Immune control of functional memory CD8 T cells in normal-appearing vitiligo skin Authors: Laure Migayron, Ribal Merhi, Chloé Avril, Simon Léonard, Atika Zouine, Fanélie Jouenne, Sophie Gilardeau, Claire Drullion, Denis Thiolat, Emilian Mertinaj, Mélanie Humeau, Adèle Mauroux, Jean-Philippe Guégan, Xavier Gauthereau, Samia Mourah, Thomas Darde, Sylvie Bordes, Brigitte Closs-Gonthier, Julien Seneschal, Katia Boniface Published: 2025 ABSTRACT Emerging evidence demonstrates that self-reactive CD8+ T resident memory cells (TRM cells) may already populate the never-lesional (NL) skin of patients with chronic inflammatory conditions, including vitiligo. However, their functional activity and regulation remain unknown. By combining cellular and transcriptomic approaches, we show that common CD8+ T cell clusters with a similar transcriptional profile infiltrate NL and peri-lesional (PL) skin of patients with vitiligo. Pathways related to immune activation were more active in T cells from PL skin, while those related to immune regulation were enriched in NL skin T cells, also evidenced by the stronger infiltration of regulatory T cells and the higher expression of programmed cell death protein 1 (PD-1) by CD8+ T cells in NL skin. Treatment of patients with baricitinib led to an increase in regulatory T cells and CD8+ T cells expressing PD-1 in repigmented skin. These findings suggest that functional T cells populate pigmented skin in vitiligo and highlight the importance of regulatory mechanisms to restrain their activation and prevent depigmentation. INTRODUCTION T resident memory cells (TRM cells) play a crucial role in the establishment of local immune memory in tissues like skin to provide a first line of defense against foreign pathogens (1, 2). TRM cells accumulate over time and are mainly anchored in barrier tissues like the skin, where they may persist for a long time under resting conditions in the absence of their cognate antigen. Tissue residency partly results from the expression of the C-type lectin CD69 and the alpha chain of the integrin αEβ7, also known as CD103 (3–5). The latter binds E-cadherin expressed by epidermal cells, explaining the epidermal location of CD103+ TRM cells in the skin (4). In addition, CD49a defines a subset of epidermal CD8 TRM cells prone to cytotoxic function (6, 7). While first reported for their protective function, self-reactive TRM cells can also be harmful to the organism and play a key role in the occurrence, maintenance, and relapse of several chronic immune and inflammatory skin diseases such as psoriasis, atopic dermatitis, graft-versus-host disease, and vitiligo (6, 8–11). Vitiligo is a CD8+ T cell–mediated depigmenting inflammatory disease characterized by the appearance of white macules resulting from the loss of epidermal melanocytes (12). Both lesional and peri-lesional (PL; the skin area located in close proximity to the lesion border) areas harbor an increased number of CD8+ TRM cells releasing elevated levels of the type 1 cytokines interferon-γ (IFN-γ) and tumor necrosis factor–α (TNF-α) (8, 13–16). In vitro and in vivo studies evidenced the role of TRM cells in disease pathogenesis (17). Hence, targeting TRM cells and the impact of their related cytokines on the epidermis, and in particular melanocytes, represents a promising therapeutic strategy for patients. Janus kinase (JAK) inhibitors have shown notable efficacy in patients with vitiligo (18–22). However, they failed to eliminate autoreactive TRM cells in a mouse model of depigmentation (23). Nonetheless, accumulating evidence highlights the role of TRM cells in the relapse of lesions in inflammatory skin diseases like psoriasis and atopic dermatitis (24). Following effective therapy, resolved psoriatic and atopic dermatitis skin still contains pathogenic TRM cell clones that have the propensity to induce disease recurrence following reactivation (25–28). Similar T cell clones have also been found in the never-lesional (NL) skin of patients, corresponding to clinically normal-appearing, pigmented skin (24, 26, 29–31). Normal-appearing vitiligo skin already displays abnormalities compared to healthy skin (HS), with an increase in epidermal thickness; altered expression of adhesion proteins important for melanocyte stabilization to the basal layer, like E-cadherin; and an increase in inflammation-associated pathways (32–34). It was recently shown that keratinocytes from vitiligo nonlesional skin are characterized by a deregulated differentiation with altered lipid composition and impaired metabolism associated with increased secretion of pro-inflammatory soluble factors (35). In addition, Gellatly et al. (36) reported an up-regulation of the IFN-γ pathway in NL vitiligo skin CD8+ T cells at the transcriptional levels, albeit at a weaker level compared to the lesional area. These data beg the question of whether the NL skin of patients with vitiligo already contains pathogenic TRM cells fully equipped to trigger depigmentation. In this study, we used multispectral immunofluorescence, flow cytometry, and single-cell RNA sequencing (scRNA-seq) analyses to show that while infiltration of CD8+ T cells was more prominent in PL vitiligo skin, NL skin displayed similar subsets of TRM cells. Common CD8+ T cell clusters were identified in both sites. However, gene ontology (GO) analysis identified the enrichment of pathways related to T cell activation in PL skin. Analysis of T cell receptor (TCR) diversity revealed the presence of similar T cell clones in NL and PL vitiligo skins, suggesting similar antigen specificity. Upon T cell activation in situ, NL skin displayed an intermediate inflammatory transcriptional profile compared to HS and PL vitiligo skin, with an increase in genes associated with cellular response to IFN-γ and chemokine activity. In addition, activated NL vitiligo skin T cells showed an increased capacity to secrete both type 1– and type 2–related cytokines. CellChat analysis by scRNA-seq of NL and PL vitiligo skins identified signaling pathways associated with immune regulation enriched in NL skin, such as CD96, transforming growth factor–β (TGFβ), and CD39, notably expressed by regulatory T cells. CD8+ T cells in NL areas showed a higher expression of the immune checkpoint receptor (ICR) programmed cell death protein 1 (PD-1), likely retaining their inflammatory function. Treatment of patients who have vitiligo with the JAK1/2 inhibitor baricitinib increased the expression of PD-1 on CD8+ T cells. These results suggest that “pathogenic” TRM cells already populate the normal-appearing skin of vitiligo and reveal the importance of regulatory mechanisms that restrain their activation and prevent depigmentation. RESULTS Single-cell profiling of NL and PL vitiligo skins reveals the presence of similar T cell clusters We first used scRNA-seq to investigate the cellular composition and transcriptional profiles of NL skin and PL skin from the arms of six patients with active vitiligo (Fig. 1A). Following enzymatic digestion of the dermis and epidermis, CD45− nonimmune cells and CD45+ immune cells were sorted by flow cytometry from both epidermal and dermal cell suspensions to subject 10,000 CD45− and 10,000 CD45+ cells to scRNA-seq with the 10x Genomics Chromium platform to generate a transcriptional profile for each individual cell (Fig. 1A and as described in Materials and Methods). The final dataset comprised 77,532 cells with a median of 7139 cells per sample and an average of 2289 genes detected per cell. Uniform manifold approximation and projection (UMAP) dimensionality reduction and unsupervised clustering identified 23 distinct cell clusters (fig. S1A). Most of these clusters were detected in all samples including PL and NL skin. Clusters were annotated as the 12 main cell types typically found in the skin using canonical markers established in the literature (Fig. 1B and fig. S1B): fibroblasts (FGF2+COL3A1+COL1A2+COLA1+), keratinocytes (KRT5+KRT10+KRT7+KRT14+), melanocytes (MITF+MLANA+PMEL+TYR+DCT+), endothelial cells (PECAM1+VWF+CD93+EGFL7+), pericytes (MCAM+ACTA2+TAGLN+TGSS+), neuron cells (NRXN1+FRMD5+CADM2+), B cells (IGKC+IGHA1+IGLC2+), T cells (PTPRC+CD3E+CD4+CD8A+), NK cells (natural killer cells; PTPRC+KLRK1+GNLY+), mastocytes (KIT+HPGDS+TPSB2+CTSG+), myeloid cells (ITGAM+ITGAX+CLEC10A+FCER1A+), and dendritic cells (DCs; CD1E+CD74+CD83+CD86+CST3+) (Fig. 1B). A decrease in melanocytes was observed in PL skin, reflecting the loss of melanocytes in PL vitiligo skin (Fig. 1C). Single-cell landscape of NL and PL vitiligo skins. scRNA-seq analysis of NL and PL skin biopsies obtained from the arms of six patients with active vitiligo. (A) Schematic view of the workflow. (B) UMAP visualization of the expression profiles of the 77,532 single cells derived from the 12 skin tissues. Unbiased clustering of the scRNA-seq data in 12 different cell types and their respective canonical markers transcripts. (C) Cell type proportion in NL skin vs PL skin. (D) Refined clustering of the T cell cluster resulted in 17 distinct clusters. TFH cells, T follicular helper cells. (E) UMAP visualization of the distribution of selected transcripts defining the different T cell clusters identified in (D). (F) Number of CD3+ T cells in NL and PL skins. Data show the means +/− SEM. (G) Proportion of NL or PL cells forming each T cell cluster. (H) Heatmap of all detected transcripts (5140) in T cell clusters for each patient sample. To better characterize and compare the T cell compartment at the transcriptional level in NL and PL vitiligo skins, we subclustered the T cell population and obtained 17 clusters: 6 CD4+ T cell clusters, 1 T peripheral helper cell–like cluster (CD4+CXCL13+), 2 regulatory T cell (Treg cell) clusters (FOXP3+), 7 clusters of CD8+ T cells, and 1 cluster including both CD4+ and CD8+ T cells (Fig. 1, D and E, and fig. S1C). PL skin contained more T cells than NL areas (Fig. 1F), suggesting a prominent T cell infiltrate in this area. Each T cell cluster was composed of a majority of T cells from PL skin (Fig. 1G). However, the transcriptomic profiles between matched NL and PL skins from the six patients were similar, as shown by the heatmap generated using an unsupervised approach based on the mean expression of genes detected in at least 10% of T cells (including 5140 genes) (Fig. 1H and fig. S1D). Similar T cell clones infiltrate NL and PL skins of patients with vitiligo We next assessed whether T cells from NL and PL vitiligo skins could present similar TCR diversity. High-throughput sequencing of the CDR3 region of the TCRβ gene was performed in both NL and PL skins of eight patients with vitiligo and four HS samples. The possibilities of sequence rearrangements of the TCRβ gene made each sequence and corresponding clone unique to each sample. T cells from vitiligo skin and HS were mostly polyclonal (Fig. 2A and fig. S2, A and B). The sample overlap, which measures the percentage of amino acid sequences shared between the samples, showed the sharing of sequences within paired PL and NL skin, distinctively from other pairs (fig. S2C). High-throughput sequencing of TCRβ revealed many common clones in matched NL and PL skins from the same patient, with no emergent clone detected, except for patient V878 (Fig. 2, A and B, and fig. S2C). The four HS samples did not share any common clone (fig. S2, A and B). The vast majority of the top 10 predominant clones observed in NL and PL vitiligo skins, as well as HS, was not predicted to recognize common viral or microbial antigens. Overall, NL and PL vitiligo skins display a similar T cell repertoire diversity that is distinct from HS, suggesting a common background between the affected lesion site and the clinically normal-appearing skin of patients, as previously observed in other skin inflammatory diseases, like atopic dermatitis (26). However, clonal sharing across skin sites cannot be considered as a disease-specific signature, as it is also observed between two proximal sites of HS (fig. S2D), possibly reflecting the normal tissue distribution and persistence of TRM cells. The number of unique TCRβ CDR3 sequences was significantly higher in PL vitiligo skin compared to NL skin and HS (fig. S2E), consistent with the increased T cell infiltration in PL skin. The number of unique T cell clones also tended to be higher in NL vitiligo skin compared to HS, whereas no difference was observed between the two sites of HS. NL and PL vitiligo skins share common T cell clones. NL and PL skin biopsies were obtained from the arms of eight patients, and T cell clone diversity was evaluated by high-throughput TCRβ sequencing. (A) Pie chart of the TCR Vβ rearrangement detected in each sample. (B) For each patient, left panels represent the dot plot of the NL and PL T cell clone frequency. Each blue or red dot represents a T cell clone detected only in NL or PL skins, respectively, and purple dots indicate T cell clones shared between NL and PL skins. Right panels represent frequencies of the top 10 T cell clones in NL and/or PL skin. Within the same patient, a common color indicates a common clone between NL and PL skins. NL and PL vitiligo skins contain similar CD8+ TRM cell subsets We then investigated the phenotype of CD8+ T cells infiltrating NL and PL skins of patients with vitiligo. As expected, immunofluorescence analyses showed an increase in CD8+ T cells in both the epidermis and dermis of PL skin compared to HS (Fig. 3, A and B). NL skin also contained a higher number of CD8+ T cells compared to HS, albeit lower than in PL areas (Fig. 3, A and B), which was confirmed by flow cytometry (fig. S3, A and B). Given the central role of TRM cells in vitiligo physiopathology, we then evaluated the presence of CD8+ TRM cell subsets, defined by the markers CD69 and/or CD103 and/or CD49a. Multispectral imaging and multiparametric flow cytometry analyses showed a similar frequency of CD8+ TRM cell subsets in NL and PL skins from the same patients (Fig. 3, C and D, and fig. S3C). We investigated the distribution of CD69+ CD103+ CD8+ TRM cells, a subset associated with more “mature” and epidermally localized CD8+ TRM cells (37, 38). While a nonsignificant trend toward higher frequencies of this subset was observed in both NL and PL skins compared to HS, interpatient variability was considerable (fig. S3D), and no consistent pattern correlated with disease activity. In addition, no significant difference was observed in the median distance between CD8+ T cells and melanocytes in NL and PL zones (Fig. 3E). We further characterized vitiligo skin CD8+ T cell heterogeneity and adopted an unbiased approach from scRNA-seq data, which showed an increase in their frequency in PL skin (Fig. 3F) and identified eight clusters of CD8+ T cells (Fig. 3G). All patients and skin sites were represented in each cluster (fig. S4A). These clusters were defined according to the expression of specific transcripts (fig. S4, B and C), as described below: cluster 1 as a mix of TCM + TRM Tc2- and Tc17-like cells (ITGAE+TCF7+LEF1+GATA3+CCR4+CCR6+), cluster 4 as KLRG1+ TEM cells (effector memory T cells; KLRG1+NKG7+EOMES+), cluster 5 as TCM cells (central memory T cells; TCF7+LEF1+), cluster 6 as effector TRM Tc1 cells (CD69+IFNG+NKG7+EOMES+), cluster 8 as a mix of TCM + TRM Tc1- and Tc17-like cells (TCF7+ITGA1+RORA+DPP4+STAT4+RUNX2+GZMB+), cluster 10 as stressed effector TRM Tc1 cells (CD69+NKG7+IFNG+TNF+HSPA1A+), cluster 12 as TCM cells (TCF7+LEF1+), and cluster 16 as ENTPD1+ TRM cells (ITGAE+IFNG+ENTPD1+). The frequency of each cluster was similar in NL and PL skins (Fig. 3, H and I). Clusters 6, 10, and 16, corresponding to TRM cells, displayed the highest levels of TNFA and IFNG transcripts (Fig. 3, J and K, and fig. S4B). Furthermore, the number of cells expressing these transcripts tended to be lower in NL skin (fig. S5A), suggesting a lower propensity of CD8+ TRM cells to release type 1 cytokines in the normal-appearing skin of patients with vitiligo. When comparing CD8+ T cell cluster distributions between NL and PL skins, we identified several differentially expressed genes (DEGs; table S1). To further explore the functional implications of these DEGs, we performed pathway analysis focusing on the Gene Ontology Biological Process database. Most clusters exhibited pathways associated with immune activation and/or antigen presentation in PL skin compared to NL skin (fig. S5B). NL and PL vitiligo skins harbor similar CD8+ T cell subsets. (A and B) Immunofluorescence analysis of CD8+ T cells in NL and PL vitiligo skins obtained from the arm (n = 7) and HS obtained from breast plastic surgery (n = 7). (A) Representative image showing the skin infiltration of CD8+ T cells (in red). Cell nuclei stained with DAPI are illustrated in gray. Scale bars: 50 μm. (B) Number of CD8+ T cells per field in the epidermis and the dermis. Data show the means +/− SEM. *P < 0.05. (C to E) Vectra Polaris multispectral imaging combining MelanA, CD8, CD69, and CD103 immunofluorescence staining in NL and PL vitiligo skins (n = 6). (C) Representative image showing melanocytes stained with MelanA (in red) and the infiltration of CD8+ (in white), CD103+ (in purple), and CD69+ (in orange) T cells in NL skin (left column) and PL skin (right column). Individual staining is shown. Scale bars: 50 μm. (D) The frequency of CD8+ T cells expressing or not expressing CD69 and/or CD103 is shown for each patient. (E) Median distance between CD8+ T cells and melanocytes (MelanA+ cells). Data show the means +/− SEM. (F) Percentage of CD8+ T cells in cells expressing PTPRC from scRNA-seq analysis. Data show the means +/− SEM. (G to K) Refined clustering of CD8+ T cell clusters from scRNA-seq analysis. (G) UMAP representation of the eight clusters obtained. (H) Proportion of each cluster in NL and PL vitiligo skins. (I) Distribution of CD8+ T cells in the eight clusters in NL and PL skins. Data show the means +/− SEM. (J) UMAP visualization of the distribution of IFNG and TNF transcripts. (K) Frequency of cells expressing IFNG and TNF genes in each CD8+ T cell cluster of NL and PL skins. Data show the means +/− SEM. *P < 0.05. In situ–activated T cells from NL vitiligo skin induce a vitiligo-like inflammatory transcriptional profile We and others previously showed that PL and lesional vitiligo skin is characterized by an up-regulation of type I and type II IFN pathways (8, 36, 39–43). In addition, we recently showed that PL vitiligo skin also displays an increase in both type 1 and type 2 T cell responses (44). We therefore sought to determine whether T cells in the NL skin of patients were able to regulate the local inflammatory response. To this end, polyclonal activation of T cells in skin explants from NL and PL vitiligo skins and HS was induced with anti-CD3 T cell activating antibody OKT-3. As expected, activated PL vitiligo skin T cells mainly produced type 1 and type 2 cytokines, while cytokine release from HS T cells remained low (Fig. 4A and fig. S6, A to C). NL vitiligo skin T cells also released significantly higher levels of these cytokines upon activation, although at lower levels than those from PL skin. Flow cytometry analysis of phorbol 12-myristate 13-acetate/ionomycin–stimulated skin cell suspensions showed that although not statistically significant, the frequency of CD8+ T cells producing IFN-γ, TNF-α, or interleukin-13 (IL-13) tended to be lower in NL and PL vitiligo skins compared to HS (fig. S6D), suggesting that the overall cytokine production is more likely related to the density of CD8+ T cell infiltration in patients’ skin rather than an enhanced intrinsic capacity of these cells to produce cytokines upon stimulation. NL and PL vitiligo skins display a similar inflammatory transcriptional profile following T cell activation. Skin explants from HS (n = 9, obtained from breast plastic surgery) and NL and PL vitiligo skins (n = 15, obtained from the arm) were incubated for 16 hours with an anti-CD3 mAb. (A) Levels of indicated cytokines were assessed by multiplex enzyme-linked immunosorbent assay in a cell-free supernatant. Data show the means +/− SEM. *P < 0.05, **P < 0.01, and ***P < 0.001 (B to E) RNA transcript analysis using NanoString technology was performed on HS (n = 6) and NL and PL vitiligo skins (n = 9). (B) PCA of 579 RNA transcripts is presented. Each dot corresponds to one patient. (C) Heatmap of 125 DEGs in PL and/or NL vitiligo skins compared to HS. (D) GO biological process enrichment analysis using up-regulated genes. Only the top 10 pathways are shown. (E) Normalized expression of selected genes of interest in HS and NL and PL vitiligo skins is shown. Data show the means +/− SEM. *P < 0.05, **P < 0.01, and ***P < 0.001. These results prompted us to investigate the impact of T cell activation in NL vitiligo skin. The expression of 579 genes related to immunity and inflammation was quantified using NanoString technology in NL and PL vitiligo skins and HS exposed to OKT-3. Principal components analysis (PCA) separated HS from NL and PL skins (Fig. 4B). PCA did not discriminate NL skin from PL skin, showing that NL vitiligo skin displayed similarities with PL skin and not HS. We identified 125 DEGs in NL/PL vitiligo skin compared with HS: 90 were up-regulated and 35 were down-regulated (Fig. 4C, fig. S6E, and table S2). In contrast, no DEG was obtained when comparing NL versus PL vitiligo skin. GO on up-regulated genes revealed several pathways related to inflammation and chemokine and cytokine signaling (Fig. 4D). A significant increase in vitiligo-associated genes, such as CXCL9, CXCL10, TNF, STAT1, MX1, and CCL18, was observed in NL skin compared to HS, albeit at a lower level than in PL skin (Fig. 4E). Together, these data suggest that the NL skin of patients with vitiligo is “armed” with an increased density of T cells that, upon activation, can induce a local inflammatory response and likely contribute to depigmentation. Our functional analyses indicate that this heightened inflammatory activity results from the greater abundance of T cells in the skin rather than an intrinsic increase in their per-cell functional capacity, highlighting T cell density as a potential determinant of disease activity between NL and PL sites. Regulatory pathways control the immune response of CD8+ T cells in NL vitiligo skin Given that functional self-reactive CD8+ T cells are likely already present in the NL skin of patients with vitiligo, we decided to look at the overall gene set enrichment analysis of biological process pathways in CD8+ T cells from PL skin versus NL skin. We found that most of the pathways up-regulated in PL skin were related to immune activation (fig. S7A). Therefore, we sought the mechanisms that could explain the absence of depigmentation. We used CellChat on our scRNA-seq data to detect signaling changes between NL and PL skin T cells from patients with vitiligo. To identify the conserved and context-specific signaling pathways in NL and PL skins, we compared their overall communication probability. Eleven of 32 signaling pathways were significantly more active in PL skin T cells, including pathways involved in inflammatory and immune responses such as ICAM, MHC-I, MHC-II, CD40, CD70, TNF, and IFN-II (Fig. 5A). These data reinforce the hypothesis of microenvironmental inflammation in PL leading to the loss of melanocytes and the development of new active lesions. NL vitiligo skin is enriched in regulatory pathways. (A and B) Cell-cell communication analysis using CellChat. (A) Signaling pathways significantly up-regulated in T cells were ranked by differences in overall information flow between NL (blue) and PL (red) vitiligo skins. (B) Heatmap showing the relative strength of these pathways in the Treg cell cluster in NL and PL skins. (C to E) Immunofluorescence analysis of CD8+ T cells and Treg cells in NL and PL vitiligo skins (n = 8, arm samples). (C) Representative image showing CD8+ T cells (red), FOXP3+ Treg cells (green), and DAPI-stained nuclei (gray). Scale bars: 50 μm. (D) Ratio of FOXP3+ Treg cells in NL skin vs PL skin per section. (E) FOXP3+/CD8+ T cell ratio in NL and PL skins. Data represent the means +/− SEM. *P < 0.05. (F) Violin plots of FOXP3 and ENTPD1 expression in the Treg cell cluster from scRNA-seq data (n = 6). (G) Dot plots showing the expression of specific ligands (left) and receptors (right) significantly up-regulated in NL skin. Treg cells are highlighted in red; Treg cell–associated interactions are connected by dark lines. The color scale indicates the expression level; the dot size reflects the percentage of expressing cells. (H) Flow cytometry analysis of T cells from NL or PL skin (n = 7). Frequency or mean fluorescence intensity (MFI) of PD-1 in CD8+CD45RO+ T cells. *P < 0.05. (I to K) Vectra Polaris multispectral imaging of CD8, PD-1, CD11c, and PD-L1 in NL and PL skins (n = 6). (I) Representative image: CD8+ (white), PD-1+ (blue), CD11c+ DCs (green), and PD-L1+ cells (yellow). Scale bars: 50 μm. (J) Ratio of CD8+ T cells expressing PD-1. (K) Ratio of CD11c+ cells expressing PD-L1. (L) NL vitiligo skin explants (n = 5, arm samples) were treated with anti-CD3 in the presence of an anti–PD-1 mAb or isotype control. IFN-γ levels were measured by enzyme-linked immunosorbent assay. Conversely, 21 signaling pathways were enriched in NL skin T cells, mainly related to pathways associated with immune regulation, such as CD96, TGFB, CD99, CD46, CD39 (ENTPD1), and NOTCH (Fig. 5A). Five signaling pathways, including IL2, ICOS, and FASLG, were specifically active in NL skin T cells. Overall signaling patterns in NL skin compared to PL skin revealed that these pathways, along with the CD39 pathway, exhibited higher expression in Treg cells within NL skin (Fig. 5B). Treg cells play an immunosuppressive role via several mechanisms, including cell-to-cell contact with other immune cells (CTLA4, TIGIT, and CD96), the production of anti-inflammatory cytokines (TGFβ and IL-10) or cytotoxic granules, and the conversion of adenosine 5′-triphosphate to adenosine initiated by the ectonucleotidase CD39 (45). In addition, a higher infiltration of Treg cells together with an increase in the FOXP3+/CD8+ cell ratio was observed in NL skin (Fig. 5, C to E). Differential expression analysis within each donor revealed variable transcriptional differences between Treg cells from NL and PL skins, with no consistent DEG shared across all patients. Nonetheless, genes involved in the Treg cell suppressive function, such as FOXP3 and ENTPD1 (encoding CD39), were up-regulated in NL skin compared to PL skin, and a higher proportion of Treg cells from NL expressed IL2RA compared to PL skin (Fig. 5F and fig. S7B). To study the specific ligand-receptor pairs in NL skin, we selected the pairs with higher interaction scores in NL skin, which revealed pathways arising from both immune and nonimmune cells (Fig. 5G). Although no difference was observed in the expression of CD96, TIGIT, and TGFB transcripts (fig. S7C), the interaction between Treg cells and DCs via TIGIT, CD96-NECTIN2, or ENTPD1-ADORA2B/ADORA2A and the interaction between Treg cells and CD8 T cells via TGFB-TGFBR were up-regulated in NL skin (Fig. 5G). These results suggest that immunosuppressive mechanisms are involved in the regulation of CD8 T cell activation in NL skin. Treg cells are not the only known regulators of immune cell activation. Inhibitory ICRs play an important role in controlling the activation of CD8+ T cells. Damo et al. (46) recently described the role of PD-1 (encoded by PDCD1) in maintaining CD8+ T cell tolerance toward neoantigens in the skin. Hence, we used flow cytometry to compare the expression of PD-1 on CD8+ T cell subsets in NL and PL skins of patients with vitiligo. The frequency of CD8+ memory T cells expressing PD-1 was higher in NL skin than in PL skin (Fig. 5H). Multispectral imaging confirmed this observation and showed a tendency toward a higher number of CD11c+ DCs expressing programmed death-ligand 1 (PD-L1) in NL skin (Fig. 5, I to K). Unsupervised analysis using the phenograph algorithm identified three PD-1+ clusters expressing TRM cell markers present in both NL and PL skins (fig. S8A). Clusters 2 and 3 showed higher PD-1 expression in NL skin relative to PL skin (fig. S8B). Clusters 1 and 2 expressed CD69 and HLA-DR, likely defining CD8+ TRM cells in the dermis. Cluster 3 was positive for CD69, CD103, and CD49a, which is characteristic of epidermal CD8+ TRM cells (fig. S8C). In addition, PD-1+CD69+CD103−CD8+ TRM cells were more frequent in NL skin compared to PL skin, whereas PD-1 expression among CD69+CD103+CD8+ TRM cells was variable across patients (fig. S8D). Concurrent T cell activation and PD-1 blockade with nivolumab in skin explants from NL vitiligo areas led to an increase in IFN-γ production (Fig. 5L), whereas the effect on IL-10 secretion was more variable and did not follow a consistent pattern (fig. S8E). These results suggest that PD-1 signaling may prevent the effector function of CD8+ T cells infiltrating NL skin in patients. Effective therapy promotes regulation in vitiligo skin We next investigated whether repigmentation of vitiligo lesions was associated with an increase in regulation in patients. We conducted a phase 2 clinical trial investigating the effect and tolerance of the association of baricitinib and phototherapy in patients with active vitiligo (BARVIT study NCT# 04822584), which demonstrated greater and faster repigmentation than phototherapy alone (47). During this trial, skin biopsies were performed on PL areas at the baseline and on the same lesion after 9 months of treatment. Immunofluorescence analyses on six patients showed a decrease in CD8+ T cell infiltration in NL skin after 9 months of treatment (Fig. 6, A and B). In addition, we observed a higher number of CD8+ T cells expressing PD-1 and DCs expressing PD-L1 (Fig. 6, A, C, and D) together with an increase in the FOXP3+/CD8+ T cell ratio (Fig. 6, E and F) following treatment. These findings suggest that effective therapy may contribute to the promotion of immune-regulatory pathways in repigmented skin. However, this hypothesis will require validation in future fundamental and translational studies. Treatment of patients with vitiligo with baricitinib and phototherapy induces an increase in PD-1+ CD8+ skin T cells. Patients with active vitiligo were treated with an association of baricitinib and phototherapy (BARVIT study). Skin biopsies were performed on PL skin at the baseline (BL) and on the same lesion after 9 months of treatment (M9). (A to D) Vectra Polaris multispectral imaging combining CD8, PD-1, CD11c, and PD-L1 staining at BL and M9 (n = 6). (A) Representative image showing the infiltration of CD8+ T cells (in gray), PD-1+ cells (in blue), CD11c+ DCs (in green), and PD-L1+ cells (in yellow) in the skin at BL (left column) and M9 (right column). Individual staining is shown in the lower rows. Scale bars: 50 μm. (B) The number of CD8+ T cells per section and (C) and (D) ratios of CD8+ PD-1+ cells with CD8+ T cells (C) and of CD11c+ PD-L1+ with CD11c+ (D) at BL and M9 are shown. *P < 0.05. (E and F) Immunofluorescence analysis of CD8+ T cells and Treg cells at BL and M9 (n = 6). (E) Representative image showing the skin infiltration of CD8+ T cells (in red) and of FOXP3+ T cells (in green). Cell nuclei stained with DAPI are illustrated in gray. Scale bars: 50 μm. (F) The ratio of FOXP3+ cells with CD8+ T cells at BL and M9 is shown. *P < 0.05. DISCUSSION Our findings shed light on the complexity of the skin immune response in vitiligo, especially in normal-appearing skin. Our data show that the NL skin of patients with vitiligo is characterized by an immunologically primed pre-lesional state, with the presence of functional T cell subsets capable of initiating the disease. Using scRNA-seq of suction blisters obtained from lesional and NL skin, Gellatly et al. (36) posited the presence of subclinical inflammation in NL skin by revealing intercellular signaling pathways involving cells located in the epidermal layer, including keratinocytes, melanocytes, and T cells. However, owing to the technique used to isolate the cells, no information was available regarding cell subsets located deeper in the dermis such as fibroblasts, endothelial cells, and/or other immune cells or potential interactions with resident T cells. NL and PL vitiligo skins exhibit elevated expression of inflammatory transcripts compared to HS. Whether this heightened inflammatory state is specific to the skin or reflects underlying basal inflammation in the context of autoimmunity remains to be fully elucidated. In support of the latter, previous studies have reported immune deregulation within T cell subsets in the peripheral blood of vitiligo (48, 49). The recognition of self-antigen by autoreactive T cells is the classical hallmark of autoimmunity. Using TCR sequencing, we now show that T cell clones in both NL and PL skins of patients with vitiligo are highly polyclonal, with no clear evidence of oligoclonal expansion. However, the top expanded clones in PL skin are also found in NL skin, suggesting that the normal-appearing skin is already armed with pathogenic resident T cells capable of inducing the disease. This hypothesis has already been suggested in other chronic inflammatory skin diseases such as psoriasis and atopic dermatitis (26, 29–31). Our findings indicate that vitiligo, like other chronic inflammatory skin conditions, is not triggered by a dominant clone. While the presence of melanocyte-specific T cells has been identified in vitiligo skin, these T cells represent a small proportion of infiltrating T cells, and it cannot be ruled out that as yet unidentified autoantigens trigger T cell activation and melanocyte loss in vitiligo. Our findings raise questions about the origins of these functional T cells in NL skin. Recently, Connors et al. (50) explored TRM cell development and memory formation and showed that tissue memory T cells progressively mature over childhood to become functional, tissue-adapted TRM cells highly able to produce pro-inflammatory cytokines that, in turn, establish tissue immune protection. Therefore, very early in life, antigen-specific T cells might populate the skin of patients genetically predisposed to developing an autoimmune disease such as vitiligo, even before its onset. These antigen-specific TRM cells might accumulate in the skin where the target cells predominate. Even if the reservoir of tissue TRM cells is flexible, antigen-specific T cells might have a selective advantage over other T cell subsets and persist in peripheral tissue chronically exposed to the antigen, as previously demonstrated (51, 52). The continuous presence of local antigen may reduce CD8+ T cell responsiveness to stimulation, as CD8+ T cells from NL and PL skins tended to produce lower inflammatory cytokines compared to those from HS. Our findings, together with a recent study from Brunner et al. (53), indicate that type 2 inflammation, particularly IL-13 production by T cells, is a consistent feature of both NL and PL vitiligo skins. This expands the traditional view of vitiligo as a predominantly T helper 1 cell (TH1 cell)–driven disease, suggesting a more complex and heterogeneous immune response involving both TH1 and TH2 cell pathways. Yet, the precise role of type 2 inflammation in vitiligo pathogenesis remains unclear. Further research is needed to determine whether this pathway contributes to disease progression and modulation of tissue inflammation or potentially influences therapeutic outcomes, such as responses to JAK inhibitors. The cross-talk between immune, epidermal, and dermal cells is a key feature of vitiligo pathogenesis. Keratinocytes are central players in the mechanism of the disease. In this line, a recent study showed that keratinocytes from normal-appearing skin displayed alterations in lipid barrier composition as well as impaired energy metabolism (35). They showed a high capacity to produce pro-inflammatory mediators in response to mechanical stress, such as IL-1β and C-X-C motif chemokine ligand 10 (CXCL10), suggesting that keratinocytes in NL skin could be involved in the initiation of the disease through the recruitment of immune cells and the activation of resident immune cells. In addition, IL-15 signaling is crucial for TRM cell persistence in healthy and vitiligo skin, and keratinocytes trans-present IL-15 to TRM cells, leading to their activation. A similar signaling mechanism could also be hypothesized in NL skin to maintain the pool of TRM cells (54, 55). Emerging data also suggest that other nonimmune cells such as fibroblasts and endothelial cells also contribute to disease pathogenesis (56–58). If the normal-appearing skin is already in a state of subclinical inflammation and is fully equipped with functional pro-inflammatory T cell subsets, an important clinical question arises: Why do patients not develop depigmentation in normal-appearing skin? Our data provide some answers, given that regulatory mechanisms such as Treg cells and ICRs likely restrain the pathogenicity of T cells infiltrating the pigmented skin of patients. Treg cells are major tolerogenic factors maintaining immune tolerance and preventing autoimmunity, in part through the inhibition of T cell immune response. Our data reinforce the potential role of Treg cells in controlling immune activation in NL vitiligo skin. The dysfunction and/or lower number of Treg cells in the blood and skin of patients with vitiligo has been previously reported (59–65). In NL skin, Treg cells might maintain tolerance against melanocytes. In this line, a higher FOXP3/CD4 ratio was observed in the NL skin of patients with psoriasis compared to lesional skin (66). CCR5 expression on Treg cells is mandatory to suppress vitiligo in vivo (36). However, the precise role of Treg cells in controlling the local activation of resident immune cells in NL skin will need further investigation. CD8+ T cells in NL skin express PD-1 at higher levels than in PL skin, suggesting that PD-1 expression may prevent activation of CD8+ T cells, a hypothesis further supported by our findings showing that PD-1 blockade enhances IFN-γ production by CD8+ T cells from NL skin. Our data are in accordance with recent findings showing that PD-1 maintains self-reactive CD8 T cells in a tolerant state, preventing their acquisition of effector functions and the elimination of cells expressing self-antigens (46). In addition, we show that CD8+ T cells expressing PD-1 are mainly located in the dermis and that this is associated with an increased proportion of myeloid CD11c+ cells expressing PD-L1 in the normal-appearing skin of patients with vitiligo. One could speculate that self-reactive PD-1+ CD8+ T cells may interact with myeloid cells expressing PD-L1 to prevent their activation. This would have potentially important clinical implications. As melanocytes and self-reactive T cells coexist in normal pigmented skin, the expression of PD-1 might maintain T cells in a tolerant state, thereby avoiding pathogenic inflammation targeting melanocytes. In addition, the use of anti–PD-1 in patients with melanoma could unleash melanocyte-specific CD8+ T cell pathogenicity, leading to the development of vitiligo lesions (67–70). This would explain the high frequency of this adverse event in patients receiving anti–PD-1 for advanced melanoma compared to patients receiving other immune checkpoint therapies such as anti-CTLA4 (71, 72). Our data thus underline the interest of activating PD-1/PD-L1 signaling in vitiligo (73). On the other hand, other immune regulation pathways might explain the absence of lesions in normal-appearing skin. Recently, Xu et al. (58) showed how fibroblasts play a key role in controlling the skin immune response. They identified a distinct regional heterogeneity of fibroblast subsets in the skin of patients that accounted for the localization of vitiligo lesions. While PL skin is populated by IFN-γ–responsive fibroblasts that are able to recruit pathogenic CD8+ T cells via the local production of CXCL9 and CXCL10, NL skin is enriched with IFN-γ–resistant fibroblasts that are unable to promote local inflammation. By interfering with signaling pathways such as JAK/STAT (signal transducer and activator of transcription) inhibition, treatments targeting the effects of pro-inflammatory mediators, such as IFN-γ, have provided positive outcomes in managing the disease. Recent data support the emerging view that extended use of topical JAK inhibitors after achieving a clinical response may modify the disease course in vitiligo, contributing to sustained repigmentation and improved long-term outcomes (74). Our data demonstrate additional mechanisms of action of these targeted therapies given that they potentiate immunoregulatory pathways in patients who show good repigmentation. Yet, whether JAK inhibitors promote Treg cell expansion and/or recruitment in the skin, as observed in graft-versus-host disease, has to be evaluated (75). However, it is now well known that after cessation of treatment, patients experience disease relapse at exactly the same site as that previously affected, suggesting that therapies specifically targeting pathogenic TRM cells are needed (24, 76). Our study has several limitations that should be acknowledged. While we describe functional differences in CD8+ T cell subsets and regulatory mechanisms in NL and PL skins, we did not directly assess melanocyte-specific autoreactive T cells, which remains challenging as melanocyte-specific T cells represent only a small fraction of the total T cell population in the skin and yet unidentified autoantigens may drive vitiligo. In addition, skin biopsies from patients with vitiligo and healthy controls were not always obtained from the same anatomical sites, which may introduce bias, given that regional differences in skin immune composition have been described. Last, although we provide initial evidence of increased immunoregulatory events in NL skin that could inhibit the activation of skin-resident TRM cells, the synergistic role of ICR expression on CD8+ T cells and the increased frequency of Treg cells remains to be investigated through functional assays and comprehensive clinical studies. In conclusion, the presence of potential pathogenic TRM cells in the clinically normal skin of patients with vitiligo provides further support for the concept of “treating early” to prevent their activation. In psoriasis, blockade of CD8 T cells can prevent the development of lesions (77). Such interventions could modify the course of vitiligo, as is currently the case in other chronic inflammatory skin diseases like psoriasis and atopic dermatitis (78, 79). MATERIALS AND METHODS Experimental design The objectives of our study were (i) to explore and compare at a single cell level the diversity and transcriptional profile of CD8 T cells infiltrating NL and PL skins of patients with vitiligo; (ii) to assess their functionality; and (iii) to investigate whether immune regulatory mechanisms may prevent their activation in NL skin. Specimens were obtained from patients visiting the Department of Dermatology, Bordeaux University Hospital, France. We collected fresh and paraffin-embedded skin biopsies to perform (i) scRNA-seq, (ii) protein analyses using immunofluorescence and flow cytometry, and (iii) TCR sequencing. Details of sample size and P values are indicated in each figure legend. Study participants and approval All studies involving human tissues were approved by the Ethics Committee and the Commission Nationale de l’Informatique et des Libertés (INFLAMMAVIT no. 2018-A01555-50) or were obtained from the BARVIT clinical trial (NCT04822584). All subjects provided written informed consent for inclusion in this study. Human skin biopsies (4-mm diameter) were obtained from the arm of patients with vitiligo from the Department of Dermatology, Bordeaux University Hospital, France. Skin biopsies were harvested from PL and NL areas, with NL samples taken 10 to 15 cm away from the lesion border. PL skin refers to the area located 1 to 2 mm from the lesion border and is characterized by the presence of residual melanocytes and a localized inflammatory environment. All 59 patients included in this study developed vitiligo de novo; none experienced vitiligo as an immune-related adverse event induced by anti–PD-1 therapy in the context of metastatic melanoma. Patients’ clinical characteristics are shown in tables S3 and S4. Unaffected healthy control skin was obtained as discarded human tissue from cutaneous breast plastic surgery, except for TCR sequencing (fig. S2) where the biopsies were obtained from the arm, breast, or abdomen (Bordeaux University Hospital, France). Biopsies were processed immediately, paraffin embedded, or cryopreserved at −80°C. Tissue dissociation The cell suspension was obtained from whole fresh skin tissues using the Whole Skin Dissociation Kit human and the gentleMACS dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions (without adding enzyme P). The cell suspension was then filtered through a 70-μm filter screen. Single-cell suspension preparation for scRNA-seq Fresh skin samples were digested to generate single-cell suspensions. Dispase (2.4 U/ml; Life Technologies, Carlsbad, CA) digestion was first performed to separate the epidermis from the dermis. The single-cell suspension from the dermis was obtained as described above. The epidermis was digested by a 15-min incubation in trypsin/EDTA (0.05%, Gibco, Thermo Fisher Scientific, Carlsbad, CA). The cell suspension was then filtered through a 70-μm filter screen. Cell sorting for single-cell analysis Cells were first incubated with an FcR Blocking Reagent (Miltenyi Biotec) and stained with an anti-CD45 monoclonal antibody (mAb). Dead cells were stained with a Viobility 405/520 Fixable Dye (Miltenyi Biotec). Live CD45− and CD45+ cells were purified by cell sorting with a BD FACSMelody instrument (Becton Dickinson, Franklin Lakes, NJ) at the FACSility platform (University of Bordeaux, France). Single-cell capture, library preparation, and sequencing Single-cell capture and library preparation were performed at the OneCell platform (University of Bordeaux, France). Single-cell suspensions were loaded into the Chromium Controller (10x Genomics, Pleasanton, CA) to obtain the gel beads in emulsions. From gel beads in emulsions, reverse transcription, cDNA amplification, and library preparation were performed using Chromium Single Cell 3’ Reagent Kits (version 3.1 Chemistry) (10x Genomics) according to the manufacturer’s instructions. The library quality and quantity were assessed with an Agilent 2100 Bioanalyzer (Agilent, Santa-Clara, CA) and a Qubit Fluorometer (Invitrogen, Thermo Fisher Scientific). Libraries were pooled and sequenced on the NovaSeq6000 S2 (IntegraGen, Evry, France) and generated about 20,000 pair-end reads per cell (IntegraGen). scRNA-seq data analysis Sequencing reads were aligned and converted to count tables using Cell Ranger count (version 7.0.0; 10x Genomics, Pleasanton, CA). The resulting expression matrices were processed individually in R using the Seurat package (version 4.3.0) (80). First, cells were filtered according to classical quality control metrics (number of unique molecular identifiers and detected genes per cell and proportion of mitochondrial and ribosomal reads) using sample-specific cutoffs (fig. S9 and table S5). Then, each sample was preprocessed using the standard Seurat workflow (NormalizeData, ScaleData, FindVariableFeatures, RunPCA, RunUMAP, FindNeighbors, and FindClusters functions). Resulting clusters were assigned to cell populations according to canonical marker genes listed in table S6, and clusters corresponding to multiple cell populations or characterized only by a low number of detected genes per cell were removed. After the preprocessing steps, all the datasets were merged and reanalyzed using the same Seurat workflow. A batch correction step was additionally performed with Harmony (version 0.1.1) on the first 100 components of the PCA coordinates (81). UMAP coordinates and cell clusters were computed from Harmony data. T cell and CD8+ T cell datasets were extracted at the cluster level from global datasets (all cells and T cells, respectively) on the basis of CD3/CD8 expression and reanalyzed as previously described. Gene set enrichments were computed with the clusterProfiler package (version 4.6.0) on GO, KEGG (Kyoto Encyclopedia of Genes and Genomes), and MSigDB databases (82–85). Cell-cell communication analysis was performed using the CellChat package (version 1.6.1) (86). Ex vivo T cell activation in skin specimens Whole-skin biopsies were incubated with an anti-CD3 mAb (1 μg/ml; clone OKT3, BioLegend, San Diego, CA) or an immunoglobulin G2a (IgG2a) isotype control (1 μg/ml; Ultra LEAF, BioLegend) for 16 hours at 37°C, as described previously (44, 87). In some experiments, an anti–PD-1 mAb (Nivolumab, 10 μg/ml; clone 5C4.B8, Ozyme, Paris, France) or an IgG1 isotype control (10 μg/ml; clone MOPC21.1, Ozyme) were added in combination to the anti-CD3 mAb. NanoString RNA was extracted from frozen skin biopsies of healthy controls or patients with vitiligo. All samples were analyzed using the nCounter Human Immunology V2 panel (NanoString, Seattle, WA) at the “plateforme analytique de recherché en santé” (PARS, Bordeaux University Hospital), as previously described (88). All statistical analyses were performed on normalized log2 data generated using nSolver Analysis Software 4.0 (NanoString). PCA was implemented using the scikit-learn Python library (89). Analysis of differential gene expression between groups was conducted using the limma R package (90). To visualize DEGs, a heatmap was generated from normalized data using z-score transformation included in the Heatplus R package (91). GO enrichment analysis of DEGs was performed with the Gseapy Python package (92). The following GO classes were used: Molecular Function 2021, Biological Process 2021, and Cellular Component 2021. Cytokine multiplex assay Skin T cell supernatants were analyzed using the LEGENDplex Human Th cytokine panel (BioLegend) according to the manufacturer’s protocol, and data were acquired with a FACS Canto II instrument (BD Biosciences, San Jose, CA). Analyses were performed using LEGENDplex software 8.0. High-throughput sequencing of TCRβ gene DNA was extracted from whole-skin biopsies using the Maxwell RSC DNA kit (Promega, Madison, WI). For experiments performed in Fig. 2 and fig. S2 (A to C), TCRβ libraries were prepared using the immunoSEQ Kit (Adaptive Biotechnologies, Seattle, WA) on a minimum of 100 ng of DNA according to the manufacturer’s instructions. TCRβ CDR3 regions were amplified and sequenced on a MiSeq system (Illumina, San Diego, CA), and high-throughput sequencing data were analyzed using the ImmunoSEQ Analyzer (Adaptive Biotechnologies), as described previously (93). Data representation was obtained using the software provided with Adaptive. For experiments performed in fig. S2D, TCRβ libraries were prepared using the Ampliseq kit for Illumina TCR beta SR (Illumina). TCRβ CDR3 regions were amplified and sequenced on a MiSeq system (Illumina), and high-throughput sequencing data were analyzed using MixCR software. Data representation was obtained using Immumarch 1.0.0. The emergence of clones was evaluated by sequence abundance. The abundance of a clone was defined by the abundance of the most frequent allele for TCRβ. The putative emergent clone was defined by the relative abundance of its unique CDR3 sequence. The percentage of T cells consisting of the emergent clone was determined by dividing the abundance of the emergent clone (number of reads) by the total number of T cell reads. Immunofluorescence studies Immunofluorescence staining was performed on 4-μm sections prepared from formalin-fixed, paraffin-embedded skin biopsies, as previously described (88). A mouse monoclonal anti-CD8 primary antibody (C8/144B, M710301-2, DAKO, Agilent, 1/100) or rabbit monoclonal anti-CD8a primary antibody (ab217244, abcam, Cambridge, United Kingdom, 1/100), mouse monoclonal anti-FOXP3 primary antibody (235A/E7, Thermo Fisher Scientific, 1/100), and Alexa 568 goat anti-mouse IgG (A11031, Invitrogen, 1/500) or Alexa 568 goat anti-rabbit IgG (A11036, Invitrogen, 1/500) and Alexa 488 goat anti-mouse IgG (SAB4600042, Sigma-Alrich, Saint-Louis, MO, 1/200) were used. Images were acquired using a Nikon Eclipse microscope (Nikon, Tokyo, Japan). Multispectral imaging Multiplex immunofluorescence staining was performed on the automated Ventana Discovery XT staining platform (Ventana Medical Systems, Oro Valley, AZ). Briefly, 3-μm slides of skin tissues were deparaffinized and heat-induced epitope retrieval was performed using standard CC1 reagent (tris-based buffer, pH 8.0; Ventana Medical Systems). The slides were then incubated with primary antibodies listed in table S7. Bound primary antibodies were detected using OmniMap HRP (horseradish peroxidase)–conjugated anti-rabbit IgG (Ventana) or OmniMap HRP-conjugated anti-mouse IgG (Ventana Medical Systems), followed by tyramide signal amplification using opal fluorophores (Opal 480, Opal 520, Opal 570, Opal 620, Opal 650, Opal 690, and Opal 780; Akoya Biosciences, Marlborough, MA). The slides were counterstained with spectral 4′,6-diamidino-2-phenylindole (DAPI; Akoya Biosciences), coverslipped, and digitized using a multispectral imaging platform (PhenoImager HT, Akoya). The multispectrally acquired images were unmixed using spectral libraries previously built from images stained for each fluorophore (monoplex) and with inForm Advanced Image Analysis software (inForm version 2.60; Akoya Biosciences). Unmixed images were analyzed using inForm software (version 2.6.0). Tissues were first segmented in “Derm” and “Epiderm” areas on the basis of MelanA staining. Cells were then segmented on the basis of DAPI and fluorescent membrane signals. Mean fluorescent marker intensities for each cell were extracted, and signal intensities were normalized using the GaussNorm function from the flowStats R package (version 4.8.2). Cells were ultimately phenotyped using a thresholding method in FlowJo software (version 10.8.0; FlowJo). Flow cytometry To analyze cell surface protein expression, cell suspensions were first incubated with an FcR Blocking Reagent (Miltenyi Biotec). Then, cells were stained with the appropriate fluorochrome-labeled mAbs listed in table S8. Dead cells were stained with a Viobility 405/520 Fixable Dye (Miltenyi Biotec). Data were acquired on Fortessa (BD Biosciences) or Aurora (Cytek Biosciences) analyzers and analyzed with DIVA (BD Biosciences), SpectroFlo (Cytek), and FlowJo (Tree Star, Ashland, OR) software. For unsupervised analysis, cytometry data were cleaned using the FlowAI plug-in on FlowJo software and by removing doublets, debris, and dead cells. Gating on CD3+CD56−CD4−CD8+ T cells was performed. The data were subsampled to include 777 CD8+ T cells from each sample and merged. Subsequently, a t-distributed stochastic neighbor embedding analysis was performed, and the phenograph algorithm was run to cluster the data. When assessing cytokine secretion, cell suspensions were activated for 4 hours with phorbol 12-myristate 13-acetate (50 ng/ml), ionomycin (500 ng/ml; Sigma-Aldrich), and monensin (Golgi stop, BD Biosciences). Cells were surface stained, fixed, permeabilized, and stained with labeled antibodies following the manufacturer’s instructions (Cytofix/Cytoperm Plus Fixation/Permeabilization Solution Kit, BD Biosciences). Statistics Statistical differences were analyzed using the nonparametric Wilcoxon test (comparison of two paired experimental groups) or the nonparametric Kruskal-Wallis test (comparison of three experimental groups). All statistical analyses were performed using Prism 8.4.2 software (GraphPad, San Diego, CA). For scRNA-seq, DEGs between clusters or conditions were inferred with the Wilcoxon rank-sum test as implemented in the FindAllMarkers function. For NanoString, analysis of DEGs between groups was conducted using the Limma R package. A linear model was fitted to the data, and resulting P values were further corrected using the Benjamini-Hochberg multiple test procedure. Differential expression with an adjusted P value <0.05 was considered as statistically significant.