Resident memory T cells in skin mediate durable immunity to melanoma

Abstract

Tissue-resident memory T cells (T_RM cells) have been widely characterized in infectious disease settings; however, their role in mediating immunity to cancer remains unknown. Here we report that skin-resident memory T cell responses to melanoma are generated naturally as a result of autoimmune vitiligo. Melanoma antigen-specific T_RM cells resided predominantly in melanocyte-depleted hair follicles and were maintained without recirculation or replenishment from the lymphoid compartment. These cells expressed CD103, CD69, and CLA, but lacked PD-1 or LAG-3, and were capable of making IFN-γ. CD103 expression on CD8 T cells was required for establishment of T_RM cells in skin, but was dispensable for vitiligo development. Importantly, CD103⁺ CD8 T_RM cells were critical for protection against melanoma re-challenge. This work establishes that CD103-dependent T_RM cells play a key role in perpetuating anti-tumor immunity.

Introduction

Tissue-resident memory T cells (T_RM cells) are crucial mediators of adaptive immunity in peripheral tissues. CD8 T_RM cells have been characterized in association with infections of the skin, gut, lung, and genitourinary tract (1–4), where they mediate long-lived protection against re-infection (5–7). Separately, T_RM cells have been implicated in the pathogenesis of certain inflammatory conditions of the skin, including psoriasis (8) and mycosis fungoides (9). Despite their widespread involvement in diseases of peripheral tissues, it remains unknown whether T_RM cells can mediate immunity to cancer.

Generation of T cell memory is paramount to ensuring durable anti-tumor immunity, although studies in cancer models have focused on lymphoid memory (10). Our work previously identified key requirements for generating lymphoid memory against melanoma (11). We showed that the autoimmune destruction of normal host melanocytes, a condition known as vitiligo, is required to sustain melanoma/melanocyte antigen (Ag)-specific T cells in lymph nodes and spleens for many months (11). This is consistent with the long-recognized role of vitiligo as an independent positive prognostic factor in melanoma patients (12, 13), and more recently in patients treated with pembrolizumab who exhibit vitiligo incidence as high as 25% (14). While our studies implicated vitiligo in sustaining lymphoid memory, a growing role for resident memory in cutaneous immune settings suggests that vitiligo might also support the generation of T_RM cells.

Seminal studies characterizing T_RM cells in infectious disease models and under steady state conditions, provide a basis to identify resident memory responses against cancer. Cutaneous infections generate skin T_RM cells with an effector memory-like phenotype that persist in the epidermis without recirculation through lymphoid tissues (1, 5, 6). Skin T_RM cells are phenotypically CD44^hi, CD62L^lo, and are distinguished from lymphoid memory by the expression of CD69, cutaneous lymphocyte antigen (CLA), and CD103 (15–17). CLA is a fucosyltransferase VII (FucT-VII)–modified derivate of P-selectin glycoprotein ligand 1 (PSGL-1), which is critical for T cell entry into skin (6, 18). CD103 is the TGF-β induced α-chain of the α_EE7 integrin, which binds to E-cadherin on epidermal cells in peripheral tissues. Thus, CLA and CD103 are thought to position and retain T_RM cells in skin as a barrier to re-infection (15, 16). Interestingly, CD8 T cells expressing CD103 have been identified in human lung and ovarian carcinoma specimens, where they are associated with significantly improved overall survival (19, 20). However, these T cells appeared functionally exhausted prior to PD-1 blockade (19). Thus, it remains unknown how tumor-specific CD103⁺ CD8 T cells are generated, whether they represent bona fide resident memory, and to what extent they contribute to anti-tumor immunity.

The present studies are based on our hypothesis that vitiligo-affected skin supports resident memory T cells that participate in the immune response to melanoma. We employ a mouse model of melanoma-associated vitiligo, induced by depletion of regulatory T cells (T_reg cells) and surgical excision of a primary dermal B16 melanoma (11, 21). The goals of this study were three-fold: (1) to define the characteristics of tumor-specific T_RM cells, (2) to illustrate T cell- and host-intrinsic requirements for generating T_RM cells against melanoma, and (3) to define a role for T_RM cells in mediating tumor protection. Our findings establish a key role for T_RM in sustaining immunity to cancer.

Results

Functional melanoma Ag-specific T_RM cells develop in the skin of mice with vitiligo

Our prior studies showed that dermal inoculation with B16 melanoma followed by T_reg cell depletion and curative tumor excision (Fig. 1A) breaks tolerance to melanocyte antigens, resulting in autoimmune vitiligo in ~60% of mice (11, 21). Consistent with our prior finding that vitiligo is CD8 T cell-mediated (11), we found that CD8 T cells were enriched in the skin of vitiligo-affected mice, as compared with unaffected mice (Fig. 1B). To track the melanoma/melanocyte antigen-specific response in skin, mice received congenically-marked naïve transgenic CD8 T cells specific for gp100_25–33 (pmel cells) (22). This small sentinel population of low-avidity T cells is primed during tumor growth (23), yet does not influence the extent or kinetics of vitiligo (11). Thirty days after surgery, pmel cell populations were greatly elevated in skin from vitiligo-affected mice as compared with unaffected mice (Fig. 1C). Pmel cells could be recovered throughout the skin of mice with vitiligo, but proportions were significantly increased in perilesional as compared with pigmented sites (Fig. 1D; depigmented and perilesional sites depicted in the representative image, were used interchangeably for subsequent analyses). As compared with lymph node and spleen, pmel cells were enriched 10- to 100-fold in skin (Fig. 1E). Thus, Ag-specific T cells exhibited strong preferential persistence throughout skin of mice with vitiligo.

Figure 1. CD8 T cells recognizing tumor/self- and tumor-specific antigens persist in vitiligo-affected skin, and exhibit a T_RM cell phenotype.

(A) Experimental scheme to induce melanoma-associated vitiligo; unaffected mice underwent identical procedures but did not depigment. (B) Mice were treated as in A, and proportions and absolute numbers of CD8 T cells were detected 65d post-surgery by flow cytometry; gated on live CD45⁺ cells. (C–E) 10⁴ naïve Thy1.1⁺ pmel cells were transferred 1d prior to treatment as in A, and pmel proportions (gated on live CD8⁺ cells) were quantified 50d post-surgery in (C) vitiligo-affected versus unaffected skin, (D) depigmented, perilesional, and pigmented skin sites of vitiligo-affected mice (sites depicted in representative image), and (E) skin versus lymphoid tissues of vitiligo-affected mice. (F) Expression of CD44, CD62L, CD103, CD69, and CLA on CD8⁺Thy1.1⁺ pmel cells in vitiligo-affected mice, treated as in panel C; percent CLA⁺ is reported for skin, gated based on unstained control. (G) 10⁴ naïve Ly5.1⁺ OT-I cells were transferred 1d prior to treatment according to A, but with B16-OVA given on d0; OT-I cells were quantified in skin and lymphoid tissues (gated on live CD8⁺) 30d post-surgery. (H) Mice were treated either as in C or G, and proportions of pmel or OT-I cells were compared in skin 30d post-surgery. Symbols represent individual mice; horizontal lines depict means. Significance was determined by t test (for B, C, and H), Kruskal-Wallis test (for D and G), or one-way ANOVA (for E and F); NS denotes p > 0.05. Data in each panel are representative of 2 independent experiments, each with n ≥3 mice/group; data in G are pooled from two experiments.

Phenotypic analysis indicated that pmel cells in skin were overwhelmingly CD44^hi and CD62L^lo, and also expressed high levels of the T_RM cell markers CD103 and CD69 (Fig. 1F). Approximately 50% of these cells also expressed the skin-specific marker CLA (Fig. 1F). CD103 and CLA expression were higher in skin as compared with lymphoid tissues; however, CD69 expression was also elevated in lymph node (Fig. 1F), as expected (11). A similar phenotype was observed for total CD8 T cells in vitiligo-affected skin (Fig. S1). Thus, T cells in skin of vitiligo-affected mice exhibited a resident memory phenotype.

To determine if persistence in skin was limited to T cells with melanocyte antigen specificity, mice were instead transferred with small sentinel populations of naïve OT-I cells, and vitiligo was induced using B16 cells expressing OVA as a model tumor-specific antigen. Similar to pmel cells, OT-I cells persisted at substantially higher levels in skin as compared with lymphoid tissues of mice with vitiligo (Fig. 1G), where they exhibited a CD62L^loCD103^hiCD69⁺CLA^+/− T_RM cell phenotype (Fig. S1). Thus, CD8 T cells with specificity for both tumor-specific and shared antigens were maintained in vitiligo-affected skin. However, pmel cells persisted at ~10-fold higher levels than OT-I cells (Fig. 1H). Therefore, subsequent analyses focused on the pmel response.

We next assessed whether pmel cells in skin maintained functional capacity. Further phenotypic analysis of the population indicated negligible expression of the exhaustion markers PD-1 and LAG-3 (Fig. 2A). Accordingly, ex vivo re-stimulation of these cells with cognate gp100 peptide elicited production of IFN-γ at levels comparable to pmel cells from lymph nodes (Fig. 2B). Additionally, in skin of vitiligo-affected mice that had not received pmel cells, IFN-γ-producing TRP-2_180–188-specific CD8 T cells could be detected, indicating a functional endogenous response (Fig. S2). To more precisely establish the localization of T cells within skin, microscopy was performed on perilesional skin sections. Consistent with their CD103 expression, CD8 T cells and pmel cells were located in proximity to E-cadherin-expressing cells at the dermal-epidermal junction, predominantly within hair follicles (Fig. 2C), the primary location of skin melanocytes in mice (24). Furthermore, T cells in perilesional skin were preferentially associated with follicles containing white hairs (Fig. 2D). Thus melanoma-specific T cells persisted in melanocyte-depleted hair follicles, and were not functionally exhausted.

Figure 2. Functional melanoma/melanocyte Ag-specific CD8 T cells persist in depigmented hair follicles.

Mice received 10⁴ naïve Thy1.1⁺ pmel cells 1d prior to treatment as in Fig. 1A, and vitiligo-affected skin was analyzed 30d later by (A) flow cytometry to detect PD-1 and LAG-3 on CD8⁺Thy1.1⁺ pmel cells, with B16 tumor-infiltrating CD8⁺ cells as a positive control; histograms are representative of n = 8 total mice in 2 independent experiments. (B) Flow cytometry to detect IFN-γ production by CD8⁺Thy1.1⁺ pmel cells from digested skin, following 14h ex vivo restimulation with cognate (gp100_25–33) or irrelevant (OVA) peptide. Symbols represent individual mice; horizontal lines depict means. Data are representative of 2 independent experiments each with n = 4 mice/group; significance was determined by Kruskal-Wallis test (skin) or one-way ANOVA (LN). (C–D) Perilesional skin sections were analyzed by fluorescence microscopy to determine localization of CD8⁺ T cells and Thy1.1⁺CD8⁺ pmel cells in association with E-cadherin-expressing epidermis and hair follicles containing white or black hairs. (C) Arrows indicate pmel cells in hair follicles containing white hairs; Scale bar; 50 μm. (D) Pie charts summarize localization of CD8 T cells (left) and pmel cells (right) in association with the indicated structures, compiled from 89 images from n = 5 mice; percentages are of total counted cells.

To conclusively determine whether CD8 T cells in vitiligo-affected skin are resident memory, we assessed their ability to persist when isolated from the lymphoid compartment. First, perilesional skin containing pmel cells was grafted onto RAG^−/− recipients, which were then rested for 50 days. During this time, skin grafts progressed to complete depigmentation but surrounding host tissue remained pigmented (Fig. S3). We found that both CD8 T cells and pmel cells in skin grafts on RAG^−/− mice were undiminished, as compared with vitiligo-affected control skin (Fig. 3A). In skin graft-draining lymph nodes, small populations of CD8 T cells were identified, however pmel cells were absent (Fig. 3B), suggesting that tumor/self Ag-specific T cells do not recirculate. Moreover, pmel cells in grafted skin retained a CD103⁺CD69⁺CLA⁺ phenotype that was comparable to those in vitiligo-affected control skin (Fig. 3C), but distinct from the CD103⁻ CLA⁻ phenotype of CD8 T cells that had migrated to lymph nodes (Fig. S4). As a second approach to address residency, wild-type vitiligo-affected mice were treated with the S1P receptor agonist FTY720 for 35 days to block T cell egress from lymph nodes. FTY720 reduced circulating CD8 T cells (Fig. 3D), but did not reduce populations of CD8 T cells (Fig. 3E; p = 0.311) or pmel cells (Fig. 3F; p = 0.999) in skin, nor did it alter T_RM cell phenotypic marker expression (Fig. 3G). These data collectively illustrate the maintenance of a self-sustaining, T_RM cell population in skin of vitiligo-affected mice.

Figure 3. Melanoma-specific CD8 T cells maintain residence in vitiligo-affected skin.

(A–C) Skin graft donor mice received 10⁴ naïve Thy1.1⁺ pmel cells 1d prior to treatment as in Fig. 1A. Fifty days post-surgery, vitiligo-affected skin was harvested and grafted onto RAG^−/− recipients, which were rested for 50d prior to analysis. (A) Quantification of total CD8 T cells (top) and CD8⁺Thy1.1⁺ pmel cells (bottom) in RAG^−/− skin grafts (Grafted) compared to skin from time-matched vitiligo-affected control mice (Control). (B) Quantification of CD8 T cells (top) and Thy1.1⁺ CD8⁺ pmel cells (bottom) in skin graft-draining lymph nodes from RAG^−/− mice. (C) Expression of CD103, CD69, and CLA on Thy1.1⁺ pmel cells in skin from mice in A; percent CLA⁺ is reported for cells in grafted skin. (D–G) Mice received 10⁴ naïve Thy1.1⁺ pmel cells 1d prior to treatment to induce vitiligo as in Fig. 1A. Vitiligo-affected mice then received FTY720, or no treatment, for 35 consecutive days, beginning 30d post-surgery (analyzed 65d post-surgery). (D) Proportions of CD8⁺ cells in blood. (E–F) Quantification of (E) total CD8⁺ cells and (F) Thy1.1⁺ pmel cells, in skin. (G) Phenotype of Thy1.1⁺ pmel cells in skin; percent CLA⁺ is reported for cells in FTY720-treated mice. Symbols represent individual mice; horizontal lines depict means. Significance was determined by t test; NS denotes p > 0.05. Data in each panel are representative of ≥ 2 independent experiments, each with n ≥4 mice/group.

Vitiligo is required for the generation of melanoma-specific T_RM cells

While the above data demonstrated the development of T_RM cells in association with vitiligo, an absolute requirement for host vitiligo remained to be shown. We previously found that vitiligo-affected mice initially prime larger Ag-specific T cell responses than those that remain unaffected (11), indicating that the establishment of resident memory could be due to enhanced priming. To determine if host vitiligo was required, we normalized T cell priming by pooling pmel cells from lymphoid tissues of multiple donor mice on the day of surgery. This uniform population was then adoptively transferred into treatment-matched recipient mice, some of which would subsequently develop vitiligo, and others of which would not (Fig. 4A). Transferred pmel cells had an overwhelmingly KLRG1^loCXCR3⁺Ki67^hi phenotype (Fig. 4B), consistent with T_RM precursor cells described in skin HSV infection (15). A proportion of these cells also expressed CD103 (Fig. S5). Thirty days following adoptive transfer, small populations of pmel cells were detected in skin of 7 out of 10 recipients that developed vitiligo versus only 1 out of 8 unaffected mice (Figs. 4C and D), supporting a requirement for host vitiligo in the development of resident memory.

Figure 4. Host vitiligo is required for the establishment of T_RM cells.

Donor mice were treated as in Fig. 1C and, on the day of surgery, CD8 T cells were harvested from pooled lymph nodes and spleens and transferred into mice treated as shown. (A) Experimental timeline; (B) Proportion of tumor-primed pmel cells (gated on CD8⁺; left) from LN on the day of harvest; expression of KLRG1 and CXCR3 (gated on Thy1.1⁺; middle), and Ki67 (gated on Thy1.1⁺KLRG1^loCXCR3⁺; right). (C) Thirty days post-adoptive transfer as shown in A; pmel cells were detected in skin of unaffected vs. vitiligo-affected recipient mice (gated on CD8⁺); (D) fractions of recipient mice from each group with any detectable pmel cells within a 2 cm² patch of skin, on d30. (E–G) Tumor-primed pmel cells were alternatively transferred as in timeline (E); RAG^−/− mice received TRP-1 Tg cells 1d prior to sham skin surgery to induce vitiligo; RAG^−/−TRP-1^−/− recipients served as unaffected controls. (F) Quantification of Thy1.1⁺ pmel cells in unaffected vs. vitiligo-affected skin 30d post-transfer as in E (gated on CD8⁺). (G) Representative phenotype of pmel cells (gated on Thy1.1⁺) from skin in panel F. Symbols represent individual mice; horizontal lines depict means. Significance was determined by Mann-Whitney test; Data are representative of 2 independent experiments each with n ≥5 mice/group, or (D, F) pooled from 2 experiments each with n ≥3 mice/group.

To confirm this finding using a second model of vitiligo, TRP-1_113–125 melanocyte Ag-specific CD4 T cells were transferred into RAG^−/− mice to become activated and induce vitiligo as previously shown (25), or into RAG^−/−TRP-1^−/− mice (which lack the target Ag) to serve as unaffected controls (Fig. 4E). Tumor-primed pmel cells (described above) were then transferred into these vitiligo-affected or unaffected recipients. Consistent with the above experiment, significantly larger pmel cell populations were identified in skin of vitiligo-affected mice thirty days later, as compared with unaffected controls (Fig. 4F; p = 0.002). Pmel cells in vitiligo-affected skin were overwhelmingly CD44^hiCD62L^loCD103^hiCD69⁺, and exhibited mixed expression of CLA (Fig. 4G). Therefore, host vitiligo was required for the generation of melanoma/melanocyte antigen-specific T_RM cells.

FucT-VII and CD103 are required for optimal formation of tumor-specific T_RM cells

To identify T cell-intrinsic requirements for the development of vitiligo-driven T_RM cells, we initially focused on CLA, a FucT-VII-glycosylated isoform of PSGL-1 (18). T cells deficient in FucT-VII lack expression of CLA, as we confirmed (Fig. S6). Indeed, melanoma-associated vitiligo was greatly reduced in FucT-VII^−/− mice (Fig. 5A), although tumor growth was also unexpectedly enhanced (Fig. S7). Regardless, vitiligo was also reduced in RAG^−/− mice that had been reconstituted with FucT-VII^−/− CD8 T cells (Fig. 5B), in which tumor growth was normal (Fig. S7), thus indicating an important role for FucT-VII in vitiligo development.

Figure 5. FucT-VII expression on CD8 T cells promotes skin access, formation of memory, and overt vitiligo development.

(A–B) Vitiligo incidence was tracked in (A) wild-type (WT) mice vs. FucT-VII^−/− mice or (B) RAG^−/− mice reconstituted with either WT or FucT-VII^−/− naïve CD8 T cells, 1d prior to treatment as shown in Fig. 1A. Significance was determined by log-rank analysis. Data are (A) representative of, or (B) pooled from 2 independent experiments, each with n ≥ 7 mice group. (C–F) Mice received 10⁴ naïve, congenically distinct WT and FucT-VII^−/− pmel cells, admixed at a 1:1 ratio, 1d prior to treatment to induce vitiligo as in Fig. 1A. Flow cytometry was performed to detect relative frequencies of WT vs. FucT-VII^−/− pmel cells (gated on CD8⁺ cells) in (C) skin of vitiligo-affected mice 50d post-surgery, (D) lymph nodes 4d prior to surgery, (E) skin on the day of surgery, and (F) lymph nodes of vitiligo-affected mice 50d post-surgery. Symbols represent individual mice, with lines joining cell populations in the same mouse. Significance was determined by Wilcoxon matched pairs test (C), or paired t test (D–F). Data in C–F are combined from 2 independent experiments, each with n = 3 mice/group. Arrows indicate mean percent difference between WT and FucT-VII^−/− population sizes.

To determine the role of FucT-VII in the establishment of skin T_RM cells in a setting where vitiligo was not impaired, congenically-distinct, naïve sentinel populations of wild-type (WT) and FucT-VII^−/− pmel cells were co-transferred at a 1:1 ratio into wild-type mice prior to vitiligo induction. Fifty days after surgery, we observed a profound reduction in FucT-VII^−/− pmel cells, relative to WT pmel cells, in vitiligo-affected skin (Fig. 5C; p = 0.004). This was not due to a defect in priming, as FucT-VII^−/− pmel cells had expanded to an even greater extent than wild type cells in tumor-draining lymph nodes prior to surgery (Fig. 5D). However, early T cell accumulation in skin was significantly impaired by FucT-VII deficiency (Fig. 5E; p = 0.013), suggesting that a defect in skin access contributed to the defect in memory. Surprisingly, FucT-VII^−/− pmel cells also demonstrated reduced lymphoid memory formation (Fig. 5F; p = 0.010). Thus, in accordance with substantially reduced vitiligo, loss of FucT-VII (and CLA) in CD8 T cells perturbed T cell access to skin, and impaired the establishment of both lymphoid and resident memory compartments.

Pmel cells in vitiligo-affected skin also expressed high levels of CD103 (see Fig. 1F), which is important for T_RM cell formation in HSV skin infection (15). In contrast to FucT-VII^−/− mice, melanoma-associated vitiligo occurred with unaltered incidence and kinetics in CD103^−/− mice (Fig. 6A), and in CD8^−/− mice that had been reconstituted with CD103^−/− CD8 T cells (Fig. 6B). We noted, however, that CD103^−/− mice and CD8^−/− mice reconstituted with CD103^−/− CD8 T cells had a lower incidence of vitiligo that had disseminated beyond the surgical site (Fig. 6C and D). Therefore, CD8 T cell expression of CD103 was not required for vitiligo, however it enhanced disease severity.

Figure 6. CD103 expression on CD8 T cells promotes skin access and T_RM cell formation.

(A–B) Vitiligo (induced as in Fig. 1A) was tracked in (A) wild-type (WT) mice vs. CD103^−/− mice or (B) CD8^−/− mice reconstituted with WT or CD103^−/− naïve CD8 T cells. Data are representative of 2 independent experiments, each with n ≥ 7 mice/group (C–D) Relative proportions of vitiligo-affected mice (from A and B, respectively) with localized vs. disseminated vitiligo; with dissemination defined by depigmentation extending beyond a 2 cm² area surrounding the surgical site. (E–J) Mice received 10⁴ naïve, congenically distinct WT and CD103^−/− pmel cells, admixed at a 1:1 ratio, 1d prior to treatment as in Fig. 1A. Flow cytometry was performed to detect relative frequencies of WT vs. CD103^−/− pmel cells (gated on CD8⁺ cells) in (E) skin of vitiligo-affected mice 45d post-surgery, (F) skin on the day of surgery, (G) lymph nodes 4d prior to surgery, (H) lymph nodes of vitiligo-affected mice 45d post-surgery, and (I) tumors on the day of surgery. (J) Analysis of lymph nodes from panel G, to detect IFN-γ production by CD8⁺Thy1.1⁺ pmel cells, following 5h ex vivo restimulation with cognate (gp100_25–33) or irrelevant (OVA) peptide. Symbols represent individual mice. (E–I) Significance was determined by Wilcoxon matched pairs test, pairing CD103^−/− and WT pmel populations in the same mouse (denoted by line-connected points). Arrows indicate mean percent difference between WT and CD103^−/− pmel population sizes. Data are pooled from 2 independent experiments, each with n ≥ 3 mice/group, with the exception of G, which is representative of 3 experiments each with n ≥ mice/group. (J) Significance was determined by Kruskal-Wallis test; data are representative of 2 independent experiments each with n = 4 mice/group. NS denotes p > 0.05.

To separately assess a role for CD103 in the development of T_RM cells, congenically-distinct populations of wild-type and CD103^−/− pmel cells were co-transferred at a 1:1 ratio into wild-type mice prior to melanoma-associated vitiligo induction. Forty-five days after tumor excision, there was a significant reduction in CD103^−/− relative to WT pmel cells in vitiligo-affected skin, although a minor CD103-independent population persisted (Fig. 6E; p = 0.005). Unexpectedly, CD103-deficiency impaired T cell accumulation in skin as early as the day of tumor excision (Fig. 6F; p = 0.005). Defects in skin access and T_RM cell formation were not due to reduced T cell priming, as CD103^−/− populations expanded normally in tumor-draining lymph nodes (Fig. 6G). Moreover, in contrast to FucT-VII-deficiency, loss of CD103 did not impair the formation of lymphoid memory (Fig. 6H). Additionally, CD8 T cell access to tumors (Fig. 6I), production of IFN-γ (Fig. 6J), and acute rejection of established B16 tumors (Fig. S8), were unaffected by CD103-deficiency. Thus, the loss of CD103 on CD8 T cells only appeared to reduce T cell accumulation in skin and the subsequent establishment of resident memory.

CD103-dependent T_RM cells are necessary for long-lived immunity to melanoma

We previously demonstrated that mice with melanoma-associated vitiligo are protected against melanoma re-challenge >1 month after surgery, in flanks contralateral to depigmented surgery sites (11, 21). However, the protective memory T cell compartment was not identified. As the above data show that CD103-deficiency impairs T_RM cell formation without perturbing lymphoid memory, this model was used to identify a role for CD103⁺ T_RM cells in mediating tumor protection. CD8^−/− mice reconstituted with CD103^−/− or WT CD8 T cells were treated to induce melanoma-associated vitiligo as in Fig. 1A. Those that developed vitiligo (equivalent in both groups; Fig. 6B) were re-challenged thirty days later with intradermal B16 melanoma on contalateral flanks. In contrast to mice reconstituted with WT CD8 T cells, which had strong tumor immunity, those reconstituted with CD103^−/− T cells demonstrated no tumor protection (Fig. 7A). Thus, CD103 expression on CD8 T cells was absolutely required for tumor protection in skin of vitiligo-affected mice.

Figure 7. CD103⁺ CD8 T_RM cells are required for long-lived tumor protection.

Mice were treated as in Fig. 1A, and those that developed vitiligo (right flank) were re-challenged in pigmented skin on the left flank by inoculation with 1.2×10⁵ B16 cells 30d post-surgery. (A) Tumor incidence (left) and average diameter of palpable tumors (right) in CD8^−/− mice that had been reconstituted with either WT or CD103^−/− naïve CD8 T cells. Naïve denotes untreated wild-type mice. Data are pooled from 2 independent experiments with n ≥ 8 mice group. B) Tumor incidence (left) and average diameter of palpable tumors (right) in vitiligo-affected mice that were either untreated, or treated with FTY720 +/− anti-CD8 mAb. FTY720 was given daily starting 14 days prior to tumor challenge and continuing to the end of the experiment. Anti-CD8 was given 1d prior to tumor challenge and weekly thereafter. (Left) data are pooled from 2 independent experiments with n ≥21 mice/group; (right) data are representative of 2 experiments with n ≥10 mice/group. Significance was determined by Gehan-Breslow-Wilcoxon analysis (left panels), or two-way ANOVA (right panels; tumor diameter given as mean +/− SEM).

To assess whether T_RM cells were capable of providing tumor protection independent of lymphoid memory, FTY720 was administered daily to wild-type mice, starting 2 weeks prior to tumor re-challenge, to sequester tissue non-resident T cells in lymph nodes, and mice were again re-challenged on contralateral flanks. Indeed, tumor protection in vitiligo-affected mice was completely unaffected by FTY720 treatment (Fig. 7B), indicating that the lymphoid compartment was not required for tumor immunity. However, treatment with a CD8 depleting antibody, which efficiently eliminated CD8 T cells from both lymphoid tissues and skin (Fig. S9), abrogated tumor immunity in FTY720-treated mice (Fig. 7B). Contralateral flanks were still pigmented at the time of re-challenge, and protection was independent of whether or not this area eventually depigmented (Fig. S10), indicating that protective T_RM responses were not limited to lesional skin of mice with vitiligo. Moreover, vitiligo-affected mice exhibited no cross-protection against the unrelated Lewis Lung carcinoma, demonstrating that protection was tumor antigen-specific (Fig. S11). Collectively, these results show that CD103⁺ CD8 T_RM cells are necessary and sufficient for long-lived melanoma protection in skin of vitiligo-affected mice.

Discussion

The present studies establish that T_RM cells provide durable immunity to cancer. We demonstrate that autoimmune tissue destruction provides a hospitable niche for tumor-specific T cell residence in peripheral tissue. Indeed, melanoma-associated vitiligo supports T_RM cells that persist and function independently from the lymphoid compartment. Thus, skin of hosts with vitiligo provides the necessary components of long-lived protection against melanoma.

Here we show that a number of concepts from studies of infectious disease extend to T_RM in cancer. Melanoma-specific T_RM cells resemble viral Ag-specific skin T_RM cells in their generation from a KLRG1⁻CXCR3⁺ precursor (15), their CD44^hiCD62L^loCD103^hiCD69⁺ phenotype (1, 17), and their long-term persistence throughout skin (1, 5, 6). Notably, we demonstrate that T_RM cells are primary mediators of long-lived cutaneous immunity to cancer, which parallels the protective requirement for T_RM cells in prevention of viral re-infection (5, 6).

We report that vitiligo supports resident memory against both tumor-specific and self-antigens, which could both participate in long-lived protection against melanoma. However, encounter with self-antigen in skin was shown to enhance the size of T_RM cell populations, consistent with a role for persistent Ag in promoting resident memory formation during viral infection (26). Our observation that gp100-specific T_RM cells reside predominantly in follicles containing white hairs, together with published findings that IL-7 and IL-15 production by hair follicles maintains T_RM cells (27), suggests that these follicles serve as an antigen-free niche. Despite this, we show that protection against melanoma re-challenge is afforded even in pigmented areas of skin, indicating widespread T_RM cell function. As melanocytes reside only in hair follicles in C57BL/6 mice, studies are still needed to assess whether non-exhausted T_RM cells develop in vitiligo-affected melanoma patients, wherein melanocytes also reside at the dermal-epidermal junction. Of note, the functional status of T_RM cells in vitiligo-affected skin is distinct from the exhausted state of CD103⁺ CD8 T cells that infiltrate human lung and ovarian carcinomas (19, 20).

Regarding T cell-intrinsic requirements for T_RM cell generation, we illustrate important yet distinct roles for FucT-VII and CD103. We show that FucT-VII is required for skin access and the establishment of T_RM cells, but also contributes to lymphoid memory formation. Accordingly, FucT-VII loss has been implicated in increased peripheral T cell apoptosis (28). In contrast, we show that CD103-deficiency impairs the establishment of skin T_RM cells without an apparent defect in lymphoid memory. This is consistent with a requirement for CD103 in resident memory formation during skin viral infection (15). However, we further highlight CD103 as a mediator of early CD8 T cell accumulation in skin, suggesting its potential role in promoting T_RM precursor cell access to skin. The specific mechanisms whereby CD103-dependent T_RM cells mediate anti-tumor immunity remain to be defined, although recent studies show that human CD103⁺ T_RM cells are superior to their CD103⁻ counterparts in production of IFN-γ and TNF-α (29). While we demonstrate a vital requirement for CD103-dependent T_RM cells in mediating anti-tumor immunity, their role in mediating vitiligo was less apparent, indicating distinct roles for CD103-dependent T_RM cells in mediating anti-tumor immunity and autoimmunity.

There remain several limitations to the present work. First, our studies do not differentiate what characteristics of the primary tumor itself are required for T_RM cell programming. One could speculate that dermal melanomas direct T cell homing to skin, and that tumor-derived TGF-β induces CD103 expression on T_RM cell precursors, however this remains to be determined. Second, the properties of autoimmune disease that promote T_RM cell responses are not yet fully understood, as we have not directly demonstrated a requirement for melanocyte killing as opposed to some other inflammatory aspect of autoimmunity in skin. Further, while we show that CD4 T cell-mediated vitiligo promotes CD8 T_RM cell responses, our results do not exclude a role for CD4 T cells in directly providing help to CD8 T_RM cells. Finally, our studies do not address the mechanism whereby vitiligo-affected mice also maintain protection against intravenously-inoculated melanoma lung metastases (11). Additional studies are needed to determine whether T_RM responses are also established in lungs of mice with vitiligo. Indeed precedence exists for the simultaneous generation of protective skin and lung T_RM populations following viral skin infection (5).

The critical role for melanoma-specific T_RM cells in tumor protection supports the induction of resident memory as a worthwhile goal for cancer immunotherapies. Forced expression of key surface molecules, such as CD103 and CLA, could potentially generate therapeutic T cells that reside in peripheral tissues. Our work underscores the equally important goal of generating a hospitable host environment for T_RM cell seeding and maintenance. In future work the depletion of T_reg cells in conjunction with checkpoint-blockade immunotherapy may further improve the host T_RM response in skin. Moving forward, T cells resident in peripheral tissues should be recognized as a vital component of the immune response to cancer.

Materials and Methods

Study design

The research objectives were (1) to characterize the melanoma-specific memory CD8 T_RM cell response in skin of mice with melanoma-associated vitiligo, and (2) to determine how T_RM cells contribute to long-lived anti-melanoma immunity. Experiments were performed using tissues from mice with vitiligo, as indicated, with cell characterization by flow cytometry or immunohistochemistry, in vitro peptide re-stimulation, grafting of skin, treatment with FTY720, or tracking of vitiligo or tumor growth. Mice were randomly distributed between experimental groups without blinding. Sample sizes and endpoints were based on prior published studies to detect similar biological effects (11) and were selected prior to each experiment. All assessed samples, including outliers, were included in analyses. Detailed descriptions of experimental parameters and data analysis can be found in figure legends and the following paragraphs. Primary source data are in Table S1.

Mice and tumor cells

C57BL/6 mice were purchased from Charles River Laboratories. All transgenic and knockout mice were purchased from The Jackson Laboratory and bred in-house (see Supplementary Materials and Methods), with the exception of Pmel-1 mice (22) on a Thy1.1 congenic background, which were a gift from Nicholas Restifo (NCI), and TRP-1 TCR transgenic mice (referred to as TRP-1 Tg), carried on a RAG^−/− TRP-1^−/− background (30), which were a gift from Paul Antony (U of Maryland). Female and male mice were used, with individual experiments using mice of a single gender, enrolled at 6–8 weeks of age. B16 melanoma, B16-OVA, and Lewis lung carcinoma (LLC) cell lines were validated as described in Supplementary Materials and Methods.

Depleting antibodies and peptides

Depleting anti-CD4 (mAb clone GK1.5) and anti-CD8 (mAb clone 2.43) were produced as bioreactor supernatants from hybridoma cell lines (American Type Culture Collection) and administered 250 μg/dose, i.p. Peptides (>80% purity) were obtained from New England Peptide: human gp100_25–33 (KVPRNQDWL), OVA_257–264 (SIINFEKL), and mouse TRP-2_180–188 (SVYDFFVWL).

Induction of melanoma-associated vitiligo

As previously described (21, 23), mice were inoculated i.d. on the right flank with 1.75 × 10⁵ B16 cells (or 5×10⁵ B16-OVA cells, where indicated) on d0, and treated with anti-CD4 mAb i.p. on d4 and 10. Tumors were surgically excised on d12. Spontaneous metastases were not observed with this B16 subline, and mice with recurrent primary tumors after surgery (<5%) were removed from the study.

Pmel and OT-I T cell adoptive transfer

Naïve TCR transgenic T cells (CD8⁺Thy1.1⁺ pmel cells or CD8⁺Ly5.1⁺ OT-I cells) were magnetically sorted by anti-CD44-PE negative selection followed by CD8 positive selection (Miltenyi). T cells were transferred i.v. at 10⁴ cells/mouse, one day prior to induction of melanoma-associated vitiligo as described above. Where indicated, Ly5.1⁺ pmel cells were admixed at a 1:1 ratio with Thy1.1⁺ FucT-VII^−/− or CD103^−/− pmel cells (see Supplementary Materials and Methods) prior to transfer.

Tissue processing and flow cytometry

Lymphoid tissues were mechanically dissociated. Skin was minced and incubated in 3 mg/mL collagenase Type IV (Worthington Biochemical Corporation) and 2 mg/mL DNase (Sigma-Aldrich) in HBSS at 37 °C for 45 minutes with stirring. Unless otherwise indicated for vitiligo-affected skin, a 2 cm² patch of perilesional or depigmented skin from the right flank was used. Single cells suspensions were stained with antibody clones specified in Supplementary Materials and Methods. Flow cytometry was performed on a MACSQuant 10 Analyzer (Miltenyi), and data were analyzed using FlowJo software V6 (Tree Star). Where given, MFI is the geometric mean of fluorescence intensity. Isotype controls and representative gating strategies are shown in Figures S12 and S13, respectively; absolute cell counts for histogram overlays are provided in Table S1.

Detection of intracellular IFN-γ

Skin was digested as above, and subjected to a percoll gradient. Cells were incubated for 5–14h at 37 °C in RPMI containing MHC I-restricted peptides (1 μg/mL; sequences given above), IL-2 (10 U/mL), and brefeldin A (10 μg/mL). Cells were fixed, permeabilized, and stained for IFN-γ, followed by flow cytometry as described above.

Immunohistochemistry and fluorescence microscopy

Paraformaldehyde-fixed sections of skin were blocked with 5% BSA, 1% goat serum, 1% rat serum and 1% donkey serum for 1h followed by staining with combinations of Anti-CD8-AF594 (clone 53-6.7), -Thy1.1-AF647 (clone OX7), and mouse E cadherin-AF488 (clone DEMA-1), and DAPI (representative control staining in Fig. S14). Sections were analyzed with a Zeiss AX10 fluorescence microscope fitted with a CoolSnap HQ2 mono14-bit camera, using an ×10 Apo objective. Images were taken with ZEN Pro 2012 (Zeiss) software and processed with Fiji software. White hairs were distinguished from black hairs in bright field images as hairs lacking black striations.

Skin grafting

Fifty days after surgery, pmel cell-adoptively transferred mice with melanoma-associated vitiligo were used as donors of 2 cm² patches of skin. Donor skin was harvested proximal to the healed tumor excision site, and was grafted onto RAG^−/− recipient mice. 50d later, skin grafts were harvested for flow cytometry. Skin from mice in the original donor group (100d post-tumor excision) was used as a positive control.

FTY720 treatment

Mice received daily i.p. injections of FTY720 (2-amino-2-(2[4-octylphenyl]ethyl)-1,3-propanediol hydrochloride, Cayman Chemical), 1mg/kg, beginning 30d post-surgery.

Adoptive transfer of tumor-primed pmel cells

Donor mice received Thy1.1⁺ pmel cells (d-1), B16 tumors (d0), and anti-CD4 (d4 and 10), as described above. On d12, spleens and lymph nodes were pooled from 10–12 mice/group. CD8 T cells (containing tumor-primed pmel cells (11, 21)) were isolated by magnetic selection (Miltenyi) and transferred i.v. into recipients that had previously been induced to develop vitiligo. Wild-type recipients had received B16 tumors, anti-CD4 mAb, and surgery, to induce melanoma-associated vitiligo. RAG^−/− mice had received 4.5×10⁴ naïve transgenic CD4 T cells from TRP-1 Tg mice, followed by sham skin surgery 1 day later. Thirty days after transfer of tumor-primed pmel cells, skin of recipient mice was analyzed by flow cytometry.

Reconstitution of RAG^−/− and CD8^−/− mice

RAG^−/− or CD8^−/− recipient mice received 8×10⁶ magnetically-isolated CD8 T cells from spleens of naïve WT, FucT-VII^−/−, or CD103^−/− donor mice. One day after transfer, melanoma-associated vitiligo was induced in recipient mice.

Tumor challenge

In mice with melanoma-associated vitiligo, 1.2×10⁵ B16 cells or 1.2×10⁵ LLC cells were inoculated i.d. ≥30 days after surgical excision of primary tumors. Cells were inoculated on flanks contralateral to surgical sites, unless otherwise indicated. Tumors were measured thrice weekly, starting on d6, and mice were euthanized at a tumor diameter of 20 mm.

Statistical analyses

Statistical analysis was performed using GraphPad Prism 5 software (GraphPad Software Inc). Data were tested for Gaussian distribution using the D’Agostino and Pearson omnibus normality test. When normally distributed, statistical differences were analyzed with unpaired t-test for comparing two groups from separate mice, paired t-test for comparing two groups from the same mouse, or one-way analysis of variance (ANOVA) with Bonferroni posttest for comparing multiple groups. When not normally distributed, the Mann-Whitney test was used to compare two unrelated groups, the Wilcoxon matched pairs test to compare two groups from the same mouse, and the Kruskal-Wallis test with Dunns posttest to compare multiple groups. Differences in tumor incidence were assessed using the Gehan-Breslow-Wilcoxon method (hazard ratios not proportional), tumor growth differences were assessed by two-way ANOVA with Bonferroni posttest, and differences in vitiligo incidence were assessed by log-rank test. t tests were two-sided, and a significance level of 0.05 was used for analyses.

Supplementary Material

Suppl Materials

Figure S1. Phenotype of CD8 T cells and OT-I cells in vitiligo-affected skin

Figure S2. IFN-γ production by endogenous TRP-2-specific CD8 T cells in skin

Figure S3. Appearance of vitiligo-affected skin grafts on RAG^−/− mice

Figure S4. Phenotype of skin graft-derived CD8 T cells that enter lymph nodes

Figure S5. CD103-expression by tumor-primed pmel cells

Figure S6. Absence of CLA on FucT-VII^−/− T cells

Figure S7. Tumor growth kinetics in FucT-VII^−/− mice

Figure S8. Adoptive immunotherapeutic efficacy of CD103-deficient CD8 T cells

Figure S9. CD8 T cell depletion from skin by anti-CD8 mAb

Figure S10. Effects of local skin depigmentation status on tumor protection

Figure S11. Absence of protection against Lewis lung carcinoma in mice with vitiligo

Figure S12. Isotype staining controls for flow cytometry

Figure S13. Flow cytometry gating strategies

Figure S14. Representative control staining for immunohistochemistry

Table S1. Primary source data

Editor’s summary.

Resident memory to cancer

Melanoma patients with vitiligo are more likely to have a positive outcome, but the mechanism behind this association has remained unclear. Now Malik et al. report that skin-resident memory T (T_RM) cells specific to melanoma antigens are maintained in vitiligo-affected skin. These cells persist and function independently of the lymphoid compartment, suggesting the vitiligo lesions provide a niche for the T_RM cells. The T_RM cells provide durable memory to the tumor, even in pigmented skin. These data suggest that skin-resident T_RM cells are critical for maintaining anti-tumor immunity.