Advancing beyond the twists and turns of T cell exhaustion in cancer

Abstract

Chronic antigen stimulation leads to T cell exhaustion, and nutrient restrictions and other suppressive factors in the tumor microenvironment further exacerbate T cell dysfunction. Better understanding of heterogeneity and dynamics of exhausted CD8 T cells will guide novel therapies that modulate T cell differentiation to achieve more effective anti-tumor responses.

One-sentence summary:

Understanding the heterogeneity of exhausted CD8 T cells will help inform new therapies.

Defining T cell exhaustion

T cell exhaustion, a dysfunctional and hyporesponsive cellular state commonly observed in response to persistent antigen exposure, is a shared feature in chronic infection, cancer, and autoimmune disease (1). Originally perceived as an undesirable outcome, it has become increasingly apparent that the molecular and phenotypic programs of exhaustion constitute a means for T cells to survive in the context of continuous antigen exposure while maintaining limited effector functions such as granzyme B expression and modest interferon (IFN)-γ production (2). Exhausted CD8 T cells (T_EX) present a distinct transcriptional program from memory or effector T cells, and a unique epigenetic landscape that becomes largely irreversible over time (3–5). T_EX are characterized by abundant expression of the transcription factor TOX, which drives the epigenetic program required for T_EX development and maintenance (6–8), and sustain high expression of multiple inhibitory receptors that restrain T cell function (1). Targeting inhibitory receptors on CD8 T_EX represents a therapeutic opportunity to reinvigorate anti-tumor immunity. Most prominently, therapies that block the programmed cell death-1 (PD-1) inhibitory pathway have shown unprecedented clinical responses in different tumor types (9) and combination therapies that block additional inhibitory receptors, such as lymphocyte-activation gene-3 (LAG-3) and cytotoxic T-lymphocyte antigen-4 (CTLA-4), can further improve response rates (10, 11).

Yet, it is important to highlight that inhibitory receptors enable T_EX survival and are not a unique feature of T_EX. Upon activation, effector CD8 T cells rapidly upregulate an array of inhibitory molecules, including PD-1, LAG-3, and CTLA-4. Many of these receptors are directly induced by T cell receptor signaling, constituting a negative feedback loop mechanism that limits T cell-inflicted immunopathology (2). Supporting this conceptual framework, murine autoreactive CD8 T cells were recently found to be restrained by an exhaustion-like program under control of the inhibitory receptor LAG-3 (12), whereas the presence of an “exhausted” gene signature was predictive of a better clinical prognosis in multiple autoimmune disorders (13). These studies suggest that therapeutic induction of an exhaustion-like state may generally be beneficial to prevent tissue damage, but also underline the delicate balance between effective and detrimental responses, best exemplified by autoreactivity triggered by T_EX-reinvigorating cancer therapies (1).

Even though CD8 T cell dysfunction has historically been observed in tumor settings, many aspects of our current mechanistic understanding of T cell exhaustion stem from experimental work in mice chronically infected with lymphocytic choriomeningitis virus (LCMV). Elegant studies have confirmed that the phenotypical and transcriptional profiles of tumor-infiltrating lymphocytes (TILs) broadly overlap with exhausted CD8 T cells in chronic LCMV infection (14). Nevertheless, there can be major differences in T_EX that have adapted to different microenvironments: chronic inflammation, hypoxic environments, and nutrient restrictions all impose substantial challenges that enhance T cell dysfunction (15). Metabolic alterations, in particular resulting in mitochondrial dysfunction, accelerate and exacerbate the development of exhaustion (16, 17). Therefore, characterization of cancer-associated CD8 T_EX dysfunction across species and in different tumors has been complicated by a lack of common defined degrees of exhaustion, as well as functional and phenotypical correlates that correspond to each degree. Longitudinal analysis, as well as functional and metabolic assays and epigenetic mapping, will therefore be key to improve assessment of T cell exhaustion in cancer.

At the population level, T cell exhaustion has been described as a progressive process in which T cells become more dysfunctional over time. Even though mouse models enable a comprehensive assessment of TILs, these analyses are temporally restricted – given that most murine tumors reach endpoint in about one month. In contrast, patients are often only diagnosed when tumors have attained a sizable stage, and tumorigenesis may have been initiated months to years before, thus increasing the likelihood of more pronounced exhaustion. Therefore, differences in time scale for T cell differentiation in mouse tumor models and patients need to be carefully considered. Additional factors that may affect T cell function, such as history of infections, diabetes, and obesity, require stringent characterization in animal models to provide mechanistic insights into potentially confounding effects on anti-tumor immunity. Finally, TIL diversity in patients is caused not only by the presence of non-tumor-specific bystander T cells, but also due to tumor heterogeneity (with regards to driver mutations, availability of nutrients, immune cell infiltrates) and asynchronous generation of neoantigens (with different affinities and expression), culminating in a mixed pool of exhausted PD-1^hi CD8 T cells at various dysfunctional stages with different functionalities.

Characterizing heterogeneity among CD8 T_EX

Despite heterogeneity among CD8 T_EX, all T_EX express PD-1 and TOX and share additional adaptations in response to chronic antigen stimulation. Perhaps the most striking finding regarding CD8 T_EX diversity was the identification of progenitor exhausted cells (T_PEX) which arise early after T cell activation and give rise to other T_EX subsets (18–20) (Fig. 1). T_PEX are characterized by preserved expression of T cell factor 1 (TCF-1, encoded by the Tcf7 gene), a key transcription factor also found in naïve and memory T cells. Previous studies have shown that T_PEX are essential for CD8 T_EX maintenance through self-renewal, as well as through differentiation into TCF-1^negCD39⁺ T_EX that exert effector/cytotoxic activity (18–20). T_PEX express numerous costimulatory molecules, such as CD28, Inducible T-cell costimulator (ICOS), and CD226, while lacking co-expression of many receptors characteristic of more differentiated T_EX subsets, such as T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), CD39, LAG-3, and CTLA-4.

Figure 1. Dynamics of exhausted CD8 T cells.

During chronic antigen exposure, PD-1⁺ TCF-1⁺ CD8 T cells (T_PEX) continuously supply the pool of exhausted T cells (T_EX) by self-renewal and conversion into more differentiated PD-1⁺CD39⁺TCF-1^neg effector-like cells, which express IFN-γ and Granzyme B (GzmB) as well as terminal T_EX, which retain GzmB expression. Upon blockade of PD-1/PD-L1 interactions, effective reinvigoration of T_EX requires re-priming of T_PEX by APCs that provide T cell receptor (TCR) stimulation as well as costimulatory signals through CD28 resulting in an increase in number and function of effector-like T_EX. CD4 T cells may directly and indirectly enhance this process through secretion of cytokines such as IL-2 and IL-21 as well as through CD40/CD40L-mediated APC activation. CREDIT: ASHLEY MASTIN/SCIENCE TRANSLATIONAL MEDICINE

Notably, although PD-1/programmed cell death ligand-1 (PD-L1) blockade may directly enhance effector function in differentiated T_EX, several murine studies have shown that T_PEX are responsible for immunotherapy efficacy by giving rise to the expanded effector pool (14, 18). Upon PD-1/PD-L1 blockade, T_PEX proliferate and differentiate into effector-like T_EX with enhanced cytotoxic capacity and ability to secrete cytokines, such as IFN-γ. Although less potent than bona fide effector T cells generated in the wake of acute infections or vaccination, effector-like T_EX have been shown to be central in controlling viral load in murine LCMV chronic infection (21, 22) as well as tumor growth in B16F10 melanomas (21). Effector-like T_EX eventually transition to terminally differentiated T_EX, with higher expression of inhibitory receptors and limited ability to produce cytokines and proliferate (although granzyme B expression is preserved). The signals that drive T_PEX differentiation and whether T_PEX can directly give rise to terminally differentiated T_EX depending on cellular interactions and molecular cues still need to be determined (Fig. 1). Importantly, dynamic changes in the frequency of effector-like and terminally differentiated T_EX subsets underlie, at least in part, the observation that the functionality of T_EX decreases over time.

Additional T_EX states or subsets have been proposed by recent studies in chronic LCMV infection. Tsui et al. further pinpointed CD62L-expressing T_PEX with a unique transcriptional program and dependent on the transcription factor MYB to be particularly endowed with long-term proliferative and self-renewing properties (23). Another report highlighted that T_PEX transition from an early active and less differentiated state to a more quiescent status that is irreversibly committed to exhaustion; this study also identified a new subset of T_EX characterized by expression of receptors usually associated with natural killer cells (5). These new studies show that T_EX diversity is not fully resolved and continues to be dissected. Future studies in both animal models and patients will need to further investigate these T_EX subsets and states and our proposed model for the dynamics of T_EX will have to be updated accordingly.

High-throughput profiling techniques. such as single-cell RNA sequencing. have revolutionized the characterization of the tumor microenvironment by highlighting human TIL diversity beyond a handful of markers. Remarkably, despite differences between studies, overall similar populations of human T cells showing signs of chronic antigen stimulation and clonal expansion have been identified across various cancer types (24–26). In patient tumors, in addition to an “exhaustion signature”, CXCL13 expression identifies tumor-specific CD4 and CD8 T cells (27, 28). Furthermore, most studies have described a “pre-dysfunctional” or “transitional” subset of human CD8 T_EX corresponding to effector-like T_EX, marked by intermediate expression of inhibitory receptors as well as GZMK expression. In addition, T_EX cells displaying abundant expression of multiple inhibitory receptors (LAG3, ENTPD1, CTLA4), matching the terminally-differentiated, more dysfunctional cells observed in mice, have also been identified in tumor tissues of patients (24–26, 29). Although many earlier reports did not report human TILs equivalent to the T_PEX subset, more recent studies have been able to identify this population in patients with cancer, either by analysis of large data sets (24, 29) or by focusing on antigen-specific CD8 T cells (30). In silico trajectory, T cell receptor sequencing analyses, and even proliferation assays have revealed relationships between different states of human T_EX consistent with studies in mouse models (24, 29, 30). Yet, an integrated characterization of T_EX phenotype and function in conjunction with clinical data, as well as additional analysis of both the tumor microenvironment and immune cell composition, is still needed to better understand the dynamics and function of T_EX subsets in patients with cancer.

Re-priming for reinvigoration of anti-tumor T cell responses

When T cell exhaustion was first described, the primary focus was on reversing this process. Now, novel strategies to improve T cell function during chronic stimulation need to reconcile that epigenetic modifications and metabolic dysregulation can restrict T_EX differentiation and functional potential, but also that, among T_EX, the small T_PEX subset can give rise to cells that retain some functionality. In the last few years, it has become evident that merely blocking individual inhibitory signals might not be sufficient to induce long-lasting, effective anti-tumor responses in a majority of patients with cancer. For instance, we previously demonstrated that CD28-mediated co-stimulation is essential for the proliferation of PD-1⁺ CD8 T cells and successful anti-PD-1/PD-L1 therapy (31). Recent studies in patients have further pointed out the role of professional antigen-presenting cells (APCs) as well as cellular interactions to promote anti-tumor CD8 T cell responses (29, 32, 33). Together, these data provide a paradigm shift in our understanding of how to reinvigorate dysfunctional T cell responses, implying that this process requires re-priming of tumor-specific T_PEX. Even though many aspects of these re-priming events share similarities with the activation of naïve T cells, major differences need to be considered with regards to expression of chemokines and cytokines, costimulatory and inhibitory receptors, and the epigenetic landscape of T_PEX. Therefore, combination therapies based on T_PEX biology and requirements for differentiation into effector-like T_EX would represent novel promising approaches.

T_PEX express high XCL-1 (lymphotactin), a chemoattractant for XCR1⁺ dendritic cells (DC1s) (18). In murine models, consistent with DC1 colocalization, T_PEX are mostly found in the T cell zones of secondary lymphoid organs. Likewise, in patients with cancer, T_PEX were identified within intratumoral niches containing major histocompatibility complex (MHC)-II^hi APCs, and the presence of these niches in kidney tumors correlated with improved progression-free survival (32). Moreover, it has been shown that APCs in ovarian tumors can provide adequate in situ CD28 costimulation to T cells during PD-1 blockade therapy (33). Finally, our recent observations in patients with hepatocellular carcinoma treated with anti-PD-1 show that, in responders, T_PEX are in close proximity to DCs enriched in maturation and regulatory molecules (“mregDC”) (34) and CXCL13⁺ CD4 T cells, suggesting that this triad in the tumor microenvironment may enhance the clinical response to PD-1 blockade (29).

In murine models, CD4 T cells play a key role for effector-like T_EX differentiation (21). Studies of TILs from patients have recently highlighted the presence of tumor-specific CD4 T cells with features of chronic antigen exposure and similarities to follicular helper CD4 T cells (24, 29). CD4 T cells may enhance the function of T_EX by providing cytokines such as IL-2 and IL-21 or by activation of APCs (Fig. 1) in a similar manner as described for CD4 help during priming of naïve CD8 T cells. Therefore, it is possible that specific cell interactions provide an adequate environment to support T_PEX survival, ensuring T_PEX self-renewal and, upon PD-1/PD-L1 blockade, differentiation into effector-like T_EX. Hence, intratumoral re-priming of T_PEX might play a key role in clinical responses to immunotherapy. In sum, the general focus has recently shifted away from reverting T cell exhaustion to instead focus on better understanding of how interventions modulate the dynamics and function of specific T_EX subsets. In addition to optimized strategies to promote T_PEX differentiation into effector-like T_EX, clinical efficacy may also rely on enhanced survival and function of effector-like T_EX. Furthermore, recruitment of T_PEX and naïve clones from draining lymph nodes may also need to be harnessed to potentiate responses, especially when tumors lack APC-rich niches.

Additional studies are required to better address these pending questions, which are central to guiding successful combination therapies. Multiple clinical trials are currently underway evaluating the role of cytokines and other agents that may promote APC recruitment and activation or that can bypass the need for some of these cellular interactions by directly activating T_EX. The impact of local cues for reinvigoration will also need to be carefully studied. In sum, we must acknowledge that tumor-specific T cells have adapted to chronic antigen stimulation and are therefore exhausted. Furthermore, in cancer, T_EX may show additional dysfunctional features due to adaptations to the specific tumor microenvironment, such as nutrient limitations and interactions with suppressive immune cells. Despite differences across diseases and patients, we need to find common patterns to better understand the biology of T cells that respond to persistent antigen. We now possess unmatched tools to comprehensively characterize T_EX; appreciating both the limitations and strengths of animal models as well as an extended analysis of patient samples will allow for more efficient progress in the development of novel therapeutic strategies against cancer.