Immune cell landscape in LP skin consists predominantly of CD8+ T cells. We collected 4-millimeter skin biopsies from lesional and nonlesional skin from 7 patients with LP (Figure 1A and Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.179899DS1). Three male and 4 female patients were included, with a median age of 62 years (mean, 57 years). All patients had clinically active disease; 5 patients were not receiving treatment, and 2 patients were on oral treatment (hydroxychloroquine and prednisone, respectively). We generated 188,607 high-quality single-cell RNA-sequencing (scRNA-Seq) profiles (Supplemental Table 2). Unsupervised cell clustering of scRNA-Seq profiles revealed 29 unique cell populations that were annotated to 10 cell types using marker gene identification and mapping to single-cell databases (Figure 1B and Supplemental Figure 1, A–E). The cell types were shared between lesional and nonlesional skin samples, and lymphoid cell populations were consistently enriched by number in all lesional LP skin samples (Supplemental Figure 1, B and C).

Immune cell landscape in lichen planus. (A) Clinical photograph of lesional and nonlesional skin biopsies from a patient with lichen planus (LP). (B) Identification of cell clusters from lesional and nonlesional LP skin (n = 13). (C) UMAP depicting subclustering of lymphoid cells. (D) Analysis of individual subclusters. Marker genes are shown. Density plots demonstrate location of subgroup within UMAP. (E) Bar plots showing relative contribution as a percentage of total cells. *P < 0.05, 1-way ANOVA. (F) Dot plot demonstrating levels and percent of cells expressing IFNG, IL17A, and IL4. Data are shown as the mean ± SEM.

Next, we subclustered the lymphoid cell population and identified 9 cell types, 8 T cell subtypes and 1 NK cell population (Figure 1, C and D, and Supplemental Figure 1, F and G). Consistent with prior studies, CD8+ T cells were the predominant type. Three populations of CD8+ T cells were identified: CD8+ T1 (CD8A, GZMA, GZMK, IFNG), CD8+ T2 (CD8A, IFNG, HSPA1A, DNAJB1), and a proliferating CD8+ population “CD8+ Pro” (MKI67, CD8A). As a percentage of total lymphoid infiltrate, the CD8+ T2 population was significantly increased in LP skin (P = 0.04, Figure 1E). This population of T cells was notable in that it expressed the highest levels of IFN-γ (IFNG) and granzyme A (Figure 1F and Supplemental Figure 1I). As a percentage of lymphoid cell infiltrate, we did not see major differences in most subpopulations, suggesting increased recruitment of all T cells. Notably, we found an increased percentage of Tregs (FOXP3, CTLA4) in LP skin (P = 0.02) (Supplemental Figure 1G), similar to other inflammatory skin diseases (12). Three additional populations of T cells were identified that consisted of both CD4+ and CD8+ T cells, including naive (CCR7, TCF7), an exhausted T cell phenotype “Texh” (CTLA4, PDCD1, HAVCR2, IFNG), and a miscellaneous T cell category (TRAC, IKZF3, PLCG2) (Figure 1D). Exhausted T cells were characterized by their expression of PD1 (PDCD1), low levels of cytokine production (i.e., IFN-γ), and high levels of CXCL13 expression (13, 14). CD4+ cytotoxic T lymphocytes (CTLs) (TCF4, IL3RA, IFNG, GZMB) represented a small population of cells in lesional LP skin. Finally, a few prior studies have suggested a role for Th17 cells in LP (15–17). However, we observed no significant IL17A production in our T cell populations (Figure 1F). Taken together, our data demonstrated that CD8+ T2 cells were the major source of IFN-γ (Figure 1F).

LP skin secretes CXCL9, CXCL10, and CCL19 cytokines. To assess how LP skin may recruit lymphoid cell populations, we subclustered epidermal cell populations and identified eleven unique clusters, which were annotated based on expression of canonical marker genes (Figure 2A and Supplemental Figure 2, A and B). We identified 4 populations of basal cells along with suprabasal cells, melanocytes, and cells of the hair follicle and eccrine glands (18, 19). LP is histologically characterized by apoptosis in the basal layer of keratinocytes. Within basal keratinocytes, CXCL9 and CXCL10 were induced as much as 8-fold in lesional skin compared with nonlesional skin, making them two of the highest induced genes (Figure 2B and Supplemental Figure 3A). Melanocytes also markedly expressed CXCL9, CXCL10, and CCL19 (all 2-fold) compared with nonlesional skin (Supplemental Figure 3A). Transcription factor analysis of the keratinocyte populations showed an upregulation of IRF7, ETV7, and STAT2 (Supplemental Figure 3B).

Lichen planus skin secretes CXCL9, CXCL10, and CCL19. (A) UMAP depicting subclustering of epidermal skin cells. (B) Volcano plots of differential gene expression from basal keratinocyte subpopulations from lichen planus (LP) lesional versus nonlesional skin. Expression of CXCL9, CXCL10, and CCL19 is labeled on the plots. A P value of less than 0.01 (Wilcox’s test) and a 2-fold expression change was used for significance. (C) UMAP depicting subclustering of fibroblasts. (D) Volcano plots of differential gene expression from fibroblast subpopulations from LP lesional versus nonlesional skin. Expression of CXCL9, CXCL10, and CCL19 is labeled on the plots. A P value of less than 0.01 (Wilcox’s test) and a 2-fold expression change was used for significance. (E) Representative immunofluorescence images depicting localization of CXCL9 (green), CCL19, (white), and DAPI (blue) in LP skin (n = 5 patient samples). The white dotted line depicts the epidermal-dermal junction. Scale bars: 100 μm. (F) scRNA-Seq data from publicly available single-cell datasets for vitiligo, psoriasis, and atopic dermatitis were used to analyze skin cells for their expression of chemokines. Dot size corresponds to percentages of cells expressing chemokine, while color corresponds to level of gene expression.

Fibroblasts subclustered into 8 unique populations, and 4 subpopulations exhibited increased expression of CXCL9 (up to 10-fold) and CXCL10 (up to 5-fold) in LP lesional skin compared with nonlesional skin (Figure 2, C and D). Interestingly, all fibroblast populations also substantially induced expression of CCL19 (average 3-fold induction) (Figure 2, D and E). CCL19 is a chemokine typically expressed in thymus and lymph nodes to regulate immune cell trafficking, but it does not have an established role in skin inflammatory diseases (20). One population of basal keratinocytes, Basal 1, also induced CCL19 significantly (Figure 2B). Transcription factor analysis of these fibroblast clusters showed upregulation of IRF7, STAT1, STAT2, and RUNX3 (Supplemental Figure 3D).

To confirm our scRNA-Seq findings, we performed immunohistochemistry on LP skin biopsies. CXCL9 and CXCL10 were primarily expressed by keratinocytes in the lower layers of the epidermis and fibroblasts in the superficial dermis (Figure 2E). CCL19 was also strongly expressed in the lower levels of epidermis and in the superficial dermis. We confirmed that CCL19+ dermal staining came from fibroblasts (CD3–, vimentin+, CCL19+) and to a lesser extent tissue-infiltrating T cells (Supplemental Figure 4A). We additionally assessed skin biopsies of patients with lichen planopilaris (LPP) and psoriasis. LPP is a scalp-restricted clinical variant of LP. Consistent with prior reports, we found increased expression of CXCL10 and CCL19 that localized to the hair follicle epithelium (Supplemental Figure 4B) (21). We did not observe staining for these chemokines in psoriasis skin samples (Supplemental Figure 4C). Taken together, CXCL9, CXCL10, and CCL19 were the major cytokines induced in keratinocytes and fibroblasts of LP skin.

Next, we compared the chemokine environment in LP to that in other T cell–mediated skin diseases. We analyzed publicly available scRNA-Seq skin datasets for psoriasis, atopic dermatitis, and vitiligo (22–24). Within fibroblasts, keratinocytes, and melanocytes, we found that LP lesional skin had the strongest expression of CXCL9, CXCL10, and CCL19 as well as the highest frequency of fibroblasts expressing these chemokines (Figure 2F). Thus, LP skin strongly secretes a unique combination of CXCL9, CXCL10, and CCL19 cytokines.

CellChat computationally identifies potential ligand-receptor interactions within a population of cells. Global analysis of lesional and nonlesional skin from patients with LP revealed upregulation of signaling pathways in the cytotoxic and Th1 lymphocyte response (TNF, IL2, OX40, LT, IFN-II) and chemokine signaling (CCL, CXCL) (Figure 3A and Supplemental Figure 5, A and B). In lesional skin, basal keratinocytes and fibroblasts secreted CXCL9 and CXCL10, which were received by CXCR3 expressed on CD8+ Pro T cells, CD8+ T1 cells, Tregs, and Texh populations in lesional LP skin only (Figure 3B). CCL19 expressed by fibroblasts and basal keratinocytes was received by CCR7 expressed on CD8+ Pro T cells, naive T cells, CD8+ T1 cells, CD8+ T2 cells, Tregs, and Texh cells (Figure 3C). Taken together, CellChat analysis suggests that CXCL9, CXCL10, and CCL19 signals converge on T cells in LP skin.

CCL19 synergizes with skin-secreted CXCL9 or CXCL10 to recruit T cells. (A) Global analysis of ligand-receptor pathways. Arrows highlight relevant signaling. (B) Analysis of cell-to-cell interactions between epidermal keratinocytes cells and immune cells in lichen planus (LP) lesional skin. Dot color illustrates communication probability, and dot size illustrates P value. (C) Analysis of cell-to-cell interactions between dermal fibroblasts and immune cells in lesional (salmon color) and nonlesional (blue color) LP skin. Dot color illustrates communication probability, and dot size illustrates P value. (D) Schematic of migration assay. (E–H) Left: migration index (no. of migrated cells in response to cytokine/no. of migrated cells in response to control) for CD4+ and CD8+ T cells in response to different conditions of CXCL9, CXCL10, and CCL19 (n = 12). Right: Combined treatment with or without CCR7 blocking antibodies (n = 9). Data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, 1-way ANOVA was used for migration assays, and 2-tailed paired Student’s t test was used for CCR7 antibody analysis.

CCL19 works synergistically with CXCL9 and CXCL10 to recruit T cells. To test whether these cytokines recruit T cells, we performed in vitro migration assays with PBMCs from healthy donors (Figure 3D). We quantified migration of CD4+ and CD8+ T cells by flow cytometry (Supplemental Figure 6A). We tested migration in response to CXCL9, CXCL10, or CCL19 as well as combinations of both CXCL9 and CCL19 and CXCL10 and CCL19 (Figure 3, D–H). The migration index is the ratio of cells that migrated in response to a chemokine stimulus divided by the number of cells that migrated in response to control media.

Compared with vehicle control, CD4+ T cells exhibited more migration toward CXCL9 (mean migration index, 9.5 ± 3.6) and CCL19 (mean migration index, 8.1 ± 1.8) alone (Figure 3E, left). Combination treatment with CXCL9 and CCL19 induced CD4+ T cells to migrate significantly more strongly (mean migration index, 36.7 ± 7.3). In fact, these cells migrated more than double the sum of the individual chemokine migration indices, suggesting a synergistic response. Blocking antibodies against CCR7, the receptor for CCL19, significantly ameliorated the combined treatment’s effect (mean migration index, 12.3 ± 4.5) (Figure 3E, right).

CD8+ T cells showed an even more pronounced response (Figure 3F). Combination treatment with CXCL9 and CCL19 induced CD8+ T cells to migrate significantly more strongly (mean migration index, 87.0 ± 21.9) than the sum of individual chemokine migration indices for CXCL9 (mean migration index, 13.4 ± 4.9) and CCL19 (mean migration index, 15.0 ± 3.1). Blocking antibodies against CCR7 also reduced the combined treatment’s effect (mean migration index, 29.0 ± 10.8) (Figure 3F, right). Taken together, combination treatment of CXCL9 and CCL19 was synergistic for T cell migration and induced almost 3-fold more CD8+ T cells migration compared with CD4+ T cells.

We repeated migration assays with CXCL10 with and without CCL19. CD4+ T cells demonstrated increased migration toward CXCL10 (mean migration index, 3.8 ± 1.0) and CCL19 (mean migration index, 17.5 ± 4.0) (Figure 3G). Combination treatment also induced CD4+ T cells to migrate more (mean migration index, 40.0 ± 9.0) than the sum of the migration indices of each individual chemokine. CCR7 blocking antibodies ameliorated this effect (mean migration index, 14.8 ± 4.6) (Figure 3G, right). CD8+ T cells again showed a more pronounced response (Figure 3H). Combination treatment with CXCL10 and CCL19 induced CD8+ T cells to migrate more strongly (mean migration index, 68.0 ± 20.3) than the sum of the migration indices for each individual chemokine for CXCL10 (mean migration index, 4.7 ± 1.1) and CCL19 (mean migration index, 26.8 ± 8.1). CCR7 blocking antibodies also reduced this effect (mean migration index, 20.8 ± 5.9), but it did not reach statistical significance (P = 0.06; Figure 3H, right panel). In summary, CCL19 worked with CXCL9 or CXCL10 to synergistically amplify the migration of CD4+ and CD8+ T cells. This effect was greater on CD8+ T cells than CD4+ T cells.

T cell–secreted CXCL13 synergizes with CCL19 to recruit CD8+ T lymphocytes. We examined our scRNA-Seq dataset to identify other cytokines that may recruit CD8+ T cells into LP lesional skin. Prior studies demonstrated that CXCL13 is specifically expressed by exhausted T cells (13, 14, 25). Indeed, CXCL13 was among the highest induced genes in the Texh (>5-fold induction) and CD8+ Pro (>10-fold induction) populations in LP lesional skin compared with nonlesional skin (Figure 4A). Immunohistochemistry of LP skin confirmed that CXCL13 was primarily expressed by infiltrating CD8+ T cells (CD3+CXCL13+CD4–) (Figure 4B).

CCL19 synergizes with T cell–secreted CXCL13 to recruit T cells. (A) Volcano plots of differential gene expression from exhausted (Texh) and CD8+ proliferating (CD8+ Pro) T cell populations from lesional versus nonlesional lichen planus (LP) skin. Expression of CXCL13, CTLA4, GZMB, and GNLY is labeled. (B) Representative immunofluorescence images of LP skin depicting localization of CD4 (green), CXCL13 (white), CD3 (red), and DAPI (blue) (n = 5 patient samples). Scale bars: 100 μm. (C) Dot plot demonstrating levels and percentages of cells expressing CXCR3 and CXCR5. (D) Analysis of cell-to-cell interactions between immune cells in lesional (salmon color) and nonlesional (blue color) LP skin. (E and F) Left: Migration index (no. of migrated cells in response to cytokine/no. of migrated cells in response to control) for CD4+ (E) and CD8+ T cells (F) in response to different conditions of CXCL13 and CCL19 (n = 9). Right: Combined treatment with or without CCR7 blocking antibodies (n = 7). Data are shown as the mean ± SEM. *P < 0.05, 1-way ANOVA was used for migration assays, and 2-tailed paired Student’s t test was used for CCR7 antibody analysis.

CXCL13 canonically signals through the CXCR5 receptor. However, we did not detect expression of CXCR5 in our dataset. CXCL13 has also been shown to bind CXCR3, and this interaction has been functionally demonstrated to induce T cell migration (26, 27). In our dataset, CXCR3 was expressed in CD8+ Pro, Treg, CD8+ T1, and CD4+ CTL populations in lesional LP skin (Figure 4C). CellChat analysis highlighted that CD8+ Pro T cells may signal to other CD8+ Pro T cells in lesional LP skin via CXCL13/CXCR3 interactions as well as CD8+ T1 and Treg populations (Figure 4D). Consistently, Texh cells may also signal to CD8+ Pro, CD8+ T1, and Treg populations (Figure 4D).

We performed migration assays to test the sufficiency of CXCL13 and CCL19 in T cell recruitment. CXCL13 induced migration of both CD4+ and CD8+ T cells (migration index, 3.3 ± 1.3 and 4.0 ± 2.0, respectively) (Figure 4, E and F). As expected, CCL19 also induced migration of both CD4+ and CD8+ T cells (migration index, 16.2 ± 4.6 and 29.0 ± 8.9, respectively). Combined treatment significantly increased migration synergistically for both CD4+ and CD8+ T cells (migration index, 36.9 ± 12.3, P = 0.045; 47.5 ± 13.0, P = 0.01 respectively) (Figure 4, E and F). Finally, CCR7 blocking antibodies reduced the combined treatment’s migration effect on CD4+ and CD8+ T cells (migration index, 34.27 ± 14.8 and 23.5 ± 9.0, respectively), but only CD8+ was statistically significant.

Taken together, Texh and CD8+ Pro T cells in lesional LP skin secrete CXCL13 that synergizes with fibroblast-secreted CCL19 to recruit more CD8+ T cells.

Circulating levels of naive CD8+ T cells are decreased in patients with LP. Tissue-infiltrating lymphocyte populations are generally recruited from peripheral blood. We assessed if there were any changes in circulating immune cell populations between patients with LP and individuals acting as healthy controls. We collected paired PBMC samples from the 7 patients with LP for scRNA-Seq (Supplemental Tables 1 and 2). We used publicly available data for 3 healthy adult controls to create a combined 116,108-cell dataset. After unsupervised clustering, we annotated cell populations using a reference atlas for human PBMCs (Figure 5A and Supplemental Figure 6B) (28, 29). We found two lymphoid populations that changed significantly in frequency between patients with LP and individuals acting as healthy controls. Patients with LP exhibited fewer circulating naive CD8+ T cells compared with individuals acting as healthy controls (3.3% ± 1.3% versus 10% ± 1.2% in individuals acting as healthy controls, P = 0.02) (Figure 5B). This decrease in circulating naive CD8+ T cells may correspond with the influx of CD8+ T cells into LP lesional skin. We also observed that CD4+ CTLs were enriched 10-fold in circulating blood of patients with LP (2.9% ± 0.7% compared with 0.3% ± 0.01% in individuals acting as healthy controls, P = 0.04) (Figure 5B). These cells expressed higher levels of granzyme B compared with those from individuals acting as healthy controls (Supplemental Figure 6C). Notably, none of the circulating cell populations exhibited increased expression of CXCR3 or CCR7 receptors, suggesting that increased recruitment to skin is mediated by changes in cytokine ligand expression in the tissue and not due to changes in receptor expression levels in LP immune cells (Supplemental Figure 6D).

Peripheral naive CD8+ T cells migrate to skin in lichen planus. (A) Identification of cell clusters from lichen planus (LP) blood (n = 7). (B) Bar plot showing the relative contribution as a percentage of total cells for CD8+ naive and CD4+ CTL T cell populations between patients with LP and individuals acting as healthy controls (n = 3). Data are shown as the mean ± SEM. *P < 0.05, 2-tailed unpaired Student’s t test. (C) Pseudotime trajectory of naive CD8+ T cells isolated from PBMCs and CD8+ T1 and CD8+ T2 populations from LP lesional skin. Each dot represents a cell. Top: Trajectory through time (expressed in blue with a pseudotime scale). Bottom: Cells colored according to cell-type origin (naive, salmon; CD8+ T1, green; CD8+ T2, blue). (D) Gene expression changes as the cells progress through the pseudotime trajectory (from naive state in peripheral blood to effector state in tissue).

Finally, we performed pseudotime analysis on our LP skin and blood datasets. Pseudotime rationally defines developmental trajectories of analyzed cells based on single-cell transcriptomes that are assumed to be individual variations of developmental states (30). Unbiased analysis of the naive CD8+ T cell subset from blood and CD8+ T cell subsets from lesional and nonlesional LP skin demonstrated naive peripheral CD8+ T cells at the initial state and CD8+ T1 and CD8+ T2 skin populations at the terminal state (Figure 5C). This trajectory correlated with the upregulation of the genes found on activated CD8+ T cells (IFNG, GZMA, GZMK, CCL4, and CCL5) (Figure 5D). Additionally, the pseudotime trajectory correlated with the downregulation of CCR7, which is known to be downregulated after binding to CCL19 (31). Thus, pseudotime analysis suggests that T cell activation in LP is not a systemic finding and only occurs after local recruitment to the skin.