The evolving role of tissue-resident memory T cells in infections and cancer

Abstract

Resident memory T cells (T_RM) form a distinct type of T memory cells that stably resides in tissues. T_RM form an integral part of the immune sensing network and have the ability to control local immune homeostasis and participate in immune responses mediated by pathogens, cancer, and possibly autoantigens during autoimmunity. T_RM express residence gene signatures, functional properties of both memory and effector cells, and remarkable plasticity. T_RM have a well-established role in pathogen immunity, whereas their role in antitumor immune responses and immunotherapy is currently evolving. As T_RM form the most abundant T memory cell population in nonlymphoid tissues, they are attractive targets for therapeutic exploitation. Here, we provide a concise review of the development and physiological role of CD8⁺ T_RM, their involvement in diseases, and their potential therapeutic exploitation.

T_RM cells are involved in the control of infections and cancer.

INTRODUCTION

Memory responses enable humans and animals to rapidly mount an effective response to previously encountered pathogens. Memory T cells (T_MEM) constitute an essential component of secondary defense to viruses and other threats to health. T_MEM are heterogeneous with unique transcriptional programs and protein expression that correspond to functionality, frequency, and localization. Reflecting growing appreciation for more complex T cell heterogeneity (1), T_MEM can be broken up into three main subsets: central memory (T_CM), effector memory (T_EM), and tissue-resident memory T cells (T_RM) (1–3). T_CM and T_EM collectively comprise a pool of CD8⁺ circulating memory cells (T_CIRCM) that move through the bloodstream and lymph as a means of providing secondary defense. T_RM have been the subject of intense investigation because of their abundance, heterogeneity in tissues, and residence in nearly all tissues examined in mice and humans including secondary lymphoid organs (SLOs), and nonlymphoid tissues (NLTs) including barrier tissues (skin, gut, etc.) and nonbarrier tissues (brain, liver, etc.) (2, 4–6). Elegant experiments have identified T_RM as nonmigratory (7, 8) and with a unique capacity of coordinating rapid immune responses (9–11). Developmentally, T_RM are thought to arise early on during the peak effector T cell expansion phase because, within tissues, these cells exhibit 90% of signature transcripts that identify and differentiate T_RM from T_CIRCM (6, 12). Despite this central transcriptional understanding, the exact mechanisms that underpin T_RM ontogeny remain poorly understood. Two models have been proposed for the T_RM lineage divergence: one supporting tissue-specific differentiation and local divergence and a second supporting systemic residence memory differentiation and systemic divergence (6). The local divergence model proposes that factors within a specific tissue microenvironment drive the memory cell population toward a T_RM fate. Experimental evidence supporting this model hinges on the entry of memory cells to NLTs expressing factors, such as transforming growth factor–β (TGF-β) and interleukin-15 (IL-15), which can promote differentiation and survival of T_RM (13). In contrast, the systemic residence memory differentiation and systemic divergence model posits that T cells are transcriptionally marked and skewed toward a specific subset fate before tissue entry. Studies have generated direct experimental evidence that naïve T cells with variable or identical T cell receptors (TCRs) can skew progeny toward either a T_RM or T_CIRCM lineage based on exposure to different encounters during priming. For example, interactions of CD1c⁺CD163⁺ dendritic cells (DCs) with naïve T cells can drive a T_RM-specific phenotype (14); monocytes have the ability to drive T_RM differentiation by IL-10–mediated TGF-β release (15), whereas migratory DC (16) and keratinocytes promote T_RM differentiation by activating TGF-β (17). In addition to providing insight into the developmental process, both these models also highlight the uniquely plastic potential of T_RM that can not only be transcriptionally influenced before entry to a tissue but also adapt to unique tissue microenvironments to enable distinct functionality.

T_RM ESTABLISHMENT

Given their unique role as mediators of localized responses within NLTs, the transcriptional changes that T_RM undergo during differentiation have been extensively studied (Fig. 1). Up-regulation of four transcription factors have been identified as important for T_RM development: Runx3, Notch, Hobit, and Blimp1 (12, 18–20). Computational analysis of an in vivo RNA interference screen identified Runx3 as a critical regulator of T_RM differentiation, homeostasis, and expression of tissue-resident genes (12). Use of an adoptive transfer mouse model of melanoma provided evidence that Runx3-deficient T cells failed to accumulate within tumors, resulting in greater growth and mortality, whereas T cells overexpressing Runx3 had greater abundance and resulted in prolonged survival, and greater functionality compared to controls (12). Hobit, a homolog of Blimp1, has been shown to be important for the development of T_RM in mice (21). Using a Hobit reporter/deleter system, it was found that virus-specific T_RM development required Hobit, a transcription factor exclusively expressed in T_RM precursors and absent from T_CIRCM. It was also found that Hobit and Eomes expression displayed an inverse correlation and that expression of Hobit in CD8⁺ T effector cells was regulated by T-bet and Eomes, which induced or suppressed Hobit expression, respectively (21). In addition, T_RM display decreased expression of Klf2, which regulates expression of CD62L and sphingosine-1-phosphatase receptor 1 (S1pr1) (22). Runx3, Hobit, and Blimp1 are all up-regulated within precursor T_RM but down-regulated within T_CIRCM, highlighting that unique developmental lineages are present during memory differentiation (20). More recently, single-cell RNA-sequencing (RNA-seq) analysis identified other essential transcription factors for T_RM generation such as Nr4a2, Junb, and Fosl2. Junb and Fosl2 are both essential to down-regulate T-bet expression, whereas Fosl2 is also a positive regulator of Smad3, a key transcription factor of TGF-β signaling, which is an indispensable cytokine for T_RM development (23). A distinct set of transcription factors has been identified in CD103⁺ lung-resident T_RM, among which Notch has a mandatory role for their maintenance and persistence (20). In the lung, Blimp1 works in conjunction with Hobit, or instead of it, to suppress the expression of genes involved in tissue egress including CCR7 and S1pr1 (19, 24).

Fig. 1. Characteristics of T_RM.

T_RM are identified by CD103 and CD69 expression. CD103 interacts with E-cadherin and promotes T_RM effector function and retention at mucosal tissues. TGF-β and IL-15 promote differentiation and survival of T_RM. T_RM can express several immune checkpoint receptors including PD-1, TIM-3, CTLA-4, and LAG-3, depending on the tissue/context. General transcriptional profile includes expression of Runx3, Hobit, Blimp1, and Notch and down-regulation of T-bet, Eomes, and Klf2. T_RM also display decreased expression of the egress receptors CD62L and S1pr1.

Within NLT, either barrier or nonbarrier, along with transcriptional changes, T_RM down-regulate expression of receptors that promote T cell recirculation including S1pr1, CD62L (L-selectin) (25, 26), and CCR7 (22, 27), thereby reducing their mobility and enabling localization within tissues. In contrast, T_RM up-regulate expression of integrins such as CD103 and CD49a (28, 29). CD103 (integrin αE; encoded by Itgae) pairs with integrin β7 to form a complete heterodimeric integrin molecule aEβ7 at epithelial barriers and enables binding to the epithelial marker E-cadherin, providing insight into the surveillance functions of T_RM in mucosal tissues.

PHYSIOLOGICAL ROLES OF T_RM

T_RM provide a unique role in orchestrating local secondary defense responses. There are three main traits by which T_RM aid in secondary defense: longevity and residence in NLT, immediate cytotoxic response, and recruitment and activation of other defense responders. The high expression of CD49a or CD103 T_RM within NLTs, such as the epidermis, is induced in response to local antigen challenge (29). In addition, some T_RM have increased expression of CXCR3 and medium to low expression of CX3CR1 (20, 29, 30). The increased expression of CXCR3 might provide a mechanistic understanding of the role T_RM play in type 1 responses in certain tissues such as the lungs, as CXCL9, CXCL10, and CXCL11 are ligands for CXCR3 and are commonly released at local inflammatory sites during type 1 responses (31). T_RM express basal levels of cytotoxic molecules such as granzyme B, which are up-regulated upon reactivation, enabling a rapid cytotoxic function. T_RM can also heterogeneously express inhibitory receptors including PD-1, TIM-3, CTLA-4, BTLA, LAG-3, SPRY1, adenosine receptor A2AR, CD39, CD101, and 2B4, although it is currently unclear whether this expression is independent of antigen persistence or is induced upon antigen reexposure (32, 33). The precise role of these inhibitory receptors in T_RM is currently unclear. However, T_RM can be reactivated and mediate effector functions despite the presence of these inhibitors (34, 35).

T_RM have unique roles within the tissues that they reside as first responders and local regulators of defense during pathogen reencounter, which is anatomically and functionally distinct from a primary T cell response. During a primary immune response, naïve T cells are primed in SLOs, such as lymph nodes (LNs) and other internal sites, which drain infected barrier tissues. Once primed, T cells proliferate and enter tissues to initiate effector responses. This can be likened to an “inside-out” response originating in SLO (inside) and migrating out to peripheral tissues where they can mediate effector responses. In contrast, during a recall response, T_MEM residing in peripheral tissues after resolution of infection respond rapidly upon rechallenge activating local immunity and, recently demonstrated in viral infections, can rejoin the circulation and repopulate lymphoid structures. This can be likened to an “outside-in” response originating in tissues (outside) and migrating in to SLO/blood. During this recall response, T_RM produce inflammatory cytokines and develop a broad spectrum of defensive responses by altering the properties of neighboring cells, thereby promoting DC maturation, activation of natural killer (NK) and T cells, and recruiting circulating innate and adaptive cells to tissues. T_RM can also participate in viral clearance by directly killing infected cells due to their cytolytic capacity (36). During recall, T_RM can up-regulate cellular egress receptors but do not undergo terminal differentiation, allowing interconversion between T_RM and T_CIRCM (37). Thus, due to developmental plasticity, T_RM may be able to replenish T_CIRCM populations after reexposure to pathogens or maintain their T_RM fate depending on the tissue microenvironment (10, 37).

T_RM IN VIRAL IMMUNITY

Function, classification, and origin of T_RM in viral immunity

T_RM are positioned in tissues to rapidly mount a defensive response upon a secondary encounter with a pathogen and restrict infection within local tissues (Fig. 2 and Table 1). Although T_RM were initially identified in pathogen entry sites such the mucosae, they are now known to have an emerging role in internal organs including the liver and brain (38–40). In these tissues, it is thought that their protective responses strike a balance between infection control and immunotoxicity. For instance, brain T_RM are sufficient to protect against intracranial lymphocytic choriomeningitis virus (LCMV) infection with minimal immunotoxicity; however, this is exacerbated by T_CM, which produce immunopathological damage in the brain (41).

Fig. 2. Examples of T_RM in viral infections.

T_RM develop in NLTs after viral infections. Brain T_RM express PD-1, and in MuPyV infection models, their recruitment is facilitated by TGF-β release by T_regs (35, 137). In the liver, Hobit⁺ effector T cells were identified as precursors of T_MEM, giving rise to T_CM or T_RM, characterized by up-regulation or down-regulation of Eomes, respectively (21). In the gut, during LCMV infection, T_EF cells are characterized by a Blimp1^hiId3^loKLRG1^hi/intCD127^lo phenotype, which switch to Blimp1^loId3^hiKLRG1^loCD127^hi in the newly generated T_RM, in response to TGF-β (46, 47). Gut T_RM may also exit tissue and convert to other types of T_MEM (37). Influenza-specific T_RM, expressing PD-1, CD103, and granzyme B (GrB), can repopulate lung-draining LNs (66). In the skin, after HSV infection, T_RM are generated from KLRG⁻ T_RM precursors, which up-regulate CD103 expression under the control of local IL-15 and TGF-β production (60). Skin T_RM expansion can be boosted by inflammatory stimuli, such as DNFB (67). IFN-γ, CXCL9, and CXCL10 recruit T cells to HSV-infected female reproductive tract (FRT) and up-regulate PD-1 expression in T_RM (10, 50).

Table 1. Role of T_RM in viral immunity.

CMV, cytomegalovirus; MCMV, murine cytomegalovirus; HSV, herpesvirus; EBV, Epstein-Barr virus; MuPyV, murine polyomavirus; LCMV, lymphocytic choriomeningitis virus; VacV, vaccinia virus; VSV, Indiana vesiculovirus; HBV, hepatitis B virus; HPV, human papillomavirus; SIV, simian immunodeficiency virus; T_RM, tissue-resident memory T cells; T_CM, central memory T cells; T_EM, effector memory T cells; NLT, nonlymphoid tissue; SLO, secondary lymphoid organ; LN, lymph node; HIF-2α, hypoxia-inducible factor–2α; PD-1, programmed cell death–1; Blimp1, B lymphocyte–induced maturation protein 1; KLRG1, killer cell lectin-like receptor subfamily G member 1; Id3, inhibitor of DNA binding 3; Nr4a2, nuclear receptor subfamily 4 group A member 2; Fosl2, FOS like 2; TGF-β, transforming growth factor–β; T-bet, T-box expressed in T cells; FRT, female reproductive tract; IFN-γ, interferon-γ; CXCL9 and CXCL9/10, C-X-C motif chemokine ligand 9/10; VCAM-1, vascular cell adhesion molecule–1; GrB, granzyme B; Hobit; homologous of Blimp1 in T cells; Runx3, runt-related transcription factor 3; S1PR1/5, sphingosine-1-phosphate receptor 1/5; CD62L, selectin L; FA, fatty acid; FABP4/5, fatty acid binding protein 4/5; mTOR, mammalian target of rapamycin; Klf2, Krüppel-like factor 2; TIM-3, T cell immunoglobulin mucin 3; CTLA-4, cytotoxic T lymphocyte antigen 4.

Virus	Host	Tissue	Phenotype	Findings	References
CMV, HSV, EBV	Human	Liver	CD8+CD69+ CD103−	CD8+CD69+CD103− TRM development depended on HIF-2α up-regulation	(44)
HSV1	Mouse	Epidermis	CD103+	Predominantly effector function	(42, 43)
		Dermis	CD103−	Showed higher proliferation
MuPyV	Mouse	Brain, spleen	PD-1	TRM in brain were PD-1+ but PD-1− in the spleen	(35)
LCMV Armstrong	Mouse	Intestine	Blimp1, KLRG1, CD127, Id3, Nr4a2, Junb, Fosl2, TGF-β	Effector phase cells were BlimphiId3loKLRG1hi/ intCD127lo and memory cells Blimp1lo Id3hiKLRG1loCD127hi; Junb and Fosl2 repress T-bet; Fosl2 up- regulates TGF-β	(23, 46)
MCMV	Mouse	Brain	TGF-β	Tregs recruit T cells by producing TGF-β	(137)
HSV-2, LCMV	Mouse	FRT	IFN-γ, CXCL9, CXCL10, VCAM-1	These molecules recruit T cells to infected zone and up-regulate PD-1	(9–11, 50)
HSV-2	Human	Genital tract, in silico	IFN-γ, GrB		(51)
HSV	Mouse	Skin	KLRG1, IL-15, TGF-β	CD8+CD103+ TRM are generated from KLRG1− TRM precursors after infection and entrance in tissue, through IL-15 and TGF-β, which up-regulates CD103	(60)
LCMV	Mouse	Intestine, lung, skin, brain, kidney, salivary glands, brain	Blimp1, Hobit, Runx3, S1PR1, CD62L	TRM development requires up-regulation of Blimp1, Hobit, and Runx3 and down- regulation of S1PR1 and CD62L	(12, 18, 22)
VacV	Mouse	Skin	FABP4, FABP5	FABP4 and FABP5 permit FA uptake for TRM metabolism switch to lipid β-oxidation	(77)
LCMV, VacV	Mouse	Liver, lung	PD-1, mTOR	PD-1 regulates mTOR ensuring T cell metabolism switches to FA β-oxidation	(138)
LCMV, VSV	Mouse	Skin, FRT, spleen	Klf2, s1pr1, s1pr5, CD62L	Antigenic stimulation in NLT enhanced TRM accumulation in SLOs, with a similar signature to TCM and FRT TRM (low Klf2, s1pr1, or s1pr5 but CD62L+).	(78)
Influenza	Mouse	Lung, LNs	CD103, CD69, GrB	CD103+CD69+GrB+ lung TRM populate lung-draining LNs after retrograde migration	(66)
LCMV, VSV	Mouse	Intestine, blood	N/A	TRM can recirculate and differentiate into TEM, TCM, or TRM	(37)
Influenza, LCMV	Mouse	Lung, brain, intestine	TGF-β	TRM accumulate due to TGF-β during lifetime and may cause chronic inflammation	(47, 65, 83, 84)
HBV	Human	Liver	CD103, IL-2, PD-1	Chronic HBV TRM had increased CD103, PD-1, and IL-2. They showed expanded TCR clonotypes and expressed low PD-1. Peptides and IL-2 reactivated dysfunctional hepatocyte-primed TRM	(86, 98, 99)
HBV, HCV	Human	Liver	PD-1, TIM-3, CTLA-4	TRM PD-1, TIM-3, and CTLA-4 up-regulated in chronic infections.	(87, 88)
HPV	Mouse	FRT	N/A	Intramuscular immunization followed by intravaginal boost with specific antigens enhanced TRM recruitment	(96)
EBV	Humanized mouse	Kidney, liver, spleen	N/A	Adoptive transfer of EBV-specific cells control EBV later infections	(139)
SIV	Macaque	Intestine, lung, LNs	N/A	Adoptive transfer of SIV-specific cells reduce SIV chronic phase viral loads	(140)

Successful retention of pathogen-specific T_RM relies on unique tissue properties such as structure and nutrient availability (3). T_RM adapt themselves in the distinct features of different tissues, thereby resulting in distinct phenotypes. This has hindered a clear classification of T_RM based on expression markers and raised a debate about their activation status. Currently, CD103 and CD69 are no longer considered exclusive identification markers for T_RM. For instance, following herpes simplex virus 1 (HSV1) infection in the skin, CD8⁺ T_RM found in the epidermis were mainly CD103⁺ and showed a high effector function, whereas dermal T_RM were CD103⁻ but presented higher proliferative capacity (Fig. 2) (42, 43). It was reported that CD8⁺ T cells specific for nonhepatotropic viruses such as cytomegalovirus (CMV), HSV, and Epstein-Barr virus (EBV) were present in the CD103⁻ subpopulation but not in the CD103⁺ subpopulation (44). Similarly, CD103⁺ T_RM exhibited a higher production of interferon-γ (IFN-γ) than their CD103⁻ counterparts after murine polyomavirus (MuPyV) infection in the brain (Fig. 2) (45). Furthermore, these CD103⁺ T_RM specific for chronic polyomavirus infection and an acute vesicular stomatitis virus infection expressed the inhibitory marker PD-1, while CD103⁺ cells from the spleen did not (34, 35). Notably, despite PD-1 expression, brain T_RM has the ability to respond effectively to antigen reexposure. For this reason, it has been proposed that PD-1 expression in brain T_RM might have a beneficial role by curbing T_RM overactivation that might lead to detrimental immunopathology, without preventing immune activation sufficient to mediate viral clearance (38).

T cells generated during antigen encounter at effector and memory phases of infection have distinct expression profile of immune markers. In the intestinal tissue, two discrete lineages of antigen-specific CD8⁺ T cells have been identified (Fig. 2) (46, 47). The first is constituted of Blimp^hiId3^loKLRG1^hi/intCD127^lo cells, representing tissue-resident T effector cells and is mainly abundant during the early phase of infection. The second is characterized by a Blimp1^loId3^hiKLRG1^loCD127^hi signature, which identifies T_RM and predominates during later infection incidents (Fig. 2).

Mechanisms of T_RM action in viral immunity

In the skin and female reproductive tract, it has been shown that even in the absence of antigen, T_RM patrol tissues extending dendrite-like arms in search of antigens during their inactivated state (48). Upon reencounter with cognate antigen, T_RM limit their motility, boost their proliferative capacity, and alert the tissue to a reinfection (42, 49). T_RM secrete cytokines that trigger rapid adaptive and innate immune responses, including local humoral responses, maturation of local DCs, activation of NK cells, and recruitment of T_CIRC cells (10, 11). These combined actions of T_RM on innate immune activation profoundly alter the local tissue environment, creating a “pathogen alert” state that is sufficient to provide immediate protection from infection, even after challenge with an antigenically unrelated virus (9–11). Moreover, IFN-γ release by activated T_RM up-regulates adhesion molecules such as VCAM-1 and chemokines such as CXCL9 and CXCL10, facilitating the entry of circulating CD8⁺ T cells to the tissues (10, 50). By using a mathematical predictive model, it was proposed that the rapid elimination of HSV-2 latent virus reactivation in genital tissues, despite the low abundance of T_RM, might rely practically on the potent antiviral response mediated by bystander cells (51). An intriguing study using intracranial LCMV infection in mice suggested that T_RM may also be capable of an autonomous cytotoxic response within the brain to mediate viral clearance. In this work, it was found that in mice depleted of circulating CD8 T cells, T_RM adopted effector cell functions upon viral rechallenge and killed infected cells through the release of granzyme B and perforin, achieving pathogen control independently of T_CIRC cells and NK cells (41). However, the extent to which this autonomous cytotoxic T_RM response occurs in other tissues/infections is unclear, and several of the studies outlined above suggest that although the recognition of infected cells by antigen-specific T_RM is necessary for mounting an active response, other cell populations also contribute to tissue-wide protection.

Maintenance of pathogen-specific T_RM in tissues

Despite extensive studies, it is still unclear when, where, and how T cells are committed to transformation into T_RM after pathogen encounter. Two main models have been proposed to explain T_RM origin in the context of viral infections. The first, “one cell, one fate,” model supports that each naïve T cell can only generate one type of memory cell, i.e., T_EM, T_CM, or T_RM. This fate determination might be based on TCR–major histocompatibility complex (MHC) interaction strength, although there is contradictory evidence for what magnitude of signal strength promotes T_RM establishment. For example, following persistent MuPyV intracerebral inoculation in mice, brain T_RM presented a 20-fold higher mean affinity than splenic CD8⁺ T_MEM for the immunodominant epitope of MuPyV (52). This is supported by a study finding that higher-affinity CD8⁺ T cells specific for chronic Toxoplasma gondii were preferentially retained in the brain (53). In contrast, during a systemic MuPyV infection, TCRs with weak affinity preferentially formed more functional brain T_RM (54). This is consistent with a recent study that found that low-affinity TCRs favored the formation of lung T_RM (55).

The second, “one cell, multiple fates,” model advocates that a single naïve T cell can differentiate into any T cell precursor, and its fate might depend on three determinants: asymmetric cell division (ACD), inflammatory/costimulatory signal strength, and antigenic decreasing potential (56). ACD implies unequal inheritance of intracellular components, which will produce two different daughter cells. A correlation between cell stemness and ability of CD8⁺ T cells to undergo ACD has been proposed (57). Using CD8⁺ T cells isolated from spleens of LCMV-infected mice and evaluating ACD based on CD8 cell surface distribution after stimulation, it was observed that both naïve and memory CD8⁺ T cells, but not short-lived effector cells (SLECs) or exhausted cells, maintained the ability to generate asymmetry during mitosis. Transient inhibition of the mammalian target of rapamycin (mTOR) pathway, which is thought to promote memory generation (58), increased ACD rates in CD8⁺ T cells. A higher ACD leads to improved memory potential and viral clearance and increased memory and homing signatures, including up-regulated expression of Il7r, Ccr7, and Cxcr3 (57).

In addition to the antigen-mediated signal strength that governs the level of activation, costimulatory signals and cytokines also determine T cell fate. In this regard, a study evaluated the effects of IFN-α/β and IL-12 signaling on the differentiation of SLEC and memory precursor effector cells (MPECs). In Indiana vesiculovirus (VSV)–ovalbumin–infected mice, adoptively transferred IFNAR^−/− and IL-12rβ^−/−IFNAR^−/− OT-I cells not only failed to generate SLEC but also displayed increased MPEC, suggesting that inflammatory signals, such as IL-12, favor development of SLECs over MPECs (59). Although this study did not investigate the role of IL-12 and type I IFN in T_RM differentiation, it provided evidence about the importance of these factors in T cell fate commitment. Last, the decreasing-potential model proposes that cumulative exposure to antigen during infections and inflammatory signals gradually guides T cell differentiation toward a terminally differentiated state with progressive restriction for T_CM-associated features such as longevity and proliferation but retains features of effector cells such as cytolytic capacity. This model explains the observation that T_RM develop by persistent local antigen exposure in the NLT microenvironment and the fact that T_RM development is supported specifically in tissues that favor T_RM retention and local antigen exposure (60, 61). Using skin infections with vaccinia virus (VacV)–expressing model antigens, it was found that activated CD8⁺ T cells trafficked to VacV-infected skin in an inflammation-dependent but antigen-independent manner (62). In contrast, after viral clearance, there was a 50-fold increase in T_RM formation when antigen remained present in the NLT. During a second skin infection with VacV, T cells developing potent localized inflammatory responses were CD8⁺ T_RM generated by local exposure to antigen and not recruited from other T cell memory pools. Although all the proposed factors might be contributing, several studies have suggested that the context of priming in SLOs might be an important determinant driving T cell commitment to T_EM, T_CM, or T_RM (56, 63, 64). For example, T cells primed in SLOs by type I classical DCs (cDC1) are characterized by a CD127^hiKLRG1^lo phenotype and form precursors that can generate T_RM, although their differentiation program will start only after they reach tissues (60, 63, 65, 66).

While it is clear that local antigen within tissues strongly promotes T_RM establishment, several studies have demonstrated antigen-independent inflammation as a driver of T_RM formation. For example, effector T cells primed with an HSV vaccine were pulled into the female reproductive tract with vaginal application of chemokine, promoting T_RM establishment (50). Consistent with this, in vitro activated T cells transferred intravenously into a naïve mouse were pulled into the skin with topical application of the inflammatory agent 1-fluoro-2,4-dinitrobenzene (DNFB), where they formed T_RM in the absence of antigen (67). Moreover, a recent study demonstrated that brain T_RM were established in mice following peripheral vaccination. While the group did not rule out the possibility that small amounts of antigen were present in the brain, this suggests that a local central nervous system (CNS) infection is not needed for T_RM seeding (68). Last, local inflammation may not even be necessary for T_RM establishment; in a lymphopenic setting, transferred naïve CD8⁺ T cells seeded several NLTs and acquired phenotypic markers of T_RM (69).

Once in tissues after an infection, different cues promote T cell conversion to T_RM. An important question is whether antigen-dependent inflammation is required to initiate such process. It has been observed that the presence of antigen is required for T_RM generation in tissues such as brain or lung but not in other tissues such as skin, gut, and mucosae (38, 39, 66, 67, 69–71). In these latter tissues, inflammatory signals mediated by cytokines seem sufficient to boost T_RM recruitment. In the skin or the gut, two of the most studied tissues where T_RM reside, the release of TGF-β by epithelial cells up-regulates the expression of CD103, which serves as an anchor for retaining T cells in the tissue by binding with E-cadherin (47, 60). The presence of other cytokines, including IL-15 or IL-7, also has a decisive role in T_RM long-term survival in the context of pathogens (72), although it has been reported that T_RM might also be maintained by IL-15–independent mechanisms (73) . Furthermore, the lack of oxygen in hypoxia conditions is considered a pivotal cue for T_RM generation (74). Under these conditions, T cells undergo a large number of changes, acquiring increased expression of CD69, Hobit, Blimp1, or Runx3, and down-regulating of CD62L, S1PR1, Tcf1, T-bet, or Eomes, acquiring properties of T_RM (12, 18, 19, 22, 75). This modulation is accompanied by enhancement of signaling pathways including Notch, Janus kinase (JAK)/signal transducer and activator of transcription 5 (STAT5), phosphatidylinositol 3-kinase (PI3K)/Akt, and Wnt, which promote T_RM survival (76). In addition, a study focused on skin T_RM reported that metabolism switches preferentially to fatty acid uptake and β-oxidation, as demonstrated by the essential role of fatty acid binding proteins 4 and 5 (FABP4 and FABP5) for the maintenance of skin T_RM (77).

T_RM were historically thought to have limited migratory capacity (3). However, this paradigm has recently been challenged by studies that observed T_RM presence in SLOs. Although the precise mechanisms remain elusive, T_RM can exit tissues and colonize draining LNs in a process named retrograde migration (37, 66, 78, 79). By this process, it is thought that T_RM can protect SLOs upon reinfection, by coordinating recall responses with other immune cells and being ready to repopulate NLTs when needed (37, 66, 78, 79). Moreover, CD4⁺CD103⁺ T_RM with transcriptional and clonal profiles of cutaneous T_RM have been found in the circulation and LNs of healthy humans (80). In the lung, where memory cells undergo a rapid contraction after infection, it has been observed that T_RM are relocated to mediastinal LNs (medLNs) via lymphatic vessels to preserve regional immunity. Although T_RM appeared much later in medLNs than in the lung, they represented a more durable memory pool (66). In this context, pulmonary antigen encounter was required for recruitment of pathogen-specific T effector cells and T_RM establishment in the lung (81). The ability of T_RM to rejoin the circulation and differentiate into T_EM, T_CM, or T_RM, while preserving the preference to return to their tissue of origin, was also documented by a different study (37). It should be noted that additional mechanisms might control local T_RM maintenance because it has been found that specific niches for lung-resident T_RM at sites of tissue injury provide a source of T_RM maintenance in a manner independent from CD69 (82). Together, these findings outdate the notion that T_RM are permanently parked within tissues, opening new directions in the studies of T_RM migration and plasticity.

T_RM in chronic inflammation

T_RM cannot be fully understood without considering their accumulation in tissues throughout lifetime. Notably, considerable differences have been observed in the function of T_RM among children, adults, and elderly individuals. For example, expansion of influenza-specific CD8 T_RM is very poor during infancy, which some investigators attribute to the higher expression of T-bet of T_RM in children than in adults, potentially explaining why children experience more frequent respiratory infections (83). In aged individuals, in contrast, excessive accumulation of malfunctioning T_RM is associated with chronic lung inflammation and fibrotic sequelae following influenza infection (84). It was found that the higher T_RM presence in aged tissues depends on the increase of TGF-β over time and that these T_RM have impaired TCR signaling and effector function.

In the brain, the enhanced production of IFN-γ by T_RM accumulating during persistent neurotropic infections can activate microglia and promote cognitive impairment (38). In that sense, T_RM play a decisive role in chronic inflammation, where antigen is repeatedly or continuously present. Cells such as keratinocytes, DCs, and fibroblasts are constantly sampling antigens and releasing polarizing cytokines like IL-7, IL-15, IL-17, or TGF-β, which are considered responsible for an aberrant accumulation of virus-specific T cells over time (76). In the context of chronic LCMV infection, however, de novo generation of T_RM is impaired by the down-regulation of TGF-β caused by the antiviral environment, which is balanced by the recruitment of other CD8⁺ T cells (47, 65). This suggests that the proportion of CD8⁺ T_RM versus recruited CD8⁺ T cells may be orchestrated by cytokine levels (47, 65). In support of this concept, acute damage in chronic hepatitis A virus (HAV) infection in the liver correlates with overactivation of bystander CD8 T cells (85).

Not all the T_RM subpopulations show the same behavior during persistent antigen exposure. A CD103⁺ subpopulation is increased in patients with hepatitis B virus (HBV) chronic infection. These cells not only produced higher amounts of IL-2 when stimulated with HBV peptides but also expressed increased levels of the inhibitory receptor PD-1 than their CD103⁻ counterparts (86). Furthermore, PD-1 up-regulation was also reported to be accompanied by higher expression of CTLA-4 and TIM-3 in HBV and hepatitis C virus (HCV) chronic infections, respectively (87, 88). In contrast to CD8⁺ T cells, CD103^hi CD4⁺ T cells can curb lung fibrosis induced by CD103^lo CD4⁺ T_RM, in the context of Aspergillus fumigatus chronic exposure (89). Nonetheless, little is known about how to prevent T_RM inflammatory exacerbation in chronic disease. To date, attempts to treat chronic inflammatory diseases, particularly psoriasis and multiple sclerosis, using drugs that inhibited T cell recruitment were unsuccessful (90, 91), whereas targeting TGF-β down-regulation by exposing skin to ultraviolet irradiation effectively reduced T_RM abundance and ameliorated chronic skin inflammation (92).

Induction of T_RM for viral vaccines and treatment of infections

Because of their ability to provide rapid and robust protection in tissues, T_RM have been considered an attractive target for vaccine development and treatment of various diseases. Several experimental approaches have been developed to generate site-specific T_RM and to elicit T_RM after vaccination, among which the most attractive are represented by the “prime and pull” and “prime and trap” strategies (50, 93–97). In the prime and pull strategy, a first vaccination step (“priming”) is followed by local application of chemokines or local inflammation that enhance T_RM development in the tissue (“pull”). This approach was first applied successfully in mouse models to protect from genital HSV by intravaginal topical application of the chemokine ligands CXCL9 and CXCL10 (50). Generation of focal skin inflammation or scarification, two other prime and pull strategies, was also found to boost responses against HSV and poxvirus, respectively (67, 93, 94). The prime and trap strategy aims in “trapping” or recruiting new T cells to become T_RM by administration of antigen to the tissue of interest that will be, eventually, presented by DCs or other antigen-presenting cells (41, 95). By using this approach, it was found that immunization followed by intravaginal boost using adenoviral vectors expressing modified E6 and E7 human papillomavirus (HPV) proteins induced generation and recruitment of IFN-γ– and tumor necrosis factor–α (TNF-α)–producing HPV-specific CD8⁺ T cells to the cervicovaginal tract (96).

The concept of T_RM enhancement may be extended to SLOs. Studies in the context of HIV vaccination showed that local restimulation within specific NLTs increased the abundance of T_RM in the draining LNs, which are a reservoir of the virus and thus considered a potential target to increase CD8⁺ T_RM immunosurveillance (78, 97). It would be valuable to identify which T_RM clonotypes populate each tissue to develop more specific strategies to enhance their effector function. Cheng et al. (98) found that HBV-specific CD8⁺ T_RM populations in the liver were formed by clonally expanded cells, and some of the clonotypically conserved αβ TCRs were present in both healthy liver margin and hepatocellular carcinoma (HCC) tissues. This finding points to the use of peptides together with IL-2 as a strategy to reactivate HCC-specific T_RM as previously done for HBV-specific T_RM in the liver (99).

Last but not least, adoptive T_RM transfer might constitute a promising therapeutic approach against viral infections. These strategies will be particularly valuable for the control of viral infections in immunocompromised hosts after allogeneic hematopoietic stem cell transplantation, where ongoing attempts using ex vivo generated virus-specific T cells have provided promising results (100).

THE ROLE OF T_RM IN CANCER

Evidence of T_RM involvement in cancer

The role of T cells in antitumor immunity is well established, and the important contributions of T effector cells (T_EF) and T_MEM in the therapeutic responses to cancer immunotherapy have been extensively studied. Because of their unique functions including retention in tissues and rapid response to rechallenge, T_RM can be actively involved in cancer immunosurveillance and antitumor immunity. Increasing evidence from experimental work in mouse tumor models and patients’ samples supports an important role of T_RM in cancer immunology and a potential therapeutic utility of T_RM in tumor immunotherapy.

Similarly to pathogen-related T_RM, cancer-related T_RM are defined by the expression of CD103, CD69, and/or CD49 (Fig. 3A) (101). The discovery that Runx3 is an indispensable transcription factor for T_RM differentiation was made in a tumor model and suggested the potential involvement of T_RM in antitumor immunity (12). Using single-cell RNA-seq, subsequent studies determined that, similarly to infections, T_RM detected in the context of cancer display high heterogeneity and identified effector-like Id3^loBlimp1^hi and memory-like Id2^hiBlimp1^lo T_RM subsets with distinct transcription programs and capacities for effector function and memory potential (46). The biological relevance of these findings is supported by multiple observations that have identified T_RM in various types of human cancer including melanoma, non–small cell lung cancer (NSCLC), urothelial cancer, squamous cell carcinoma of head and neck, ovarian cancer, and breast cancer (102–106).

Fig. 3. T_RM in cancer illustrated by melanoma models.

(A) T_RM express immune checkpoint markers and cytokines. Tumor-infiltrating T cells expressing CD49a or coexpressing CD49a and CD103 secrete IFN-γ and granzyme B (134). T_RM in melanoma patients express immune checkpoint receptors such as PD-1, PD-L1, TIM-3, LAG-3, and CTLA-4 (33, 102). (B) T_RM mediate antitumor immunity in melanoma. CD103⁺ CD8 T_RM produce IFN-γ and play a key role against melanoma rechallenge (107). T_RM promote melanoma immune equilibrium (108). T_RM-like VHL KO CD8⁺ TILs up-regulate CD103 and promote antitumor activity (109). (C) T_RM correlate with prognosis in melanoma. The abundance of CD103⁺ T_RM correlates with improved 5-year survival rate (102). In advanced-stage melanoma, T_RM expressing high levels of the retention integrin, CD49a, or high P selectin imply better median overall survival rate (134). Patients whose metastatic LNs are composed of LN-T_RM have longer overall survival (123).

Elegant studies provided evidence for the causative role of CD103⁺CD69⁺ T_RM in antitumor immunity (Fig. 3B) (107). Malik et al. used a mouse model of melanoma-associated vitiligo (MAV) induced by depletion of regulatory T cells (T_regs) and surgical excision of primary dermal B16 melanoma, in which they infused congenic pmel T cells, which carry a TCR that recognizes the melanoma antigen gp100, to study the function and phenotype of antigen-specific responses in the skin. In this model, skin T_RM were generated naturally, expressed CD44^hi CD62L^lo CD103⁺ CD69⁺, and lacked PD-1 and LAG-3, but half of them also expressed cutaneous lymphocyte antigen. T_RM were highly enriched at the skin compared with spleen and LNs. Ex vivo stimulation of pmel cells from the skin produced higher levels of IFN-γ than LN pmel cells. Notably, although CD103⁺ and CD103⁻ T_RM-like cells could induce vitiligo, only CD103⁺ T_RM could protect from rechallenge with implanted melanoma (107), providing the first experimental evidence for the indispensable role of CD103 in mediating the antitumor function of T_RM.

A different study used an epicutaneous melanoma mouse model to investigate the function of T_RM in cancer-immune equilibrium (Fig. 3B) (108). In this study, 40% of mice did not develop melanoma in the skin and were defined as nondevelopers. Comparison among tumor area, peritumoral area, and nondeveloper skin revealed that CD69⁺CD103⁺ T_RM were highest in nondeveloper skin, peritumoral area, and tumor, respectively. The investigators suggested that controlling B16 melanoma growth required dynamic interaction between T_RM and cancer cells. Moreover, by using CD69 knockout (KO) and CD103 KO mice, this study provided evidence for the causative role of these molecules in the generation of T_RM that regulate cancer-immune equilibrium, because CD69 KO and CD103 KO were more susceptible to melanoma formation (108).

Using mice with von Hippel–Lindau (VHL) deficiency, it was shown that up-regulation of hypoxia-inducible factor–1α (HIF-1α) induced constitutive elevation of CD103 expression and promoted cytokine production and cytotoxic function of T_RM-like CD8⁺ tumor-infiltrating lymphocyte (TIL) in tumor (Fig. 3B) (109). VHL-deficient TILs expressed core T_RM transcription factors including Egr2, Runx3, and Prdm1 and down-regulated expression of T-bet and Eomes, whereas the CD103⁺ subset displayed decreased Klf2 expression preventing T_RM egress. Treatment of VHL-deficient mice with an anti-CD103 antibody resulted in inability to suppress tumor growth. These findings suggest that HIF-1α functions in a manner dependent on CD103 integrin, which enhances retention of T_RM in tissues, and provide evidence for the beneficial role of T_RM in cancer (109).

Phenotype and function of T_RM in cancer

As mentioned above, T_RM have transcriptional and immunophenotypic features that distinguish them from circulatory memory T cells and other TILs. Studies in tumor-reactive CD8⁺ memory T cells demonstrated the effector function of T_RM in cancer (12, 108, 110). In the context of cancer, TGF-β can induce CD103 expression through binding of Smad2/3 and NFAT-1 transcription factors to promote and enhance elements of the ITGAE gene and ITGB7 gene that encode CD103 (αE) and β7 subunits of the aEβ7 integrin, respectively (103, 106, 111–114). Furthermore, TGF-β can activate and strengthen CD103–E-cadherin adhesion (115). CD103 may be an important integrin that mediates T_RM residence and potential effector functions in TGF-β–rich tumor microenvironments (101). Interaction between CD103 and E-cadherin promotes phosphorylation and triggers activation of extracellular signal–regulated kinases 1 and 2 (ERK1/2), lytic granule polarization, and T cell effector function (116, 117).

Similarly to their profile in other locations, T_RM residing in tumors express several immune checkpoint receptors such as PD-1, TIM-3, CTLA-4, and LAG-3, and transcription factors such as Blimp1, Hobit, and Runx3 that are indispensable in T_RM differentiation and function (12, 18, 117–119). In some contexts, these cells retain polyfunctionality and cytolytic capacity despite expressing high level of immune checkpoint markers and respond to PD-1–blocking immunotherapy (119). In addition, it was reported that the expression of PD-1 and TIM-3 is correlated with IFN-γ levels and cytotoxicity, as cancer-infiltrating PD-1⁺CD103⁺ T_RM were capable of inducing robust cytokine production after pharmacologic stimulation (120). Consistent with these observations, T_RM in breast cancer patients express high levels of PD-1, although in this context TIM-3 and LAG-3 were not detected (121). Furthermore, T_RM in lung cancer express cytolytic proteins granzyme A and granzyme B (122), providing additional evidence that T_RM in cancer patients have effector functions.

To investigate in-depth the role of T_RM in antitumor immunity at metastatic locations, Molodtsov et al. (123) generated a MAV mouse model in which CD8⁺ T cells were accumulated in vitiligo skin region. In MAV mice, melanoma protection is sustained in the dermis and against intravenous rechallenge in the lungs (124). After adoptive transfer of congenic pmel T cells, which recognize the melanocyte/melanoma antigen gp100, and eradicating T_reg using anti-CD4 antibody treatment, pmel cells strongly expressing the CD103⁺CD69⁺CD62L^lo T_RM signature accumulated in the skin and tumor-draining LNs. Pmel cells expressing CD103⁺CD69⁺CD62L⁻ cells also accumulated in lung and liver, albeit in the small numbers. This study provided evidence that in the context of cancer, tumor-specific T_RM can populate not only nonlymphoid but also lymphoid organs and localize throughout multiple tissues. Single-cell RNA-seq of pmel-T_RM showed distinct phenotypic markers and transcription factors depending on their localization. T_RM from skin expressed CD103, CD69, Cxcr6, and Nr4a1 and lacked T_CIRCM markers. A subpopulation of LN T_RM expressed high levels of CD103, Cxcr6, and low Klf2 and S1pr1 T_CIRCM markers, similarly to skin T_RM. In contrast, T_RM-like pmel cells isolated from lungs expressed Gzma, Itgax (CD11c), and Fabp3. In immunofluorescence microscopy, many CD103⁺ LN T_RM were located in the T cell zone and subcapsular region that is characterized by cells producing IFN-γ. Phenotypic analysis of these cells depicted the high expression of IL7r(CD127), Cxcr6, and Cxcr3, which can distinguish CD103⁺ from CD103⁻ T cell subsets. The study further illustrated the antitumor specificity of T_RM using single-cell RNA-seq and single-cell TCR-seq, which identified clonotype signatures of endogenous tumor-specific CD8⁺ T cells in T_RM populations located in tumor-draining LNs. These LN cancer-specific T_RM provided protection against melanoma metastasis in regional LNs in the MAV model and were correlated with improved survival in patients with melanoma. While T_RM have been demonstrated to reside in SLO in the context of viral infections (78, 125), these studies are the first to demonstrate that this occurs in the context of antitumor immunity. Together, these results challenge the dogma established in systems of viral infections that T_RM reside only in primary NLTs and provide evidence that in the context of cancer T_RM can be found not only in regional LNs but also in several distal organs such as liver and lung.

However, not all tumor T_RM are specific for tumor antigens. Virus-specific T_MEM expressing phenotypic markers of T_RM have been found in at least 14 different human tumor types, including brain, endometrial, lung, colorectal, and breast cancer (106, 126–128). Harnessing the potent immune activating functions of T_RM, preclinical studies have identified these intratumoral virus-specific T cells as promising therapeutic targets to trigger antitumor immune responses (128–131). A recent study in patients with HCC demonstrated a link between activated HBV-specific T_RM and infiltration of bystander CD8⁺ T cells into the tumor, providing evidence that therapeutically activating virus-specific T_RM may promote immune recruitment (98).

Transcriptional program of T_RM in response to anti–PD-1 blockade in cancer

A correlation between responses to checkpoint immunotherapy and T_RM expression in tumors is currently emerging (119). This observation raises the tentative clinical utility of T_RM detection as a biomarker of favorable therapeutic response to PD-1–blocking immunotherapy. In immunotherapy-naïve melanoma patients, CD103⁺ T_RM showed early expansion during anti–PD-1 treatment in the responder group (102). In patients with NSCLC treated with neoadjuvant anti–PD-1 immunotherapy, mutation-associated neoantigen-specific TIL expressed hallmark T_RM transcriptional factors and coordinately up-regulated checkpoint inhibitory receptors and T cell activation markers (119). In a VHL-deficient mouse model, T_RM produced high IFN-γ levels in response to anti–PD-1 therapy and resulted in complete regression of B16 melanoma tumors (109). Ex vivo experiments showed that TILs in lung carcinoma are T_RM and can mediate cytolytic activity after PD-1 blockade (103). Together, these findings in experimental tumor models and patients’ samples emphasize the important role of T_RM in enhancing efficacy of PD-1–blocking cancer immunotherapy.

T_RM in cancer prognosis

CD8⁺ T_RM have been reported to correlate with tumor size, tumor grade, and overall survival in melanoma and solid cancers (Fig. 3C and Table 2) (108, 110, 112, 122, 132, 133). In melanoma patients, the abundance of CD103⁺T_RM provides the strongest association with 5-year survival with 50% survival in the T_RM high group compared with 20% in the group with lower T_RM number (102). In advanced-stage melanoma, patients with T_RM expressing high levels of the retention integrin VLA-1 (CD49) or high P selectin display better median overall survival rate (134). Moreover, patients whose metastatic LNs are composed of LN-T_RM have overall survival at least 670 days longer than other patient groups (123). In primary triple-negative breast cancer (TNBC) patients, CD8⁺CD103⁺ T_RM were significantly correlated with improved relapse-free and overall survival rates after standard chemotherapy (108, 110). T_RM numbers also correlate with prognosis for relapse-free and overall survival rate in basal-like subtype of breast cancer (104). In lung cancer (103), ovarian cancer (135), and other solid tumors (106), CD103⁺ TIL correlates with improved patient survival rate. In patients with cholangiocarcinoma, high proportion of CD69⁺CD103⁺ cells expressing coinhibitory receptors in intrahepatic cholangiocarcinoma tissues displayed a significant correlation with response to immune checkpoint inhibitors (136). In addition, analysis of samples from melanoma patients from The Cancer Genome Atlas by multiparameter flow cytometry and multiplex immunofluorescence staining provided evidence that T_RM abundance is a strong predictor of survival (102). These extensive studies provide compelling evidence that T_RM are important prognostic biomarker in cancer patients and provide evidence for the important role of T_RM in antitumor immunity.

Table 2. Role of T_RM in tumor immunity.

Cancer	Host	Tissue	Phenotype	Findings	References
Melanoma	Human	Tumor	PD-1, TIM-3, PD-L1	TRM express PD-1, TIM3, and PD-L1 in the tumor.	(33)
Melanoma	Human	Tumor	PD-1, LAG-3, 2B4, CD137, granzyme B, CD137, HLA-DR	CD69+CD103+CD8+ TRM express high PD-1, LAG-3, 2B4, and TIM-3 and moderate expression of granzyme B.	(102)
		Anti–PD-1–treated tumor	IL-15	IL15 levels may influence CD103+ tumor-resident CD8+ T cells. The number of CD103+ TRM tends toward response of PD-1 inhibitor on melanoma patients.
Melanoma	Mouse	Tumor, peritumoral skin, non- developer mice skin after epicutaneous injection	CD103+ CD69+	Nondevelopers’ skin and peritumoral skin have higher number of CD103+CD69+ TRM than in tumor area. TRM control melanoma growth.	(108)
Melanoma	Melanoma- associated vitiligo mouse (MAV) Human	Skin, Lung, lymph node, liver, spleen	ITGAE (CD103), CD69, CXCR6, CXCR3, IFN-γ, IL-7, Klf2, S1p1	Parabiosis shows skin TRM and lymph node TRM share high ITGAE and CXCR6 and low expression of Klf2 and S1p1. TRM profiles differ between tissues Lymph node TRM highly express IL7R(CD127), CXCR6, CXCR3, and IFN-γ. Regional lymph node TRM strongly correlate with prognosis benefit in metastasis melanoma patients.	(123)
Melanoma	Human	Tumor, skin, and blood of long-term survival metastatic melanoma patients	CD69, RGS1, NR4A1, and CXCR6a	Subpopulation of TRM from skin and tumor highly expresses cytotoxic transcripts (IFN-γ, TNF, CCL3, and CCL4) and immune checkpoint transcripts (TOX, LAG-3, PDCD1, and CTLA-4). TRM clonotypes, which express IFN-γ/TNF, have strong prognostic value for patients.	(141)
Melanoma	VHL deficiency Mouse	Tumor	Granzyme A, granzyme B, Blimp1, Egr2, and Runx3	Increased expression of granzyme A and granzyme B in VHL-KO CD69+CD103+ TILs. VHL-KO TRM express Prdm1, Egr2, and Runx3 and down-regulation of Bcl6, Tcf7, and Eomes.	(109)
Breast cancer (TNBC)	Human	Tumor	TIM-3, PD-1, CTLA-4, LAG-3, granzyme B, and perforin	CD103+CD8+ TRM express higher TIM3, PD1, CTLA-4, TIGIT, and LAG-3 and increased expression of granzyme B and perforin as compared with CD103−CD8+ T cells.	(110)
Lung cancer (NSCLC)	Human	Tumor	PD-1, TIM-3	CD8+CD103+ TILs display characteristics of tissue-resident memory T cells and express PD-1 and TIM-3.	(103)
Lung cancer	Human	Tumor	Granzyme B, perforin, CD107a, and IFN-γ	CD103+ TRM express 4-1BB, PD-1, TIM-3, and link to cytotoxicity function such as granzyme B, granzyme A, perforin, and CD107a, and produced IFN-γ.	(112)
Lung cancer (NSCLC)	Human	Anti–PD-1– treated tumor	HOBIT, BLIMP1, PD-1, CTLA-4, TIM-3, TIGIT, CD39, and IL-7lo	Transcriptional programming of mutation-associated neoantigen (MANA)– specific TIL after immune checkpoint therapy expresses TRM transcription program. They highly express HOBIT and BLIMP1 and up-regulate PD-1, CTLA-4, TIM-3, TIGIT, and CD39.	(119)
Cholangiocarcinoma (ICC)	Human	Blood and tumor	PD-L1, Wnt/β-catenin, TGF-β	Tumor margin and core density have a higher density of CD103+ CD8+ TILs. ICCs with high proportions of CD69+CD103+ cells display higher levels of PD-L1. ICCs with lower proportions of CD69+CD103+ CD8+ TILs are enriched for genes related to Wnt/β-catenin and TGF-β pathways.	(136)
Ovarian cancer	Human	Tumor	PD-1, TIM-3, CTLA-4, LAG-3	PD-1 and CD103 coexpress within CD8+ TIL compartment. Ex vivo PD-1+CD103+ CD8 TILs produce cytokine after pharmacologic stimulation and express TIM-3, CTLA-4, and LAG-3.	(135)
Cervical cancer	Human/mouse	Tumor	ITGAE (CD103) CD137, CTLA-4, PD1, and PD-L1	Low number of TRM is associated with poor prognosis, and CD103+CD8+ T cells express CD137, CTLA-4, PD-1, and PD-L1.	(142)
Head neck cancer (HNSCC)	Human	Tumor	CD39+, CD103+, PD-1, PD-L1, CTLA-4, TIM-3, CD127, KLF2, CD62L, S1PR1	TRM are characterized as CD39+, CD103+ T cells. TRM express PD-1, CTLA-4, TIM-3 Higher frequencies of TRM in patients are associated with better overall survival.	(106)

CONCLUDING REMARKS

CD8⁺ T_RM under steady-state conditions are excluded from the circulation and reside in tissues. T_RM reside preferentially in mucosal tissue, such as the lung, gut, and skin, and are typically identified by CD103⁺CD69⁺CD62L^lo phenotype. T_RM differentiation is driven by the expression of the transcription factors Runx3, Hobit, and Blimp1 and simultaneous down-regulation of Klf2, which drives expression of S1pr1 and Ccr7 that promote T cell egress from NLTs. T_RM have an essential role in immune defense against pathogens and cancer and are currently emerging as key mediators of responses to checkpoint immunotherapy and as biomarkers with strong correlation with favorable prognosis in cancer. Therapeutic exploitation of T_RM might improve the efficacy of cancer immunotherapy.