Tissue-resident memory T cells in gastrointestinal cancer immunology and immunotherapy: ready for prime time?

Abstract

Tissue-resident memory T (T_RM) cells have emerged as immune sentinels that patrol the tissue microenvironment and orchestrate localized antitumor immunity in various solid cancers. Recent studies have revealed that T_RM cells are key players in cancer immunosurveillance, and their involvement has been linked to favorable responses to immunotherapy as well as general better clinical outcome in cancer patients. In this review, we provide an overview of the major advances and recent findings regarding T_RM cells phenotype, transcriptional and epigenetic regulation in cancer with a special focus on gastrointestinal tumors. Finally, we highlight the exciting clinical implication of T_RM cells in these types of tumors.

Introduction

The tumor microenvironment (TME) represents a complex and dynamic structure with several actors from tumor-fighting cells such as cytotoxic CD8⁺T cells, T helper 1 (TH1) cells to immunosuppressive ones such as regulatory T cells (Treg), myeloid-derived suppressive cells, tumor-associated macrophages to cancer-associated fibroblasts.1 The Composition of the TME dictates not only the patient’s prognosis but also treatment response. Based on the immune contexture of the TME, cancers are now classified into two major groups: ‘cold’ tumors (with low to absent T cells infiltration) and ‘hot’ tumors (with high T cell infiltration).2 Gastrointestinal (GI) cancers are one of the most prevalent and death-leading cancers followed by lung and breast cancers (BC). The majority of GI tumors are essentially classified as cold tumors with the minor exception of tumors with microsatellite instability (MSI). Understanding the interplay between the immune system and the tumor cells has helped tremendously better cancer stratification as well as therapeutic decisions for each patient. An important proof of concept observation in highlighting the key role of the immune infiltrates in the control of tumor, is the positive correlation between tumor-infiltrating lymphocytes (TILs) levels and cancer prognosis observed in a large spectrum of solid cancers including GI tumors. One example is the tremendous success of the ‘immunoscore’ in colon cancer treatment. Based on CD3⁺ and CD8⁺ T cells density in the tumor, Galon et al were able to outline the immune fitness of a given tumor, to better stratify patients according to their prognosis, and to predict their response to therapies.3 4 Remarkably, the immunoscore exceeded MSI status in predicting patients’ survival and disease recurrence.5 Taking together, these observations highlight the key role played by the immune system in shaping cancer progression, and suggest that a comprehensive characterization of the composition of the immune infiltrate in GI cancers is ever-growingly advocated, and is crucial for better treatment decision making.

Maintaining a memory from a previously encountered antigen is one of the most fundamental features of the immune system. On encountering their cognate antigen, naive T cells go through clonal expansion and become effector T cells. Subsequently, these cells undergo a bifurcation in their differentiation leading to the generation of a major subset of short-lived terminally differentiated effector T cells and a minor subset of memory T cell precursors. Generally, memory T cells are subdivided into two major subpopulations: central memory T cells (T_CM) and effector memory T cells (T_EM). Classification of these subsets relies on the expression levels of the lymphoid homing molecules CD62L and CCR7.6 While T_CM express both CD62L and CCR7, T_EM on the other hand express neither of these markers. Lately, a novel subpopulation of memory T cells named ‘Tissue-resident memory T cells’ (T_RM) has been described. In contrary to T_CM and T_EM that are mainly found recirculating between the blood and peripheral tissues; T_RM cells persist in peripheral tissues.7 8

T_RM cells can induce potent imminent localized immune responses against diverse pathogens. In the central nervous system (CNS), peripheral immunizations led to the generation of robust T_RM cells conferring long-term protection against CNS infection.9 Additionally, in the brain during a secondary challenge, T_RM cells autonomously cleared lymphocytic choriomeningitis virus (LCMV) independently of the presence of circulating memory T cells.10 Along with mouse models, T_RM cells specific to a wide spectrum of viruses such as CMV, hepatitis B virus (HBV), Epstein-Barr virus, and HIV were also detected in diverse human tissues.11–15 Such findings identify T_RM cells as a potent, first-line, self-sufficient adaptive immune response in non-lymphoid tissues.

T_RM cells have been reported in several human tissues such as the pancreas, liver, intestine, skin, kidney, brain, lung, as well as lymph nodes, adipose tissues, and salivary glands.16–26 Various studies reported a core transcriptional residency program of T_RM cells characterized by high expression of adhesion and retention biomarkers as well as down-regulation of tissue egress ones (S1PR1, CCR7, CD62L).23 27 Expression of CD103 and/or CD69 are considered bona fide markers of T_RM cells. CD103 is essentially restricted to CD8⁺ T_RM cells, with little expression on CD4⁺ T_RM cells. CD103 or Integrin α_E pairs with integrin β₇ to form a heterodimeric receptor that binds to E-cadherin and enables tissue retention of CD103-expressing T cells.28 29 Similarly, CD69, a type II C-lectin counteracts tissue egress by down-regulating the expression of sphingosine-1-phosphate receptor-1 (S1PR1).30 31 Consequently, the expression of these biomarkers by T_RM cells sustains their peripheral tissue retention. Furthermore, a recent study conducted on over 50 000 activated and resting human T cells from lung, lymph nodes, bone marrow, and blood revealed that both resting and activated CD4⁺ and CD8⁺ T_RM expressed the canonical biomarkers CXCR6 and CD49a.32 CXCR6 is a chemokine that promotes extra-lymphoid tissue homing for lymphocytes and was reported to regulate T_RM cells’ localization to the airways.33 CD49a (also known as Integrin α₁ subunit) pairs with integrin β₇ to constitute VLA-1 (Very Late Antigen 1) which binds to collagen IV.34 35 CD49a was described as a regulator of cutaneous T_RM cells persistence and is a biomarker for highly cytotoxic T_RM cells.36 37 This T_RM signature was largely enriched in lung tissue followed by the lymphoid site.38 Interestingly, although T_RM cells are defined as memory T cells expressing tissue residency biomarkers, these cells display phenotypic as well as functional tissue-specific heterogeneity dictated by the tissue environment, which will be discussed in detail later in this review.

One of the major limitations in studying T_RM cells in human tissues is the exclusive use of the phenotypic characterization for T_RM cells identification, which reliability is highly questioned as certain of the tissue residency markers mentioned above (eg, CD69, CD49a) could also be expressed by circulating T cells. Results from solid-organ transplantation and antibody-depletion studies allowed the identification of human T_RM cells based on their functional characteristics. In a study conducted on patients who had received pancreatic-duodenal transplantation (Tx), Zuber et al reported the presence of donor CD8 T cells for more than 600 days in some non-rejected transplants.39 In parallel, by studying normal and transplanted human small intestine (SI), Bartolomé-Casado et al demonstrated that the majority of both intraepithelial (IE) and lamina propria (LP) T cells were tissue-resident and persisted over a year in the graft.40 Interestingly, the IE and LP T_RM cells were phenotypically, clonally, and functionally distinct, with LP CD8⁺ T_RM cells exhibiting higher polyfunctionality and more potent cytokine production (GZMB, perforin) (table 1). Another study showed that the use of low-dose alemtuzumab, an anti-CD52 antibody that depletes circulating blood T cells and effectively treated patients with leukemic cutaneous T cell lymphoma, induced depletion of T_CM cells but sessile, non-circulating skin T_EM cells were spared and showed higher IFN-γ production.41 Additionally, alemtuzumab-treated patients presented lower marked infections despite the total absence of circulating T cells. These observations provided evidence of the existence of non-recirculating skin resident T cells, that are capable of providing frontline protection against infection even in the absence of circulating T cells recruitment.

Table 1.

Phenotypic and functional heterogeneity of T_RM cells across organs

Localization		Phenotype	Functional characteristics	Reference
CD8 TRM
Gut	mice	Id3low Blimp1high		50
		Id3high Blimp1 low
	Human	CD103+CD69+	Highly polyfunctional (IFN-γ+/IL-2+/TNF-α+)	51
		β2integrin+CD103−CD69+	Highly cytotoxic (GZMB+)
	IE	CD103+ KLRG1- CD28 low 2B4 high CD161high CD127high PD-1low	Damped polyfunctionality Damped cytotoxicity	40
	LP	CD103+ KLRG1- CD28low 2B4 high CD161high CD127 high PD-1low	Highly polyfunctional (IFN-γ+/IL-2+/TNF-α+) GZMB+, Perforin+
		CD103- KLRG1- CD28high CD161low CD127low PD-1 high NKG2D high	Polyfunctional (IFN-γ+/IL-2+/TNF-α+)
		CD103- KLRG1+ CD28 high CD161 low CD127 low PD-1 high NKG2D high	Highly cytotoxic (GZMB+, Perforin+)
Liver	mice	CD69+ CD103-	Highly polyfunctional (IFN-γ+/IL-2+/TNF-α+)	55
	Human	CD69+CD103+ CXCR6+ CXCR3+	High IL-2 production	11
Pancreas	Human	CD69+ CD103+ CD49a+ CD101+ PD-1+	GZMB+ Highly polyfunctional (IFN-γ+/IL-2+)	16
Skin	mice	CD69+ CD103+	Damped polyfunctionality (IFN-γ+/IL-2+/TNF-α+)	55
		CD69+ CD103+ CD49a+	Increased IFN-γ production	37
Salivary gland	mice	CD69+ CD103- CD69+ CD103+	Highly polyfunctional (IFN-γ+/IL-2+/TNF-α+) Damped polyfunctionality (IFN-γ+/IL-2+/TNF-α+)	55
Lymph nodes	human	CD69+ CD103-	IFN-γ+, IL-2+	16
CD4 TRM
Liver	human	CD69 high CXCR6+ CD49a+ PD-1+ S1PR1- CD69 int CXCR3+ CX3CR1+ CXCR1+	TH1 cytokine production (IL-2, IFN-γ, IL-21) TH2 cytokine production (IL-4)	56
Small intestine	human	CD69+ CD103+ 2B4+ CD49a+ CXCR6+ CD101+ CD161+ CD28low CD69+ CD103- KLRG1+ CD49a+ CXCR6+ CD101+ CD161+ CD28low	TH1 cytokine production (IFN-γ+/IL-2+/TNF-α+) GZMB+, Perforin+, IL-17+ TH1 cytokine production (IFN-γ+/IL-2+/TNF-α+) Perforin+	57

IE, Intra-epithelial; LP, lamina propria.

Most of our understanding of T_RM cell biology stems from studies conducted on T_RM cells discovered in healthy tissues. Understanding the underlying mechanisms of T_RM cells’ differentiation, maintenance, and function in the context of cancer is an ongoing process. Here, we provide an overview of major advances and recent findings regarding T_RM cells phenotype, transcriptional and epigenetic regulation in cancer with a special focus on GI tumors. Finally, we highlight the exciting clinical implication of T_RM cells in these cancers.

Hallmark features of T_RM cells in the GI tract

T_RM cells’ presence was described in a wide range of GI tissues.16 42–45 In the intestine, T_RM cells have been widely studied in mice and were reported to be essentially lodged in the IE niches, expressing CD103 and CD69.46–49 Recently, single-cell RNA sequencing (scRNAseq) analyses identified distinct intestinal antigen-specific T_RM cells subsets based on the expression of transcription factors (TF) Blimp1 and Id3 at different memory time points post LCMV infection.50 These subsets shared transcriptional features of effector like T_EM cells and T_CM cells and they strikingly presented inter and intratemporal heterogeneity. While Id3^low Blimp1^high cells were present at high levels early after LCMV infection, the frequency of Id3^high Blimp1 ^low cells increased by 10-fold during the course of LCMV infection and was associated with an enriched memory gene expression signature.

In addition, the Id3^high cell subset had a more long-lived transcriptional profile and expressed higher levels of CD69 corresponding to a T_RM phenotype, suggesting that this Id3^high Blimp1^low cell subset represents T_RM cells or T_RM precursors. In contrast, in the human SI, T_RM cells represented the majority of CD8⁺ T cells accumulated in both the LP and the epithelium, expressing CD103 and CD69 and demarcated by long-term persistence and polyfunctionality (IFN-γ⁺ IL-2⁺TNF-α⁺).40 Using scRNAseq, Fitz Patrick et al recently identified two major CD8⁺ human intestinal T_RM subsets with distinct phenotypic and functional features; CD103⁺CD69⁺ T_RM cells and β2-integrin⁺CD103⁻CD69⁺ T_RM cells; CD103⁺CD69⁺ T_RM displayed greater polyfunctionality, whereas β2-integrin⁺CD103⁻CD69⁺ T_RM cells had higher granzyme expression.51

In the liver, T_RM cells patrol the hepatic sinusoids, which allow them to interact with a vast variety of circulating cells and antigens.52 Contrarily to gut T_RM cells, liver T_RM were essentially identified as CD69⁺ cells lacking CD103 expression. Liver T_RM do not require CD103 for tissue adhesion but rather LFA-1 expression.53 However, a small subset of liver T_RM cells was identified as CD69⁺CD103⁺CXCR6⁺CXCR3⁺ cells poised for non-cytolytic antiviral effector functions.11 54 In line with the observed heterogeneity in the intestine, a novel study revealed that liver and skin T_RM illustrate two extremes of phenotypic variations.55 While liver T_RM were essentially CD103^- CD69⁺, skin T_RM were exclusively CD103⁺ CD69⁺ cells exhibiting high PD-1 expression. Strikingly, this observed phenotypic variation was associated with a functional heterogeneity mirrored by high polyfunctionality and an increased proliferative potential of liver T_RM cells, while skin T_RM were less polyfunctional but with higher longevity.55 Interestingly, the investigation of T_RM cells phenotype and function in the salivary glands revealed the presence of two T_RM subsets CD103^-CD69⁺ and CD103⁺CD69⁺ endowed with the same functional characteristics as liver and skin T_RM, respectively. Hence, the salivary gland model proved the existence of an intraorgan phenotypic as well as a functional heterogeneity of T_RM cells in addition to the previously discussed inter-organ variability. CD8⁺ CD69⁺CD103⁺PD-1^high T_RM cells were also reported in the pancreas, showing an almost exclusive lodgment in the exocrine areas.16 Consistently with previously discussed results, pancreatic T_RM cells were reported to be phenotypically and functionally distinct from neighboring GI sites (jejunum) and lymphoid tissues. T_RM cells in the pancreas showed higher expression of PD-1 and CD49a compared with T_RM cells in the jejunum and pancreas-draining lymph nodes, in addition to the significantly higher expression level of GZMB.16

Although little is known about hepatic CD4⁺ T_RM characteristics, a recent study has identified two phenotypically and functionally heterogeneous intrahepatic CD4⁺ T_RM subsets according to CD69 expression.56 CD69^high T_RM cells occupied sinusoidal and periportal niches and were characterized by the expression of CXCR6, CD49a, PD-1, and the production of TH1 cytokine. CD69^int T_RM cells had a distinct profile and were characterized by the expression of CX3CR1, CXCR3, and CXCR1, as well as by the production of IL-4. In transplanted duodenum in humans, CD4 +T_RM cells also presented intraorgan heterogeneity with the vast majority of CD4⁺ T_RM being CD69⁺CD103^-, however, a small fraction expressed CD103⁺ and was exclusively lodged in the LP.57 Although both CD4⁺ T_RM subsets displayed a TH1 cytokine profile, CD103⁺ cells produced significantly higher levels of GZMB and contained a small fraction of T_RM cells that produced IL-17.

Although T_RM cells share a common transcriptional foundation to establish tissue residency (figure 1), recent studies gave great insight into their phenotypic heterogeneity which was completely independent of the infection model but stringently tied to their tissue of lodgment. Interestingly, whether T_RM are localized in the GI tract or the skin or the salivary glands, they did exhibit a phenotypic heterogeneity that conferred heterogenous functional properties to these cells (summarized in table 1).

Figure 1.

Summary of the most upregulated/ downregulated genes in T_RM cells identified by Bulk/Sc-RNA sequencing. This spin chart summarizes the core residency signature specific to T_RM cells compared with T_EM and T_CM identified by bulk and/or single-cell analysis. T_RM signature is mainly characterized by high expression of genes encoding for tissue residency and immune checkpoints (in red), both quite tissue-specific, in this figure, we also focused on those GI cancers specific such as CD103, CXCR6, PD-1, and LAG-3. Another remarkable feature of T_RM cells signature is the upregulated expression of cytotoxicity and functionality encoding genes compared with their memory counterparts T_EM and T_CM. Along with this upregulation, certain genes signaling are either completely shut down (in blue) such as tissue egress ones (SELL, CCR7) and TCF1 which inhibit ITGAE (CD103) expression or downregulated such as TBX21 (T-bet) for IL-15R expression maintenance. GI, gastrointestinal; T_CM, central memory T cells; T_EM, effector memory T cells; T_RM, tissue-resident memory T cells.

Finally, recent findings unveiled tissue-specific metabolic requirements of T_RM cells. Although fatty acid-binding proteins ‘FABP4’ and ‘FABP5’ were reported to be important for skin T_RM cells persistence, a novel study by Frizzell et al demonstrated that the expression of FABP encoding genes in T_RM cells is dictated by their tissue of lodgment.58 While skin T_RM cells showed high expression of Fabp4 and Fabp5, T_RM cells lodged in the epithelium of the SI (SI-IEL) showed high expression of Fabp1, Fabp2, and Fabp6, whereas liver T_RM cells expressed high levels of Fabp1. Interestingly, the expression of FABP encoding genes was driven by tissue-specific mediators rather than a residency program.

T_RM phenotype in GI cancer: a field of ongoing investigations

Several independent studies reported the presence of T_RM cells in various cancer types including colon cancer,44 pancreatic cancer,42 melanoma,59 lung cancer,60 head and neck squamous cell carcinoma,61 and bladder cancer (figure 2).62 63 Remarkably, in the majority of these studies, T_RM cells were reported to express the canonical tissue residency biomarkers CD103 and/or CD69. However, it is important to note that CD103 has also been reported to be a hallmark for tumor-infiltrating Treg and is expressed on dendritic cells (DCs)64 65 and that CD69 is a well-established T cell activation marker.31 Therefore, taken solely these biomarkers does not represent an exclusive hallmark of tissue residency. Consequently, the use of a single biomarker to identify T_RM cells is overly simplistic, and in-depth phenotypic and functional investigations are warranted to identify novel, more refined biomarkers for T_RM characterization.

Figure 2.

Phenotypic heterogeneity of CD8 T_RM cells across cancer types. An overview of the reported CD8 T_RM cells phenotype in GI and non-GI cancers showing an inter and intra-organ heterogeneity. HNSCC, head and neck squamous cell carcinomas; GI, gastrointestinal; PDAC, pancreatic ductal adenocarcinoma.

Although T_RM cells phenotype differs between tissue sites and organs, in GI cancers, T_RM cells were essentially identified as CD103⁺, notably in colorectal cancer (CRC), pancreatic cancer, gastric cancer, esophageal cancer, and hepatocellular carcinoma (HCC).42 44 66–68 The high expression of CD103 by T_RM cells is possibly resulting from the epithelial origin of these cancers. A subset of CD8⁺CD103⁺CD39⁺ T_RM has also been described in CRC.69 Transcriptomic profiling of this subset unveiled a T_RM signature characterized by high expression of HAVCR2, CTLA4, and LAYN, while still exhibiting high expression of cytotoxic proteins (GZMA/GZMB/GZMH, TNF, Perforin 1, and IFN-γ). Luckily, the field of T_RM phenotyping is gaining huge momentum since the emergence of sc-RNAseq technology, which helps to deciphering the underlying heterogeneity of T_RM cells in the context of cancer as well as identifying novel T_RM subsets. A recent single-cell study conducted on T_RM cells in lung cancer enabled the identification of a novel uncharacterized tumor-associated antigen specific CD8⁺ T_RM subset defined as CD103⁺TIM-3⁺IL-7R^-.38 Data analysis of this T_RM subset, showed enrichment in transcripts encoding for high cytotoxic activity, as well as inhibitory molecules such as CTLA4 and especially PDCD1 transcripts. Expression of these biomarkers is known to be the most emblematic hallmark of ‘T cell exhaustion’, a state of functionally compromised T cells (reviewed in70 71). Contrarily to effector T cells, expression of these checkpoints wasn’t related to an exhausted state in T_RM, but rather to a functionally active one.38 In line with these observations, single-cell profiling of TILs in BC revealed a CD8⁺ CD103⁺ T cell subset with a core transcriptional profile of tissue residency.72 This subset also exhibited high expression levels of immune checkpoints and presented high levels of transcripts encoding for cytotoxic molecules.

Despite the emergence of few studies investigating the transcriptional landscape of TILs in certain GI tumors such as CRC and liver cancer by single-cell sequencing, transcriptional profiling of T_RM cells has not been completely elucidated on a single cell level yet. Nevertheless, dimensionality reduction and clustering analysis of CD8⁺ TILs in HBV-induced HCC revealed five CD69⁺PD-1⁺ T_RM subsets that could be separated according to CD103 and CD57 expression.73 Consistently, ScRNAseq of HBV specific CD8⁺ T cells revealed two major T_RM subsets coexpressing CD103 and CD69, and despite PD-1 expression, these cells do not display characteristics of exhausted T cells.73 Strikingly, the presence of these intratumoral HBV-specific T_RM cells was correlated with a better prognosis of HBV-induced HCC patients, which might strongly be attributable to their role in controlling tumor growth. Taken together, these transcriptomic analyses helped to mature our understanding of the transcriptional landscape of T_RM cells in certain types of solid tumors as well as the identification of novel tumor-specific subsets of T_RM sharing the expression of residency biomarkers as well as a sustained activation, and tumor reactivity (figure 1). However, further investigations on the single-cell level of GI ones are warranted to decipher the intra-tumoral phenotypic heterogeneity of T_RM cells which may help to improve T_RM subsets identification according to tumor type.

T_RM cells differentiation and maintenance

Cues promoting T_RM cell ontogeny and differentiation remain poorly understood. Two models explaining the generation of committed precursors for memory T cells populations including T_RM cells were proposed: The ‘One cell, one fate’ model, which speculates that naïve T cells are predetermined to give rise to either memory or effector T cells. In this model, TCR-MHC interaction strength may dictate differentiation into T_CM, T_EM, or T_RM cells. The second proposed model is the ‘One cell, multiple fates’ model during which asymmetric cell division occurs allowing the generation of effector and memory T cells from a single naïve T cell. (Reviewed by Enamorado et al).74

On cognate antigen recognition, naïve T cells priming in the secondary lymphoid organs (SLOs) leads to their differentiation into T_RM precursors. Priming-wise, certain cell types have been described to be involved in T_RM cells generation. DC3s, a subset of human DCs CD1c⁺CD163⁺ CD88^- drove the differentiation of CD103⁺CD8⁺ as well as CD103⁺CD4⁺ T cell in vitro via TGF-β production.75 76 Crosspriming by murine DNGR-1⁺ Batf3-dependent DC was reported to be required for committed T_RM precursors generation in the LN (lymph node). DNGR-1 mediated cross-presentation provides type 2 (CD24) and 3 (IL-15, IL-12) signals that lead to T-bet induction promoting T_RM cells precursors generation and retention in the LN.77 Additionally, migratory αV-expressing DCs activate and present TGF-β to naïve T cells in the LN, leading to epigenetic preconditioning of these cells to differentiate into epithelial T_RM.78 Monocytes also contribute to T_RM cells generation. In an autocrine manner, monocytes produce IL-10 that will be fixed on their cell surface IL-10 receptor.79 80 Consequently, IL-10 stimulation will lead to TGF-β release resulting in CD103 expression on T cells (figure 3). Recently, novel findings revealed the importance of type 1 Treg in T_RM cells development.81 CXCR3⁺T-bet⁺ Treg generates bioactive TGF-β necessary for T_RM cells differentiation. However, it’s important to note that a minor proportion of T_RM cells still could be generated in type 1 Treg deficient mice which leads to the hypothesis that other cells may play a role in TGF-β production.81

Figure 3.

From Zero to Hero: T_RM cells in the cancer-immunity cycle. (A) T_RM cells precursors are primed in the tumor-draining lymph node by DCs. (B) Once in the tumor microenvironment and in the presence of TGF-β and IL-15, T_RM cells precursors will differentiate into T_RM cells. Induction of a core residency signature such as expression of tissue retention molecules (CD103, CD69, CD49a, CXCR6) and acquisition of high cytotoxic activity (GZMB, perforin) is under epigenetic regulation and key transcription factors instructions (Blimp1, Hobit, Runx3, Bhlhe40, and Notch). (C) Once fully differentiated, tumor-specific T_RM cells can enact cancer cells eradication via an arsenal of cytotoxic molecules such as perforin and granzyme. CD103 expression enables T_RM cells biding to E-cadherin expressing tumor cells which sustains the immunological synapse, hence triggering lytic granules polarization and exocytosis. DC, dendritic cell; T_RM, tissue-resident memory T cells.

An emerging proof of concept suggests that cytokine milieu might drive T_RM cell signature. TGF-β and IL-15 were reported to be key regulators of T_RM cells’ development and survival (figure 3). IL-15 was essentially reported to assure T_RM cell maintenance and persistence.82 Additionally, IL-15 was shown to promote T_RM cells recruitment from SLOs into mucosal sites through the mTOR signaling pathway activation.83 TGF-β was reported to be a key regulator of T_RM cell differentiation and migration (figure 3). TGF-β induces CD103 expression on T_RM cells precursors in the gut, SI, lung, and skin.84–87 During T_RM cells development, TGF-β downregulates T-bet expression to a residual level, which is necessary for maintaining IL-15 receptor β-chain (CD122) expression and therefore essential to their long-term persistence.82 Falling in line with these observations, sequential exposure of T cells to IL-15 and TGF-β induced the development of CD103⁺ CD8⁺ T_RM cells in the liver.11 Interestingly, notwithstanding the evident role of TGF-β in the differentiation and retention of T_RM cells in peripheral tissues, it has been demonstrated that in SLOs, TGF-β inhibited T_RM cells migration and gut homing via α4β7 expression downregulation.84 Apart from these well-studied cytokines, other inflammatory stimuli such as pro-inflammatory cytokines IL-33, TNF-α, and type I IFN were reported to induce T_RM cells differentiation in the peripheral tissues.74 Interestingly, these stimuli are highly organ-specific, dictated by the inflammatory microenvironment leading to the generation of T_RM subsets with different phenotypes and functions. In the GI tract, IL-21 drove the differentiation of highly activated and cytotoxic T_RM cells from circulatory precursors only during graft-versus-host disease but not under homeostatic conditions.88 In murine polyomavirus (MuPyV) brain infection, IL-21 produced by MuPyV high-affinity TCRs CXCR5^high PD-1^high CD4⁺ T cells drove CD8⁺ T_RM cells differentiation in response to viral infection.17 Additionally, although T_RM cells in the skin required TGF-β signaling for their differentiation, the development of T_RM cells in the liver was TGF-β independent.55 Stimuli driving T_RM cells’ survival and maintenance are also tissue-specific, although IL-2/IL-15 signaling is important for intrahepatic T_RM survival, long-term persistence of antigen-specific T_RM cells in the epidermal niche is TGF-β dependent.89

Taken together, seeing from afar, T_RM cells share a core residency program as well as the same differentiation mechanisms regardless of their location, nonetheless, it is now evident that stimuli driving their differentiation and maintenance are dictated by their tissue microenvironment which explains the observed phenotypic, functional, and metabolic heterogeneity as well as persistence across organs.

Transcriptional regulation of T_RM cells

Gene expression profiling conducted on T_RM cells revealed a unique transcriptional program implicated in the differentiation and maintenance of T_RM cells.19 Six major transcriptional factors have been reported to play a key role in T_RM cells transcriptional regulation; Hobit, Blimp1, Runx3, Notch, T-bet, and recently Bhlhe40 (figure 3). Hobit and Blimp1 have been identified as T_RM-specific TF that instructs T_RM cells differentiation and maintenance in the liver, gut, kidney, and skin by silencing tissue egress genes such as S1pr1, Ccr7, and Tcf7.27 90–92 Runx3 was also reported to play a role in T_RM cells maintenance in a wide range of tissues including the SI. Both in human and mouse, upregulation of Runx3 expression lead to the induction of the core residency signature including expression of tissue retention biomarkers such as CD103 and downregulation of tissue egress ones.93 Another important TF in T_RM regulation is the T-box family member T-bet. Remarkably, T-bet was reported to be down-regulated in T_RM cells, which is needed for optimal CD103⁺CD8⁺ T_RM maturation. Further, the total shutdown of T-bet leads to decreased survival of T_RM cells, which highlights its implication in T_RM maintenance. Lung CD103⁺ T_RM cells exhibited an upregulated Notch signaling signature, reported to be required for their differentiation and maintenance.27 94 95 Interestingly, Blimp1, Runx3, T-bet, and Notch are also key TF implicated in the terminal differentiation of effector T cells and the exhibition of effector functions. This highlights the importance of maintaining an effector function through differentiation into a memory state for T_RM cells.

In contrast to T_CM, T cell factor 1 (TCF1) is downregulated in T_RM cells. Recently, it has been demonstrated that TCF1 binds to the Itgae locus and inhibits CD103 expression.96 Using a global gene expression analysis of Tcf7 sufficient and Tcf7 deficient P14 CD8⁺ T cells, WU et al reported an increased expression of CD103 in Tcf7 deficient cells associated with a decreased expression of tissue egress genes such as Sell and Ccr7.96 Finally, Bhlhe40 is another TF that has been recently described.97 Bhlhe40 or Basic helix-loop-helix family member E40 is a stress-responsive TF that fosters the development, commitment, fitness, and polyfunctionality of T_RM cells via metabolic and epigenetic programming.97 Li et al reported that Bhlhe40 drove the expression of several residency genes such as Cxcr6 and Itgae and that loss of Bhlhe40 expression decreased T_RM cells in mice as well as TILs survival without impacting circulating T cells function following B16 melanoma challenge.97 98 Taken together, these observations demonstrate that T_RM cells’ differentiation and maintenance are controlled by a hybrid transcriptional program of memory and effector cells.

Epigenetic regulation of T_RM cells

Changes in the gene expression profile of T cells are mirrored by a modification in their functional properties. It is well established that epigenetic mechanisms regulate gene expression patterns and dictate whether a gene is expressed or silenced. Of relevance, investigating T_RM cells’ epigenetic profile will offer a deeper understanding of their phenotype, function, differentiation, and maintenance.

In line with transcriptomic analysis, Assay for Transposase Accessible Chromatin analysis of T_RM cells in lung cancer revealed greater chromatin accessibility of CD103 gene promoter (ITGAE).38 Notably, TIM-3⁺IL-7R^- T_RM subset exhibited increased chromatin accessibility of genes encoding the effector molecules IFN-γ as well as granzyme B along with increased accessibility of TIM-3 (HAVCR2) and PD-1 (PDCD1) loci.38 Moreover, a study investigating the DNA methylation profile of PRF1 gene on CD8⁺ T_RM cells in urinary bladder cancer revealed that these cells exhibit signs of exhaustion yet are epigenetically cytotoxic, explained by a low DNA methylation in the reporter CpG site located in the enhancer of the PRF1 locus.63

Genome-wide DNA methylation analysis of tumor-reactive CD8⁺CD103⁺CD39⁺ T cells in CRC has been recently conducted.69 Interestingly, these cells exhibited hypomethylated regions (HypoMRs) that affected both ENTPD1 and ITGAE along with exhaustion markers encoding genes (PDCD1, LAYN, and HAVCR2). Paired with these observations, TFs binding motifs enrichment analysis of CD103⁺CD39⁺ cells revealed 85 significantly enriched TFs binding motifs. Notably, five TFs were particularly overrepresented: BATF, NR4A1, RUNX1, EGR2, and VDR. Unsurprisingly, these TFs are largely associated with T cell exhaustion and CD103 expression. In summary, these epigenetic observations fall in line with previous transcriptomic results, showing hypomethylation of the highly expressed genes encoding for tissue residency, activation, and T cell exhaustion markers, coupled with hypermethylation of tissue egress genes. Altogether, this demonstrates plausible evidence that epigenetic regulation shapes T_RM cells molecular features.

T_RM are strictly non-recirculating cells: time to think twice!

T_RM cells are a population of immune cells defined as permanent resident cells in non-lymphoid organs without recirculation through SLOs or the blood. However, over the last few years, certain studies have demonstrated that some CD8⁺ T_RM cells in the LN are derived from cells that exit non-lymphoid tissue (NLT),99 thereby enhancing the accumulation of antigen-specific CD8⁺ T_RM cells in the draining LN. In addition, a recent study conducted on a murine model reported that reactivated T_RM cells can indeed rejoin the circulating pool.100 Strikingly, they reported that T_RM cells are not terminally differentiated, and are endowed with certain developmental plasticity that allows them to give rise to T_CM and T_EM cells. These exciting results give rise to a new ‘outside-in’ model of protective immune response and reveal an inter-conversion between T_RM and T_CM cells. In line with these observations, using Hobit lineage tracer mice, Behr et al demonstrated that on antigen reencounter, intestinal T_RM cells downregulated Hobit giving rise to an ex-Hobit⁺ circulatory memory subset.101 These T_RM cells-derived offspring referred to as ‘ex-T_RM’ acquired a KLRG1⁺ CX3CR1⁺ T_EM phenotype that was shown to be transcriptionally and functionally distinct from non-T_RM-derived T_EM cells. Interestingly, ex-T_RM secondary T_EM cells presented higher protective potential compared with their non-T_RM-derived counterparts. This work gave new insight into the role of T_RM cells in shaping not only local but also systemic T cells responses on reinfection. Another intriguing work by Mackay’s group showed that T_RM cells’ plasticity is intertwined with the tissue-specific microenvironment which eventually dictates T_RM cells’ malleability.55 In fact, in contrast to CD103⁺CD69⁺ skin T_RM cells, CD103^-CD69⁺ liver T_RM cells were able to trans-differentiate into skin T_RM cells on relocation. Additionally, ex-liver T_RM cells also showed the capacity to repopulate the circulatory memory pools on restimulation. This is indicative of a restrained trans-differentiation capacity of skin T_RM cells and higher plasticity of liver T_RM cells. Whereas liver ex-T_RM cells showed the capacity to differentiate to both skin and liver T_RM, skin ex-T_RM cells were less malleable, showing the ability to only differentiate into skin T_RM on adoptive transfer. TGF-β was the major driver of this lack of plasticity of skin T_RM cells since as discussed above, liver T_RM cells do not exhibit TGF-β imprinting. Finally, this T_RM cells’ developmental plasticity further supports the ‘one cell multiple fate’ hypothesis previously described. However, these interesting findings need to be validated in the cancer setting, meanwhile, they should be taken with a grain of salt.

T_RM cells: local key players in GI cancer immunosurveillance

Cancer immunosurveillance is a process whereby the immune system suppresses cancer development and progression.102 T_RM cells’ presence has been described in the majority of solid tumors as a sub-population of TILs. Evidence merging from animal models as well as studies of human cancers supported a central role of T_RM cells in cancer immunosurveillance. Data from a recent study conducted in a transplantable epicutaneous melanoma mouse model showed that around 40% of mice didn’t develop macroscopic lesions for a long period following epicutaneous inoculation.103 Since the epidermis is hardly accessible by circulating T cells (Tcirc), Park et al speculated that T_RM cells are responsible for controlling tumor growth independently of Tcirc cells presence. Indeed, they reported the generation of CD8⁺CD103⁺CD69⁺ T_RM cells following epicutaneous inoculation in macroscopic lesions-free mice that was correlated with disease control. Whereas depletion of these T_RM cells resulted in tumor growth highlighting the key role of T_RM cells in keeping cancer cells dormant and maintaining immune equilibrium.

As previously described, T_RM cells are endowed with a cytotoxic activity reported to be higher than non T_RM cells. A growing body of evidence supports the hypothesis that T_RM cells are able to enact tumor eradication via an arsenal of cytotoxic molecules such as perforin and granzyme (figure 3). In line with these speculations, in lung cancer, ex vivo analysis of T_RM cells showed higher co-expression of PD-1 and GZMA and GZMB compared with non-T_RM cytotoxic T cells (CTLs).38 Similarly, the transcriptomic signature of CD8⁺CD103⁺CD39⁺ T_RM cells in CRC exhibited high expression of genes encoding for cytotoxic proteins such as GZMA/GZMB and PRF1, probably echoing T_RM cells capacity to kill CRC cells and thus controlling tumor growth.69 In HCC, single-cell analysis revealed that T_RM cells represented up to 90% of intrahepatic and intratumoral HBV- specific CD8⁺ T cells.73 Interestingly, these subsets were clonally expanded, showed no enrichment in T cell exhaustion signaling pathways, and most importantly their presence was correlated with prolonged relapse-free survival in HCC patients. These results suggest the existence of an anti-tumor immune response imposed by T_RM cells. Broadly, these observations evidenced the role of T_RM cells in cancer elimination and in maintaining immune equilibrium in certain solid tumors including CRC and HCC.

Prognostic value of T_RM cells in GI cancers

The prognostic value of T_RM cells is now well established in several cancer types including melanoma, NSCLC, bladder, ovarian and cervical cancer. The presence of CD8⁺CD103⁺ T_RM cells in the TME was associated with a higher overall survival (OS) rate in these cancers.72

Unfortunately, the clinical implication of T_RM cells in GI cancers remains rudimentary at best. Notwithstanding, a positive correlation between T_RM cells infiltration and a favorable prognosis has been reported in certain GI cancers. In a study conducted on a cohort of 165 pancreatic ductal adenocarcinoma patients, IE CD8⁺CD103⁺ over total CD8⁺CD103⁺ cells ratio was significantly associated with improved disease-free survival (DFS) (p=0.022) as well as an improved OS (p=0.009).42 In CRC, a higher number of CD8⁺CD103⁺ T_RM cells was reported to be significantly associated with a better OS rate. In addition, a higher number of CD8⁺CD103⁺ cells was reported to be inversely and significantly associated with distant metastasis.104 Investigations conducted on HBV-related HCC revealed an enrichment of CD8⁺CD103⁺ T_RM cells in the TME, associated with improved OS.68

Paired with these results, a recent study conducted on three independent CRC cohorts’ datasets revealed that ITGAE⁺CD8⁺ infiltrating lymphocytes were associated with a significantly improved OS (p<0.001). Interestingly, uni and multivariate analysis identified ITGAE as an independent prognostic factor for both OS and DFS in CRC patients.105

Taken together, these results highlight the key role of T_RM cells in mediating improved clinical outcomes of numerous solid cancers including GI ones.

T_RM cells in GI cancer immunotherapy: new hope or more hype

T_RM cells: a novel target for checkpoint inhibitors

As outlined above, one common hallmark of T_RM cells in healthy tissue as well as in cancer independently of tumor type is the high expression of a wide range of immune checkpoints, such as PD-1, TIM-3, CTLA-4, or LAG-3. This suggests that T_RM cells could be a prominent target of immune checkpoint immunotherapy (ICI) therapy and rightfully so. Evaluation of gene expression dataset of nivolumab treated melanoma patients revealed that responders’ samples were significantly enriched in T_RM signature at baseline.72 Interestingly, a significant increase of the T_RM signature was observed during nivolumab treatment in these patients. Results of the phase 2 NEOSTAR trial conducted on non-small cell lung cancer patients treated with nivolumab or nivolumab plus ipilimumab in the neoadjuvant setting, revealed a considerable increase of CD103⁺CD8⁺ as well as CD103⁺CD4⁺ T_RM cells in the patient’s group receiving nivolumab plus ipilimumab compared with nivolumab alone.106 Similar results were observed in the retrospective DISCOVERY cohort involving advanced NSCLC patients treated with anti-PD-1 therapy; accumulation of CD103⁺CD69⁺CD8⁺ T_RM cells was noticed following anti-PD-1 administration only in responders associated with an improved overall outcome.107 In line with these observations, analysis of 19 lung cancer biopsies treated with anti-PD-1 therapy showed a significant increase of CD8⁺CD103⁺TIM-3⁺IL-7R^- T_RM cells in responders compared with non-responders and treatment-naive patients.38 Paired single-cell transcriptomics with TCR analysis of cytotoxic lymphocytes of responders to anti-PD-1 in post-treatment samples revealed enrichment of ITGAE expression as well as CD38, GZMB, and GZMH along with TCR clonal expansion. Parallelly, coupled scRNA-seq-TCR-seq of in vitro expanded mutation-associated neoantigens (MANA) specific TILs revealed that 90% of MANA-specific T cell clones in NSCLC treated with neoadjuvant anti-PD-1 were T_RM cells.108 These MANA-specific T_RM were characterized by high expression of CD103, and HOBIT, however, they showed low expression of IL-7R. These results strongly support the hypothesis that anti-PD-1 therapy leads to the expansion of T_RM cells endowed with high tumor-neoantigen specificity.

Although the use of ICI is less common in GI cancers compared with melanoma and lung cancers, some emerging studies highlight the predictive value of T_RM cells density to ICI. Remarkably, a recent study conducted on esophageal squamous cell carcinoma (ESCC) revealed that tumor-reactive CD8⁺CD103⁺ T_RM cells secreted higher levels of IFN-γ and IL-2 after anti-PD-1 blockade when cultured with tumor cells.67 Additionally, using an ESCC mouse model, they demonstrated a higher infiltration with CD8⁺CD103⁺ T_RM in the TME following anti-PD-1 blockade compared with the control group, which indicates a possible resurrection of T_RM cells by the use of checkpoint inhibitors. T_RM cells metabolism relies on the oxidation of fatty acids, thus in the TME tumor cells outcompete T_RM cells for lipid uptake which induces their apoptosis. In gastric adenocarcinoma the use of anti-PD-L1 in a murine model, not only increased CD8⁺CD103⁺ T_RM cells cytotoxicity, it also decreased FABP4 and FABP5 expression on tumor cells, while increasing their expression on T_RM cells leading to a metabolic switch and enhanced lipid uptake.66 This leads us to speculate that along with enhancing T_RM cells’ anti-tumor effect in gastric adenocarcinoma, PD-L1 blockade reverses the metabolic reprogramming of T_RM dictated by tumor cells leading to their improved survival.

All in all, checkpoint inhibitors administration increases T_RM cells cytotoxicity and induces their accumulation in several cancer types. There is a growing body of evidence suggesting that TCF1⁺ PD-1^high TILs represent progenitor subsets that proliferate following ICI. However, TCF1 is downregulated in T_RM cells, which leads us to question whether T_RM accumulation following anti-PD-1 administration stems from T_RM cells reinvigoration or a clonal replacement due to de novo T_RM induction remains unclear and needs to be investigated.

Noteworthy, T_RM cells’ presence could represent a double-edged sword when it comes to ICI use. Recent studies revealed that T_RM cells play a key role in ICI-induced colitis (ICI-colitis), one of the most common immune-related adverse events. Investigating normal and inflamed colon of anti-PD-1/CTLA-4 treated melanoma patients, Luoma et al reported a striking accumulation of highly cytotoxic CD8⁺ T cells in ICI-colitis.109 Interestingly, the use of coupled scRNA-seq-TCR-seq revealed that ICI-colitis is associated with a shift of T_RM cells to effector T cells with high immune checkpoints expression and high cytotoxicity. In addition, a recent study revealed that activated CD8⁺ T_RM cells represented the key effector cells in ICI-colitis.110 The presence of T_RM cells was correlated with endoscopic and clinical ICI-colitis severity. These T_RM cells were characterized by high production of IFN-γ associated with high expression of immune checkpoints such as PD-1, CTLA-4, LAG-3, and TIM-3.

In summary, ICI use is emerging as a promising therapeutic strategy for ‘hot’ GI cancers mainly characterized by having a MSI and anti-PD-1 treatments are currently FDA approved in the metastatic setting.111 We previously discussed that T_RM cells represent an important subset of TILs in CRC, HCC, and pancreatic cancer. Therefore, T_RM cells can represent an interesting population to target in these cancers and the presence of T_RM cells might be used in a predictive setting in order to stratify responders and non-responders to this therapeutic strategy (figure 4).

Figure 4.

T_RM cells as a tool for GI cancers immunotherapy. T_RM cells are emerging as long persisting potent cytotoxic T cells destined for tissue residency, making them a hot target for cancer immunotherapy in general and GI cancers in particular. This is a representation of the potential therapeutic strategies where T_RM cells could be incorporated into GI cancers standard treatment paradigms. For immune infiltrated or the so-called ‘hot’ GI tumors, T_RM cells could serve as a predictive factor as well as a target for checkpoint inhibitors. For the therapeutically challenging ‘cold’ GI tumors, in vitro expanded or generated CXCR6⁺ tumor-specific T_RM cells could be adoptively transferred in combination or not with checkpoint inhibitors. GI, gastrointestinal; T_RM, tissue-resident memory T cells.

T_RM in adoptive cells transfer

GI cancers are considered cold tumors with limited access to checkpoint inhibitors.112 Therefore, adoptive cell transfer (ACT) is being considered as one of the most promising therapeutic strategies for patients with cold cancers. One of the major challenges for ACT is the ability of T cells to infiltrate the tumor and to be able to persist long-term, both of which constitute the hallmark features of T_RM cells. Additionally, due to their observed in vitro functionality and cytotoxicity, T_RM cells are gaining the researcher’s attention as a potent candidate for ACT. A recent study showed that in vitro generated NY-ESO1/SSX2 specific CD8⁺CD103⁺ presented a faster cancer recognition as well as cytotoxicity compared with CD8⁺CD103^- non T_RM cells.113 Interestingly, they were also endowed with high energetic potential. These data gave new evidence for T_RM cells’ antitumor efficacy compared with non T_RM cells which makes them a promising candidate for ACT, especially in GI cancers (figure 4). Although ACT of in vitro expanded TILs showed promising results, only a minor fraction of treated patients achieved durable responses. As previously discussed, TILs represent an heterogenous T cells populations encountering tumor-specific as well as bystander T cells which may explain the limited success of the use of the unfractionated TILs populations for ACT.114 Recent studies revealed that whether in HBV-induced liver cancer or NSCLC, T_RM cells represented the tumor antigen-specific population of TILs, therefore tetramer sorting and in vitro expansion of these cells with potent anti-tumor properties may represent a better therapeutic strategy for solid cancers.114 115 One of the major challenges impeding the use of CAR-T cells in solid tumors is their restrained trafficking and persistence in the TME. T_RM on the other hand represent highly cytotoxic, long-lived T cells endowed with the expression of tissue homing biomarkers and the downregulation of tissue egress ones, which represent highly desired properties when designing T cell-based therapies for solid tumors, GI included. Additionally, to overcome tissue homing challenges, beyond intravenous transfusion, intratumoral injection of T_RM cells could be considered. Last but not least, since our understanding of T_RM cells generation has evolved substantially over the last years, in vitro generation of tumor antigen-specific T_RM cells from the patient’s peripheral blood could represent an easily accessible therapeutic strategy. Furthermore, since T_RM cells phenotype is tissue-specific, the generation of T_RM cells with organ-specific homing biomarkers (CXCR6⁺ T_RM for liver cancer/metastasis, CD103⁺ T_RM for CRC…) would represent an intriguing approach for improving clinical responses. T_RM cells have emerged as the predominant tumor antigen-specific TILs population, they are endowed with high cytotoxicity, tissue-homing ability, and long-term persistence. The success of ACT is built on these duly warranted hallmarks, clinical studies using tumor antigen-specific T_RM cells are critically needed especially for cold tumors such as GI cancers.

Conclusions and future perspectives

T_RM research has been gaining huge momentum lately. It’s now well established that T_RM cells patrol the TME, and they can promote both cancer elimination and equilibrium. Remarkably, immune infiltrated GI tumors were enriched in T_RM cells which represented a better indicator of the patient’s prognosis. Moreover, they were predictive of checkpoint inhibitors’ response. However, the importance of leveraging this proof of concept into cancer treatment will require further understanding of T_RM function, the molecular network that drives their development and maintenance in the TME, the transcriptional regulation behind their exhaustion status, and first and foremost identification of solid biomarkers that enable to distinguish between T_RM cells from conventional TILs subsets. Of note, single-cell analysis helped to gain a more comprehensive understanding of T_RM cells heterogeneity, therefore, the use of this technology should be widespread to a larger panel of GI cancers. Once we got exhaustive answers to these questions, capitalizing on T_RM cells will be an exciting therapeutic approach to turning ‘cold’ GI tumors ‘hot’.

Ethics statements

Patient consent for publication

Not applicable.