Dermatomyositis is characterized by JAK1-mediated monocyte-driven vasculopathy and inflammation

Abstract

Dermatomyositis is a rare yet devastating autoimmune disease characterized by inflammatory and vasculopathic changes in skin and muscle. Dermatomyositis and systemic lupus erythematosus (lupus) skin lesions have overlapping clinical and histopathological features, but show disparate responses to available therapeutics, with dermatomyositis skin disease often relapsing and being recalcitrant. To investigate dermatomyositis immunopathogenesis, non-lesional skin, lesional skin, and circulating immune cells from patients with dermatomyositis were analyzed using single-cell RNA-sequencing. Samples were analyzed in parallel with lesional and non-lesional lupus skin, healthy control skin, and peripheral blood from all three patient groups. We demonstrate a pervasive type I interferon (IFN) signature in dermatomyositis stroma that persisted in culture and was distinguished from lupus by up-regulation of vascular endothelial growth factor and interleukin-18 signaling in dermatomyositis keratinocytes. Furthermore, endothelial cells in lesional dermatomyositis exhibited decreased proliferation, which was not observed in lupus skin. Using cell communication networks, we identified a population of dermatomyositis-specific monocytes interacting with non-proliferating dermatomyositis endothelial cells. Co-culture of monocytes from patients with dermatomyositis with endothelial cells resulted in increased endothelial cell apoptosis, which was inhibited by Janus kinase 1 (JAK1) blockade. JAK1 inhibition also resulted in reversal of dermatomyositis stromal and inflammatory signatures. Together, our data provide a comprehensive cross-disease characterization of lesional and non-lesional skin in dermatomyositis and implicates monocyte-mediated endothelial cell dysfunction in dermatomyositis vasculopathy. Moreover, these results suggest that JAK inhibition may offer a suitable therapeutic intervention for refractory skin disease.

ONE SENTENCE SUMMARY

A comparison of dermatomyositis and lupus skin revealed that dermatomyositis is characterized by monocyte-mediated endothelial cell dysfunction.

INTRODUCTION

Dermatomyositis is a rare, chronic autoimmune disease that predominantly affects the skin and muscle but can also involve other organs such as the lung. Skin disease is often relapsing and recalcitrant to therapy, even when systemic disease is well controlled (1, 2). Lack of knowledge regarding the pathogenesis and drivers of dermatomyositis has prevented the development of effective therapies.

Similar to the pathology in the skin of systemic and cutaneous lupus erythematosus (lupus), type I interferons (IFNs) are up-regulated in lesional dermatomyositis (3–5) and are believed to play a pathogenic role in the destruction of muscle tissue (6, 7). Recent studies have suggested myeloid-derived dendritic cells may be an important source for IFN-β in dermatomyositis lesional skin (5); however, other cell populations have not been comprehensively studied. Moreover, our understanding of differences between dermatomyositis and lupus skin lesions remains under-explored. These differences are likely important to understand disease drivers and to develop better therapies, as there are still no FDA approved therapies for skin manifestations for either condition.

Inflammation-associated vascular damage also contributes to the pathogenesis of dermatomyositis. Loss of intramuscular microvessels accompanied by endothelial cell (EC) activation and poor regenerative capacity are hallmarks of dermatomyositis-related endothelial cell dysfunction (8, 9). Early disease stages of dermatomyositis have been associated with higher vascular endothelial growth factor (VEGF) production by fascial endothelial cells (10), which may reflect ongoing vascular injury and attempts at regeneration (11). Additionally, circulating markers of endothelial cell injury are increased in patients with dermatomyositis and correlate with disease activity (12, 13). Thus, although endothelial dysfunction has been demonstrated in the muscle compartment, the role of ECs in the skin has not been well-defined.

In this paper, we utilized single-cell RNA sequencing (scRNA-seq) to examine the cellular composition of paired lesional and non-lesional skin samples from patients with dermatomyositis with active skin disease and compared the findings to healthy control and patients with lupus. The aim of this study was to comprehensively define the cellular makeup of the skin and to characterize mediators of inflammatory changes that contribute to disease. We also examined paired peripheral blood from the same dermatomyositis patients to determine the origin of immune cell populations that contribute to EC activation and cell death. Overall, we found an IFN-rich stromal environment in dermatomyositis that was distinct from lupus and was characterized by up-regulated interleukin (IL)-18 and VEGF pathways. Furthermore, we demonstrated increased EC senescence JAK-1-dependent EC and apoptosis in dermatomyositis, which appeared to be driven by a dermatomyositis-specific population of inflammatory monocytes. Together, our results provide detailed insights into dermatomyositis pathogenesis and support a role for the use of JAK inhibition in refractory dermatomyositis skin disease.

RESULTS

Comparison of dermatomyositis and lupus epidermal and stromal cell populations identifies shared type I IFN signatures but VEGF and IL-18 signaling in dermatomyositis epidermis

To investigate the cellular and molecular differences between lupus and dermatomyositis, we performed scRNA-seq on paired lesional and non-lesional skin (upper thigh, sun-protected). A total of 5 patients with active discoid lupus erythematosus, 6 patients with active subacute cutaneous lupus erythematosus, and 8 dermatomyositis patients with active skin disease were assessed (see Fig. 1A for study design and table S1 for patient details). Samples were analyzed in parallel with skin from 8 healthy controls (HC). The final dataset comprised 123,150 cells, with an average of 2,226 genes and 8,960 transcripts per cell. Cells were clustered based on differentially expressed genes and visualized on a uniform manifold approximation and projection (UMAP) plot (Fig. 1B). Clusters were annotated based on canonical lineage markers reported in previous skin disease scRNA-seq studies (14–18) (fig. S1A). We defined 11 major cell types across lesional dermatomyositis (LDM), non-lesional dermatomyositis (NDM), lesional lupus (LLE), non-lesional lupus (NLE), and HC (H) skin biopsies (Fig. 1C and D). The cell types identified were not patient-specific and spread across disease and lesional skin state (fig. S1B).

Figure 1. scRNA-seq reveals differences in cell populations from non-lesional and lesional skin of patients with dermatomyositis or lupus.

(A) Schematic of human sample acquisition and experimental approaches throughout the study. Figure made with BioRender. (B) UMAP plot of 123,150 cells colored by cell type. (C) UMAP plot of cells colored by disease state. H, HC skin n=20,488 cells from 8 donors; NDM, non-lesional dermatomyositis n=22,370 cells from n=8 patients; NLE, non-lesional lupus n=36,762 cells from 11 patients; LDM, lesional dermatomyositis n=20,456 cells from 8 patients; LLE, lesional lupus n=23,074 cells from 11 patients. (D) Bar plot of disease state proportions across cell types.

Splitting cell clusters by disease and skin state (i.e., lesional compared to non-lesional) revealed enrichment for keratinocytes (KC), fibroblasts (FBs) and ECs in dermatomyositis and lupus lesions (Fig. 1C and D). Immune cell populations, including both T cell and myeloid lineages, were specifically enriched in LLE; there was also an enrichment in endothelial cells in LDM (Fig. 1D). KCs were the most abundant cell type (61,008). Sub-clustering revealed 7 KC subtypes: annotated as Basal, Spinous, Supraspinous, Granular, Follicular, Basal Inflammatory, and Spinous Inflammatory (Fig. 2A). Basal Inflammatory and Spinous Inflammatory KC subpopulations were dominant in lesional samples from patients with lupus and dermatomyositis (Fig. 2B and C). Consistent with previous work, Basal Inflammatory KC populations were also present in NLE (19) and, to a lesser extent, in NDM (Fig. 2C). Differential expression analysis was performed to identify differences in Basal Inflammatory KCs between lupus and dermatomyositis, non-lesional and lesional skin. In non-lesional skin, Basal Inflammatory KCs from patients with lupus showed elevated expression of type I IFN-induced genes (IFI27, IFI6), whereas Basal Inflammatory KCs in dermatomyositis skin displayed high expression of genes related to cell cycle, protein degradation, and ribosomal function (RPS26, UBE2S, EGR1) (Fig. 2D). Comparison of Basal Inflammatory KCs in lesional skin revealed a stress response signature (S100A9, S100A8, S100A7) in lupus and enrichment of heat shock proteins involved in mitochondrial signaling and stress responses in dermatomyositis (DNAJA1, DNAJB1) (Fig. 2E). Basal Inflammatory KCs in lupus skin lesions also showed elevated expression of KRT6B and KRT16, indicating incorporation of KCs from the hair follicles. The total DEG analysis can be found in tables S2 and S3.

Figure 2. scRNA-seq analysis highlights differences in cytokine signature of dermatomyositis and lupus stroma that persists in culture.

(A) UMAP plot of 61,008 keratinocytes (KCs) colored by subtype. (B) UMAP plot of KCs colored by disease state. (C) Bar plot of the distribution of KC subclusters by disease state. (D) Dot plot of the top differentially expressed genes (DEGs) up-regulated and down-regulated in basal inflammatory non-lesional disease states. Color scale indicates average marker gene expression. Dot size indicates percentage of cells expressing marker gene. (E) Dot plot of the top DEGs up-regulated and down-regulated in basal inflammatory lesional disease states. (F) Violin plots of KC IFN-α scores split by disease state for each KC subtype. (G) Bar graph of top up-regulated pathways in LLE and LDM KCs. (H and I) IFN gene expression by RT-qPCR following stimulation of primary CTL, lupus, or dermatomyositis (n=3 each) basal keratinocytes from skin samples with PolyI:C (10 ug/ml) (H) or UVB (50 mJ/cm²) (I). Data were analyzed by two sided unpaired t test for panels in H and I. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

We then wanted to compare cytokine exposures across diseases and lesional skin states to understand the relevance of inflammatory signals. The low sequencing depth of 10X RNA-seq precludes direct examination of cytokine expression due to high dropout. Thus, we calculated KC module scores derived from genes induced in cultured KCs upon stimulation with the indicated cytokines. Inflammatory cytokines relevant to skin and autoimmune diseases were selected based on our previous findings (19–22) (fig. S2), IFN-α module scores were elevated in Basal Inflammatory KCs. The score was highest in NLE, followed by LDM, LLE, NDM, and, lastly, HC (Fig. 2F). Follicular LLE KCs also exhibited a high type I IFN-α signature, raising the question of whether IFN-α signaling plays a role in the pathogenesis of periadnexal inflammation and follicular plugging in discoid lupus subtypes (Fig. 2F). LDM exhibited a high IL-1β score in Basal, Supraspinous, Follicular, and Basal Inflammatory subclusters c (fig. S2).

To obtain a better overall understanding between LLE and LDM, we performed Ingenuity Pathway Analysis (IPA) to compare the top 250 uniquely expressed genes in the Basal Inflammatory KC sub-cluster of LDM vs LLE. VEGFA-VEGFR2 signaling (p= 1.35×10⁻¹⁹) and IL-18 signaling (p=8.08×10⁻¹⁷) emerged as the most significantly enriched pathways in LDM, whereas cytoplasmic ribosomal proteins, translation factors, mediators of mitochondrial oxidative phosphorylation and glycolytic metabolism were enriched in LLE Basal Keratinocytes. Enriched pathways in Basal Inflammatory KCs in LLE and LDM were associated with cytokine signaling, RNA processing, and type I IFN signaling responses (Fig. 2G).

Given the limitations of 10X scRNA-seq to detect type I IFNs, we examined IFN expression in primary Basal Keratinocytes of NLE and NDM through RT-qPCR and compared it to HC. Consistent with our previous findings (23, 24), NLE KCs exhibited a chronic baseline up-regulation of IFNK (p=0.0005) compared with HC KCs. In contrast, there was no notable increase in baseline IFNK or IFNB in NDM KCs compared to HC KCs (Fig. 2H). Stimulation with Poly(I:C) (10 µg/ml), which activates both TLR3 and RNA cytosolic sensing pathways, induced IFNB and IFNK in HC, NLE, and NDM KCs. Furthermore, a preference for IFNK was seen in NLE, whereas IFNB production was more prominent in NDM, mimicking results from tissue mass cytometry performed on dermatomyositis immune populations (3). Since dermatomyositis and lupus are both photosensitive disease, we assessed the response to UV light exposure in NLE and NDM. Following UVB stimulation, NLE KCs increased IFNK and IFNB expression. NDM KCs also exhibited increased production of IFNs compared with HC (Fig. 2I). Together, these results implicate type I IFN-driven epidermal changes in dermatomyositis pathology.

We next analyzed FB populations. Published FB markers (25) were used for cell type annotation which revealed populations characterized by SFRP2, APOE, TNN, or CLDN1 expression (fig. S3A to D). The SFRP2⁺ cell population included two subpopulations, defined by PI16 and COMP expression. SFRP2⁺PI16⁺ fibroblasts showed high expression of CCN5, a repressor of transforming growth factor-β signaling (26). There were no notable differences of FB populations between disease states; however, a shift was noted in UMAP placement between cells across non-lesional and lesional LE and dermatomyositis disease states, indicating small differences in gene expression (fig. S3B). To determine the drivers of lesional and non-lesional disease states, module scores for FB cytokines were calculated (fig. S3E). Both SRFP2⁺ FBs and APOE⁺ FB subpopulations from LDM showed high IFN-α, IFN-γ, and TNF-α module scores (fig. S3E). A list of DEGs between non-lesional and lesional SRFP2⁺ FB populations for samples from patients dermatomyositis and lupus are shown in tables S4 to S7. These results indicate that education of inflammatory fibroblast occurs in both dermatomyositis and lupus and that lesional FBs may be more affected in dermatomyositis versus lupus skin.

Endothelial cells are activated and demonstrate increased cellular senescence in LDM

Patients with lupus and dermatomyositis both exhibit endothelial dysfunction (9, 27). and patients with dermatomyositis also show nailfold capillary changes (9, 28). This is an indication that endothelial cells (ECs) may play a unique role in dermatomyositis pathogenesis. We thus compared differences in gene expression between endothelial cells of HCs, patients with lupus, and patients with dermatomyositis. Sub-clustering analysis of 5,611 dermal blood ECs identified 11 sub-clusters (Fig. 3A to D). We defined three EC subgroups based on anatomical gene expression signatures: arteriole (GJA4, SEMA3G), venule (SELP) and capillary (RGCC) (Fig. 3B to D). EC0 shared gene expression with the capillary and venule anatomical gene signature and thus is of interest, as this is the site of immune cell migration and interaction. LDM ECs were mainly EC0, EC8, and EC10 (Fig. 3E). These EC subtypes showed evidence of activation (high SELE and ICAM1 expression), up-regulation of IFN-induced genes (ISG15) and inflammatory regulators (TNFAIP3, encoding A20) (Fig. 3D).

Figure 3. Endothelial cells are activated and demonstrate increased cellular senescence in lesional dermatomyositis compared to other disease states.

(A) UMAP of ECs colored by disease state. (B) UMAP of ECs colored by sub-cluster. (C) Schematic of EC classification and representative gene markers. (D) Dot plot of the representative marker genes for each sub-cluster. Color scale indicates average marker gene expression; dot size indicates percentage of cells expressing marker gene. (E) Bar plot of the EC subtypes within each disease state. (F) Bar plot of up-regulated genes of lesional dermatomyositis vs lesional LE in endothelial subcluster 0. (G) Violin plot of senescence score of combined endothelial cell by disease state. Data analyzed by Welch’s two-sided t test. ****=p<0.0001. (H) Violin plot of senescence score of endothelial cells by subcluster. (I) Representative immunofluorescence of frozen HC and lesional dermatomyositis skin biopsies (n=3). Scale bar, 20 µm. (J) Quantification (mean ± SD) of immunofluorescence for p16 and Ki67 using mean fluorescence intensity (MFI) across VWF⁺ cells; dots represent average of 3 different areas per vessel. Mean and SEM are shown. Data were analyzed by Student’s t test, **P < 0.01 and ****P < 0.0001.

Given that both dermatomyositis and lupus exhibited expansion of capillary-associated cell populations, we performed IPA of the top 200 up-regulated genes in LDM EC0 and compared them to the up-regulated genes of LLE EC0. In LDM, IL-10 signaling, antigen presentation, and macrophage activation were the most robustly up-regulated pathways compared with LLE EC0 (fig. S4A). Comparing EC0 cells in LDM to LLE revealed major histocompatibility complex II (MHC)-associated genes HLA-DRB1 (p=7.8×10⁻¹⁷), HLA-DQB1 (p=3.34×10⁻⁸), and HLA-DRB5 (p=1.22×10⁻⁸) as most differentially expressed. IFN response genes ISG15 and IFI6 were also up-regulated in EC0 cells of LDM compared with LLE, as was HES1 and TAGLN, genes involved in NOTCH signaling and fibrosis, respectively (Fig. 3F). IPA was performed to identify upstream regulators up-regulated in EC0 cells of LDM compared with LLE. HSF1, a driver of endothelial prostaglandin production (29) and anti-coagulation phenotypes (30), and SATB1, involved in chromatin remodeling during active cell death (31), were identified as upstream regulators up-regulated in LDM EC0. In contrast, TNF signaling was more dominant in LLE EC0 (fig. S4B).

Senescent phenotypes have been identified to play a strong role in vasculopathy in other autoimmune diseases, such as scleroderma (32). Up-regulation of superoxide dismutase (SOD) in EC0 was indicative of elevated oxidative stress (Fig. 3D). We thus assessed enrichment for senescence-associated genes across EC subclusters. A score was generated from a list of 68 genes associated with cellular senescence which were selected based on published work (33–36) or cellular senescence annotation in the Reactome database. The full gene list can be found in Shi et al. (33). The non-lesional and lesional dermatomyositis ECs exhibited an increased senescence score when compared with non-lesional and lesional lupus ECs, respectively (Fig. 3G). The senescent signature was highest in EC0 and EC9 (Fig. 3H). To examine the underlying drivers of these changes, we performed IPA to identify upstream regulators for the genes induced in lesional ECs taken as a whole. Consistent with known inducers of endothelial dysfunction and senescent phenotypes, LDM samples had higher Z-scores for enrichment of signaling from type I and II IFNs, IL-1β, IL27, TNF, and NONO, a regulator of transcription and RNA splicing, compared with LLE (fig. S4C). These results suggest that, in LDM, inflammatory signaling contributes to the different cutaneous EC phenotypes observed, including a propensity towards senescence that could lead to vasculopathy.

To further examine the senescent phenotype at the protein level, we performed immunofluorescence on skin sections derived from three HCs and three patients with LDM, respectively (Fig. 3I). The senescent and proliferation marker p16 and Ki67 were selected and combined with the Von Willebrand Factor (VWF) to stain ECs. Consistent with our gene expression analysis, we identified increased and decreased EC co-staining with p16 and Ki67, respectively, in LDM ECs compared with HC (Fig. 3J).

T peripheral helper cells are expanded in LDM

We next examined lymphoid cells in the skin. Sub-clustering followed by annotation based on established marker genes identified 7 T cell and innate lymphoid cell (ILC) subsets: CD4⁺ T cells, T regulatory (T_reg) cells, T peripheral helper (T_ph) cells, CD8⁺ T cells, tissue resident memory T (T_RM) cells, natural killer (NK) cells, and Innate lymphoid cells (ILCs) (Fig. 4A to C). CD4⁺ T cells were the predominant T cell subset (42 to 60%) across all disease states, followed by CD8⁺ T cells, making up 13 to 14% of all T cells (Fig. 4D and E). One subset was defined by the expression of interferon stimulated genes (ISGs) and hence labelled IFN T cells (Fig. 4C). More IFN T cells were present in LDM compared with LLE. IFN T cells were also recovered from NLE, but not NDM (Fig. 4D and E).

Figure 4. T peripheral helper cells are expanded in lesional dermatomyositis compared to lupus and healthy control.

(A) UMAP of T cell subsets, natural killer (NK) cells, and innate lymphoid cells (ILCs) from skin biopsies by subtype. (B) UMAP of T cells from skin biopsies by disease state. (C) Dot plot of representative marker genes for each T cell subtype. (D) Bar graph showing distribution of cellular subsets by disease state. (E) Percentage of T cells by subtypes within each disease state. (F) UMAP of activation, cytotoxicity, exhaustion and ISG scores. (G to J) Violin plots of ISG (G), exhaustion (H), cytotoxicity (I) and activation (J) scores in T cell, NK cell, and ILC subsets by disease state.

T_ph cells were expanded in lesional dermatomyositis skin compared with other disease states. Proportions of these cells of total T cells varied by disease state, with 1% for HC, 1% for NDM, 11% for LDM, 2% for NLE, and 4% for LLE (Fig. 4E). Furthermore, this subcluster was distinguished by expression of CXCL13, encoding a B cell-attracting chemokine that has been implicated in the loss of immune tolerance of B cells in idiopathic inflammatory myopathies (reviewed in Zanframundo et al. (37)). To investigate the characteristics of T_ph cells, we assessed activation, cytotoxicity, exhaustion, and interferon-stimulation module scores. T_ph cells were activated but not enriched for cytotoxic or exhaustion scores. CD8⁺ T cells were activated as well but demonstrated increased exhaustion and cytotoxicity scores. Cytotoxicity scores were highest in NK cells, which also showed modest exhaustion. ISG scores were enhanced in IFN T cells, validating this phenotype (Fig. 4F). ISG scores were elevated across all immune cell populations in LDM and LLE, with NLE exhibiting scores higher than NDM in NK cells, T_reg cells, and CD4⁺ T cells (Fig. 4G). More pronounced exhaustion was observed in LDM CD8⁺ T cells, T_RM cells, and NK cells (Fig. 4H). Cytotoxicity and activation scores did not vary much between diseases (Fig. 4I and J). In summary, LDM showed an increase in T_ph cells, which were similarly activated as compared to LLE T_ph cells. Cytotoxic populations seem to be equivalent in dermatomyositis versus lupus lesions.

Langerhans cells are expanded and inflammatory in non-lesional dermatomyositis

We then evaluated myeloid cells, the other major immune cell type detected in our skin samples.

Sub-clustering and annotation of myeloid cells revealed 11 myeloid sub-clusters and a B cell subset with largely distinct marker genes (Fig. 5A to C). B cells subcluster with the myeloid cells likely secondary to MHC class II expression. Subsets included Langerhans cells (LC; characterized by expression of CD207 and CD1A), classical type 1 dendritic cells (cDC1; characterized by expression of DNASE1L3 and C1orf54), classical type 2 dendritic cells subset A (cDC2A; characterized by expression of CCL17, CCL19, and BIRC3) and subset B (cDC2B; characterized by expression of CLEC10A and IL1B), CD16⁺ DCs, interferon-educated dendritic cells (IFN-DC){Duong, 2022 #7585}, lipid-associated macrophages (LAM; characterized by expression of APOE and APOC1), perivascular macrophages (PVM; characterized by expression of SELENOP and STAB1), monocytes (Mono; characterized by expression of EREG and THBS1), plasmacytoid dendritic cell-like cells (pDC-like; characterized by expression of PPP1R14A), and plasmacytoid dendritic cells (pDCs; characterized by expression of GZMB and JCHAIN). B cells were primarily derived from LLE, consistent with the literature (38, 39). LCs were primarily present in dermatomyositis (Fig. 5D), with 52% and 10% of all LC deriving from non-lesional and lesional dermatomyositis, respectively. Consistent with a known reduction of LC in lupus (40), only 2% of LC derived from LLE, with similar abundance in NLE and HC (16% vs14%) (Fig. 5E). Immunofluorescence staining for CD207 in frozen NDM skin confirmed robust LC staining in dermatomyositis non-lesional skin (Fig. 5F). In NDM LCs, there were 139 significantly up-regulated genes [Log₂FC > 0.25, false discovery rate (FDR) adjusted p-value < 0.05] compared with HC LC. Pathway analysis revealed up-regulation of IL-1 signaling, TCR signaling, and antigen processing and presentation (Fig. 5G, table S8). A total of 156 significantly up-regulated genes (Log₂FC > 0.25, FDR adjusted p-value < 0.05) were found in LDM LC vs HC LCs (table S9), and IPA revealed the top signaling pathways to be associated with type I IFN (Fig. 5H). These results indicate that the LC influx in dermatomyositis skin may be more consistent with monocyte-derived inflammatory LCs in patients with dermatomyositis (41).

Figure 5. Dermatomyositis skin harbors increased numbers of inflammatory Langerhans cells whereas cutaneous lupus erythematosus skin exhibits Langerhans cell loss.

(A) UMAP of skin derived myeloid cells and B cells by subtype. (B) UMAP of skin derived myeloid cells by disease state. (C) Dot plot of the representative marker genes for each myeloid subtype. (D) Bar graph of the percentage of myeloid cells by disease state within each subtype. (E) Percentage of B cell and myeloid cell subtypes within each disease state. (F) Representative immunofluorescence of LC marker CD207 in frozen dermatomyositis lesional skin biopsies (n=3); scale bar, 200μm. (G) Bar graph indicating the top significant predicted up-regulated pathways in NDM vs HC within LCs. (H) Bar graph indicating the top significant predicted up-regulated pathways in LDM vs HC within LCs.

The peripheral blood of patients with dermatomyositis contains an inflammatory classical monocyte population.

We next examined peripheral blood mononuclear cells (PBMCs) by scRNA-seq from five patients with dermatomyositis, seven patients with lupus, and four HC (Fig. 6A and B, fig. S5A). Monocytes were then selected for sub-clustering. Intermediate and classical monocytes were in subclusters 1–4 and nonclassical monocytes in cluster 5 (based on high expression of FCGR3A) (Fig. 6C to E). Subcluster 0 (Mono0) was dominantly observed in dermatomyositis PBMCs (DMP), with little presence in HC and lupus peripheral blood (HP and LEP) (Fig. 6C to F, fig. S5B). Numerous dermatomyositis samples contributed to this monocyte population, including both treatment naïve and treated (table S1, fig. S5C and D). Overall, we identified Mono0 as a subset of classical monocytes as it expressed CD14 (fig. S5E). Multiple inflammatory cytokines and chemokines were highly expressed in Mono0, including CXCL1 and CXCL5, as well as expression of FGF2, an important activator of EC and neovascularization (42)(Fig. 6E). IPA revealed IL-1β, TNF-α, IL-17A, IL-6, and IL-18 were enriched cytokine regulators in Mono0 (fig. S5F), suggesting circulating monocytes in patients with dermatomyositis are predominantly educated by inflammatory conditions. To further compare this, we examined the top canonical pathways enriched in dermatomyositis Mono0 compared with other monocyte subclusters and identified IL-10 signaling, IL-4 signaling, IL-13 signaling, IL-8 signaling, the Pathogen Induced Cytokine Storm Signaling Pathway, and neutrophil degranulation to be enriched in Mono0 (Fig. 6G, table S10). Relationships between circulating and skin myeloid cells were then further explored by integrating both datasets (Fig. 6H). Sub-clustering of the aggregated myeloid cells revealed that Mono0 cells were also present in the skin of patients with dermatomyositis (Fig. 6H, red arrowhead), suggesting that these cells are not only in circulation.

Figure 6. Dermatomyositis peripheral blood monocytes exhibit a unique pro-inflammatory classical monocyte population that migrates into skin of lesional dermatomyositis.

(A) UMAP of PBMC cells by cell type. (B) UMAP of PBMC cells by disease state. (C) UMAP of PBMC monocytes by subclusters. (D) UMAP of PBMC monocyte subclusters by disease state. (E) Dot plot of representative marker genes by monocyte subclusters. (F) Bar graph of the percentage of monocyte subclusters by disease state. (G) Bar graph of the top significant up-regulated pathways in Mono_0 vs .Mono_1–5. (H) UMAP of combined skin myeloid cells and PBMC monocytes by cell subtype, split by disease state, red arrow indicates abundance of Mono0 in skin.

Ligand-receptor analysis demonstrates cell-cell interactions among peripheral monocytes and dermal ECs are enriched in dermatomyositis

Since Mono0 cells are found in dermatomyositis skin, we wanted to examine whether they could interact with any other population in the skin. We thus used CellChat to examine potential cellular crosstalk. When acting as senders (expressing ligands), Mono0 showed prominent interactions with ECs (Fig. 7A). When acting as receivers (expressing receptors), Mono0 showed strong interactions with fibroblasts (Fig. 7B). The Mono0/EC interactions were of interest given the inflammatory and senescent changes in dermatomyositis EC (EC0,6,8,9,10) (Fig. 3). We used Cellphone DB to specify critical ligand-receptor interactions between Mono0 and ECs that were up-regulated in LDM versus HC skin. Crosstalk was robust for Mono0-dependent production of VEGF-A and -B, TNF-α, IL-6, IL-1β, TNFSF14 (LIGHT) and TNFSF12 (TWEAK) (fig. S6). Given that capillary changes, including dropout, are prominent feature of dermatomyositis nailfolds (28) and muscle (43), Mono0 production of inflammatory mediators such as IL-6 and TNF family members, known to promote EC apoptosis and senescence (44) is intriguing. We thus tested whether monocytes from patients with dermatomyositis exhibited an enhanced ability to kill ECs. We isolated dermatomyositis and HC CD14⁺ monocytes through negative selection from PBMCs (n=3, each) and co-cultured them with ECs. Apoptosis was assessed through a caspase3/7 dye. dermatomyositis monocytes induced EC death at a higher rate than HC monocytes (age and sex-matched), suggesting that dermatomyositis monocytes may contribute to vascular dysfunction and inflammation in dermatomyositis skin and possibly other organs (Fig. 7C).

Figure 7. Peripheral monocytes from patients with dermatomyositis increase apoptosis of dermal blood vascular endothelial cells, which is reversed by JAK1 inhibition.

(A) Receptor ligand analysis showing circulating Mono_0 ligand to EC receptors. (B) Receptor ligand analysis showing circulating Mono_0 receptor and EC ligands. (C) Plot of EC apoptosis in monocyte killing assay. Statistical comparison made by 2-way ANOVA with Sidak’s multiple comparison testing. 16 replicates per patient were completed****=p<0.0001. (D) Representative plot from Patient 2 of EC apoptosis in monocyte killing assay with inhibition of JAKSTAT signaling with 0.1 µM Upadacitinib (Upa). (E) Quantification (mean ± SD) of apoptosis cell counts at 42 hours of incubation n=2 patients with DM and 2 HC, 8 technical replicates per sample. Comparisons made by Welch’s unpaired 2-tailed t test. (F) Schematic of lesional skin explant study: Explants were treated with JAK inhibition and processed for scRNA-seq. (G) Senescence scores in untreated (CTL) and JAK inhibitor-treated (JAKi) capillaries and venules within AKAP12+ ECs. *p<0.05. Analysis via one-way Welch’s t test. (H) Dot plot of the top 25 predicted upstream significant cytokine regulators enriched among DEGs in untreated EC versus JAKi-treated EC. Color scale, −log₁₀(P-value) from the enrichment analysis. Dot size indicates number of DEGs corresponding to each upstream regulator; a negative z-score indicates enrichment in CTL cells; a positive score indicates enrichment in JAK inhibitor-treated cells.

Various medications have been proposed for treatment of dermatomyositis, including JAK inhibitors, which are useful for refractory skin disease (45) and are known to block many cytokines including IL-6 (46). We thus tested the JAK1 inhibitor upadacitinib in our culture system and found that it attenuated the increased apoptosis of ECs (Fig. 7D and E). A TWEAK inhibitor had similar benefits on monocyte-mediated killing (fig. S7A). We then took a broader look at the effects of these drugs on the skin as a whole. We studied explanted lesional skin from two untreated patients with active dermatomyositis and treated the biopsies overnight with upadacitinib or a TWEAK inhibitor, followed by scRNA-seq (Fig. 7F). Cell populations were similar to our initial UMAP (fig. S7B–E). Upon Upadacitinib treatment, AKAP12⁺ ECs exhibited a reduction in capillary senescence scores (Fig. 7G, fig. S7C, F, table S11). This was accompanied by reduced inflammatory signaling, as defined by negative Z scores for predicted upstream regulators, including IFN-γ, type I IFN, TNF, and IL-6 signaling in upadacitinib treated skin samples(Fig. 7H). In contrast, no reduction in monocyte inflammatory pathways was seen with upadacitinib treatment (table S12). The reduction in capillary senescence and inflammatory signaling was not observed upon treatment with the TWEAK inhibitor (fig. S7F). Thus, endothelial dysfunction is partly reversed by JAK1 inhibition, potentially through inhibition of monocyte interactions such as IL-6 rather than direct inhibition of monocytes themselves (fig. S7G).

DISCUSSION

In this study, we have provided a comprehensive understanding of the cellular similarities and differences between dermatomyositis and lupus non-lesional and lesional skin using scRNA-seq. Dermatomyositis skin samples harbored dysfunctional states of keratinocytes, an increase in inflammatory LCs, and a unique, proinflammatory circulating monocyte population that may contribute to a propensity for EC death and senescence in the tissue. We further identified that type I IFN signaling is a relevant pathway in both diseases, but KC IFN signaling may be less prominent in dermatomyositis than in cutaneous lupus erythematosus skin. Dermatomyositis keratinocytes exhibited a propensity for IFN-β responses over IFN-κ, which has also been observed in other cell types in previous studies (3, 47). Notably, IFN-β correlates with cutaneous disease activity in dermatomyositis(48). The reasons behind the preference for IFN-β versus other type I IFNs in dermatomyositis are unclear, but could include epigenetic changes to the IFNB promoter, genetic polymorphisms, or other yet undefined mechanisms. Importantly, trials using antibodies targeting IFN-β in dermatomyositis have completed early phase studies, with promising results (49).

Besides the preference for type I IFN-signaling, dermatomyositis keratinocytes also showed enhanced IL-18 signaling, which is in line with previous studies (5). IL-18 is a pleiotropic cytokine and its expression in epithelial cells is promoted by multiple upstream signals, including type I and II IFN(50). Secretion of IL-18 is driven by inflammasome activation, which has not been studied in dermatomyositis skin disease, but IL-18, IL1β and NLRP3 expression in muscle tissue have been shown to be elevated in dermatomyositis(51). Canonical downstream effects of IL-18 are MyD88-dependent, leading to activation of MAPK and NF-κB signaling(50). Unlike IL-1β, IL-18 can be easily detected in peripheral blood, which may be of benefit when dissecting the role of IL-18 signaling in dermatomyositis.

Other differences in populations between dermatomyositis skin and cutaneous lupus erythematosus were observed. Both lupus and dermatomyositis skin had a pronounced inflammatory fibroblast compartment with SRFP2⁺PI16⁺, APOE⁺, and SRFP2⁺COMP⁺ fibroblasts from LDM exhibiting robust up-regulation of genes predicted to respond to inflammatory cytokines. Our analysis revealed communication between fibroblasts and monocytes as they enter the tissue, which suggests an important role of fibroblasts in inflammatory education of monocytes in the skin. Whether these fibroblasts are contributing to the development of calcinosis or tissue atrophy requires further investigation.

Vasculopathy represents a life-threatening manifestation in dermatomyositis and clinically apparent nail fold capillary changes occur in up to 89% percent of patients with dermatomyositis (52). Our evaluation of ECs identified pronounced inflammatory changes in dermatomyositis capillary-associated EC clusters driven by type I IFN and HSF1 signaling, whereas lupus ECs were dominated by TNF signaling. This suggests that dermatomyositis capillaries have different inflammatory signaling networks compared with lupus, and these differences may be reflected in clinical features and treatment responses. Endothelial dysfunction has been observed in circulating endothelial progenitors and in dermatomyositis-associated myositis (9, 53). Our results highlight pronounced senescence and apoptosis of dermatomyositis skin ECs, which does not occur in lupus. Multiple cytokines and other external stressors, such as UV light, promote senescence, including IL-6, TNF and type I IFNs(54–56). In dermatomyositis, we identified type I IFNs and IL-6 as potential drivers. In addition, the Mono0 population was enriched for multiple inflammatory pathways including IL-18. This is in agreement with the literature where IL-18 was shown to contribute to endothelial cell progenitor dysfunction in lupus and dermatomyositis (9). Whether monocyte-derived IL-18 is a contributor to vascular dysfunction in dermatomyositis skin needs to be further explored. It is also unclear whether the senescence phenotype will respond to treatment in vivo. However, as nailfold capillary changes correlate with skin disease activity (33), one could speculate that improvement in disease may also encourage endothelial dysfunction. More research is needed in this regard.

The senescent phenotype in dermatomyositis ECs may have similarities with ECs in patients with scleroderma, where activation and senescence of ECs has been identified (33, 57). Expression of p16 is elevated in scleroderma skin, but it is unclear whether this increase in senescence is associated with ECs (33). However, given the differences in fibrotic and vasculopathic outcomes in dermatomyositis compared to scleroderma, a dedicated study to compare the EC expression, endothelial to mesenchymal transition (58), and cellular crosstalk between these diseases should be considered.

The most prominent difference in cell populations between dermatomyositis, lupus, and HC samples was observed in circulating PBMCs, where hyperinflammatory CD14⁺ monocytes (“cluster 0” monocytes) were highly enriched for in dermatomyositis. These cells exhibited strong expression of inflammatory cytokines and chemokines, were found to infiltrate dermatomyositis lesional skin, and had strong interactions with ECs. This stands in contrast to what has been previously reported in lupus skin, where CD16⁺ DCs played a more dominant role (19). Tissue mass cytometry has found that CD14⁺ monocytes in dermatomyositis skin correlate with cutaneous disease severity (3). Another study in juvenile dermatomyositis found that mitochondrial dysfunction in CD14⁺ peripheral blood monocytes contributes to type I IFN signaling (59). Of note, Type I IFN signaling was also elevated in our adult dermatomyositis CD14⁺ Mono0 cells. Our data also suggests that this population of highly inflammatory CD14⁺ monocytes contributes to endothelial dysfunction in dermatomyositis, possibly through apoptosis or inflammatory cytokine production, which can be targeted through JAK inhibition. This is in line with a previous report showing that JAK1/2 inhibition using ruxolitinib restored type I IFN-mediated vascular network disruption in dermatomyositis (60). Clinical trials showed efficacy of JAK inhibition in dermatomyositis (45), but assessment of the underlying cellular and mechanistic changes have not been explored. Our results show that JAK1 and TWEAK blockade can prevent dermatomyositis monocyte-mediated killing but that only JAK1 inhibition reverses EC senescence ex vivo. Myeloid cell gene expression was not changed in this ex vivo assay. This could suggest that monocytes mediate cytokine production, which induces senescence but renders cells susceptible to JAK1 inhibition. These results corroborate the importance of therapeutically targeting monocyte-derived signals in dermatomyositis. Whether the ex vivo findings can be replicated in vivo has to be tested. Longitudinal follow-up with monitoring of circulating and cutaneous myeloid populations in patients with dermatomyositis should be performed to understand fully how these cells vary in disease and in treatment response.

Our study has limitations. First, scRNA-seq does not capture every cell population. This is true especially for neutrophils. Others have shown that neutrophils may be important instigators of systemic inflammation through triggering in the skin and that neutrophils may be important for certain types of cutaneous lupus (61–63). The role of neutrophils in cutaneous lupus erythematosus and dermatomyositis could not be assessed by our study, as neutrophils were not captured in our data. In addition, given the relatively small size of our study populations, we were not able to analyze data based on concomitant medications at the time of sampling. Further, we intentionally took all non-lesional biopsies from sun-protected, upper thigh/buttock skin, since we were not able to control for location or sun exposure of lesional skin. In addition, our data may be limited by the enrollment of only White patients in this study. Hence, our data may not fully represent disease manifestations in non-white patients. Future studies could consider addressing whether there are underlying differences in diverse patient populations with dermatomyositis

Together, our results provide single cell resolution of dermatomyositis skin, an understanding of type I IFN signaling as it compares to lupus, and functional evidence that targeting monocyte-EC interaction with JAK inhibition may represent a new avenue to treat vasculopathy and potentially improve other features of disease. This is important as new treatment modalities are studied in DM as future clinical trials should evaluate not only which patients benefit from JAK inhibition but the specific disease manifestations and cellular phenotypes that benefit. Further, this data raises the question of whether early intervention with JAKi might prevent endothelial dysfunction in dermatomyositis.

MATERIALS AND METHODS

Study Design

The goal of this study was to compare gene expression at single cell resolution between different dermatomyositis and lupus disease states, and healthy controls. Patients were recruited sequentially in clinic when agreeable to the study. Inclusion criteria included active dermatomyositis lesions without use of >10 mg of prednisone daily. Skin biopsies and PBMCs from HC, NDM, LDM, NLE and LLE were subjected to scRNA-seq.

Human sample acquisition

Skin biopsies for single cell RNA-seq were collected from 8 patients with dermatomyositis with active skin disease (table S1), all of whom contributed lesional and non-lesional 6mm punch skin biopsies (sun-protected skin of the buttock). No exclusions based on concurrent medications were made (see table S1 for listing). A diagnosis of dermatomyositis was confirmed for all patients based on the 2017 European League Against Rheumatism/American College of Rheumatology classification criteria (64). Eleven patients with active cutaneous lupus erythematosus with or without systemic lupus (10 with associated systemic lupus and 1 cutaneous lupus erythematosus only) were included for comparison. A total of 8 HC were also recruited for skin biopsy. For EC death assays, additional dermatomyositis and HC (age and sex-matched) participants were recruited for blood draws only. An additional two patients with active dermatomyositis were recruited for explant studies with upadacitinib. The study was approved by the University of Michigan Institutional Review Board (IRB), and all patients gave written, informed consent. The study was conducted according to the Declaration of Helsinki Principles.

Single-cell RNA library preparation, sequencing, and alignment

Generation of single-cell suspensions for scRNA-seq were performed as follows: Skin biopsies were incubated overnight in 0.4% dispase (Life Technologies) in Hank’s Balanced Saline Solution (Gibco) at 4°C. Epidermis and dermis layers were separated. Epidermis samples were digested in 0.25% Trypsin-EDTA (Gibco) with 10U/mL DNase I (Thermo Scientific) for 1 hour at 37°C, quenched with fetal bovine serum (FBS; Atlanta Biologicals), and strained through a 40μM mesh. Dermis samples were minced, digested in 0.2% Collagenase II (Life Technologies) and 0.2% Collagenase V (Sigma) in DMEM for 1.5 hours at 37°C, and strained through a 40μM mesh. Epidermal and dermal cells were combined at a 1:1 ratio, and libraries were constructed by the University of Michigan Advanced Genomics Core on the 10x Chromium system with chemistry v3. Libraries were then sequenced on the Illumina NovaSeq 6000 sequencer to generate 150 bp paired end reads. Data processing including quality control, read alignment to the reference genome (hg38), and gene quantification was conducted using the 10x Cell Ranger software (v3.1.0) with default parameters. The samples were then merged into a single expression matrix using the CellRanger aggr pipeline.

Seurat object generation, dimensionality reduction, clustering and cell type annotation

The Seurat (v4.0.3) pipeline was selected for the analysis of the single cell transcriptomic data (65). Quality control was performed by removing cells with less than 500 and more than 50,000 transcripts, less than 100 and more than 8,000 features, and with a mitochondrial gene percentage above 10% of total genes. The NormalizeData, FindVariableFeatures, and ScaleData functions were applied using default parameters. Principal component (PC) analysis was performed on the variable genes as denoted for each comparison, and the top 30 PCs were used in the RunHarmony function from the Harmony (v0.1.0) package (66) to remove potential batch effects among samples processed through different libraries. Specifically, confounding donor-specific variations were corrected by setting group.by.vars=“donor” in the RunHarmony function so that the universal biological signals will be discovered across all samples. UMAP dimensional reduction was performed using the RunUMAP function. The clusters were obtained using the FindNeighbors and FindClusters functions with default settings and the resolution set to 0.6. DEGs associated with each cluster were found using the FindAllMarkers function. Out of the 31 clusters, three low-quality clusters were identified that likely represented apoptotic cells: one cluster with extremely low UMI counts, extremely low detected genes, and yet a very high mitochondrial percentage, and the other two clusters with extremely high expressions of hemoglobin genes without otherwise definable lineages (HBB and HBA2). The cell types of the remaining 28 clusters were annotated based on enrichment for canonical cell type signature genes. Cells were then grouped into 11 known cell types, including KC, FB, EC, Lymphatic EC, Myeloid, T cell, Melanocyte, Eccrine gland, Pericyte, Mast cell, and Schwann cell (fig. S1A). We performed sub-clustering within each cell type to remove doublets (high UMI counts) and low-quality cells (low UMI counts and detected genes, and high proportions of mitochondrial genes). After removing all these doublet/low-quality cells, we performed a second-round analysis, re-running the previously described workflow using the same parameter values. The UMAP plot by cell types is presented in Fig. 1B.

For PBMCs, cells with more than 30,000 transcripts, less than 100 genes or more than 15% mitochondrial genes of the total gene expression were filtered out as low-quality cells. After quality control, the afore-mentioned scRNA-seq analysis workflow was applied and the same parameter values were used. We grouped PBMCs into Monocytes, B cells, Plasma cells, T cells, NK cells, Dendritic cells, Granulocytes, Erythrocytes, Proliferating cells, doublets, and donor-specific cells. For Monocytes, B cells, T cells, and NK cells, we performed sub-clustering analyses. After removing low-quality cells, doublets, and cells that clustered only with a specific donor and thus did not represent disease state (donor-specific cells), monocytes, B cells, T cells, and NK cells were merged, and we performed a second-round analysis. The UMAP plot highlighted by cell type is presented in Fig. 6A.

PBMC-derived monocytes and skin myeloid cells (excluding B cells, Fig. 5A) were merged for further analyses. The Seurat package and pipeline (v4.3.0.1), as described above (65), was used to re-cluster myeloid cells in the combined dataset. The six subsets from the PBMC-derived monocytes (Fig. 6C) were annotated as Mono_0 – Mono_5 within the combined myeloid UMAP (Fig. 6H), and the monocytes within the skin myeloid cells (Fig. 5A) were labelled Mono_S within the combined myeloid UMAP (Fig. 6H). Key steps in the scRNA-seq analysis workflow included using 43 PCs in the FindNeighbors function and setting the resolution to 1.2 in the FindClusters function. The donor-specific batch effects were removed with the Harmony (v0.1.1) package. The number of cells for each cell type across each disease state was counted, divided by the total number of cells within each disease state, and then scaled to 100 percent for each cell type.

Module score analysis

Module scores for FBs and KCs (IFN-α score, IFN-γ score, IL-1-β score, TNF-α score), ECs (senescence score) and T cells (activation score, cytotoxicity score, exhaustion score, and IFN-stimulated gene score) were calculated using the AddModuleScore function on previously published gene signatures: For FBs and KCs, the gene signatures are available within the supplementary data files S3 and S2, respectively, in Billi et al. (19). For ECs, the full gene list for generation of the senescence score can be found in the supplemental table s2 of Shi et al. (33). For T cells, the activation, cytotoxicity, exhaustion, and IFN-stimulated gene signatures are available within the table S3 in Dunlap et al. (67).

Cell type sub-clustering

Sub-clustering was performed on the most abundant cell types. The same pipeline as described above was used to define sub-clusters. Sub-clusters that were characterized exclusively by mitochondrial gene expression, indicating low quality, were removed from further analysis. The subtypes were annotated based on expression of canonical subtype signature genes.

Cell-Cell communication analyses

CellChat (v2.1.2)(68) was used to identify interactions between PBMC monocytes (Mono_0) with the different cell types present within the skin. CellPhoneDB (v2.0.0)(69) was used for Ligand-Receptor analysis. Cells were split by disease state (H, LDM, NDM, LLE, NLE). Cellular interactions with P-value > 0.05 were filtered out from further analysis. To compare among disease states, each cellular interaction was assigned to the condition in which it showed the highest interaction score. The number of interactions for each cell type were then calculated. In fig. S6, the top interactions from Mono_0 (Ligands) to skin AKAP12⁺ ECs (EC0, EC6, EC8, EC9, EC10) were plotted in Circos plots using the R package circlize version 0.4.15.

Cell Culture

Primary human keratinocytes were established from non-lesional lupus and dermatomyositis punch biopsies and healthy control punch biopsies as previously described (15). Cells were maintained in culture with Epilife medium and were passaged at 50% confluency to avoid differentiation prior to study. For functional assessment, cells were treated with 10 ug/ml poly (I:C) or were irradiated with 50mJ/cm² UVB (310nm) with a UV-2 irradiator (Tyler Industries). Cells were incubated in fresh medium for 6 hours or 24 hours, followed by harvest and RNA isolation for real-time polymerase chain reaction run at the University of Michigan Advanced Genomics Core on the QuantStudio Real-Time PCR system (PCR) for 30 cycles as previously reported(70) (IFNB1 FW 5’3’: ACGCCGCATTGACCATCTAT RV 5’3’: GTCTCATTCCAGCCAGTGCT, IFNK FW 5’3’: GTGGCTTGAGATCCTTATGGGT, RV 5’3’: CAGATTTTGCCAGGTGACTCTT).

Immunofluorescence

Immunofluorescence was performed on frozen sections obtained from HC skin biopsies, non-sun exposed and non-lesional lupus skin biopsies, or non-lesional dermatomyositis skin biopsies. Thawed sections were fixed at −20 °C in acetone, washed, and blocked with 5% bovine serum albumin (for CD207) or 5% goat serum (for vwf, p16, and ki67 antibodies) for 1 hour at room temperature. Slides were stained with primary antibody overnight at 4°C using the following concentrations: Anti-vwf antibodies (Novus NBP2–33003, clone VWF635, 1μg/ml), anti-p16 antibodies (Novus NBP2–98881, 1μg/ml), or anti-ki67 antibodies (Abcam 15580, 1μg/ml); anti-CD207 (BioLegend clone 929F3.01, 1:100 dilution The slides were then washed and the Alexa Fluor secondary antibodies (Thermo Fisher Scientific, A-11001 Alexa Fluor 488 or A-11012 Alexa Fluor 594, 1:200) were added for 2 hours at room temperature. The slides were again washed with PBS and mounted using antifade mounting medium with DAPI (Thermo Fisher Scientific). Images were acquired using Zeiss Axioskop 2 microscope and analyzed by SPOT software. Images are representative of n=3 control samples, n=3 lupus samples, and n=3 dermatomyositis samples.

Monocyte isolation and killing assay

Human dermal microvascular endothelial cells (Lonza) were seeded in 96-well plates (Corning) at a density of 2 × 10⁴ cells per well and grown overnight. On the day of the assay, monocytes (40,000 cells/well) were isolated from HC and dermatomyositis from PBMCs using StemCell negative selection kits (StemCell technologies, #19359) following Ficoll-based separation of PBMCs. Monocytes were added to each well of endothelial cells together with caspase-3/7 reagent (1:2000 dilution, Essen Bioscience) in the presence or absence of 100 nM upadacitinib (Selleckchem) or 100 nM Fn14 (TWEAK receptor) antagonist (Millipore Sigma). Endothelial cells were imaged at 10-fold magnification in an IncuCyte S3 Live Cell Analysis System (Sartorius) at 37°C with 5% CO₂. Images were acquired every 3 hours with 4 images per well. Results were analyzed using the IncuCyte analysis software to detect and quantify the number of apoptotic (green) cells per image. Results were plotted using the GraphPad Prism 10 software. Results are presented as mean ± SEM at 57 hours of incubation.

Statistical analysis

Individual-level data are presented in data file S1. Statistical analysis was performed in GraphPad for functional assays using one or two-way ANOVA with Sidak’s multiple comparison testing or Welch’s or Student’s t-test where indicated, with p-values less than 0.05 were considered significant. F test to compare variance was used to ensure normal data distribution. For scRNA-seq, genes were considered differentially expressed with an FDR-adjusted p-value less than 0.05. Ingenuity pathway analysis was applied to the differentially expressed genes to determine the canonical pathways and the potential upstream regulators, and those with a Z-score of ≥ 2 or ≤ −2 were considered significant.

Supplementary Material

Fig. S1 to S7

Table S1 to S12

MDAR Reproducibility Checklist

Data file S1

Data, Code, and Materials Availability

All data associated with this study are in the paper or supplementary materials. The single cell RNA-seq data from skin biopsies is publicly accessible on GEO: GSE303609 contains 8 HC samples, 8 DM samples, and 5 of 11 CLE samples; the remaining 6 of 11 lupus samples are available from GSE186476, where the lesional and non-lesional lupus skin samples from donors 1, 2, 4, 5, 6, and 7 have been used within this study. No MTAs are required to access the publicly available databases. No custom code was created to run the listed analyses.