Tissue-resident memory T cells in tumor immunity and immunotherapy

Tissue-resident memory T cells (T_RM) reside in peripheral tissues, patrol their surroundings, and rapidly respond to alarming signals (Jiang et al., 2012; Schenkel et al., 2014; Park and Kupper, 2015; Clark, 2015; Dijkgraaf et al., 2019). These features enable them to potentially serve as critical players in antitumor surveillance and immunity. Early studies in viral infections have revealed that T cells were retained in tissues, including human and murine skin as well as murine small intestine and lung tissues (Zhu et al., 2007; Gebhardt et al., 2009; Masopust et al., 2010; Teijaro et al., 2011). In support of T_RM tissue retention, T_RM down-regulate the expression of markers for tissue egress and express surface molecules that enable their retention in tissues (Kumar et al., 2017; Park et al., 2019; Raphael et al., 2020; Byrne et al., 2020). T_RM differentiation is guided by transcription programs common to both effector and memory T cells (Milner and Goldrath 2018), consistent with their persistence in tissues as memory T cells yet retaining rapid effector function for immune protection. T_RM are defined in multiple infection models to be retained in tissues, as shown in parabiosis experiments in mice, antibody in vivo labeling, and in other approaches that mark and monitor tissue T cells (Szabo et al., 2019; Masopust and Soerens, 2019). Of note, the transcriptome signature of T_RM suggests a “tissue-tailored” retention model rather than a “one size fits all” model, highlighting tissue- and organ-specific immune regulation, also named tissue- and organ-specific “immunostats” (Amsen et al., 2018; Pao et al., 2018). Tissue-specific adaptations may enable T_RM to maintain homeostasis in a site-specific fashion. Intriguingly, transcriptionally and functionally distinct T_RM subsets and T_RM precursors have been observed in the murine small intestine during infection (Kurd et al., 2020). This suggests a high heterogeneity and complexity of T_RM populations.

Compared with their protective role in infectious diseases (Wu et al., 2018; Wilk and Mills, 2018; Muruganandah et al., 2018; O’Hara et al., 2020), the importance and significance of T_RM in tumor immunity are not adequately characterized. Nevertheless, recent studies in cancer-bearing mouse models have indicated a pivotal role of T_RM in antitumor immunity in several tumor types (Nizard et al., 2017; Milner et al., 2017; Malik et al., 2017; Enamorado et al., 2017). Moreover, human studies show an association between tumor T_RM and improved responses to immunotherapy and favorable clinical outcome in patients with cancer (Table 1). These findings fuel interest in T_RM research in cancer immunology and immunotherapy. Here, we review the current understanding of T_RM from mouse infection models and human studies and how these findings have been used to guide studies of T_RM in tumor immunity, as well as implications for the role of T_RM in immune surveillance and immunotherapies.

T_RM are defined based on the expression of specific markers, a distinct transcriptional profile, and their functional retention in tissues, all properties that distinguish T_RM from circulating memory T cells, including effector memory T cells (T_EM) and central memory T cells (T_CM). Based on this definition, CD8⁺ T_RM are found to be a component of tumor-infiltrating lymphocytes (TILs) in patients with solid tumors (Byrne et al., 2020). However, it is important to note that T_RM are loosely defined in different studies in the context of tumor immunity.

T_RM express tissue retention markers, including CD69 (C-type lectin), αE(CD103)β7 (E-cadherin receptor), and α1(CD49a)β1 (VLA-1), and exhibit reduced expression of migration and tissue egress markers, such as CCR7, CD62L, and sphingosine-1-phosphate receptor 1 (S1PR1; Schenkel and Masopust, 2014; Kumar et al., 2017; Szabo et al., 2019). Other core markers identified in both human and mouse T_RM include the chemokine receptor CXCR6, the inhibitory receptor programmed death receptor 1 (PD-1), and CD101, which have inhibitory function (Mackay et al., 2016; Kumar et al., 2017; Wein et al., 2019). The expression of T_RM-associated markers differs between organs and tissue sites. CD103 is most highly expressed by CD8⁺ T_RM in mucosal and barrier sites, is variably expressed by T_RM in lung and skin, and is not expressed by T_RM in lymphoid sites (Kumar et al., 2017; Steinert et al., 2015). PD-1 is highly expressed by human T_RM in exocrine sites, such as the pancreas, and to a lesser extent by lymphoid T_RM, but is not highly expressed by intestinal T_RM (Weisberg et al., 2019). In human tumors, T_RM phenotype can also vary according to the tumor type (Table 1). Interestingly, human T_RM associated with multiple tumor types in different sites express CD103, possibly due to the epithelial origin of many solid tumors, while lung tumor–associated T_RM also express CXCR6 mRNA (Ganesan et al., 2017). PD-1 is expressed by T_RM in tumor tissues and healthy tissues. Tumor-associated myeloid APCs, including macrophages and myeloid dendritic cells (DCs), highly express programmed death-ligand 1 (PD-L1; B7-H1), engage PD-1⁺ T cells, and mediate immune suppression in spontaneous and immunotherapy-induced tumor immunity (Curiel et al., 2004; Zou et al., 2016; Lin et al., 2018). It is unknown if PD-L1⁺ APCs interact with PD-1⁺ T_RM, resulting in T_RM functional alteration in the tumor-draining LNs and tumor microenvironment. Additionally, whether tumor-associated T_RM in a particular tissue site differ phenotypically from T_RM in the identical healthy tissue site requires further investigation.

Transcriptome profiling of the population and single-cell RNA sequencing (scRNAseq) have elucidated core features of T_RM in mouse models and humans and further revealed T_RM heterogeneity and tissue-specific signatures. Population-level RNAseq of mouse and human T_RM has revealed conserved signatures that include genes encoding the surface and intracellular molecules described above, as well as transcriptional regulators (Mackay et al., 2016; Kumar et al., 2016; Hombrink et al., 2016). In mice, the transcription factors (such as Hobit, Blimp-1, Runx3, and Id2 family members) have been reported to play a role in T_RM biology (Mackay et al., 2016; Milner et al., 2017; Milner et al., 2020). Human T_RM do not express elevated levels of Hobit, although Notch expression is up-regulated and is related to T_RM establishment (Hombrink et al., 2016; Kumar et al., 2017). A “master regulator” defining human T_RM development has not yet been identified. However, single-cell transcriptome profiling of T_RM has begun to reveal new insights into distinguished features of T_RM, including their heterogeneity and tissue-specific variations. In humans, scRNAseq analysis of T cells from lung and lymphoid sites has revealed a T_RM-associated tissue gene signature, including cell–cell communication, cell structure, and cell–cell matrix interactions (Szabo et al., 2019). This suggests that cell structure and cell interaction may regulate T_RM formation and maintenance. Transcriptome analysis has also indicated considerable heterogeneity among mouse T_RM populations (Milner et al., 2020).

Transcriptome analysis has likewise identified tumor-associated T_RM subsets in the tumor microenvironment. For example, single-cell sequencing suggests different T_RM subsets in human breast cancer (Savas et al., 2018). Interestingly, PD-1–expressing T_RM may possess superior functionality when compared with PD-1–expressing non-T_RM, as suggested in transcriptome profiling of human lung cancer (Clarke et al., 2019). Based on our current understanding of PD-L1 and PD-1 in tumor immunity (Curiel et al., 2004; Zou et al., 2016; Lin et al., 2018), this finding is unexpected. Transcriptionally distinct subsets of T_RM, including Blimp1^hi and Id3^hi subpopulations, were also identified in a mouse melanoma model (Fig. 1; Milner et al., 2020). Although transcriptome studies have generated some insight into T_RM in tumors, it is important and critical to phenotypically and functionally define the different subsets of tumor-associated T_RM, including PD-1⁺ T_RM. Notably, although tumor-infiltrating lymphocytes show a T_RM-like signature, there are no specific phenotypic and functional markers to define T_RM among different memory T cell subsets in the tumor microenvironment (Sasson et al., 2020). Thus, functional studies following single-cell phenotyping are critical to resolve discrepancies regarding different tumor-infiltrating T cell subsets, including T_RM.

T_RM were originally identified in mice using parabiosis models and in vivo antibody labeling (Anderson et al., 2014; Szabo et al., 2019). Multiple molecules (including CD44, CD69, integrins [CD49a, CD103]), some transcription factors (i.e., Notch, Runx3, Blimp1, Hobit), fatty acid–binding proteins (FABPs), and microbiome-derived metabolites (such as short-chain fatty acid [SCFA]) are reported to be involved in T_RM retention in the tissue. CD44 is a receptor for extracellular matrix and assists T_RM interaction with epithelial cells and collagen (Amsen et al., 2018). CD69 restrains the function of S1PR1 signaling and blocks T_RM egress from tissue (Skon et al., 2013; Mackay et al., 2015). Similarly, CD49a and CD103 function as “anchors” to arrest T_RM within the tissues (Chen and Shen, 2020; Byrne et al., 2020).

The pathways controlling T_RM retention in tumors are poorly defined. Different levels of CD8⁺CD103⁺ T_RM are found in tumor epithelium and tumor stroma in different tumor types in patients (Cresswell et al., 2001; Ling et al., 2007; Djenidi et al., 2015; Wang et al., 2015; Workel et al., 2016; Nizard et al., 2017), suggesting that CD103 may be particularly important for T_RM retention in tumors. Antibody blockade of CD103 or genetic deficiency of CD103 results in reduced tumor-infiltrating T cells and accelerated tumor progression in mice (Sandoval et al., 2013; Murray et al., 2016; Sun et al., 2016; Malik et al., 2017), supporting a role for CD103 in T cell tumor retention. Notch plays a role in controlling maximal CD103 expression in tumor-associated T_RM (Hombrink et al., 2016). Transcription factors, including Runx3, Blimp, Hobit, and KLF2, have been shown to down-regulate homing receptor expression for egress in T_RM and promote T_RM tissue retention in mouse infection models (Milner et al., 2017; Mackay et al., 2016; Skon et al., 2013). The potential role of these transcription factors in T_RM tumor retention remains to be established.

It appears that T_RM retention is subject to their adaptation to and regulation by their tissue of residency. Consistent with this concept, tissue-tailored, variable, and malleable profiles of FABP isoforms are found in murine T_RM after viral infection (Frizzell et al., 2020). Interestingly, microbiota-derived SCFA favors CD8⁺ T cell long-term survival and memory (Bachem et al., 2019). Altogether, these early studies suggest that it is crucial to decode molecular mechanisms by which T_RM gain tissue-tailored “labels” and to characterize the “area code” to control T_RM memory, survival, and function in the tumor microenvironment in different tumor types.

Several models (including linear, asymmetrical, self-renewal, simultaneous, “one cell, one fate,” and “one cell, multiple fates” programs) are proposed to explain T_RM differentiation (Enamorado et al., 2018; Raphael et al., 2020). Moreover, T_RM differentiation is influenced by different factors at the earliest priming stage in the LN, the cytokine environment during differentiation and activation, and finally through tissue-specific influences. At the level of priming, murine Batf3-dependent DCs and human CD1c⁺/CD163⁺ TGF-β–producing DCs can prime T cells for T_RM generation in lymphoid tissues (Mami-Chouaib et al., 2018; Amsen et al., 2018; Bourdely et al., 2020). DC-specific DNGR-1 (CLEC9A) provides optimal signal for murine T_RM generation (Iborra et al., 2016). Even before antigen encounter, naive T cells can undergo “training” in the LNs via interacting with migratory αVβ8⁺ DCs. These DCs activate and present TGF-β to naive CD8⁺ T cells, resulting in T_RM-like features, including up-regulation of CD103 expression and epigenetic modifications of T_RM-related genes (Mani et al., 2019). It is speculated that DCs in the tumor-draining LNs may similarly affect T_RM development in the tumor microenvironment.

T_RM undergo a unique, hybrid effector cell–memory cell differentiation program driven by transcription factors associated with both memory and effector cell characteristics. For example, Blimp1 and Notch are required for T_RM and favor T_EM, whereas Runx3 and Nr4a1 promote T_RM and support T_CM (Milner and Goldrath, 2018). Conversely, T-bet and Eomes inhibit T_RM formation but promote T_EM and T_CM differentiation, respectively. Furthermore, T_RM seem to be reminiscent effector cells via expression of PD-1, IFNγ, perforin, and granzyme B (GzmB) on both mRNA and protein levels (Szabo et al., 2019; Milner and Goldrath, 2018; Ganesan et al., 2017), and they share properties of stem cells, as they may be long-lived and not terminally differentiated (Milner and Goldrath, 2018). Overall, coexistence of these memory-, effector-, and stem-like properties may reinforce the antitumor functionality of T_RM.

Notably, TGF-β promotes CD103 expression and is critical for formation of T_RM in the gut, skin, and lungs (Zhang and Bevan, 2013; Raphael et al., 2020). TGF-β can be highly expressed in the tumor microenvironment and may promote T_RM establishment. Tissue-specific factors are also important for T_RM establishment at specific sites. The cutaneous lymphocyte antigen and chemokine receptors, including CCR4, CCR8, CCR10, and CCR19, are expressed on T_RM resident in the skin, whereas CCR9, CXCR3, and integrin α₄β₇ are revealed in intestinal-resident T_RM (Farber et al., 2014; Amsen et al., 2018; Sun et al., 2019). Murine T_RM in the kidney have enhanced expression of E-selectin and P-selectin (Ma et al., 2017). CXCR6 controls T_RM trafficking to murine lungs (Wein et al., 2019) and is expressed in T_RM in human lung cancer (Ganesan et al., 2017), while murine T_RM migrate to the brain via CCR5 and CXCL10 (Glass et al., 2005; Klein et al., 2005). However, α₄β₇⁺ T_RM are also present in the murine skin (Ohmatsu et al., 2010). CXCR3 can “navigate” T_RM to the lungs, skin, and vagina, whereas P-selectin plays a role in migration of CD4⁺ T_RM to the intestine (Haddad et al., 2003; Jeyanathan et al., 2017; Chen and Shen, 2020). Future studies are required to precisely characterize T_RM-related functional states in different tumors.

Once established in the tissues, there is evidence that T_RM can be maintained for a long period of time. Mouse studies in a parabiosis model show that CD4⁺ and CD8⁺ T_RM generated from site-specific infection do not emerge into the circulation and enter peripheral organs during homeostasis for several weeks (Teijaro et al., 2011; Jiang et al., 2012; Steinert et al., 2015). Additionally, T_RM are identified from intravenous antibody labeling due to their presence in tissues, not in the vasculature (Anderson et al., 2012; Turner et al., 2014). In humans, long-term persistence of CD4⁺ and CD8⁺ T_RM has been demonstrated in transplanted organs in which donor T_RM in lungs and intestines were maintained for over a year after transplant but were not detected in blood (Snyder et al., 2019; Bartolomé-Casado et al., 2019; Bartolomé-Casado et al., 2020; Pallett et al., 2020). Moreover, T_RM frequencies in multiple mucosal barriers and lymphoid sites are stably maintained throughout decades of adult life (Kumar et al., 2018). Together, these findings suggest that T_RM maintenance is tissue specific and integral for tissue homeostasis.

Lung CD8⁺ T_RM generated after infection tend to wane over time, unless persistent antigen stimulation is provided (Uddbäck et al., 2021). Over time, lung T_RM were also found to migrate out of the lung to the associated lymphoid tissue (Stolley et al., 2020). Recent studies have found low levels of potential T_RM in healthy human peripheral blood (Klicznik et al., 2019; Guggino et al., 2019) and in the synovial fluid of individuals with spondyloarthritis (Guggino et al., 2019; Qaiyum et al., 2019). In line with this, scRNAseq analysis identifies a tissue gene signature in a minor fraction of peripheral blood T cells (Szabo et al., 2019). Thus, it seems that T_RM emergence into the circulation may occur, but it may be a rare event. In mice, reactivation of T_RM in secondary infection can result in migration of effector progeny to the local draining LNs (Fonseca et al., 2020; Stolley et al., 2020). Adoptive transfer of purified mouse T_RM can enable T_RM to migrate to and populate different tissue sites in response to systemic virus challenge (Fonseca et al., 2020). While these results suggest T_RM plasticity for reactivation, further studies are needed to investigate the trafficking and retention properties of T_RM in different sites; this is an important research area in T_RM biology.

Tumor-associated T_RM may consist of T_RM already in the original tissue site before tumorigenesis and T cells that migrate from the periphery, which become T_RM due to certain tumor environmental factors such as TGF-β (Fig. 1). Based on results in murine infection models, we propose that tumor-specific T_RM reside primarily in the tumor milieu, where they may locally proliferate in response to antigen encounter in situ without or with minimal exiting from the tumor site. This concept is in agreement with the data that tumor-associated antigen-specific T cells are largely found in tumor tissue rather than in the circulation (Webb et al., 2014; Smazynski and Webb, 2018). Recall immune responses could be initiated at nonlymphoid tissues, such as tumor tissues (Zou, 2005). If so, tumor-infiltrating T_RM may be locally activated to proliferate and combat tumor cells in the tumor microenvironment. Subsequently, they may exit the primary tumor niche and inhabit new sites within the tissue of origin—for instance, metastatic tumor tissues. Given that the majority of cancer patients die from tumor metastases, boosting T_RM-mediated long-term local and systemic memory would be meaningful. Nonetheless, the potential mechanisms controlling tumor-associated T_RM maintenance, replenishment, and function remain to be dissected.

Due to their long-term retention in multiple tissues, T_RM can play an important role in both tumor immune surveillance and immunity at diverse sites, analogous to their role in immune protection to pathogens (Amsen et al., 2018; Gebhardt et al., 2018; Byrne et al., 2020). Recent studies in mice have pointed toward the functional importance of T_RM in immune responses (Park et al., 2018; Beura et al., 2018a; Klicznik et al., 2019; Fonseca et al., 2020). One intriguing possibility is that T_RM may eliminate transformed cells in situ, thereby preventing tumor initiation. However, once a tumor is established, tumor cells outcompete tumor-infiltrating T cells for nutrients, resulting in impaired T cell functionality (Zhang et al., 2017; Bian et al., 2020). Correspondingly, human gastric cancer cells outcompete T_RM for lipid uptake, resulting in T_RM death (Lin et al., 2020). The data suggest that fatty acids may be required for T_RM survival in the tumor niche.

T_RM mediate antitumor immunity directly through production of effector and cytolytic mediators and through the release of cytokines and chemokines for immune cell recruitment and activation. CD8⁺ T_RM can be reactivated by both hematopoietic and nonhematopoietic APCs within the sites, which can shape their functionality (Low et al., 2020). Once stimulated, T_RM release lytic granules containing perforin and GzmB and kill tumor cells, similarly to effector CD8⁺ T cells (Amsen et al., 2018). Furthermore, T_RM can support immune equilibrium in a melanoma mouse model and contribute to tumor control (Park et al., 2019a).

Interestingly, tumor-infiltrating CD8⁺CD39⁺CD103⁺ T_RM elicit more potent cytotoxic and effector functions compared with CD103⁻ counterparts (Franciszkiewicz et al., 2013; Djenidi et al., 2015; Enamorado et al., 2017; Malik et al., 2017; Nizard et al., 2017; Duhen et al., 2018; Park et al., 2019; Sasson et al., 2020). In line with this, CD8⁺CD103⁺ T_RM in tumor, but not in tumor stroma, are a better prognostic factor in patients with cancer (Koh et al., 2017; Dhodapkar, 2018). Human tumor-infiltrating T helper type 17 (Th17) cells are affected by TGFβ in the tumor microenvironment, express CD49, are long-lived memory cells, and mediate potent antitumor immunity (Kryczek et al., 2009; Kryczek et al., 2011). These features suggest that human tumor-infiltrating Th17 cells exhibit and/or gain a T_RM phenotype. Correspondingly, CD4⁺ T_RM with a Th17 signature have been observed in patients with autoimmune disease (Krebs et al., 2020). Interestingly, CD49a⁺ T_RM are the most effective tumor killers among T cells in a melanoma mouse model (Le Floc’h et al., 2007; Djenidi et al., 2015; Murray et al., 2016). It is tempting to speculate that T_RM may possess superior antitumor activity compared with other tumor-associated lymphocytes.

In addition to tumor killing via their direct cytotoxic activity, T_RM function as an immune stimulator. T_RM-derived effector cytokines stimulate local DCs, natural killer cells, and T cells to boost antitumor immune responses (Schenkel et al., 2013; McMaster et al., 2015; Hombrink et al., 2016; Glasner et al., 2018). Furthermore, T_RM more rapidly respond to antigen reexposure compared with circulating memory T cells (Mackay et al., 2012; Schenkel et al., 2014; Ariotti et al., 2014). Therefore, T_RM may play a crucial role against tumor recurrence. Collectively, activated T_RM initiate a system of a rapid, tissue-wide state of alarm for optimal immune protection. Unexpectedly, although T_RM express inhibitory checkpoint receptors, their cytotoxic and effector functionalities are maintained (Ganesan et al., 2017; Savas et al., 2018; Boddupalli et al., 2016). Treatment with PD-1 and PD-L1 blockade results in T_RM proliferation in patients with melanoma (Edwards et al., 2018), and production of high levels of GzmB, TNF-α, and IFN-γ (Djenidi et al., 2015; Ganesan et al., 2017; Behr et al., 2019). However, it has also been reported that T_RM isolated from normal human lung tissue may be more effective in effector cytokine production than their counterparts isolated from tumor lung tissue (Bengsch et al., 2018). A tolerogenic signature has been observed in CD8⁺CD103⁺ T cells with high expression of IL-10 and CTLA-4 and low expression of TNF-α, IFN-γ, and GzmB in a melanoma mouse model (Gabriely et al., 2017). Nevertheless, T_RM are associated with favorable prognosis in many types of human cancer (Table 1). Intriguingly, mouse infection models have recently shown that lung T_RM can migrate to the draining LNs (Beura et al., 2018b) and that stimulation of lung T_RM led to enhanced responses in the lung-draining LNs (Paik and Farber, 2021). These studies suggest that T_RM may coordinate local immunity through fortification of the immune response in the neighboring LNs. If T_RM existed in the tumor-draining LNs, T_RM would be an ideal cell population to inhibit tumor lymphatic spread and metastases. Collectively, all of the above multifunctional T_RM activities make them ideal effector T cells in antitumor immune responses.

There is evidence that T_RM are involved in tumor immunotherapy. Checkpoint therapy boosts T_RM formation in melanoma-bearing mice (Enamorado et al., 2017). PD-1 blockade in combination with T_CM transfer results in 10-fold increase in T_RM and inhibits B16 and MC38 tumor growth (Enamorado et al., 2017). Consistent with mouse studies, PD-1 blockade of T_RM from human non–small cell lung cancer promotes ex vivo cytolysis of T_RM to autologous tumors (Djenidi et al., 2015). Anti–PD-1 therapy leads to potent proliferation of CD8⁺CD103⁺ T_RM in patients with melanoma, and the levels of CD8⁺CD103⁺ T_RM are associated with improved patient survival (Edwards et al., 2018). Therefore, checkpoint blockade supports the notion that targeting T_RM may be therapeutically meaningful. Multiple strategies have been suggested to modulate T_RM to enhance cancer therapy efficacy (Fig. 2 and Table 2).

DC priming is essential for T_RM formation in different models, including tumors (Yu et al., 2013; Wu et al., 2014; Iborra et al., 2016; Shin et al., 2016; Enamorado et al., 2017). CD103⁺ DCs are required for optimal generation of T_RM (Iborra et al., 2016). Both CD1c⁺ DCs (Yu et al., 2013) and CD301b⁺ DCs (Shin et al., 2016) promote T_RM generation from effector CD8⁺ T cells. Given the importance of DCs in T_RM formation, DCs may be used as a vaccine to induce T_RM. Furthermore, IL-15 promotes CD8⁺ T_RM in situ (Sowell et al., 2017). Cytokine nanogel allows transference of high doses of IL-15 to the tumor microenvironment (Tang et al., 2018; Xie et al., 2019) and enhancement of the T_RM pool in the tumor. Interestingly, Bhlhe40 (a transcription factor) orchestrates T_RM survival and functionality and is critical for immunotherapy efficacy (Li et al., 2019). Thus, targeting Bhlhe40 may be an option for cancer immunotherapy. In addition, TGF-β supports T_RM formation (Zhang and Bevan, 2013) and promotes radiation resistance of T_RM (Arina et al., 2019). Blockade of TGF-β results in reduced numbers of T_RM after vaccination (Nizard et al., 2017). Given the general immune-suppressive role of TGF-β, it is challenging to specifically target TGF-β-signaling in T_RM to improve cancer therapy. Future studies will determine whether critical and specific potential TGF-β downstream gene(s) for T_RM formation can be identified.

Adoptive transfusion of preprogrammed T_RM may be a cancer immunotherapy strategy. Adoptive transfer of tumor-infiltrating T cells with overexpression of Runx3 promotes T_RM development, inhibits tumor growth, and improves mouse survival in a melanoma murine model (Milner et al., 2017). T memory stem cells (T_SCM) differentiate into T_RM (Kondo et al., 2017). Adoptive transfusion of T_SCM and chimeric antigen receptor T_SCM (Kondo et al., 2020) may lead to increased T_RM in cancers. Furthermore, in murine models of melanoma, the presence of both T_RM and circulating T cells offers improved protection against tumor challenge compared with only one alone (Enamorado et al., 2017). Thus, adoptive T_RM transfusion may be a reasonable approach in combination with other cancer therapies.

T cells compete with tumor cells for glucose, amino acids, and free fatty acids (FFAs; Zhao et al., 2016; Molodtsov and Turk, 2018; Chen and Huang, 2019; Bian et al., 2020, Wang and Zou, 2020). T_RM express high levels of FFA-binding proteins FABP4 and FABP5 (Pan et al., 2017; Lin et al., 2020). Peroxisome proliferator–activated receptor agonists promote FFA catabolism, accelerate T cell–mediated antitumor immunity, and sensitize anti–PD-1 treatment (Zhang et al., 2017; Chowdhury et al., 2018). The antitumor effect may be partially attributed to T_RM responses. Checkpoint blockade promotes FABP4 and FABP5 expression in T_RM, resulting in higher lipid uptake by T_RM and enhancing T_RM survival (Lin et al., 2020). As such, targeting FFA and FABP can affect T_RM and improve antitumor immunity. Future studies will generate insight into T_RM metabolic features and how to target T_RM metabolism for cancer therapy.

Current evidence shows that T_RM can evoke potent antitumor immune responses. However, our understanding of their phenotype, differentiation, trafficking, tissue retention, and effector function remains in its infancy. This is particularly the case in patients with cancer. Substantial human studies rely on expression of CD49, CD69, and CD103 to define T_RM. A number of mouse tumor-bearing models and human samples have revealed the presence of T_RM phenotype cells within tumors. These T_RM may arise de novo from the tissue or infiltrate as part of the antitumor immunosurveillance. Evidence thus far shows that T_RM presence in a tumor or in response to immunotherapy can be a useful prognostic indicator for improved outcomes. Using high-throughput technologies, including single-cell sequencing, microscopic tissue spatial analysis, and multi-omics studies, we may be able to fruitfully study limited clinical materials to gain critical and novel information on T_RM, leading to developing new cancer treatments via targeting T_RM.

This work was supported in part by research grants from the National Cancer Institute of the National Institutes of Health (CA248430, CA217648, CA123088, CA099985, CA193136, and CA152470 to W. Zou), the National Institutes of Health through a University of Michigan Rogel Cancer Center support grant (P30CA46592), and other National Institutes of Health grants (AI106697, AI128949, HL145547, and AI150680 to D.L. Farber). K. Okła was supported in part by the ETIUDA 7 Research Scholarship of the National Science Center in Poland.

Author contributions: Conceptualization, writing, review, and editing: K. Okła, D.L. Farber, and W. Zou.