion have been shown to limit the survival and effector functions of engineered T cells [72–75]. However, these studies previously used cell surface markers such as PD-1, Tim3, and LAG3 to define T cell exhaustion and thus, overlooked a molecular definition of the cell’s differentiation status, such as its epigenetic program. Several current approaches for preserving CAR T cell function include modulating the signaling domains of CARs and additional genetic modifications. For example, human CD19.41BBz CARs induce a Tcm (central memory) phenotype post stimulation in comparison to CD19.CD28z CARs, resulting in their improved persistence during ex vivo culture [76]. Additional studies have focused on calibrating the CAR T cell activation potential by mutating immunoreceptor tyrosine-based activation motifs (ITAMs) in the CD3z signaling domain [77]. Modifying ITAMs of human CAR T cells to induce effector function while maintaining memory programs resulted in persistence of functional long-lived CAR T cells in NOD/SCID/IL-2Rγ^null (NSG) mouse model systems [77]. Besides optimizing CAR signaling domains, investigators have focused on transgenic expression of cytokines, such as IL12, IL15, IL18, and IL23 to prevent erosion of T cell effector function [78–81]. Additionally, efforts are underway to knock out inhibitory receptors in CAR T cells such as PD-1 and LAG3 [82, 83]. Lastly, using chemical inducers to fine tune the activation of toll like receptor (TLR) pathways in CAR T cells at the antigen site, is actively being explored to ideally prevent terminal T cell differentiation [84, 85]. Although further investigation is warranted, these studies highlight the importance of maintaining CAR T cells in a less differentiated state to preserve long term functionality and prevent T cell exhaustion.

T cell exhaustion is often reinforced by stable changes in gene regulation, and several groups have reported on epigenetic mechanisms used to maintain exhaustion-associated gene expression programs [34, 42, 60, 86, 87]. Moving forward, we posit that an epigenetic definition of T cell differentiation is necessary, especially since markers such as CD95 (used to define the Tscm (stem cell memory) subsets in humans) and PD-1 (used to assess T cell exhaustion), and which are currently being used as surrogates for specific differentiation states, are both upregulated post-T cell activation; indeed, this could potentially lead to misinterpretation regarding an exhausted state [88]. In addition to providing a more accurate characterization of T cells used in cellular therapies, we argue that a defined and standardized epigenetic program will also be an invaluable tool to assess the benefit of improved culture conditions and/or additional genetic modifications currently being developed (Key Figure, Figure 2C). Purposefully modifying the epigenetic profile of human T cells might allow for engineered resistance to T cell exhaustion and yield a long-lived pool of CAR T cells that can maintain anti-tumor responses in the setting of chronic tumor antigen exposure. This is probably best exemplified by a recent case report on the vector integration-mediated disruption of the Tet methylcytosine dioxygenase 2 (TET2) gene (encoding an enzyme that regulates DNA demethylation) in CAR T cells; this disruption resulted in clonal expansion of a single CAR T cell that induced leukemia remission [89]. While this case report also raises safety concerns, it clearly highlights the therapeutic potential of manipulating epigenetic programs in CAR T cells. Thus, safely manipulating the activity of de novo DNA methyltransferases (DNMTs) and demethylating enzymes that play a crucial role in establishing exhaustion and memory associated methylation programming may hold a key to developing exhaustion-resistant CAR T cells for a broad range of human malignancies.

CONCLUDING REMARKS

Memory CD8⁺ T cell differentiation is now being defined on an epigenetic basis. During CD8⁺ T cell memory differentiation, transient exposure to antigen imparts these cells with effector and memory-associated functions coupled to epigenetic programs; these enable the cells to preserve such acquired functions during T cell homeostasis, ultimately maintaining cell fate decisions. Likewise, chronic antigen exposure such as persistent tumor burden also imparts CD8⁺ T cells with epigenetic programs that can be long-lived; however, such programs are associated with suppressing T cell effector functions and proliferative capacity, also correlating with limited responses to immunotherapeutic approaches. The characterization of epigenetic modifications, including DNA methylation, acquired during acute and chronic antigen exposure has yielded an epigenetic fingerprint unique to each specific CD8⁺ T cell subset. This signature can be used to not only identify the differentiation status of CD8⁺ T cells but also to guide the rational engineering of T cells. Building upon the findings that stem-like T cells exhibit enhanced anti-tumor properties, are more responsive the ICB, and are more potent in the setting of CAR T cells [14, 57], we argue that future efforts should be focused on maintaining T cells in a less terminally differentiated state for T cell-based immunotherapies (Key Figure, Figure 2C). However, given the potential for off-target effects and malignant transformation of epigenetically-modified cells, additional investigations and precautions must be taken before being able to translate these exciting findings into the clinic (see outstanding questions). Moving forward, we expect that the epigenetic characterization of T cells might be used to predict responders versus non-responders in multiple immunotherapy indications. This may ultimately facilitate our efforts to transition from a one-size-fits-all T cell based immunotherapy approach, to tailored treatment modalities for individual patients.

Outstanding questions box.

At what point during CD8⁺ T cell differentiation does the progenitor subset bifurcate into developmental trajectories of functional memory and exhaustion?

Given that T cell infiltration is essential for eliciting a clinical response from T cell based immunotherapy, how can we program T cell trafficking and homing to the tumor? This question is particularly pertinent for solid and brain tumors.

Once a T cell reaches a tumor, how does the microenvironment influence its epigenetic programs and functions? Does tumor microenvironment-mediated suppression play a role in imprinting epigenetic programs that limit T cell effector functions?

Can we use epigenetically-based biomarkers to further define the therapeutic window of T cell differentiation necessary to achieve a clinical response?

Highlights.

• Epigenetic programs are increasingly being used to define the differentiation status of mammalian CD8⁺ T cells.

• Developmentally permissive CD8⁺ T cells have correlated with clinical responses to ICB and CAR T cell therapy in certain cancers.

• Modification of epigenetic programs in experimental systems and clinical settings have documented that epigenetic engineering approaches can enhance or potentiate existing T cell-based immunotherapies.

Glossary

Immunotherapy

Therapeutic modalities utilizing the endogenous immune system, or adoptive transfer of immune cells into a host, to eliminate diseased cells or pathogens

Immune checkpoint blockade

Therapeutic procedure involving the infusion of antibodies capable of blocking ligand-receptor interactions of inhibitory receptors such as CTLA-4 and PD-1 on T cells

Chimeric antigen receptor (CAR) T-cell therapy

form of immunotherapy whereby T cells are engineered to express an antibody-derived surface fragment specific for a particular antigen associated with a tumor or pathogen, subsequently transferred into the afflicted individual to treat the disease

Cytotoxic T-lymphocyte-associated protein 4 (CTLA-4)

cell surface receptor serving to repress the T cell effector response when ligated, often described as an immune checkpoint

Programed cell death receptor 1 (PD-1)

cell surface receptor serving to repress the T cell effector response when ligated, often described as an immune checkpoint

Terminal effector T cell

KLRG-1⁺ CD127⁻ activated cells forming in the acute phase of an infection with potent effector functions, but committed to die following the removal of the infected target

Memory precursor effector T cell

KLRG-1⁻ CD127⁺ cells forming in the acute phase of an infection, capable of persisting in the long-term; give rise to long-lived memory cells

T cell exhaustion

differentiation state of T cells induced by persistent antigen stimulation; results in a functional impairment of the cell; can be reinforced by epigenetic modifications

Epigenetic modification

Covalent modifications to DNA and/or histones that can impact on chromatin accessibility and ultimately reinforce cell type-specific gene expression programs

DNA methylation

epigenetic modification whereby a methyl group is added to a DNA base, most commonly observed at cytosines in mammalian cells; can result in changes in transcriptional activity when occurring at promoter of enhancer elements

T cell multipotency index

novel index based on the DNA methylation status of T cells; assigns a score ranging from 0–1, predictive of the general differentiation status of the T cell

TCF1

Among a group of transcription factors associated with the Wnt signaling pathway that contribute to the preservation of the multipotent developmental potential of T cells

CD101

cell surface protein identified on T cells at distinct stages of differentiation during the development of T cell exhaustion

Naïve T cell

has exited thymic selection but has not yet undergone antigen-driven differentiation

Effector T cell

has recently or is actively engaged with its cognate antigen and is now expressing effector molecules

Central memory T cell (Tcm)

subset of memory T cells expressing CD62L and CCR7; able to recirculate between the blood and secondary lymphoid organs

Effector memory T cell (Tem)

subset of memory T cells lacking CD62L and CCR7; generally excluded from secondary lymphoid organs and able to circulate between the blood and peripheral tissues

Stem cell memory T cell (Tscm)

Bears properties allowing its development into other effector and memory T cell subsets