CD103 integrin is a heterodimeric transmembrane receptor formed by α_E (CD103) and β₇ subunits, with the epithelial cell marker E-cadherin as a unique known ligand [52]. This integrin is expressed on T cells residing in tissue microenvironments, where TGF-β is abundant, such as mucosal CD8⁺ T lymphocytes and, mainly, IEL [53], but it is also expressed on CD4⁺ and CD8⁺ regulatory T (Treg) cells [54, 55] and on a large proportion of CD8⁺ effector T cells infiltrating epithelial tumors, including bladder [56], pancreatic [57], colorectal [28], ovarian [26] and lung cancers [27, 38, 58, 59]. It is induced on tumor-specific CD8⁺ T cells by concomitant signals from the TGF-β receptor (TGFBR) and the T-cell receptor (TCR) triggered by TGF-β and major histocompatibility complex class I (MHC-I)/tumor peptide complexes, respectively (Fig. 2) [33, 58, 60]. In this regard, adoptive transfer of tumor-specific CD8⁺CD103^+-T cells in the cognate tumor engrafted in nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice and subsequent coengagement of TCR and TGFBR trigger CD103 expression on T-cell surface associated with acquisition of a strong cytotoxic capacity toward autologous tumor cells. In contrast, adoptive transfer of these CD8⁺CD103⁻ T cells in allogeneic tumor does not result in expression of CD103 [33, 58]. Along the same line, CD103 is induced on tumor-specific T cells upon engagement of TCR with anti-CD3 mAb and TGF-β treatment [58, 60, 61]. In contrast, TGF-β alone had only a slight effect on CD103 expression and anti-CD3 mAb alone had no effect [58]. It is well known that only 1 to 3% of human circulating T cells expressed CD103, which implies that tumor-specific T cells need to encounter the cognate antigen within a TGF-β-rich tumor microenvironment to induce expression of the integrin and to become T_RM.Fig2Fig. 2

Role of CD103 integrin in anti-tumor T_RM functional activities. Engagement of TCR with specific peptide-MHC-I (p-MHC-I) complexes in the presence of TGF-β, abundant within the tumor microenvironment, induces expression of CD103 on the CD8⁺ T-lymphocyte surface. Phosphorylation of integrin-linked kinase (ILK) by TGFBR1, and its subsequent binding to the CD103 intracellular domain promotes inside-out signaling resulting in an increase in the affinity of the integrin for its ligand E-cadherin on tumor cells. Activated CD103 is recruited at the immune synapse formed between stimulated T_RM cells and epithelial target cells; its interaction with E-cadherin triggers phosphorylation of extracellular signal-regulated kinases 1 and 2 (ERK1/2) and the paxillin adaptor protein. Binding of phosphorylated (p)-paxillin to the α_E (CD103) subunit tail triggers an outside-in signal that promotes CD8⁺ T-cell effector functions such as cytokine production and polarized release of cytotoxic granules, leading to TCR-mediated target cell death. Intra-tumoral T_RM cells express very low levels of CD28 co-stimulatory receptor. Moreover, expression of LFA-1 on TIL is downregulated by TGF-β. Finally, cancer cells often downregulate expression of the LFA-1 ligand ICAM-1 to escape from immune effector cells

Lung T_RM cells display high expression levels of mRNA encoding effector molecules, such as granzyme B, IFN-γ and TNF, without the need for ex vivo stimulation [14]. In human lung tumors, CD8⁺CD103⁺ T_RM cells were also found to express mRNA encoding molecules associated with cytotoxic activities of killer cells, such as IFNG, GZMA, GZMB and RAB27A [27, 38, 64]. Expression by CD8⁺CD103⁺ TIL of granzyme B, perforin and the degranulation marker LAMP-1 (CD107a) was confirmed at the protein level, further supporting their cytotoxic potential [27, 38, 64]. Moreover, in ovarian and lung cancer, T_RM cells express the activation marker HLA-DR and the proliferation marker Ki67 [38, 64]. These cells may be functionally exhausted within the tumor microenvironment by the induction of T-cell inhibitory receptors including PD-1 and Tim-3 [27].

Mueller and Mackay [77] revealed that T_RM cells are mainly present in non-lymphoid tissues and express CD69 and the CD103 integrin. These cells are also found in various tumors, including melanoma [78], lung cancer [27, 39], urothelial cell carcinoma [29] and endometrial adenocarcinoma [79]. Tumors with a high density of CD8⁺ T cells showed enrichment for transcripts linked to tissue-T-cell-residency, such as CD103 [38]. However, there exists phenotypic heterogeneity in T_RM cells according to their location and tumor histological subtype. In all subtypes of endometrial adenocarcinoma, CD8⁺ TIL were present in both the tumor epithelium and stromal areas, but the frequency of CD8⁺CD103⁺ T cells was significantly higher in the tumor epithelium than in the stroma [27, 29, 79]. Most CD103⁺ cells in the tumor microenvironment co-express the CD8 molecule, whereas CD8⁺ TIL located in the stroma were mainly negative for CD103 integrin. Nizard et al. found that 70% of intra-tumoral CD8⁺ T cells expressed CD103, whereas the integrin is found on only 41% of stromal CD8⁺ T cells [39]. Conversely, CD103⁺ cells in the healthy endometrium were negative for CD3 and CD16, suggesting a non-T-cell origin [29].

CD103 binds to E-cadherin expressed on the surface of epithelial cells [52]; this binding may be involved in retention of these cells in the epithelial tissue, as well as in solid tumors [80]. Interestingly, in some studies, distribution of CD103⁺ TIL was positively associated with E-cadherin expression on tumor cells [29]. However, in other studies, there was no obvious correlation between E-cadherin staining intensity and the presence of CD103⁺ TIL, suggesting that other factors are also determinant in their infiltration [26, 81].

In a preclinical model of spontaneous breast cancer, a natural immune response involving resident innate lymphoid cells (ILC) close to ILC1 and TCR-positive cells was described [82]. These cells do not recirculate and delay tumor growth. Tumor growth control can be explained by the fact that CD49a⁺ and CD103⁺ cells are highly activated and exhibit more satisfactory effector functions than conventional CD8⁺ T cells [27, 38, 83]. These resident cells produce more IFN-γ and granzyme B than their integrin-negative counterparts. Accordingly, significantly impaired tumor control was observed in mice treated with either anti-CD49a or anti-CD103 antibodies [36, 83].

Following vaccination against orthotopic tumors, it has been shown that T_RM cells are required for the efficacy of a cancer vaccine. Indeed, in a preclinical head and neck cancer model expressing E6-E7 proteins from HPV, the mucosal (intranasal) delivery of a vaccine (the B subunit of Shiga toxin coupled with the E7 protein from HPV16) was efficient at eliciting local T_RM cells and control of tumor growth [36]. Interestingly, a body of experiments demonstrated the role of T_RM cells in the efficacy of this cancer vaccine. Indeed, depletion of CD49a⁺ T_RM cells with an antibody hampered infiltration of T_RM in mucosal tumors and partially inhibited the efficacy of intranasal vaccination to control mucosal tongue tumors. Similar results were obtained by Murray et al., using anti-CD49a mAb in a melanoma model [83]. In one of our group's study, the co-administration of an anti-TGF-β antibody with the vaccine reduced the number of T_RM cells and control of tumor growth by the vaccine [39]. By blocking recruitment of effector T cells arising from the lymphoid organs with the FTY720 drug, which downmodulates the S1PR1 molecule, it has been demonstrated that T_RM cells induced by intranasal vaccination are able to control tumors [39]. Lastly, in a parabiosis mouse model, one of our groups showed that T_RM cells induced by intranasal vaccination are required to control orthotopic head and neck tumor growth [39]. These results obtained with parabiosis experiments have been reproduced by the group of Sancho [10]. Together, these observations suggest that the absence of local T_RM induction correlates with lower vaccine efficiency; they highlight the crucial role of these cells in tumor control. In line with these results, it was reported that cervico-vaginal boost with an HPV vaccine after a systemic (intramuscular) prime was more efficient at eliciting local cervical T_RM cells, which was correlated with better mouse survival than that observed with an intramuscular boost [67]. In another model, it was shown that the number of T_RM cells in tissues gradually increased after each boost [20], underlining the need for repeated injections.

At the present time, anti-tumor vaccine protocols almost exclusively use systemic administration with no significant clinical results, whereas various cancers are located in mucosal sites (lung, head and neck and urogenital). Therefore, it is important to reevaluate the advantage of local delivery by better understanding T_RM cell physiology [84]. It should be mentioned that other non-T_RM effector cells might also play a role in control of mucosal tumors [85], and the presence of T_RM is not always sufficient to cure high grade cervical dysplasia after vaccination, likely due to the presence of immunosuppressive mechanisms or their insufficient local number [86].

The CD8⁺ T_RM subset has emerged as a predictive marker of survival in several human epithelial cancers [26, 27, 29, 38, 87]. In this regard, one of our groups first demonstrated that an enhanced CD103⁺ TIL subset correlates with improved early-stage non-small-cell lung carcinoma (NSCLC) patient survival [27]. The predictive value of T_RM was also demonstrated in ovarian, breast and bladder cancers [26, 28, 64, 88]. Indeed, in a large cohort of high-grade serous ovarian cancers, CD103⁺ TIL were associated with improved patient survival [26]. Moreover, the expression of CD103 on TIL was associated with improved overall and recurrence-free survival in a retrospective cohort of urothelial cell carcinoma patients [29]. This integrin also appeared to be a biomarker of favorable prognosis in a large cohort of breast cancer patients [89]. However, the CD103 biomarker could also be expressed by CD4⁺ T cells and DC, which introduces bias in interpretation of results without double immunostaining with anti-CD8 mAb. The epithelial location of CD103⁺ TIL is an even more significant prognosis marker compared to the stromal location, suggesting that intraepithelial CD8⁺CD103⁺ cells encompass a higher proportion of tumor-specific T_RM cells [27, 89]. This intratumoral infiltration of CD103⁺ TIL was associated with expression of E-cadherin on tumor cells in bladder cancer [29], but not in ovarian or breast cancer [26, 89].

Since it is well known that CD8⁺ T-cell infiltration is associated with better clinical outcome in many cancers [90], comparative analysis of the prognostic value of T_RM and CD8⁺ T cells has been lacking. Two recent studies, including one from one of our groups, demonstrated that, in two independent cohorts of lung cancer patients, T_RM cells were correlated with patient survival in both univariate and multivariate analysis, and this effect was independent of CD8⁺ T cells [38, 39].

With respect to adoptive cell transfer, Milner et al., identified the transcription factor Runx3 as critical for the establishment of T_RM cell populations in various normal tissues and in cancer [15]. In a preclinical model of melanoma, adoptive transfer of CD8⁺ TIL lacking expression of Runx3 and which did not exhibit a T_RM cell phenotype resulted in uncontrolled tumor growth and low animal survival. In contrast, when anti-tumor CD8⁺ T cells overexpressing Runx3 were transferred in vivo, tumor growth was inhibited, and mouse survival improved [15]. Thus, adoptive cell therapy with anti-tumor CD8⁺ tumor-infiltrating T lymphocytes displaying a T_RM phenotype improves the efficacy of this immunotherapy approach. Although T_RM cells lacking the expression of CD103 integrin have been observed, the transfer of CD103-deficient T cells has also been used to demonstrate the role of T_RM cells in tumor progression control. In this setting, it has been shown that T_RM cells are required for animal protection [78].

One of our groups was the first to report preferential expression of immune checkpoint receptors (PD-1 and TIM-3) and costimulatory molecules (ICOS) in T_RM cells from lung cancer patients [27], extending similar results observed in T_RM from normal tissues [37, 91]. These results have been confirmed in other cancers, both in mice and in humans [27, 38, 39, 83, 92]. In human cervical cancer, a strong correlation between expression of CD103 and exhaustion molecules such as PD-1, TIGIT, LAG-3 and Tim-3 has been observed using the Cancer Genome Atlas [93]. This result has been confirmed at the protein level in ovarian and endometrial adenocarcinomas [26, 79]. Other checkpoint receptors (NKG2A, CD39, adenosine receptor A2AR and SPRY1) may also be preferentially expressed by T_RM cells [14, 38]. Remarkably, one of our groups showed that blockade of PD-1 on T_RM cells freshly isolated from human lung carcinomas strongly promotes cytolytic activity toward autologous tumor cells ex vivo [27]. Moreover, anti-MHC-I and anti-CD103 neutralizing antibodies dramatically inhibited target cell killing by autologous TIL pretreated with anti-PD-1, further emphasizing that CD8⁺CD103⁺ T_RM cells were exhausted tumor-specific T lymphocytes, which could be rescued by blocking PD-1 signals resulting in T-cell activation and autologous tumor cell killing [27]. In line with these results, after infection in mucosal tissues, T_RM cells can proliferate and generate a second pool of T_RM, strongly suggesting that they have the ability to be activated in situ for better control of local danger [94]. Therefore, their exhausted phenotype does not preclude their sensitivity to reactivation and invigoration [95]. Consistently, recent results revealed expansion of T_RM cells in melanoma patients responding to anti-PD-1 immunotherapy [96].

Taken together, their expression of checkpoint receptors, their strategic location in close tumor contact and their ability to proliferate in situ after a local stimulus suggest that T_RM cells are enriched in tumor-specific CD8⁺ T cells, making them possible effectors of anti-PD-1/anti-PD-L1 cancer immunotherapy.