Memory CD8⁺ T cell responses to cancer

Abstract

Long-lived memory CD8⁺ T cells play important roles in tumor immunity. Studies over the past two decades have identified four subsets of memory CD8⁺ T cells - central, effector, stem-like, and tissue resident memory - that either circulate through blood, lymphoid and peripheral organs, or reside in tissues where cancers develop. In this article, we will review studies from both pre-clinical mouse models and human patients to summarize the phenotype, distribution and unique features of each memory subset, and highlight specific roles of each subset in anti-tumor immunity. Moreover, we will discuss how stem-cell like and resident memory CD8⁺ T cell subsets relate to exhausted tumor-infiltrating lymphocytes (TIL) populations. These studies reveal how memory CD8⁺ T cell subsets together orchestrate durable immunity to cancer.

1. Introduction

The use of CD8⁺ T cells has been a focus of clinical cancer therapy for over 20 years [1]. Based on their ability to clonally expand and exert cytotoxic function, T effector (T_EFF) cells have long been recognized as important mediators of tumor protection [2, 3]. However, studies in mice and humans have also revealed that the presence of tumor infiltrating T_EFF cells is not sufficient for tumor rejection because of apoptosis induced clonal contraction [4–6]. Cancer is a chronic disease and, unlike immunity to acute infections, curative immunity to cancer is thought to require long-lived T cell immunity. Indeed, the persistence of adoptively transferred T cells in peripheral blood has been shown to correlate with tumor regression in cancer patients [7]. In contrast to T_EFF cells, memory CD8⁺ T cells are capable of durably persisting and functioning throughout host tissues and tumors, making them a gold standard for anti-tumor immunity.

The earliest evidence for the generation of protective immune memory to cancer comes from pre-clinical studies showing that mice cured from primary tumors could maintain protection against secondary tumor rechallenge [8–10]. This persistent tumor immunity was mediated by CD8⁺ memory T cells, supported by the finding that depletion of CD8⁺ T cells eliminated tumor protection [8, 11]. Early work involving peptide vaccination in humans separately showed that melanoma antigen specific CD8⁺ T cells could expand dramatically, acquire a CD27⁻ CD28⁻ CD57⁺ CD45RA^lo memory phenotype, and demonstrate robust recall function in vitro[12]. An additional study examining melanoma patients after peptide vaccination revealed Melan-A/MART-1 specific memory T-cell responses capable of producing IL-2, IFN-γ and GM-CSF[13]. These early clinical and preclinical studies collectively revealed the possibility of generating functional T-cell memory against cancer, and the importance of such responses in curative immunity. However, the dynamic and heterogeneous nature of memory CD8⁺ T cell responses, throughout host lymphoid and peripheral tissue compartments, was only beginning to be explored.

Priming of naïve CD8⁺ T cells induces clonal expansion and the acquisition of peripheral tissue homing capacity, effector function and cytolytic activity. After contraction of the T_EFF response, memory T cells continue to recirculate but also become durably resident in tissues throughout the body. Memory CD8⁺ T cells persist as a heterogenous population including circulating memory (T_CIRC) subsets- central memory (T_CM), effector memory (T_EM) and stem cell memory T cells (T_SCM), and a non-circulating subset known as resident memory T cells (T_RM) (Figure 1). T_CM and T_SCM cells express molecules such as CD62L and CCR7 that enable their access to bone marrow (BM) and secondary lymphoid organs (SLO), whereas T_EM cells lack lymphoid tissue homing receptors, and preferentially recirculate through blood and peripheral tissues[14–16] . T_RM cells lack markers of circulating subsets, instead expressing tissue adhesion markers such as CD103 and CD69[17].

Figure 1. Tumor-specific memory T cells distribute throughout circulation and tumors.

T_CM, T_EM, T_EFF, and T_SCM populations are found in circulation. In contrast, T_RM cells reside in tissues and tumors, but are excluded from the blood. Whereas T_EFF cells readily undergo apoptosis, T_EM and T_RM cells persist more durably. The relationships between different subsets are indicated using thin black arrows. Upon antigen re-encounter, T_CM cells are capable of self-renewal and differentiating into T_EM, T_EFF and T_RM cells. T_SCM cells are capable of self-renewal and differentiation to T_CM, T_EM and T_EFF in lymph node. Moreover, T_RM cells can become T_CM and T_EM cells following local reactivation. The big arrow at the bottom left indicates the thoracic duct that drains to the circulation system. Image created using BioRender (https://biorender.com).

The maintenance and differentiation of these memory subsets are mediated by both cell extrinsic and intrinsic factors. Cytokines such as IL-2 promote cell proliferation during priming, and is essential for functional CD8⁺ memory T cell recall responses [18], whereas IL-7 and IL-15 promote survival, differentiation and maintenance of memory [19]. Expression of various transcription factors also controls the differentiation and function of distinct memory CD8⁺ T cell subsets. For instance, KLF2 promotes the expression of CD62L and S1PR1 which are essential for T_CIRC tissue egress and homing to SLOs[20]. In contrast, Hobit, Blimp1 [21, 22] and Runx3 [23] govern T_RM cell differentiation by limiting tissue egress. Regardless, memory T cell subsets are capable of differentiating and interconverting in vivo (Figure 1). Whereas T_SCM cells are considered the most pluripotent, T_CM cells are capable of giving rise to T_EM and T_RM subsets[24–26]. Moreover, recent data suggest that T_RM cells can become T_CM and T_EM cells following local reactivation [27]. Collectively, extrinsic and intrinsic factors within a given microenvironment regulates the distinct roles and differentiation pathways for memory T-cell subsets.

The growing knowledge of these subsets from studies in infectious disease models continues to inform the field of cancer immunology. Indeed, each of these subsets plays a unique unique role in immunity to cancer, and recent studies have revealed the heterogeneity that governs governs their differentiation and maintenance. The goal of this review article is to summarize the characteristics and anti-tumor function of different subsets of memory CD8⁺ T cells. We will discuss how these populations can coordinate to provide long-term immunity to cancer while underscoring the association between the presence of memory CD8⁺ T cells and improved outcomes for cancer patients.

2.1. Central memory and effector memory T cells

As described above, two major circulating memory CD8⁺ T cell subsets are T_CM and T_EM cells. First described in humans in 1999, these subsets were initially differentiated by their expression of CD62L [28] and CCR7 [29] which are essential for T cell migration to lymph nodes. However, these populations are now well distinguished by their unique phenotypes, functions, and migratory properties. T_CM cells exhibit a CD45RA⁻ CD45RO⁺ CCR7⁺ CD62L⁺ phenotype in humans, and are defined by their lymph node homing properties [30, 31]. They exhibit a higher proliferative capacity than their T_EM counterparts which exhibit a CD45RA⁻ CD45RO⁺ CCR7⁻ CD62L⁻ phenotype and, accordingly, are excluded from lymph nodes, while expressing integrins and chemokine receptors necessary for localization to inflamed tissues [30, 32]. In mice, the phenotype and localization of T_CM and T_EM subsets largely overlap with those in humans, with the exception that CD44⁺ is used as a memory marker instead of CD45RA/CD45RO [33]. T_CM and T_EM cells each respond to TCR stimulation in a different manner. Upon anti-CD3/CD28 stimulation in vitro, T_EM cells secrete large amounts of effector molecules including IFN-γ and perforin, whereas T_CM cells predominantly express high levels of IL-2, clonally expand, and are capable of differentiating into both T_EFF and T_EM cells [30, 34–36]. Thus, T_CM and T_EM each play an important and distinctive role in the memory response.

Memory T cell fate is determined by distinct transcriptional regulators. KLF2 has been shown to directly activate the promoters of S1P1 and CD62L, which are important for tissue egress and homing to lymph nodes [20]. T-box transcription factors T-bet and its homolog Eomes are master regulators for the differentiation of memory CD8⁺ T cells [37]. While low levels of T-bet are required for the development of KLRG1⁻ IL-7R⁺ memory precursor cells, high levels of T-bet have been shown to suppress the expression of IL-7R, which preferentially drives the differentiation of T_EM cells [38, 39]. On the other hand, Eomes has been found to be essential for T_CM differentiation, as well as maintaining the longevity and rapid recall responses of T_CM cells [40]. The transcription factors Blimp-1 and Id2 induce terminal differentiation thus promoting CD8⁺ T cells differentiation into T_EM cells, whereas Id3 correlates with the upregulation of Il-7r, Sell, Il2, Bcl2, and Ccr7, genes essential for the differentiation of CD8⁺ T cells into a T_CM phenotype [41–43]. Thus, T_CM and T_EM cells represent transcriptionally distinct lineages.

2.2. T_CM and T_EM responses to cancer: Mouse studies

T_CM cells have been shown to have more potent anti-tumor activity than T_EFF cells. Using the transplantable B16 murine melanoma model, the adoptive transfer of IL-15 cultured transgenic pmel CD8⁺ T_CM cells was shown to be significantly more effective in controlling established tumors, as compared with IL-2 cultured T_EFF/T_EM-like cells [44]. Transferred T_CM pmel cells also gave rise to 14-fold larger populations in blood and spleen [44]. In a similar preclinical model, the adoptive transfer of IL-21 cultured pmel cells, which also had a T_CM phenotype, displayed enhanced antitumor responses compared to those cultured with IL-2 [45]. Similar observations were found using CMV-pp65 specific T cells taken from human peripheral blood, with T_CM cells providing more robust antitumor immunity than T_EM cells upon transfer into immunodeficient mice bearing CMV-pp65-expressing leukemia [46]. These data establish the superior anti-tumor function of T_CM cells compared to T_EM and T_EFF cells, consistent with their improved capacity to engraft and clonally expand [44, 46].

Despite the potency of T_CM cells on a per-cell basis, T_EM cells also represent a vital component of the anti-tumor response, particularly when generated in high frequencies. In vitro-generated pmel T_EM cells clonally expand, secrete high levels of IFN-γ, and are capable of regressing large, established B16 tumors [47]. Several studies have shown that in vivo vaccination can promote tumor-Ag specific T_EM cells which sustain the ability to secret IFN-γ and induce tumor regression [48–51]. Interestingly, a persisting mouse gamma herpesvirus type 68 infection, which preferentially induces a T_EM response [50, 52], was shown to be a more effective tumor vaccine than an acute gamma herpesvirus which induces a T_CM response [50]. Similarly a cytomegalovirus-based cancer vaccine, which is known to generate large frequencies of T_EM-like inflationary memory CD8⁺ T cells [53], was shown to be more potent against cancer as compared to a vesicular somatitis viral vaccine that does not induce inflationary memory [54]. In the context of B16 melanoma-bearing mice that develop autoimmune vitiligo, large populations of melanoma-specific T_EM cells persist in the spleen, in conjunction with long-lived protective immunity against melanoma in the lungs [55]. Such studies collectively show that tumor-specific T_EM responses can be generated in large frequencies, and can remain functional against cancer, even in settings of persistent antigen exposure.

2.3. T_CM and T_EM responses to cancer: Human studies

The presence of T_CM and T_EM cells has been associated with better prognosis in cancer patients. Indeed anti-CTLA-4 immunotherapy has been shown to augment CD8⁺ memory T cell formation and maintenance in mouse models of infection, and may thus also promote memory to cancer [56]. In advanced melanoma patients on immune checkpoint blockade (ICB) therapy with anti-CTLA4, a higher proportion of CD45RO⁺ memory cells out of total CD8⁺ T cells in blood was a positive predictor for both response rate and overall survival [57]. PD-1 is more highly expressed on T_EM cells than T_CM cells[58], although both populations were able to predict patient responses to anti-PD1 immunotherapy[59, 60]. In advanced melanoma and non-small cell lung cancer (NSCLC) patients, the frequency of CD45RA⁻ CCR7⁺ T_CM cells in blood was shown to be a positive predictor for both response to anti-PD1 therapy and survival [59]. Responders that had a higher T_CM/T_EFF ratio in blood also had a higher inflammatory gene expression profile in their tumors [59], suggesting the importance of T_CM cells in anti-tumor immune responses in patients. In other studies, T_EM cells were specifically associated with better responses to ICB therapy and improved survival. In advanced melanoma patients, the number of highly expanded CD8⁺ T cell clones (clonal frequency>0.5%) in blood 21 days after the initiation of ICB treatment was found to be associated with better clinical response; these clones mainly possessed a CD45RA⁻ CD27⁻ T_EM phenotype [60]. A detailed study of the T_EM population in advanced melanoma patients after anti-CTLA-4 found that, instead of a more terminally differentiated CD45RO⁺ CD62L⁻ CD27⁻ CD28⁻ T_EM population, it was the less differentiated CD45RO⁺ CD62L⁻ CD27⁺ CD28⁺ T_EM subset that was associated with both better clinical response and overall survival [61]. Thus, the above studies show that in addition to T_CM, T_EM cells are also likely to play a key role in anti-tumor immunity.

3.1. Stem cell memory T cells

T_CM cells were considered the ‘stem-like’ memory population until a less differentiated memory stem cell (T_SCM) population was identified in 2005. T_SCM cells were first described in a mouse graft versus host disease (GVHD) model as a CD44⁻ CD62L⁺ naïve-like population, even though antigen experienced, and expressed high levels of stem cell antigen 1 (Sca-1), IL-2Rβ/IL-15Rα, and the anti-apoptotic transcription factor Bcl-2 [62]. These naïve-like memory cells exhibited a greater ability to self-proliferate, differentiated into T_EM, T_CM and T_EFF cells, and induced stronger GVH responses in mice, compared to T_CM and T_EM cells [62]. The presence of long-lived T_SCM cells was subsequently identified in humans within the naïve-like CD8⁺ T-cell compartment[63]. These cells maintained a CD45RA⁺ CD45RO⁻ CCR7⁺ CD62L⁺ CD27⁺ CD28⁺ IL-7Rα⁺ naïve-like phenotype, while also expressing memory markers CD95 and IL-2Rβ [63]. Additional transcriptomic analyses revealed that these human T_SCM cells had a transcriptional profile more closely related to human T_CM than to naïve T cells, with similar levels of TCR rearrangement excision circles (TRECs) compared to T_CM cells, indicating that T_SCM cells are not recent thymic emigrants [63]. T_SCM cells were subsequently identified in non-human primates where they served as precursors for T_CM cells [64]. Additionally, T_SCM cells were shown to exhibit enhanced proliferation when re-exposed to antigen while continuing to self-renew as T_SCM, demonstrating a true stemness[65] (Figure 1). T_SCM cells are extremely durable as evidenced by a yellow fever vaccine study wherein induced T_SCM cells were shown to persist for decades [66]. These features of self-renewal, persistence and multipotency could lead to the strong anti-tumor function of T_SCM cells.

3.2. T_SCM responses to cancer

The anti-tumor function of T_SCM was first demonstrated by studies involving in vitro cultured pmel T cells against mouse B16 melanoma. By enhancing the Wnt-β-catenin signaling pathway, the expression of Tcf1 and Lef1 was induced, thus promoting the differentiation of pmel T_SCM cells with a CD62L^hiCD44⁻ phenotype [67]. Adoptive transfer of these cells resulted in greater regression of melanoma tumors compared with mice transferred with T_CM or T_EM cells [67].

The T_SCM features of self-renewal and plasticity have led to investigations of human chimeric antigen receptor (CAR) T_SCM cells as a potential therapeutic strategy for blood cancers. Human anti-CD3/anti-CD28 activated CD8⁺ T cells cultured with IL-7 and IL-15 and transduced with a γ retroviral vector encoding a CD 19-targeting CAR resulted in a higher proportion of T_SCM cells compared with IL-2 induced CAR-T cells [68]. Multiple strategies have been used to generate clinical-grade CAR T_SCM cells. Activation of naïve T cells using the glycogen synthase-3β (GSK-3β) inhibitor TWS119 in the presence of IL-7 and IL-21 prior to transduction with a CD19-CAR, resulted in a primarily T_SCM population [69]. In another study, previously engineered CD19-CAR-T cells were expanded by TCR stimulation and then cocultured with Notch ligand Delta-like 1-expressing cells in the presence of IL-7, which resulted in conversion into T_SCM cells [70]. The enhanced NOTCH signaling promoted telomere elongation, leading to enhanced in vivo persistence as compared to conventional CAR-T cells [70]. In preclinical studies, CAR T_SCM mediated improved anti-tumor immunity compared to conventional CAR-T cells [68–70]. Multicenter, phase 1 and 2 trials are underway to test the safety and efficacy of CAR-modified T_SCM cells in patient with blood malignancies.

3.3. Stem-like tumor infiltrating lymphocytes (TILS)

Naïve-like T_SCM cells are a circulating T-cell population and, accordingly, are excluded from peripheral tissues and solid tumors [64, 71]. Despite this, several studies have demonstrated that solid tumors are occupied by populations of TCF1⁺ (encoded by TCF7) stem-like memory T cells with an antigen experienced phenotype [71–73]. Mouse and human tumors were found to contain TCF1⁺ PD-1⁺ GzmB⁻ CD8⁺ T cells [72]. In a mouse B16-gp33 melanoma model, TCF1⁺ TILs expressed memory markers including CXCR5, Ly108(Slamf6), and CD62L, as well as PD-1, CTLA-4, and Lag3 [72]. The expression of these checkpoint molecules on stem-like TILs suggest that this subset could be an important target for anti-PD1 and anti-CTLA4 treatment. Indeed, after adoptive transfer, the gp33-specific TCF1⁺ TILs were able to expand and differentiate into TCF1⁻ cells while maintaining secretion of IL-2 and TNF-α, and the durable anti-tumor effects of anti-PD1 and anti-CTLA4 therapy in mice were found to be dependent on Tcf7 expression in T cells [72].

High-dimensional flow cytometry also confirmed stem-like CD8⁺ TILs in lung cancer patients. Human stem-like CD8⁺ TILs upregulated TCF7 CXCR5 EOMES BCL6, FOXO1 and CD28, and also expressed PDCD1 and TIGIT. The CXCR5⁺ TILs rapidly produced IFN-γ, IL-2 and TNF-α upon PMA/ionomycin stimulation and extensively proliferated and differentiated into a CXCR5⁻ population. Furthermore, the abundance of CXCR5⁺ CD8⁺ T cells in tumor correlated with better disease control [71]. Concordantly, in melanoma patients, whereas TCF1⁺ CD8⁺ TILs express lower levels of PD-1 than their TCF1⁻ TIL counterparts [63], the presence of TCF1⁺ CD8⁺ TILs is associated with better clinical response to ICB therapy [74]. The above studies highlight the role of stem-like TILs in tumor control and response to immunotherapy.

The features of tumor microenvironments that support stem-like T cells are just beginning to be understood. Interestingly, an abundance of potassium in the tumor microenvironment (TME) has recently been implicated as an important factor. Tumor necrosis, which has been shown to be inversely correlated with patient survival [75, 76], leads to increased levels of extracellular potassium in the TME which suppresses T cell effector function [77]. However, treatment of pmel CD8⁺ TILS with elevated potassium concentrations resulted in differentiation of TCF1⁺ T cells, which resulted in regression of established tumors [77]. This study further shows the importance of stem-like TILS in an immune-suppressive TME.

Stem-like CD8+ TILs are phenotypically similar to tumor-infiltrating populations that have been termed progenitor exhausted (Tex^prog) cells [78], and these populations are likely to be overlapping [79]. Tex^prog is a heterogenous population as revealed by a recent study in a LCMV mouse model and human melanoma tumors [78]. Four developmentally related subsets were identified, of which two subsets were the less differentiated Tex^prog cells that expressed high or intermediate levels of TCF1. Functionally, these subsets had the ability to interconvert, and both populations could give rise to Tex^prog , as well as terminally and intermediate exhausted subsets [78]. In human kidney tumors, TCF1⁺ stem-like TILS had a high degree of TCR overlap with TCF⁻ terminally differentiated populations, thus supporting the idea of stem-like T cell precursors [80]. In another study, a similar Tcf7, Cxcr5, Ccr7, Sell, Il7r, Cd28 expressing Tex^prog population was found in the B16-OVA model [73]. Upon treatment with anti-PD-1, these Tex^prog cells expanded to a larger extent, and expressed higher levels of TNF-α and IL-2, compared to terminally exhausted cells. This study also showed that Tex^prog cells in the spleen and lymph nodes could persist in the absence of tumor antigen, indicating that they are true memory T cells. As with prior studies focused on TCF1+ stem-like cells [72], anti-PD-1 treatment expanded the number of Tex^prog cells in tumors [73]. Interestingly, in human tumors, TCF1⁺ CD8⁺ TILs preferentially reside in regions of APC dense zones which resemble T-cell zones of lymphatic tissue, or intra-tumoral immune niches [80]. In patients who had disease progression, there were >10-fold fewer intra-tumoral immune niches [80]. Thus, the critical importance of Tex^prog cells for anti-tumor immunity has emerged. However, the durability of these stem-like cells in tumors of cancer patients remains to be elucidated.

4.1. Resident memory T cells

Pioneering work in the early 2000s discovered memory CD8⁺ T cells in peripheral tissues, which were initially thought to be T_EM in recirculation from the blood [81, 82]. Early findings in models of vesicular stomatitis virus and listeria monocytogenes infections [82] revealed that these T cells were actually durably resident in tissues, and such resident memory (T_RM) responses have since been documented in response to multiple pathogens [83–90]. T_RM cells lack expression of CD62L (L-selectin); which differentiates them from naïve T cells and T_CM cells that require CD62L for entry into secondary lymphoid organs [31]. In order to differentiate T_RM cells from effector and T_EM cells, tissue retention markers, CD49a (VLA-1) and CD103, are typically used. CD49a promotes tissue retention and survival through binding to collagenase type IV [91], while CD103 is a TGF-ß induced molecule that promotes T_RM cell tissue retention by binding to e-cadherin [92]. T_RM cells are also known to express CD69, a marker of T cell activation, which blocks T cell expression of S1PR1 [83], and lack the expression of CCR7, which cooperates with S1PR1 for tissue egress, thereby promoting tissue retention and residency [93, 94]. While CD49a , CD103, and CD69 are typical markers of tissue residency, mouse studies have shown that they are not absolute nor required [88, 95–98]. However, CD69 appears to be a reliable T_RM marker in humans, with tissue T_RM cells in healthy individuals overwhelmingly expressing CD69, but not CD103[21].

A unique transcriptional program regulating T_RM development has also been well established [23, 83, 99]. As previously described, Hobit, Blimp1[22] and Runx3[23] were found essential for the development of T_RM cells in mice, while in humans, Hobit was dispensable[21]. Loss of KLF2, which regulates S1PR1, has been shown to enhance tissue residency [100]. Taken together, these findings have revealed T_RM cells as a distinct memory lineage.

T_RM cells comprise the largest memory compartment in human adults and have been identified in almost all epithelial and mucosal tissues where cancers arise [101, 102]. In some cases, T_RM cells accumulate at sites of antigen persistence [103] and adapt to their surroundings where they reside. In anatomical regions with a high cell turnover rate, such as the lamina propria of the gut, immune cells such as macrophages support the formation of T_RM aggregates [104]. In other barrier tissues, T_RM cells occupy de novo niches, such as repair-associated memory depots in the lung, and mucosa-associated lymphoid tissue in the female reproductive tract [105, 106]. Localization to barrier sites of mucosal tissues exerts a metabolic burden that typically limits the persistence of T cells. However, T_RM cells utilize fatty acid beta-oxidative phosphorylation to support their longevity [107–109]. In contrast to conventional memory T cells which conduct their own fatty acid synthesis, T_RM cells in the skin have been shown to express high levels of fatty acid binding proteins FABP4 and 5, to facilitate the necessary uptake of fatty acids [107]. These metabolic characteristics of T_RM cells that support their function in peripheral tissues may also support their function in tumors. T_RM cells in the tumor microenvironment must compete for nutrients with tumor cells, which use high levels of glucose and glutamine [110]. Reliance on fatty acid catabolism has recently been shown to be essential for CD8⁺ TIL function [111]. The PPAR agonists fenofibrate, and bezafibrate, which both promote fatty acid oxidation, have each been shown to improve T cell anti-tumor activity [111, 112], although it remains to be seen if these effects are due to improved T_RM responses. Regardless, the metabolic requirements of T_RM cells may make them ideally suited to persist and function in a metabolically hostile TME.

4.2. T_RM responses to cancer: Mouse studies

A substantial body of work from murine tumor models has revealed a central role for T_RM cells in immunity to cancer. A study in 2010 first introduced the concept of tissue-localized tumor immunity in response to infection through various routes with recombinant vaccinia virus expressing OVA_257-264 [113]. Optimal protection against dermal B16-OVA rechallenge was only afforded by prior infection in the skin [113]. This study further identified OVA-specific T cells in the skin, and referred to them as “skin-resident T_EM cells” [113].

Studies in 2017 established that tumor Ag-specific memory T cells can become durably resident in the skin, where they are necessary for long-lived protection against B16 melanoma [114]. Following Treg cell depletion and surgical excision of dermal melanoma, mice that developed autoimmune vitiligo harbored large skin populations of CD69⁺ CD103⁺ CD8⁺ T_RM cells specific for the melanoma/melanocyte shared antigen gp100 [114]. The absence of CD103 on CD8⁺ T cells impaired the establishment of T_RM cells and eliminated tumor protection [114]. The concept of generating skin T_RM responses against tumor/self-antigens was subsequently illustrated in studies involving intradermal DNA vaccination against gp100 [115].

Additional studies solidified a role of T_RM cells in immunity directed against tumors expressing viral antigens. T_RM cells were implicated in a human papilloma virus E7 vaccine-induced response against orthotopic TCI head and neck tumors [116]. Use of the drug FTY720, which blocks T cell egress from lymph nodes, in conjunction with elegant parabiosis studies, revealed a minimal role for circulating T cells in tumor protection [116]. In mice bearing B16 tumors expressing a viral antigen, T_RM cells were shown to sustain tumor equilibrium and maintain microscopic tumors in the skin [117]. This was in contrast to mice lacking CD69 or CD 103 in which T_RM responses could not be generated, and macroscopic tumors grew, illustrating that T_RM cells can also mediate cancer immune surveillance [117]. Despite this, T_RM cells may not function in isolation, as a thorough investigation of VACV-OVA vaccination strategies showed that the interplay between T_RM and T_CIRC cells was required for optimal tumor immunity in the skin [26]. Indeed, more recent studies have shown that CD8⁺ T_RM cells facilitate antigen shedding by directly killing tumor cells, thereby allowing dendritic cells (DCs) to relay tumor antigens to the draining lymph node to further prime CD8⁺ T cells [118]. Collectively, these mouse studies illustrate a crucial and multi-faceted role of T_RM cells in mediating cancer immunity and immune surveillance.

4.3. T_RM responses to cancer: Human studies

T_RM cells have recently been shown to exist in multiple human solid tumors. Prior to use of the term “resident memory,” CD 103 expressing CD8 T cells were identified within multiple types of human tumors [119–122]. With CD103 now recognized as a common T_RM cell marker, these studies can be viewed as the earliest evidence of tumor infiltrating lymphocytes (TILs) having T_RM-like properties. Additional characterization of CD8⁺CD103⁺ TILs in high grade serous ovarian cancer (HGSOC) and endometrial adenocarcinoma showed elevated expression of the exhaustion marker PD-1 [123–125]. CD8⁺ T cells in pediatric glial tumors exhibited a CD45RO⁺ CD69⁺ CCR7⁻ T_RM-like phenotype, in addition to multiple inhibitory checkpoints including PD-1, and TIGIT [126]. Using an immunofluorescence technique to visualize T_RM in NSCLC, co-expression of CD49a (VLA-1) was identified on CD103⁺CD8⁺ T cells [116]. This study also showed elevated levels of PD-1 and TIM-3 on CD103⁺CD8⁺ TILs as compared to CD103⁻ TILs. Similarly, in CD103⁺ CD8⁺ T_RM cells in mesothelioma tumors showed higher expression of PD1 and TIM-3 as well as more IFN-γ production compared to CD103⁻ TILS [127]. Accordingly Cy-TOF analysis of melanoma-infiltrating T cells showed that a CD69⁺ subset (among which ~50% expressed CD103), co-expressed high levels of inhibitory checkpoint molecules CTLA-4 and PD-1 [128]. The above phenotypic analyses demonstrated the presence of T_RM-like TILS within the tumor microenvironment.

Transcriptional profiling has further supported the identification of T_RM cells in tumors. NSCLC tumors with a high TIL infiltration score were shown to have more pronounced gene expression characteristics of T_RM cells including higher transcript levels for ITGAE (CD103), CD69, ITGA1 (CD49a), CXCR6, PDCD1 (PD-1), HAVCR2 (TIM3), LAG3, and TIGIT, but lower expression of KLRG1, CCR7, SELL (CD62L), and S1P1 [129]. With the growing availability of single cell RNA sequencing (scRNA-seq), populations of T_RM cells have been more precisely characterized. scRNA-seq of TILs isolated from two patients with triple negative breast cancer (TNBC), coupled with bulk RNA-seq data on sorted TIL populations, showed that CD8⁺ CD103⁺ TILs exhibited multiple features of T_RM differentiation [130]. Another study sequenced TILs from twelve NSCLC tumors and identified clusters with high expression of the T_RM specific transcription factors CD69, CD 103 and CXCR6 [131], and a recent study sequenced TILS from fourteen treatment naïve patients across four different tumor types and similarly identified CD 103⁺ ZNF683 (Hobit)⁺ T_RM cells that also had expression of exhaustion markers [132], Interestingly, when looking at a clonally expanded PDl-expressing T_RM subset in tumors from lung cancer patients, the T_RM, but not non-T_RM cells, were transcriptionally and epigenetically enriched for superior cytotoxicity and functionality [133], suggesting that T_RM are an important component of the anti-tumor immune response.

Evidence that CD8⁺CD103⁺ T cells had prognostic value for patients did not appear in the literature until 2014, when it was shown that this subset was strongly associated with survival in HGSOC[123]. Interestingly, the presence of CD8⁺CD103⁻ cells conferred no benefit when compared to tumors devoid of all CD8⁺ T cells, suggesting that the CD103⁺ CD8⁺ subset dominates the protective response in HGSOC [123]. Similar correlation between CD103⁺CD8⁺ T cells and improved prognosis has also been shown in urothelial cancer [134], breast cancer [135], endometrial cancer [125], and lung cancer [116, 136, 137], Even when CD69⁺ CD103⁺ PD1⁺ LAG3⁺ T_RM cells with high levels of immune checkpoint molecule expression were found in treatment-naïve melanoma tumors, these cells significantly expanded early during anti-PD1 immunotherapy, and the number of CD69⁺ CD103⁺ T_RM cells was associated with improved patient survival [138]. However, this same association has not been true for pancreatic cancer. While CD103⁺ T_RM TILS have been identified in pancreatic ductal cell adenocarcinoma (PDAC) [139, 140], CD103⁺CD8⁺ TIL numbers did not predict survival, nor did CD103⁺ CD8⁺ TILs in intratumoral regions [139]. Interestingly though, a high ratio of CD103⁺ TILs in intratumoral versus stromal locations was predictive of prognosis, potentially indicating the importance of spatial relationships between T_RM-like cell subsets in PDAC [139]. Moreover, highlighting the importance of spatial localization, increased densities of CD103⁺ CD8⁺ TILs in cancer islands within breast tumors was found to be more significantly associated with recurrence free survival than CD8⁺ TILs within stroma [141].

With increased granularity of transcriptional profiling by scRNA-seq, multiple studies have directly correlated the identification of T_RM in patient tumors with improvement in survival. A T_RM gene signature from TNBC patients was predictive of survival in TNBC patients from the METABRIC consortium, and could be used to distinguish breast cancer responders to ICB therapy [130]. As in TNBC, a scRNA-seq derived gene signature was used to stratify patients in the TCGA lung adenocarcinoma dataset; patients with a higher expression the CD8⁺ HOBIT^hi gene signature had a significantly greater overall survival compared to other TIL derived signatures [131]. The studies above clearly show that TILs with features of T_RM portend improved patient survival across multiple tumor types, emphasizing the importance of these cells in human cancer immunity.

4.4. T_RM heterogeneity in tumors

T_RM cells have now been established as an important memory subset in the human tumor microenvironment. New evidence reveals that T_RM-like cells in the tumor microenvironment are a complex population, with multiple subsets [132, 140, 142]. In lung cancer patients, although the T_RM signature including CD69, CD103 and CXCR6 was broadly enriched in all tumor T_RM cells, these cells comprised three clusters: (1) Hobit-enriched CD103^hi PD1^lo, (2) Granzyme K-enriched CD103^lo PD1^hi, and (3) CD103^hi PD1^hi CTLA^hi LAG3^hi TIGIT^hi, the latter of which was most consistent with exhaustion [142]. This work showed that not all T_RM cells express high levels of negative checkpoint molecules. Additionally, scRNA-seq analysis of CD8⁺ T cells from lung, endometrial, colon, and renal cancers identified a CD103⁺ZNF683(Hobit)⁺ T_RM subsets that similarly contained subpopulations with both high and low PD1 expression [132]. In a mouse melanoma model, two populations of TILS with tissue-resident gene signatures were identified, in which the long-lived Id3^hiBlimp1^lo memory-like CD8⁺ population had higher CD69 expression and exhibited features of progenitor exhaustion, while the Blimp1^hi Id3^lo effector-like population had features of terminal exhaustion [143]. Thus, within T_RM subsets in tumors, some exhibit a more stem-like progenitor exhaustion profile and others are more terminally exhausted.

One could speculate that circulating CD8⁺ T_SCM cells, unlike T_EX^prog cells which reside in tumors, may continuously replenish T_RM cells in tumors. Indeed, a recent study showed clonotypic expansion of effector-like T cells in tumor and blood, and surmised that circulating counterparts may replenish the existing TILs with non-exhausted cells [132]. Future studies examining the clonal relationship between different T_RM populations, as well as other memory T-cell subsets, within matched tissues from the same patient will be needed to broaden our understanding T_RM response to human cancers.

5. Conclusions

In summary, the field has identified four memory CD8⁺ T cell subsets that play crucial roles in anti-tumor immunity (Figure 1). In the past few years, a vital role for CD8⁺ T_RM cells in providing durable anti-tumor immunity has become more appreciated. Studies using single cell transcriptomic approaches have broadened our understanding of the heterogeneity within TILs across multiple different cancer types (Figure 2). Prior definitions of dysfunctional T cells in tumors, solely based on their expression of negative checkpoint molecules, was challenged due to evidence that such TILs can also have T_RM or stem-like properties and be capable of potent antitumor function. It is now clear that checkpoints are not on or off, but rather expressed at varying levels across subsets, further leading to heterogeneity. The clonal and developmental relationships and homeostasis among the progenitor/ stem-like cells, T_RM cells, and terminally exhausted cells in tumors require further study. Additional understanding of these subsets will be necessary to improve cancer treatments with ICB therapy.

Figure 2. Transcriptional profiles and functions of memory CD8⁺ T cell subsets in the tumor microenvironment.

T_EM, stem-like memory/T_EX^prog, and T_RM cells are each found in tumors. Black circular arrows indicate self-proliferation of stem-like memory/T_EX^prog cells. Two different subsets of T_RM cells are shown, with one subset exhibiting features of exhaustion. Image created using BioRender (https://biorender.com).

Since each memory T-cell subset (T_EM, T_CM, T_SCM, T_RM) contributes to cancer immunosurveillance with its specific function, developing strategies to promote the persistence of all populations may broaden the clinical benefit of immunotherapy. Circulating memory subsets may serve to guard against metastatic tumor cell seeding through circulation, whereas T_RM cells may be better suited to immunity against cancers growing in peripheral tissue. Future studies should also examine a potential role for recently reported lymph node T_RM cells[144] against cancers in SLOs. Overall, understanding how tumor-specific memory T cells persist and function in circulation, in tumors, and in tissues where tumors arise, should guide clinical strategies to induce curative cancer immunity.

Highlights.

• Memory CD8⁺ T cell subsets mediate long-lived anti-tumor immunity.

• Stem-like memory T cells infiltrate tumors and are capable of self-renewal.

• T_RM cells are comprised of transcriptionally functional and exhausted subsets.

• Stem-like memory T cells and T_RM cells in tumors portend improved patient prognosis.

Abbreviations:

T_EFF

Effector T cell

T_CM

Central memory T cell

T_EM

Effector memory T cell

T_SCM

Stem Cell memory T cell

T_RM

Tissue resident memory T cell

T_EX^prog

Progenitor exhausted T cell

Treg

regulatory T cell

Ag

antigen

TIL

Tumor infiltrating lymphocyte

TME

tumor microenvironment

APC

antigen presenting cell

GVHD

Graft Versus Host Disease

T_SCM

Stem Cell Memory T cell

Sca-1

Stem cell antigen-1

NHP

Non-human primates

CMV

Cytomegalovirus

NSCLC

Non-small cell lung cancer

TREC

TCR rearrangement excision circles

CAR

Chimeric antigen receptor

VACV

Vaccinia Virus

TNBC

Triple negative breast cancer

HGSCOC

High-grade serous ovarian carcinoma

PDAC

Pancreatic ductal cell adenocarcinoma

LUAD

Lung adenocarcinoma

ICB

immune checkpoint blockade

SLO

secondary lymphoid organ