Spatial transcriptomics identifies immune-stromal niches associated with cancer in adult dermatomyositis

Ksenia S Anufrieva; Neda Shahriari; Ce Gao; Stephanie Pei Tung Yiu; Shideh Kazerounian; Rochelle L Castillo; Jessica Liu; Sean A Prell; Kartik Bhamidipati; Khashayar Afshari; Lindsay Parmelee; Anastasia N Kazakova; Erin Theisen; Teresa Bowman; Avery LaChance; William J Crisler; Kimberly Hashemi; Ilya Korsunsky; Sizun Jiang; Mehdi Rashighi; Ruth Ann Vleugels; Kevin Wei

doi:10.1038/s41467-026-74871-7

. 2026 Jun 26;17:8000. doi: 10.1038/s41467-026-74871-7

Spatial transcriptomics identifies immune-stromal niches associated with cancer in adult dermatomyositis

Ksenia S Anufrieva ^1,^#, Neda Shahriari ^2,^#, Ce Gao ¹, Stephanie Pei Tung Yiu ^3,⁴, Shideh Kazerounian ¹, Rochelle L Castillo ^1,², Jessica Liu ¹, Sean A Prell ¹, Kartik Bhamidipati ¹, Khashayar Afshari ⁵, Lindsay Parmelee ³, Anastasia N Kazakova ⁶, Erin Theisen ^1,², Teresa Bowman ⁷, Avery LaChance ², William J Crisler ², Kimberly Hashemi ², Ilya Korsunsky ^4,^8,⁹, Sizun Jiang ^3,^4,^7,^10,¹¹, Mehdi Rashighi ⁵, Ruth Ann Vleugels ², Kevin Wei ^1,^✉

PMCID: PMC13448213 PMID: 42362560

Abstract

Adult-onset dermatomyositis (DM) is an autoimmune inflammatory myopathy with distinct cutaneous manifestations and a strong association with malignancy. Through comparative analysis with cutaneous lupus erythematosus (CLE), our integrated spatial and single-cell transcriptomics analysis reveals unique immune and stromal niches associated with DM subtypes. We find that cancer-associated DM skin lesions are distinguished by the presence of dispersed immune infiltrates enriched with macrophages or organized lymphoid aggregates with dense B cell cores surrounded by CD4 + /CD8 + T cells, accompanied by preserved vascular architecture. In contrast, non-cancer-associated DM skin is characterized by dense myeloid cell infiltrates, harbouring elevated expression of IL1B and CXCL10 localizing near injured vascular endothelia. Cytokines produced by these myeloid infiltrates, together with local tissue hypoxia, trigger dramatic stromal remodelling, leading to loss of vascular-associated fibroblasts. In addition to the CXCL10+ myeloid signature, non-cancer-associated DM skin is characterized by specific cellular pairs: PD-L1-expressing mature dendritic cells enriched in immunoregulatory molecules (mregDC) and activated regulatory T cells (Treg) expressing NFKB2 and TNF receptors. While both DM and CLE show strong interferon signatures, DM uniquely displays IFNβ expression. Together, our study provides a comprehensive spatial mapping of immune and stromal cells in adult-onset DM.

Subject terms: Autoimmunity, Skin diseases, High-throughput screening, Inflammatory diseases

Adult-onset dermatomyositis (DM) is an autoimmune inflammatory condition manifesting muscle and skin lesions, often accompanying cancer. Here authors show by integrated spatial and single-cell transcriptomics analysis that the immune infiltrates that are dominating the skin lesions, together with the surrounding stromal niches, show characteristic differences between the cancer-associated and cancer independent subtypes of the disease regarding their cellular composition and spatial organisation.

Introduction

Adult-onset dermatomyositis (DM) is a rare autoimmune disease characterized by chronic inflammation affecting the skin, muscles, and lungs¹. Characteristic cutaneous manifestations of the disease include heliotrope rash, Gottron’s papules and sign, holster sign, shawl sign, V-neck sign, and periungual changes, including dilated capillary loops and ragged cuticles². A critical association of adult-onset DM is malignancy³, with previous studies demonstrating a lifetime risk of 7–30%⁴ and a cancer diagnosis occurring either before, concurrent with, or after DM onset⁵. Malignancy commonly manifests within the first year of diagnosis, but the risk remains most elevated within the three years from disease onset⁵. Remarkably, adult-onset DM is associated with a wide range of malignancy types, most commonly ovarian, breast, lung, and hematologic cancers, necessitating an extensive and often costly screening workup^6,7. While autoantibodies against transcriptional intermediary factor-γ (TIF1γ) and nuclear matrix protein-2 are associated with increased cancer risk in adult DM patients⁸, the molecular and cellular features underlying cancer association in DM remain inadequately understood.

DM skin pathology is characterized by vacuolar interface dermatitis, increased lymphocytic infiltrate near vasculature, enhanced mucin deposition in the dermis, and basement membrane thickening⁹. While these histologic findings establish the fundamental pathologic features of DM, they provide limited insight into the complex cellular dynamics of disease pathology and progression, such as the precise organization of immune infiltrates, their molecular programs, and their spatial relationship with surrounding vascular structures. Recent advances in single-cell and spatial transcriptomics have revolutionized our ability to describe complex tissue environments¹⁰. These technologies enable comprehensive profiling of cellular ecosystems while preserving crucial information about spatial organization and cell-cell interactions, providing insights into the molecular and cellular mechanisms underlying DM pathogenesis.

To address this knowledge gap, we conduct a comprehensive, multi-omic integrated analysis of adult-onset DM skin by combining single-cell RNA sequencing, high-resolution spatial transcriptomics, and multiplexed antibody-based CODEX imaging. By comparing DM lesions with both healthy skin and cutaneous lupus erythematosus (CLE) lesional skin, we identify disease-specific immune and stromal cell states that distinguish cancer-associated from non-cancer-associated DM at the cellular and molecular level. These findings provide a new framework for understanding DM pathogenesis that has significant implications for clinical disease stratification and treatment selection based on skin biopsy analysis.

Results

High-resolution spatial and single-cell atlas reveals interferon-driven tissue reorganization in lesional skin biopsies from DM and CLE

To comprehensively characterize cellular niches in DM, we conducted high-resolution spatial and single-cell analysis using skin biopsies from DM patients. As a comparison group, our study design included skin biopsies from patients with CLE since both DM and CLE exhibit shared pathological features, including interface dermatitis, perivascular lymphocytic infiltration, and mucin deposition. By incorporating CLE for comparison, we sought to identify shared and DM disease-specific pathological mechanisms and better define the unique molecular features of DM within the broader context of other interferon-driven inflammatory skin diseases. Using the Chromium single-cell gene expression flex, we profiled single-cell transcriptomes of 44 samples, encompassing 22 DM patients, 9 CLE patients, 8 non-lesional samples from a subset of these patients, and 6 healthy controls (Fig. 1A and Supplementary Data 1).

A Overview of experimental design and analysis pipeline. FFPE skin biopsies were analyzed using two complementary approaches: probe-based single-cell transcriptomics (n = 44 samples) and spatial transcriptomics (n = 19 samples). B Uniform manifold approximation and projection (UMAP) visualization of major cell states identified by single-cell and spatial transcriptomics. C Spatial localization of major cell states. Left, H&E-stained FFPE section. Right, spatial distribution of cell states (same section) colored according to the cell state legend shown in Fig. 1B. D Differential composition of major cell states between disease conditions (single-cell data). Cell-type abundance differences were assessed using sccomp at the sample level. Red indicates significant changes (FDR, false discovery rate <0.1) and blue shows non-significant differences in cell state proportions. Arrows indicate which condition shows enrichment for each cell state in the respective comparison. E Major cell state composition of individual samples analyzed by probe-based single-cell data. F Representative samples from healthy, DM, and CLE showing the spatial distribution of six identified compartments by hoodscanR. G Compartment composition of individual samples analyzed by spatial single-cell transcriptomics (n = 19). H PROGENy pathway analysis across different skin conditions (single-cell data). I, J Density of interferon response scores calculated using UCell for all cells from all samples obtained through single-cell transcriptomics. Color gradient represents cell density at each UCell score value, with horizontal lines indicating score distribution percentiles. Spatial density maps of immune cells (K), type I interferon response (L), and type II interferon response (M) in the lesional DM sample from Fig. 1C. Correlation between immune cell proportion and *CXCL10*⁺ keratinocytes (N) or *CXCL10*⁺ fibroblasts (O). Source data are provided as a Source data file.

To further delineate the spatial architecture and cell-cell interactions within the tissue microenvironment, we performed high-resolution spatial transcriptomics (Xenium prime 5k, 10x Genomics) on a subset of biopsies (n = 12 lesional DM, n = 4 lesional CLE, and n = 3 healthy controls), with several patients contributing samples to both the single-cell and spatial transcriptomics analyses to enable direct comparison (Fig. 1A and Supplementary Data 1). Unlike previous studies¹¹ that either employed flow cytometry enrichment or analyzed the dermal and epidermal components of the skin separately, our methodology, which utilized intact FFPE tissue sections, enabled comprehensive profiling of the complete cellular ecosystem in its native context, including traditionally challenging-to-capture cell states in tissue, such as neutrophils and mast cells. Through integrated analysis of both technologies, we identified 16 major skin cell types that were consistently detected across both platforms (Fig. 1B, C), with detailed annotation methodology described in the Methods section.

To quantify differences in cellular composition between disease states, we employed sccomp¹², a computational method designed for differential abundance analysis, applying it to the single-cell dataset to take advantage of the greater sample size compared to our spatial cohort. Comparison of lesional samples (both DM and CLE) to non-inflamed (NI) tissues (healthy and non-lesional) revealed a shared pattern of immune infiltration in both diseases (Fig. 1D, E). DM lesions were characterized by prominently myeloid and lymphoid cell infiltration, while CLE lesions showed a more diverse immune infiltrate characterized by myeloid, lymphoid and a particularly striking accumulation of B cells and plasma cells (Fig. 1D, E).

To map the spatial organization of major cell states within the tissue architecture, we employed hoodscanR¹³, a computational tool that uses the approximate nearest neighbor algorithm and probability-based clustering to identify cellular neighborhoods in spatial transcriptomics data (Fig. 1F). After separating keratinocytes into follicular and interfollicular populations to better distinguish skin regions (Fig. 1F and Supplementary Fig. 1A), this analysis revealed six distinct tissue compartments: (1) epidermal compartment, primarily composed of interfollicular keratinocytes; (2) immune-vascular compartment, containing B cells, plasma cells, pDCs, lymphocytes, myeloid cells, and endothelial cells; (3) dermal compartment, consisting of fibroblasts and myeloid cells; (4) follicular compartment, containing follicular keratinocytes; (5) sweat gland compartment, containing sweat gland cells; and (6) vascular compartment, comprising endothelial cells, pericytes, fibroblasts, and myeloid cells. The immune-vascular compartment occupied a substantial area only in lesional samples, predominantly in superficial dermis and localizing near the epidermal compartment (Fig. 1F, G). This pattern of immune cell infiltration is observed in both CLE and DM, consistent with well-documented histopathologic findings in DM and CLE^14,15. Pathway analysis of single-cell data using PROGENy gene signatures across all conditions revealed significant activation of the JAK-STAT pathway in both DM and CLE lesional skin compared to healthy and non-lesional skin (Fig. 1H). Since JAK-STAT signaling is downstream of interferon receptor signaling, we analyzed type I and II interferon response gene sets from the Reactome database using UCell (Fig. 1I, J). While both diseases exhibited strong interferon signatures, CLE samples showed higher interferon response scores compared to DM samples. However, we hypothesized that this difference might be attributed to varying levels of immune cell infiltration in our cohort. Indeed, our spatial analysis demonstrated that elevated interferon signaling was predominantly localized to regions of immune infiltration, indicating a direct relationship between local immune cell infiltration and interferon response (Fig. 1K–M).

Fine-grained analysis of major cell states in single-cell data, including keratinocytes (Supplementary Fig. 1B–E) and fibroblasts (Supplementary Fig. 1H–K), revealed significant enrichment of interferon-response states (CXCL10_basal, CXCL10_spinous, and CXCL10_fib) in lesional samples from DM and CLE compared to non-inflamed tissues (Supplementary Fig. 1C, I). These interferon-response clusters were characterized by high expression of chemokines CXCL10, CXCL11, and CXCL9 (Supplementary Fig. 1D, J). Moreover, the relative abundance of CXCL10⁺ cell states within both keratinocytes and fibroblasts showed a strong positive correlation with the proportion of immune cells in the tissue (Fig. 1N, O). This suggests a cellular crosstalk loop where stromal interferon signaling recruits immune cells through upregulation of chemokine expression, leading to expansion of infiltrating immune cells and further activating stromal cells through interferon signaling. Among keratinocytes, we identified an expanded population of proliferative basal keratinocytes (cyclic_basal) in lesional samples that express both proliferation markers (MKI67) and DNA damage-response genes (FANCA, FANCI, FOXM1, and CHEK1) (Supplementary Fig. 1C, E). Detailed characterization of fibroblast heterogeneity revealed that beyond CXCL10-expressing fibroblasts, lesional skin in both conditions showed elevated numbers of CCL19-expressing fibroblasts (CCL19_fib) and interferon-associated IFIT2-expressing fibroblasts (IFIT2_fib) (Supplementary Fig. 1I).

Single-cell analysis reveals immune cell heterogeneity in DM

Since immune infiltration in lesional skin is a shared feature in both DM and CLE, we queried our single-cell data to identify immune cell composition with disease-specific association. We first identified fine-grained cell states for lymphoid cells, myeloid cells, B cells, and plasma cells, with the detailed annotation process described in the “Methods” section (Fig. 2A and Supplementary Fig. 2A–G). Unsupervised analysis of immune cell composition of single-cell samples using Bray–Curtis dissimilarity metrics revealed two distinct patterns of immune composition in DM (Fig. 2B, C). The proportion of immune cells relative to total cells represented the primary pattern of variation (PCoA1, 44.73%) (Fig. 2B). Surprisingly, DM patients’ cancer status emerged as the second independent pattern distinguishing DM samples, explaining 15.25% of the compositional variance (PCoA2) (Fig. 2C and Supplementary Fig. 2H). Cancer-associated DM was defined as the presence of a confirmed malignancy diagnosed within three years before or after DM onset, whereas non-cancer-associated DM was defined as the absence of detectable malignancy for at least three years following disease onset. This dual-pattern separation suggested that both the extent of immune infiltration and patients’ cancer status independently shape the skin immune microenvironment.

Fig. 2 — A UMAP visualization of fine-grained cell state analysis of lymphocytes, myeloid cells, B cells/plasma cells/pDC, colored by cell states identified during the analysis of single-cell data (n = 44). B, C Principal coordinates analysis of Bray–Curtis dissimilarities of immune cell state composition across lesional samples (single-cell data). B Colors represent the proportion of immune cells relative to total cells in each lesional sample. C Colors represent clinical information. D Differential composition analysis of immune cell states across disease conditions in lesional tissue (single-cell data). Cell-type abundance differences were assessed using sccomp at the sample level. Red indicates significant changes (FDR < 0.1) and blue shows non-significant differences in cell state proportions. E Heatmap of associations between MOFA factors and clinical variables (single-cell lesional data). The heatmap shows −log10 transformed adjusted p-values from Kruskal–Wallis tests examining relationships between MOFA factor scores (x-axis) and clinical variables (y-axis). F The x-axis shows MOFA Factor 1 scores, and the y-axis shows the percentage of immune cells relative to total cell number in each sample. G Distribution of MOFA factor scores across disease subtypes. The left panel displays Factor 6 stratification by disease and TIF1γ status. The right panel shows Factor 4 stratification by disease and cancer status. Each dot represents a patient sample. H The stacked bars represent the percentage of gene expression variance explained by different MOFA factors for each cell type. Source data are provided as a Source data file.

To characterize how DM patients’ cancer status stratifies the skin immune microenvironment, we performed targeted differential abundance analysis comparing cancer- versus non-cancer-associated DM samples using single-cell data (Fig. 2D). Non-cancer-associated DM demonstrated a highly distinctive immune signature characterized by dramatic enrichment of innate immune cells: mature regulatory dendritic cells (mregDC, BIRC3+LAMP3+), activated Tregs, neutrophils, monocytes, CXCL10⁺ dendritic cells, and conventional type dendritic cells (cDC2, cDC1). In contrast, cancer-associated DM samples exhibited adaptive immune predominance with increased B cells, plasma cells, and cytotoxic CD8⁺ T cells (Fig. 2D). We also examined whether myopathic versus amyopathic status could explain similar immune differences (Supplementary Fig. 2H, I). However, myopathy-based stratification revealed considerably weaker immune compositional differences, with substantial clinical overlap between cancer association and myopathy status (Supplementary Data 1).

Intriguingly, when we compared these DM subtypes to CLE samples, we observed that CLE immune composition more closely resembled cancer- than non-cancer-associated DM, particularly in B cell, plasma cell and CD8⁺ T cells enrichment (Fig. 2D). However, CLE samples showed distinctively higher levels of plasmacytoid dendritic cells (pDC) and plasma cells compared to both DM subtypes (Fig. 2C, D). This pattern suggests both shared and unique immunopathological mechanisms between CLE and DM. Further analysis revealed several additional disease-specific features. Most notably, we observed a near-complete absence of neutrophils in CLE biopsies, identifying neutrophil infiltration as a distinctive feature of DM pathology. CLE samples also showed near-complete depletion of Langerhans cells (Fig. 2D).

Multicellular factor analysis uncovers distinct stromal and immune cell signatures in non-cancer-associated DM skin

The striking immunological differences between cancer- and non-cancer-associated DM skin (Fig. 2B–D) led us to investigate transcriptional programs underlying cancer association across all skin cell types. We employed multicellular factor analysis (MOFA)¹⁶ to single-cell data, which provides an unsupervised approach to studying transcriptional programs across all cell populations in disease states at the same time. Our MOFA model effectively captured the majority of gene expression variability across all studied cell types, constructing a new latent space composed of eight factors (Fig. 2E). Factor 1 emerged as the primary driver of gene variation, explaining variability across all cell populations and containing numerous interferon response genes, including IFIT1, IFIT3, ISG15, and SOD3. Kruskal–Wallis analysis of the MOFA factors revealed its strong association with immune infiltration percentage (fdr = 0.00261) (Fig. 2E, F). Consistent with the idea that interferon pathway activation in dermatomyositis is primarily driven by the magnitude of local immune infiltration rather than malignancy-associated mechanisms, the intensity of interferon response exhibited a robust correlation with the extent of immune infiltration across samples, yet displayed no meaningful relationship with cancer presence (Fig. 2F).

Unexpectedly, we found Factor 4 to be strongly associated with the distinction between cancer-associated DM, non-cancer-associated DM, and CLE samples (fdr = 0.0141) (Fig. 2G). Factor 4 explained cell type-specific gene expression variation, primarily in myeloid cells and fibroblasts (8% each) and endothelial cells (6%) (Fig. 2H). Interestingly, the expression of ABCA8, ABCA9, and ABCA10 genes within the fibroblasts emerged as the primary drivers of Factor 4 (Supplementary Fig. 2J). Factor 6 distinctly represented keratinocyte-specific variability (16%) and showed a modest association with TIF1γ-positive DM status (p = 0.0152, fdr = 0.121), though this association did not reach conventional statistical significance after multiple testing correction (Fig. 2E, G). We confirmed that none of the MOFA factors were significantly associated with the patients’ myopathy status (Fig. 2E). To exclude the contribution of CLE samples in this analysis, we performed additional MOFA analysis using only DM samples, which yielded identical factor associations and gene expression patterns, confirming the robustness of our cancer-associated transcriptional signatures (Supplementary Fig. 2K, L).

Unique spatial organization of immune cell niches reveals disease-specific archetypes in DM and CLE lesions

Our spatial transcriptomics data demonstrated that infiltrating immune cells form structured aggregates, particularly concentrating in superficial dermis near interfollicular keratinocytes and vascular structures (Supplementary Fig. 3A and Fig. 1F). Importantly, we observed that even in healthy skin, a small number of immune cells appeared around vascular endothelial cells (Supplementary Fig. 3B), likely reflecting physiological immune surveillance. To distinguish disease-associated immune niches from normal skin organization, we implemented a comprehensive computational pipeline that integrates spatial transcriptomic profiles of healthy, DM, and CLE skin (Supplementary Fig. 3A). Using HDBSCAN clustering and Bray–Curtis dissimilarity analysis, we identified four major immune niche archetypes (for more details, see “Methods” section).

In healthy skin, we found two low-density immune niches (Fig. 3A–D): Mac_enr_Healthy (clusters of non-inflamed macrophages and mast cells in the dermis) and Dermis_Healthy (non-inflamed macrophages, mast cells, cDC2, and CD4⁺ T cells near the interfollicular regions). In DM and in CLE, Mac_enr_Healthy niches acquire a distinct composition (Mac_enr_Disease clusters) and became characterized by expansion of additional cell types including CD8⁺ T cells, CXCL10⁺ dendritic cells and macrophages, and proliferative (MKI67+) immune cells (Fig. 3C). In non-cancer-associated DM and CLE, these clusters remained primarily in the dermal compartment, while in cancer-associated DM, they extended to the interfollicular border while maintaining sparse architecture (Fig. 3C, D).

The most unique pathological immune archetype was Dermis_Disease niches, found exclusively in diseased but not healthy skin (Fig. 3E). These niches are characterized by highly dense immune aggregates, positioned at the interfollicular border with significantly fewer non-inflamed macrophages (Fig. 3D), which contain the majority of immune cells across all disease subtypes and reveal striking compositional differences amongst disease subtypes. Importantly, the highest interferon response scores localize to the Dermis_Disease niches, indicating that type I interferon activation is spatially concentrated within these immune niches (Supplementary Fig. 3C). To further explore inter-subtype variation, we performed Bray–Curtis dissimilarity analysis of Dermis_Disease niches across disease conditions (Fig. 3F).

Non-cancer-associated DM exhibited two distinct patterns (Fig. 3F, G): one characterized by CXCL10⁺ myeloid cells (neutrophils, monocytes, macrophages, and dendritic cells) and mregDCs, uniquely containing CXCL8+ dendritic cells (Fig. 3G–I); the second was enriched in CXCL10⁺ macrophages, CD4⁺ T cells, activated Tregs, mregDCs, cDC2, and pDCs (Fig. 3G–I). A key distinguishing feature of non-cancer DM compared to cancer DM was increased Treg-mregDC co-localization, with Tregs showing elevated TNFRSF18, TNFRSF4, NFKB2, and NFKBIA expression, suggesting Treg activation via mregDCs (Supplementary Fig. 3D). While cancer-associated DM generally exhibited sparse macrophage-enriched architecture with scattered disease-associated immune cells, we identified a specific cancer-associated DM cluster distinguished by high levels of CD8⁺ T cells, CD4⁺ T cells and elevated B cells (Fig. 3G–I). CLE niches showed less variability between discoid lupus erythematosus (DLE) and subacute cutaneous lupus erythematosus (SCLE) subtypes (Supplementary Fig. 3E, F), displaying immune composition that combines features of both non-cancer and cancer-associated DM clusters. However, potential differences between DLE and SCLE may not have been detected in this cohort due to the limited number of CLE samples analyzed. Compared to DM, CLE niches were distinctly enriched for pDCs, plasma cells, and plasmablasts (Supplementary Fig. 3B, F), while containing high levels of CD4⁺ T cells, activated Tregs, mregDCs, and CXCL10⁺ macrophages similar to non-cancer DM, yet sharing elevated CD8⁺ T cells and B cells similar to cancer-associated patterns. Consistent with single-cell analysis, CLE immune aggregates notably lacked neutrophils (Supplementary Fig. 3F).

To validate these spatial immune niche archetypes at the protein level, we performed multiplexed antibody-based CODEX imaging on serial sections from the same biopsies used for spatial transcriptomic and single-cell analyses, including three cancer-associated DM samples, two non-cancer-associated DM samples, and two CLE samples. A panel of 15 antibodies targeting key immune and stromal cell populations was used for quantitative CODEX imaging, enabling the identification of nine major cell types, including epithelial cells (KRT14), vascular endothelial cells (CD31), pericytes (αSMA), and major immune lineages, such as B cells (PAX5), CD8⁺ T cells (CD8, CD3), CD4⁺ T cells (CD4, CD3), myeloid cells (CD14, CD68), and pDC (IDO1) (Supplementary Fig. 4A–C). Consistent with spatial transcriptomic findings, CODEX imaging revealed dense immune aggregates preferentially localized at the interfollicular dermal–epidermal junction and in close proximity to vascular structures (Supplementary Fig. 4D, E). Importantly, multiplexed protein imaging recapitulated distinct, disease-specific immune niches identified by spatial transcriptomics (Supplementary Fig. 4C). Cancer-associated DM lesions were enriched in CD8⁺ T cells and B cells, whereas non-cancer-associated DM lesions exhibited increased abundance of myeloid cells (Supplementary Figs. 4C and 5A–C). CLE lesions displayed a hybrid architecture, combining features of both DM subtypes (Supplementary Fig. 5B, C).

Disease-associated transcriptional programs define distinct immune niches

To characterize molecular programs within disease-associated immune cell niches, we performed pseudobulk analysis of immune cells that came from Dermis_Disease niches. Principal component analysis (PCA) revealed robust separation between conditions (Fig. 4A). Differential expression analysis across CLE, non-cancer, and cancer-associated DM lesions identified distinctive immune signatures (Fig. 4B). For example, non-cancer-associated DM immune niches showed significant upregulation of pro-inflammatory mediators specifically expressed by CXCL10⁺ myeloid cells (such as neutrophils, monocytes, and CXCL10⁺ dendritic cells and macrophages), including inflammatory calcium-binding proteins S100A8/S100A9, pro-inflammatory cytokine IL1B, monocyte chemoattractant CCL2, and T-cell recruiting chemokine CXCL10 (Fig. 4B and Supplementary Fig. 6A, B). Interestingly, in these non-cancer-associated DM immune niches with elevated CXCL10 expression, we also observed increased IL1B expression, consistent with published reports indicating that CXCL10 can be co-induced by both interferon and IL-1β signaling¹⁷. This suggests that within these immune niches, CXCL10 expression likely reflects the combined activity of IFN- and IL-1β-driven inflammatory programs (Supplementary Fig. 6C).

A Principal component analysis of pseudobulk RNA-seq of immune cells data from disease-associated (Dermis_Disease) immune niches (n = 147). Each point represents an immune niche colored by disease condition. B Differential gene expression analysis of disease-associated (Dermis_Disease) immune niches between different disease conditions. Differential expression was assessed using MAST. Genes were considered significant at FDR < 0.1 and |log2FC| > 0.3. Red dots indicate significant genes, blue dots indicate non-significant genes. C UMAP visualization of myeloid cell states identified through analysis of spatial transcriptomics with mregDCs (red) and *CXCL10*⁺ myeloid cells (blue) highlighted among all myeloid populations (gray). D UMAP visualization showing the density of *CD274*-expressing cells from two types of non-cancer disease-associated (Dermis_Disease) niche. Cell positions correspond to the UMAP shown in Fig. 4C. E Spatial localization of mregDC and Treg cell interactions in disease-associated (Dermis_Disease) immune niches. The left panel shows the distribution of immune cell state, and the right panel displays the spatial distribution of specific transcripts. F Spatial organization of disease-associated (Dermis_Disease) immune niches in cancer-associated DM and CLE. The left panel shows the distribution of immune cell state, and the right panel displays the spatial distribution of specific transcripts. G Expression levels of *IFNB1* from spatial transcriptomics data across immune cell populations. The top panel shows *IFNB1* expression in major cell states, stratified by disease conditions. The bottom panel displays the density plot of *IFNB1*+ cells across the UMAP embedding for myeloid cell types visualized in Fig. 4C and Supplementary Fig. 3B. Source data are provided as a Source data file.

We found elevated CD274 expression, encoding PD-L1, in non-cancer-associated DM and CLE immune niches compared to cancer-associated DM samples (Fig. 4B). While CD274 expression was restricted to mast cells in healthy skin, lesional skin showed broad expression across mregDCs, CXCL10⁺ myeloid cells, reflecting its critical role in immune regulation¹⁸. Notably, CD274 expression patterns varied across non-cancer-associated DM niches (Fig. 4C, D). Non-cancer-associated niches enriched with neutrophils showed expression limited to CXCL10⁺ myeloid populations (Fig. 4D), while pDC-enriched niches displayed CD274 expression in both mregDCs and CXCL10⁺ myeloid cells (Fig. 4D). Importantly, CD274 expression in mregDCs coincided with activated Treg signatures (Fig. 4E).

Cancer-associated DM demonstrated significant upregulation of genes related to tertiary lymphoid structures (TLS), including chemokine receptors CCR6, CXCR6, CCR5 (Fig. 4B and Supplementary Fig. 6A). Spatial transcriptomic analysis revealed that a subset of cancer-associated DM immune aggregates exhibited organized lymphoid aggregate structures, characterized by a dense B cell core surrounded by CXCL13-expressing T peripheral helper (Tph) cells, with adjacent CD4⁺ and CD8⁺ T cells and peripheral plasma cells (Fig. 4F). Interestingly, we observed the presence of pDCs and mregDCs within these lymphoid aggregates, features that are suggestive of lymphoid organization but do not qualify as definitive TLS.

Despite this organized architecture, cancer-associated DM aggregates exhibited fewer plasma cells compared to immune aggregates identified in CLE (Supplementary Figs. 3F and 4F). This expanded plasma cell abundance in CLE correlated with high expression of B cell survival factors TNFSF13 (APRIL) and TNFSF13B (BAFF) within CLE immune niches (Supplementary Fig. 6B, C). These cytokines are essential for B cell survival during the germinal center reaction and subsequent differentiation into long-lived plasma cells, suggesting that enhanced survival signaling facilitates more efficient B cell-to-plasma cell differentiation and contributes to the increased plasma cell accumulation observed globally across CLE samples.

Also, our spatial transcriptomics data revealed differential expression of IFN-β within immune niches, with higher levels observed in both cancer- and non-cancer-associated DM compared to CLE immune niches (Fig. 4G and Supplementary Fig. 6C), supporting a critical role of IFN-β in DM pathology^19,20. CXCL10⁺ myeloid cells represented the primary cellular source of IFNB1 expression within these immune aggregates. Furthermore, single-cell analysis of DM and CLE samples confirmed IFN-β expression specifically in myeloid cells in 3 DM samples, suggesting that the presence of IFN-β expression may distinguish DM from CLE. Due to the low expression level of certain cytokines, both single-cell and spatial single-cell sequencing have limited ability to assess the abundance of critical immune mediators, such as IL2, IL10, and IFN-α.

Disease-specific signatures were further characterized by elevated expression of interferon response genes (STAT1, IFI44) and innate immune components (IRAK4, TLR7) in CLE samples (Supplementary Fig. 6A, C), consistent with lupus pathogenesis^21,22. Cancer-associated DM samples uniquely showed upregulation of cancer-associated genes, including SNHG15 (Fig. 4B and Supplementary Fig. 6D), a lncRNA frequently elevated across multiple malignancies ^23,24.

Vascular remodeling and metabolic reprogramming underlie distinct immune niche phenotypes

Building on our observation that immune cells organize into distinct niches centered around skin vasculature, we next characterized endothelial and fibroblast populations within disease-associated immune niches (Dermis_Disease) and their surrounding areas, defined as regions within 50 micrometers of niche boundaries (Fig. 5A). Using Bray–Curtis dissimilarity analysis of endothelial cells within and adjacent to immune niches, we identified four distinct vascular niches (Fig. 5B, C, and Supplementary Fig. 7A, B).

One niche, the lymphatic_cl, which comprised primarily lymphatic endothelial cells, showed no disease-specific association (Fig. 5C and Supplementary Fig. 7A, B). In contrast, we identified three endothelial niches with disease-specific associations. The CXCL10_ven_cl niche, predominantly found in CLE, consisted of CXCL10-positive venous endothelial cells (Fig. 5C and Supplementary Fig. 7A, B), while the CXCL10_cap_cl niche containing CXCL10-positive capillary endothelial cells was primarily observed in non-cancer DM samples. Similar analysis of fibroblast populations within and adjacent to immune niches revealed increased abundance of CXCL10-expressing cell states in non-cancer-associated DM and CLE samples compared to cancer-associated DM (Supplementary Fig. 7C, D). The venous_cl niche, enriched in cancer-associated DM samples, consisted primarily of venous endothelial cells lacking CXCL10 expression (Fig. 5C–E and Supplementary Fig. 7A, B). Furthermore, endothelial vasculature in cancer-associated DM samples displayed more rounded morphology with open lumens (Fig. 5D).

Since vasculopathy is a known feature of DM lesions, we sought to characterize vascular alterations specific to DM skin lesions. First, we quantified the tissue vascularity by measuring endothelial cell density within 600-μm regions extending from the basal keratinocyte layer through the papillary dermis (Fig. 5F). Quantitative analysis revealed no statistically significant endothelial cell density in both cancer- and non-cancer-associated DM compared to healthy samples (Fig. 5G). However, DM samples showed a particularly marked increase in endothelial tip cells and proliferative endothelial cells, indicating a sprouting angiogenic response and active vascular remodeling in DM (Fig. 5G and Supplementary Fig. 7E).

To understand the cellular molecular mechanisms driving distinct immune and vascular remodeling in DM, we examined differential gene expression across disease-associated immune niche populations. This analysis revealed that stromal cells within non-cancer-associated niches, compared to cancer-associated aggregates, exhibit pronounced upregulation of hypoxia- and inflammation-related pathways. Specifically, endothelial cells from non-cancer-associated DM niches displayed increased expression of hypoxia-inducible genes (HIF1A, PKM, NOS3, and PGK1) and inflammatory mediators (CXCL10, CXCL9, and CXCL11) (Fig. 5H). Additionally, these endothelial cells exhibited upregulation of genes involved in vascular injury (ICAM1, THBS1, and SELE) and angiogenesis (VEGFC, CCN1, and CCN2), indicating a coordinated transcriptional response to altered oxygen tension within the lesional microenvironment (Fig. 5I).

Consistent with a regional response to hypoxia, fibroblasts from non-cancer-associated niches, compared to those in cancer-associated aggregates, exhibited significant upregulation of myofibroblast activation genes (POSTN, COMP, and COL6A1) along with enhanced transcription of hypoxia-responsive (HIF1A, PKM, PGK1) and inflammation-related genes (CXCL10, CXCL9, CXCL11, and IL6) (Fig. 5I). Notably, fibroblasts in non-cancer-associated niches displayed significant downregulation of ABCA10, ABCA8, C7, C3, and ABCA6, with expression levels approaching near-zero, contrasting with cancer-associated and CLE-associated fibroblasts (Fig. 5I, J, and Supplementary Fig. 7G). These genes, predominantly expressed in vascular-associated fibroblasts, serve as key molecular markers of this stromal population (Supplementary Fig. 7F). Furthermore, analysis of vascular-associated fibroblasts from healthy skin (located near Healthy_Dermis immune niches) confirmed high expression of ABCA10, ABCA8, C7, C3, and ABCA6, suggesting that their loss in non-cancer immune niches represents a fundamental shift in the stromal microenvironment (Supplementary Fig. 7G, H).

Since all stromal populations associated with non-cancer-associated immune aggregates exhibit elevated expression of hypoxia- and inflammation-related genes, we performed a spatial correlation analysis to investigate their relationship with other gene expression patterns. Our analysis revealed that angiogenesis, hypoxia, inflammation, and vascular injury signatures were directly correlated with the expression of cytokines highly expressed by monocytes and neutrophils, including CCL2, CCL8, S100A8, S100A9, CXCL10, and IL1B (Fig. 6A, B). In contrast, the vascular-associated gene signature (ABCA8, ABCA9, C7, and C3) displayed an inverse correlation with these cytokines, suggesting that hypoxia and inflammatory signals actively suppress the vascular-associated fibroblast transcriptional program (Fig. 6A, B), resulting in loss of vascular fibroblast identity and injury to the steady-state perivascular compartment. However, this analysis does not clarify the main cause of the suppression of vascular-associated fibroblasts—whether the drivers are the cytokines highly expressed by monocytes and neutrophils or hypoxia. To address this question, we performed an in vitro experiment using human dermal fibroblast cultures exposed to both inflammatory cytokines and hypoxia. Fibroblasts were cultured under either normoxic or hypoxic (0.5% O₂) conditions and treated with IL1-β (1 ng/mL), IFN-β (5 ng/mL), or IFN-γ (5 ng/mL) for 24 h. Our qPCR analysis revealed that hypoxia significantly reduced the expression of vascular-associated fibroblast markers (C7, ABCA6, and ABCA8) compared to normoxic conditions, regardless of cytokine treatment (Fig. 6C and Supplementary Fig. 7I), while cytokine treatment had a mild reduction in the expression of vascular fibroblast markers. Flow cytometry analysis of apoptosis markers confirmed that this downregulation of the vascular-associated gene signature in fibroblasts was not associated with increased cell death, indicating transcriptional reprogramming driven by hypoxia (Supplementary Fig. 8A, B). Together, the dramatic reduction observed under hypoxic conditions suggests that oxygen deprivation is the primary driver of vascular-associated fibroblast identity loss in non-cancer DM skin.

Fig. 6 — A Spatial correlation analysis of gene expression within all identified immune niches. B Spatial distribution of key gene signatures in immune niches across disease conditions. C qPCR analysis of vascular fibroblast markers in dermal fibroblasts treated with IL-1β (1 ng/mL), IFN-γ (5 ng/mL), and IFN-β (5 ng/mL) under normoxic (blue) and hypoxic (purple) conditions for 24 h (n = 4 replicates per condition). Relative expression levels (normalized to *GAPDH*) are shown, with error bars representing standard deviation. Statistical significance was determined using a two-tailed Student’s t test. D Schematic representation of immune-stromal organization and molecular signatures in cancer-associated DM, non-cancer-associated DM, and CLE niches. Source data are provided as a Source data file.

Discussion

The pathology of autoimmune skin diseases has traditionally relied on histopathologic assessment of skin biopsies, which identifies general features of inflammation, but is insufficient to identify specific cellular and molecular features unique to each disease entity^15,25. Our comparative analysis of DM and CLE provides several key advantages as both conditions share similar histopathologic features—interface dermatitis, perivascular infiltrate and mucin deposition—and prominent interferon signatures, allowing us to identify disease-specific mechanisms rather than general inflammatory pathways (Fig. 6D).

First, we identified DM-specific expression of IFN-β in myeloid cells, while CLE showed broader interferon response through TLR/IRAK signaling, consistent with prior studies demonstrating myeloid-specific IFN-β production and predominant innate immune signatures in DM compared to CLE^26–31. Notably, IFN-β, but not IFN-α protein levels, correlate specifically with cutaneous disease activity scores in DM¹⁹. The clinical efficacy of IFN-β inhibition, with dazukibart–an IFN-β specific monoclonal antibody, has shown promise in phase 2 clinical trials for DM treatment²⁰. Our spatial and single-cell analysis, in addition, revealed increased pDC abundance in CLE lesions, organized in distinct immune niches with elevated TLR/IRAK signaling. The co-localization of pDCs, known producers of IFN-α through TLR7/9 activation³², suggests a coordinated inflammatory program distinct from the IFN-β-driven pathology in DM. Secondly, DM skin biopsies exhibit prominent perturbation in the vascular niches. The expansion of endothelial tip cells and increased proliferating endothelial cells in DM lesions suggests ongoing sprouting angiogenesis, likely in response to vascular damage and hypoxia in DM lesions. Specifically, non-cancer-associated DM lesions showed evidence of rarefaction, loss of perivascular fibroblast identity, and regional hypoxia, suggesting cell state changes secondary to vasculopathy, a known feature of DM lesions³³. Although we did not assess nailfold capillary morphology, published studies show that capillary abnormalities correlate with DM disease activity^34–36. Limited data exist comparing nailfold changes between cancer- and non-cancer-associated DM³⁷. Consistently, our findings of hypoxic microenvironment and vascular remodeling in non-cancer DM skin suggest microvascular dysfunction, though direct correlation studies are needed. Third, we identified lesional neutrophil infiltration as a DM-specific feature absent in CLE lesions. Studies have shown neutrophils and their extracellular traps (NETs) correlate with interstitial lung disease in anti-MDA5-positive DM and calcinosis in juvenile DM^38,39, suggesting their broad pathogenic role in organs affected by DM. In non-cancer-associated DM, neutrophil infiltration was particularly prominent, with these cells showing high IL1B expression and forming part of CXCL10⁺ myeloid aggregates. This pattern suggests that neutrophil-derived IL-1β may amplify inflammatory activation in coordination with interferon-responsive pathways within the lesional microenvironment⁴⁰. Fourth, we observed selective depletion of Langerhans cells in CLE, consistent with reports of reduced Langerhans cell density in non-lesional lupus skin and murine lupus models⁴¹. A recent study suggested type I interferon signature has a role in Langerhans cell dysfunction, which has been proposed to lead to photosensitivity observed in lupus patients⁴¹. Therefore, their loss may contribute to disease progression, as Langerhans cells also play crucial roles in maintaining skin immune homeostasis through IL-10 production and regulatory T cell activation ⁴².

Patients with adult-onset DM have a significantly elevated risk of malignancy^4,43. A major gap in the management of dermatomyositis is enhanced risk stratification of patients as cancer-associated versus non-cancer-associated DM at the time of diagnosis. Current guidelines for cancer screening in DM patients suggest utilizing a combination of autoantibody status and clinical features to determine risk for malignancy⁴⁴. Given the low threshold for a DM patient to be considered high-risk per guidelines, many patients undergo extensive screening that is repeated annually for 3 years⁴⁴, leading to increased health care costs, radiation exposure to patients, and substantial patient distress and/or anxiety. Therefore, understanding the underlying immune mechanisms that distinguish cancer-associated DM from non-cancer-associated DM can yield discernible markers that can enhance early risk stratification of DM, providing the opportunity for expedited detection and management of malignancies in these patients. Among DM patients, those with anti-TIF1γ have the highest risk of cancer, followed by anti-NXP2⁴⁵. Here, our study utilizing high-resolution spatial transcriptomics revealed a clear distinction between cancer and non-cancer DM at a cellular and molecular level. Non-cancer DM exhibited dense immune aggregates with elevated CXCL10⁺ myeloid populations, including neutrophils, monocytes, macrophages, and dendritic cells. Moreover, these lesions were characterized by abnormal vascular endothelial and perivascular stromal remodeling, leading to loss of vascular-associated fibroblast identity and upregulation of hypoxia-responsive genes in both endothelial cells and fibroblasts.

In contrast, cancer-associated DM displayed two distinct patterns of immune organization. The majority showed sparse immune infiltrates resembling healthy tissue architecture with preserved vasculature and scattered immune cells. However, a subset of cancer-associated DM samples exhibited organized immune aggregates enriched in B cells and CD8⁺ T cells, structurally resembling TLS. These TLS-like aggregates featured defined B cell follicles surrounded by CXCL13-expressing T cells, with adjacent CD4⁺ and CD8⁺ T cell zones and peripheral plasma cell positioning, an architecture commonly observed in anti-tumor immune responses^46,47. The sparse infiltrates in cancer-associated DM become particularly intriguing in light of the potential hypothesis that some DM cases may be fundamentally cancer-associated but do not manifest, given the patient’s ability to mount a successful early anti-tumor response⁴⁸. While our findings of organized B cell and CD8⁺ T cell responses in cancer-associated DM might support this adaptive anti-tumor immunity framework, the distinct immune architecture in non-cancer DM suggests an alternative pathway. The predominance of neutrophils and other CXCL10⁺ myeloid cells in non-cancer DM, coupled with hypoxic tissue remodeling, points to a heightened innate immune response. A larger study with longitudinal cancer screening data is needed to determine if a heightened innate immune response in DM is correlated with early tumor elimination. A recent study⁴⁹ showed that increased diversity of immune response in TIF1γ-positive DM patients is associated with decreased risk of cancer emergence. These fundamentally different immune niches and molecular programs suggest distinct pathogenic mechanisms, challenging the notion that non-cancer DM simply represents resolved cancer-associated disease. In this context, it is noteworthy that, according to our single-cell analysis, keratinocytes from anti-TIF1γ-positive patients showed a nonsignificant trend toward a distinct transcriptional phenotype. This observation raises the possibility that tumor-derived circulating factors, such as cytokines, metabolites, or extracellular vesicles, could influence epithelial differentiation programs in distant tissues and potentially modulate keratinocyte phenotypes in TIF1γ-positive DM.

Moreover, in our analysis, we found that lesional immune infiltrates are also quite distinct between cancer and non-cancer DM. In addition to a significantly higher number of CXCL10⁺ myeloid cells, non-cancer DM demonstrated the appearance of unique cellular neighbors: PD-L1-expressing mregDCs and activated Tregs expressing NFKB2 and TNF receptors. In contrast, cancer-associated DM immune niches showed fewer Treg and lacked both PD-L1 expression and mregDC-Treg colocalization. Interestingly, it has been repeatedly shown in cancer research that this mregDC-Treg pair is associated with immunotherapy treatment failure^50,51. While the functional role of these interactions in autoimmune skin disease remains unclear, the observed colocalization of PD-L1-expressing mregDCs and activated Tregs may reflect an attempt to establish local immune regulation that is insufficient to fully suppress chronic inflammation ^52–54.

Our analysis of spatial single-cell data reveals new insights into the multiplicity and complexity of immune cell organization across different disease entities and disease subtypes, highlighting the potential of utilizing high-resolution spatial transcriptomics in reconstructing complex immune interactions in autoimmune skin diseases. Our study has several limitations. First, our DM cohort was predominantly composed of anti-TIF1γ-positive patients and those without detectable autoantibodies, potentially influencing our observed cancer/non-cancer associations. Second, despite observing consistent immune patterns across our DLE and SCLE samples, we recognize the inherent heterogeneity of CLE may require a larger patient cohort to fully characterize and stratify CLE.

Methods

Patient recruitment and tissue sample collection

Skin tissue for analysis was derived from three independent sources—Brigham and Women’s Hospital (BWH), University of Massachusetts Memorial Medical Center (UMASS), and BWH Pathology Archives, enhancing the external validity. Fresh skin tissue acquisition was obtained through patient recruitment and sampling performed in autoimmune skin disease clinics and general dermatology clinics at BWH and UMASS. Patients were recruited with active DM and CLE (all subtypes) to provide lesional and non-lesional skin samples, and healthy patients were included to provide control tissue. Written informed consent was obtained from all participants prior to any study procedures in accordance with institutional guidelines and the Declaration of Helsinki. The study protocol was approved by the Brigham and Women’s Hospital Institutional Review Board (IRB 2024P000430, PI: Ruth Ann Vleugels; IRB 2022P001168, PI: Neda Shahriari) and the UMass Chan Medical School Institutional Review Board (IRB H00021295, PI: Mehdi Rashighi). Regions of interest on the skin were marked in preparation for tissue acquisition. Then, 1% lidocaine with epinephrine was injected into the skin to provide local anesthesia. A 4-millimeter skin punch was utilized to core out skin tissue of interest and subsequently placed in formalin. Hemostasis was achieved through suture placement. The formalin-fixed samples were then embedded in paraffin for subsequent analysis. DM and CLE formalin-fixed paraffin-embedded (FFPE) skin tissue in the BWH Pathology Archives collected within the past 12 years were also included for analysis. Samples were localized using the search terms “dermatomyositis,” “lupus,” and “interface dermatitis.” Each case was analyzed by two board-certified dermatologists (N.S. and R.A.V.) to ensure patient histology was concordant with a clinical diagnosis of DM or CLE before inclusion.

For dermatomyositis patients, cancer-associated DM was defined as the presence of a confirmed malignancy diagnosed within 3 years before or after the onset of DM. Non-cancer-associated DM was defined as the absence of detectable malignancy for at least three years following DM onset. All patients underwent systematic malignancy screening according to institutional clinical protocols, including imaging studies, such as CT (chest, abdomen, and pelvis), colonoscopy, transvaginal ultrasound, mammogram and pap smear. Tumor blood markers and serum protein electrophoresis were also obtained. Patients were monitored with ongoing malignancy surveillance during the first three years following DM onset.

FFPE single-cell RNA sequencing

For each sample, four FFPE curls (25 μm each) were dissociated following the FLEX protocol (CG000632 RevA, 10x Genomics) with the Miltenyi Biotech FFPE Tissue Dissociation Kit. Cells from each sample were hybridized with a unique Probe Barcode, as per the instructions in the “Chromium Fixed RNA Profiling Reagent Kits for Multiplexed Samples” user guide (CG000527, 10x Genomics). Post-hybridization, cells from 16 samples were washed, counted, and pooled in equal proportions. Approximately 200,000 cells from the pooled suspension were loaded onto a Chromium Q chip (PN-1000422, 10x Genomics). Sequencing libraries were prepared and sequenced on an Illumina NovaSeq platform using paired-end dual-indexing (28 cycles for Read 1, 10 cycles for i7, 10 cycles for i5, and 90 cycles for Read 2).

Xenium slide preparation

FFPE blocks were sliced onto Xenium slides according to the “Xenium in Situ for FFPE-Tissue Preparation Guide” (CG000578 Rev C, 10x Genomics) protocol. The Xenium slides were then prepared following the “Xenium in Situ for FFPE-Deparaffinization and Decrosslinking” protocol (CG000580 Rev D, 10x Genomics). In brief, the slides were baked at 60 °C for 30 min and then sequentially immersed in xylene, ethanol, and nuclease-free water to deparaffinize and rehydrate the tissue. Immediately after, the Xenium slides were incubated in the decrosslinking and permeabilization solution at 80 °C for 30 min, followed by a wash with PBS-T.

Next, the slides were prepared according to the “Xenium Prime in Situ Gene Expression” user guide (CG000760 Rev A, 10x Genomics) for the remaining steps. Probes from the Xenium Prime 5K Human Pan Tissue & Pathways Panel (PN-1000671, 10x Genomics) and the custom add-on panel were hybridized to the samples at 50 °C for 18 h. Post-hybridization, the slides underwent washing, incubation with a ligation reaction mix, another wash step, and DNA amplification. Finally, cell segmentation staining was conducted, and the slides were treated with an autofluorescence quencher and DAPI before being loaded into the Xenium instrument.

Xenium analyzer setup and data acquisition

Processed Xenium slides were loaded in the Xenium Analyzer and imaged, following the guidelines in the “Xenium Analyzer User Guide (CG000584 Rev F, 10x Genomics).” After scanning, the Xenium slides were removed from the Xenium Analyzer and processed with post-run H&E staining according to the “Xenium in Situ Gene Expression—Post-Xenium Analyzer H&E Staining” protocol (CG000613 Rev A, 10x Genomics).

Single-cell and spatial single-cell data preprocessing

The sequencing data was demultiplexed using bcl2fastq (Illumina). The resulting FASTQ files were processed with Cell Ranger using the multi-pipeline and the GRCh38-2020-A reference genome.

Data analysis was performed using the Seurat⁵⁵ v5 R package. Quality control and preprocessing steps included doublet removal using scDblFinder⁵⁶ R package and ambient RNA correction using SoupX^55,57 R package for each sample individually. For downstream analysis, we utilized Seurat v5’s functions by splitting the data by sample, which created separate layers for each sample in a single Seurat object. This approach enabled sample-specific normalization, identification of 2000 variable features per sample, and scaling. We then performed PCA with 20 principal components and applied Harmony batch correction (lambda = 1) using the sample as covariate using harmony⁵⁸ R package. Using these batch-corrected embeddings, we constructed a shared nearest neighbor graph based on the first 20 Harmony dimensions and identified cell clusters with the Leiden algorithm (resolution 1.2). Cell type annotation was performed using established marker genes, resulting in the identification of 16 major cell states.

For spatial single-cell data, we followed a similar analytical pipeline but omitted the scDblFinder and SoupX processing steps. The only notable difference was in the normalization step, where we applied a scale factor of 5000 during normalization to account for the specific sequencing depth and transcript detection characteristics of spatial transcriptomics data.

Major and granular cell states identification

Major cell states were first manually classified into 16 major states based on expression of known literature markers^59–61: keratinocytes (KRT1, KRT5, KRT17, and FLG), fibroblasts (COL6A1, FBN1, PDGFRB, and PI16), eccrine glands (AQP5, AQP7), apocrine glands (KRT77, GABRP), sebaceous glands (MGST1, PLIN2, and FASN), melanocytes (MLANA, DST), myeloid cells (CD163, C1QA, and CD74), lymphoid cells (CD3E, CD8A, and CD4), B cells/plasma cells (CD19, MZB1), Schwann cells (MPZ, SCN7A), adipocytes (PLIN1, PLIN4, and ADIPOQ), mast cells (GATA2, KIT, and MS4A2), vascular endothelial cells (PECAM1, VWF), smooth muscle cells (MYH11, TAGLN), pericytes (RGS5, TAGLN), and lymphatic endothelial cells (FLT4, CCL21).

For fine-grained cell state annotation, we employed the same computational approach used for major cell states but applied it to each individual cell population separately. We maintained the same parameters for normalization, variable feature identification (1000 genes), dimensionality reduction (20 PCs), and Harmony batch correction. For clustering, we adjusted the Leiden algorithm resolution parameters (1.0–1.5) to better capture the finer distinctions within each major cell type. During this process, we identified cells in the spatial data expressing markers from multiple cell populations, which we manually classified as mis-segmented and removed to define cleaner cell states.

We identified multiple granular cell states within each major cell state based on their marker gene expression profiles. For clusters without established nomenclature, we assigned names according to one of their specifically expressed markers. Cluster-specific marker genes were determined using the presto R package with an area under the curve threshold of 0.6⁶².

Keratinocytes were classified into several states (Supplementary Fig. 1D): basal keratinocytes expressed KRT14, KRT15, and KRT5. E2F1_cyclic_basal cells showed additional expression of E2F1 with proliferation markers (MKI67, TOP2A, and FOXM1), while cyclic_basal cells expressed similar proliferation markers without E2F1. CXCL10_basal and CXCL10_spinous states expressed interferon-response genes (CXCL10, CXCL9, CXCL11, IFI6, IRF1, and IFI44L) at different differentiation stages. More differentiated states included spinous cells (KRT10, KRT1, and NOTCH3), granular cells (KRT2, KRT23), cornified cells (LOR, FLG, and AZGP1), and CALM3_granular cells characterized by calcium-binding proteins (CALM3, CALM5) alongside KRT2 and KRT23.

Fibroblast cell states (Supplementary Fig. 1J) included papillary fibroblasts (APCDD1, NKD2, COMP, and POSTN), vascular fibroblasts (C7, C3, VEGFA, F3, and HIF3A), and CCL19_vascular_fib with additional CCL19 expression. CXCL10_fib expressed interferon-response genes (CXCL10, CXCL9, CXCL11, MX1, OAS1, and IFIT2), while IFIT2_papillary_fib showed interferon-response genes (MX1, OAS1, and IFIT2) without CXCL10 along with papillary markers. Reticular fibroblasts expressed IGFBP6 and PI16, with a FOS_reticular subset (FOS, JUNB) identified specifically in single-cell data, consistent with dissociation-associated stress responses. Spatial transcriptomics exclusively revealed ASPN_fib (COCH, ASPN) and SFRP1_fib (SFRP1) cell states adjacent to sweat glands, as well as proliferative fibroblast (cyclic_fib) expressing MKI67 and TOP2A.

Myeloid cells comprised various macrophage subsets (Fig. 1A): those expressing C1QA, CD163, and MRC1; Mac2 (additional LYVE1, MARCO); C1Q_Mac (high C1QA, C1QB, and C1QC); CXCL10_Mac (chemokines CXCL9/10/11); and TREM2_Mac (TREM2, APOE, SPP1, FABP4, and FABP5). Dendritic cell populations included cDC2 (CD1C, CLEC10A, and FCER1A), cDC1 (CLEC9A, XCR1, and BATF3), mregDC (LAMP3, BIRC3), CXCL10_DC (CD1C, CLEC10A, and CXCL9/10/11) and Langerhans cells (CD1A, CD207, and FCGBP). Other myeloid populations included neutrophils (FCGR3B, CXCR2, S100A9, S100A8, and IL1B) and monocytes (CX3CR1, VCAN, S100A9, S100A8, CCR2, and SELL), CXCL8_DC (CXCL8). Notably, CXCL8_DC populations were identified specifically in spatial transcriptomic data. In single-cell data, proliferative myeloid cells were captured as a single cyclic_DC cluster, whereas spatial data enabled further resolution of these proliferative states into distinct subsets, including cyclic_cDC2, cyclic_cDC1, cyclic_macrophages, and cyclic_pDC.

Endothelial cells included tip cells (PGF, CXCR4), venous endothelial cells (CLU, VWF, and CCL14), arterial endothelial cells (SEMA3G, ELN, SULF1, and HEY1), and capillary endothelial subtypes (CD36, RGCC)—FABP4_capillary (FABP4) and HOXD10_capillary (HOXD10). Spatial data revealed DLL4-expressing variants (DLL4_venous, DLL4_lymphatic, DLL4_artery, and DLL4_FABP4_capillary) and interferon-responsive subtypes (CXCL10_venous, CXCL10_capillary). Lymphatic endothelial cells expressed FLT4, SEMA3A, and FOXC2, while cyclic vascular cells showed proliferation markers MKI67 and TOP2A.

Lymphocyte populations (Fig. 1A) included CD4⁺ T cells (CD4, CD40LG) with a FOS_CD4 expressing the FOS gene, T peripheral helper cells (Tph, CXCL13+), Tregs (FOXP3, TIGIT) with act_Treg (TNFRSF4, NFKB2, TNFRSF18, and NFKBIA) and FOS_Treg variants, gamma-delta T cells (LTK, SCART1), and cyclic_lymphocytes (MKI67, TOP2A). In single-cell data, proliferative lymphocytes were captured as a broader cyclic_lymphocyte population, and FOS-associated states (including FOS_CD4 and FOS_Treg) were identified. In contrast, spatial transcriptomics allowed more refined resolution of proliferative lymphocyte subsets, including cyclic_CD4, cyclic_CD8, and cyclic_Treg populations. Natural killer cells expressed GNLY, KLRD1, KLRC3, and KLRC4, while innate lymphoid cells showed Nk cells markers with CX3CR1 and KLRG1.

B cells (Fig. 1A) expressed CD19, plasma cells showed high MZB1 expression, and plasmacytoid dendritic cells were characterized by IRF7, LILRA4, TCF4, and GZMB expression. In spatial data, we additionally identified plasmablast populations expressing plasma cell markers together with proliferative genes (MKI67, TOP2A), consistent with actively cycling antibody-secreting cells.

Spatial analysis of cell states organization—compartment analysis

For global tissue organization and skin compartment identification, we employed the hoodscanR¹³ R package to analyze cellular neighborhoods based on spatial proximity. This approach involved a two-step process: first, we identified k = 100 nearest neighboring cells for each cell and calculated neighbor association probabilities; second, we merged these probabilities by cell type and assessed neighborhood colocalization patterns using Pearson correlation of probability distributions between different cell states. Based on these probability distributions, we applied k-means clustering with 6 centers to define distinct tissue compartments.

Cell composition analysis

To visualize differences in cellular composition between samples, we first calculated the percentage of each cell type per sample. We then used Bray–Curtis dissimilarity to measure how different samples were from each other based on their cell type proportions and visualized these relationships using principal coordinate analysis (PCoA) implemented in the labdsv R package. To assess significant differences in cell type composition across groups, we employed the sccomp¹² R package. Sccomp implements a Bayesian framework using sum-constrained independent beta-binomial distributions to analyze differential composition. Statistical significance was determined using two criteria: (1) a 95% credible interval of the slope, where positive and negative intervals indicated cluster enrichment or depletion, respectively and (2) an FDR threshold of 0.05. The model assessed compositional changes against a fold-change threshold of −0.1 to 0.1.

Sample stratification using multicellular factor analysis

To investigate transcriptional programs underlying cancer-associated DM, we employed MOFA, implemented in the MOFAcellulaR¹⁶ R package. MOFA is an unsupervised method that extends PCA by identifying latent factors that capture major sources of variation in gene expression across different cell types. For data preprocessing, we generated pseudo-bulk profiles by aggregating single-cell transcriptomes within each cell type and sample combination. We filtered cell states to include only those with at least 25 cells per sample (B cells, plasma cells, and pDCs, melanocytes) and excluded cell states with inconsistent representation across samples (sweat gland and hair follicular cells). Genes for MOFA analysis were filtered through multiple steps: first, retaining genes with a minimum count of 100 in at least 25% of samples. Data were then normalized using TMM with a scaling factor of 1,000,000. To minimize the impact of noisy mRNA background contamination, we implemented a strict filtering strategy for highly variable genes. For each cell type, we excluded any highly variable genes that were identified as markers for other cell types. Cell-type-specific markers were identified through a quasi-likelihood framework (FDR < 0.01, logFC > 1) implemented in the edgeR⁶³ R package. These filtered and normalized matrices were then converted to MOFA format using the pb_dat2MOFA function. We fit the MOFA model with eight factors using default parameters for all lesional samples and fit a separate MOFA model with six factors using default parameters for DM lesional samples only. Factor importance was quantified through variance explained (R2) across cell types. Associations between factors and clinical variables were assessed using the Kruskal–Wallis test for categorical variables (disease status, cancer, myopathy, gender, site of biopsy) and Pearson correlation for continuous variables (age, immune infiltration ratio), with Benjamini–Hochberg correction for multiple testing.

Identification and analysis of immune niches in lesional samples

To identify and characterize immune cell niches within the tissue, we implemented a multi-step computational approach. First, we performed spatial clustering using HDBSCAN (implemented in the dbscan R package) to identify immune cell niches. The clustering included both immune and endothelial cells since immune cells typically aggregate around blood vessels. Clusters containing fewer than 20 immune cells or those dominated by vascular cells (>90% of total cells) were excluded from further analysis. For samples with technical replicates on the same slide, we first separated the replicates using k-means clustering (k = 2) and then analyzed each replicate independently.

After identifying immune niches across all samples, we calculated the relative abundance of each immune cell state as a percentage of total immune cells within each niche. To classify distinct immune niche subtypes, we first computed Bray–Curtis dissimilarities between all niches based on their immune cell state composition. We then performed PCoA and used the first six components for UMAP dimensionality reduction. Finally, hierarchical clustering was applied to define immune niche subtypes. We excluded from niche analysis one sample, which on H&E visibly showed an absence of interfollicular keratinocytes and extensive neutrophil invasion nearby, morphological features consistent with acute wound response rather than chronic inflammatory pathology (patient_DM24).

To examine whether gene expression programs differ within disease-associated immune niches, we performed pseudobulk analysis by aggregating expression counts from all immune cells within the Dermis_Disease clusters. Prior to differential expression analysis, we filtered out lowly expressed genes (present in <10% of aggregates with counts ≥5) and subsequently conducted PCA analysis. To identify differentially expressed genes in disease-associated immune niches, we used MAST⁶⁴ implemented in the Seurat R package. To account for sample-specific effects, we included sample identity as a latent variable in the model. Genes were considered differentially expressed if they met the significance threshold of adjusted p-value < 0.1 and |log2 fold change| > 0.3.

Finding stroma associated with immune niches

After identifying immune niches, we determined their precise boundaries and associated stromal cells. For each immune aggregate, we delineated border cells using the identify_bordering_cells function from the SPIAT⁶⁵ R package, which implements an alpha-hull algorithm. The alpha parameter, which determines the hull shape precision, was dynamically adjusted based on niche size: alpha = 60 for niches with <200 cells, alpha = 90 for >5000 cells, and scaled linearly for intermediate sizes (alpha = (n_cells − 300)/160 + 60). We identified two stromal cell populations around each immune aggregate: stromal cells located within the aggregate boundary and those outside the aggregate within a 50-micrometer radius. To characterize stromal niches, we applied the same pipeline used for immune niche classification. To identify differentially expressed genes in stromal cells associated with immune aggregates, we used MAST⁶⁴ implemented in the Seurat R package. To account for sample-specific effects, we included sample identity as a latent variable in the model. Genes were considered differentially expressed if they met the significance threshold of adjusted p-value < 0.1 and |log2 fold change| > 0.4.

Cell interactions analysis

To analyze local cell-type interactions, we used the crawdad⁶⁶ R package, examining spatial relationships at multiple distances (5, 20, 50, 100, 250, 500, and 700 μm). Statistical significance of cell-type associations was determined using permutation tests (n = 3) with Bonferroni correction. Z-scores were calculated to measure the likelihood of cell-type co-localization, where higher Z-scores indicated stronger spatial associations between cell populations.

Multiplexed CODEX imaging with antibody-based protein detection

FFPE tissue slides were baked in an oven (VWR, 10055-006) at 70 °C for 1 h and thoroughly deparaffinized by immersing in xylenes for 2 × 5 min. The slides were then rehydrated by a series of graded solutions using a linear stainer (Leica Biosystems, ST4020), with each step proceeding for 3 min: 3× xylene, 2× 100% EtOH, 2× 95% EtOH, 1× 80% EtOH, 1× 70% EtOH, 3× UltraPure water (Invitrogen 10977-023), and finally left in UltraPure water (Invitrogen 10977-023). Unless otherwise specified, all washes mentioned below were with gentle agitation. Antigen retrieval was performed at 97 °C for 20 min with pH 9.0 Target Retrieval Solution (Agilent, S236784-2) using a PT Module (ThermoFisher, A80400012), after which the slides were cooled to room temperature on the benchtop. The slides were then equilibrated by washing in S2 Buffer (2.5 mM EDTA, 0.5× DPBS, 0.25% BSA, 0.02% NaN₃, 250 mM NaCl, 61 mM Na₂HPO₄, 39 mM NaH₂PO₄) for 20 min, then washed in 1× TBS-T for 5 min. Tissue regions were then circled with a hydrophobic barrier pen (Vector Laboratories, H-4000). Tissue slides were blocked with Blocking Buffer that contains BBDG (5% normal donkey serum, 0.05% NaN₃ in 1× TBS-T wash buffer (Sigma, 935B-09)) supplemented with 50 μg/ml mouse IgG (diluted from 1 mg/ml stock (Sigma, I5381-10mg) in S2), 50 μg/ml rat IgG (diluted from 1 mg/ml stock (Sigma, I4141-10mg) in S2), 500 μg/ml sheared salmon sperm DNA (ThermoFisher, AM9680) and 50 nM oligo block (diluted from stock with 500 nM of each oligo in 1× TE pH 8.0 (Invitrogen, AM9849). Blocking was performed in a humidity chamber on ice during which tissue slides were photobleached using Happy Lights (Verilux, VT22) for 1 h, with the temperature kept below 40 °C.

Antibodies used in the CODEX platform were obtained from their respective commercial sources and were conjugated either in-house by maleimide-based chemistry^67,68 or through collaboration with Cell Signaling Technologies (CST). Each conjugated antibody in the panel was first centrifuged at 12,500 rpm for 8 min at 4 °C to remove aggregates. They were then diluted in a single tube containing 300 μl Antibody Diluent (5% normal donkey serum, 0.05% NaN₃ in 1× TBS-T) at a concentration based on prior titration. The antibody mixture was then filtered through a S2 Buffer pre-wetted 50 kDa filter (Sigma Millipore, UFC5050BK) at 12,500 rpm for 8 min at 4 °C. The filtered mixture was adjusted to the desired final volume by adding 25% v/v total volume with FFPE Block (50 μg/ml mouse IgG (Sigma, I5381-10 mg), 50 μg/ml rat IgG (Sigma, I4141-10 mg), 500 μg/ml sheared salmon sperm DNA (ThermoFisher, AM9680), 50 nM oligo block (diluted from stock with 500 nM of each oligo in 1× TE pH 8.0 (Invitrogen, AM9849) in S2) and topping up the remainder with Antibody Diluent. The antibody mixture was then filtered through a S2 Buffer pre-wetted 0.1 μm pore size filter (Millipore, UFC30VV00) at 12,500 rpm for 1 min at 4 °C. Blocked/photobleached tissues were then incubated with the filtered antibody mixture at 4 °C in a humidity chamber overnight.

After incubation, tissues were washed twice in S2 buffer for 5 min each at RT. The slides were then fixed with 1.6% PFA (diluted from 16% stock (EMS Diasum, 15740-04) in 1× PBS (diluted from 10× stock) in a humidity chamber protected from light, at RT for 10 min. The slides were then rinsed quickly with 1× PBS twice and washed once with 1× PBS for 2 min. The slides were fixed with ice-cold methanol for 5 min on ice, followed by rinsing with 1× PBS twice and washed once with 1× PBS for 2 min. The slides were then fixed in 4 μg/μl of BS3 Final Fixative (diluted from 200 μg/μl stock (ThermoFisher, 21580) in 1× PBS) twice for 15 min each at RT protected from light. The slides were washed once in 1× CODEX Buffer (10 mM Tris pH 7.5, 0.02% NaN₃, 0.1% Triton X-100, 10 mM MgCl₂-6H₂O, 150 mM NaCl). A cotton-tipped applicator was used to scrape off the hydrophobic barrier on the slides, and the slides were stored in 1× CODEX Buffer. Details regarding the antibody clones, vendors, conjugated channels, titers, exposure times, and assigned channels throughout the study are in Supplementary Table 1.

CODEX imaging was performed via the automated PhenoCycler Fusion platform (Akoya Biosciences). Each tissue slide was mounted with a flow cell (Akoya Biosciences, 240205) and was incubated in 1× CODEX Buffer for 10 min at RT. A reporter plate was prepared for each tissue slide, with each well corresponding to each imaging cycle. Briefly, each well of a 96-well black reporter plate (BRAND Tech, 781607) was added with Plate Buffer (500 μg/ml sheared salmon sperm DNA in 1× CODEX Buffer) supplemented with 54.11 mM Hoechst 33342 (ThermoFisher, H3570) and 100 nM each of ATTO550 or AlexaFluor647-conjugated lectin/antibody corresponding complementary reporter oligos (GenScript, HPLC purified). The wells were then sealed using aluminum plate seal (ThermoFisher, AB0626) and stored at 4 °C until use. Low DMSO (80% 1× CODEX Buffer, 20% DMSO) and High DMSO (10% 1× CODEX Buffer, 90% DMSO) buffers were prepared fresh by mixing 1× CODEX Buffer with DMSO (Sigma, 472301-4L). After imaging, the slides were stored in 1× PBS.

Quantification and analysis of CODEX data

Single-cell segmentation of CODEX images was performed using Cellpose⁶⁹ with the cyto3 model. Following segmentation, protein fluorescence intensity was quantified for each segmented cell using the SPACEc pipeline⁷⁰, generating a cell-by-marker expression matrix representing protein abundance across individual cells. Downstream analysis was conducted using Seurat. Each slide was internally normalized using Seurat’s implementation of centered log-ratio normalization, applied independently to each feature (margin = 1). The data were subsequently scaled prior to dimensionality reduction. PCA was performed using all measured protein markers, and the first eight principal components were retained for downstream analysis. To correct for batch effects across slides, Harmony integration was applied using slide identity as the batch variable. The resulting Harmony embeddings were used to construct a nearest-neighbor graph, followed by UMAP dimensionality reduction for visualization and Louvain clustering for unsupervised identification of cell populations. After the initial clustering, clusters lacking clear marker expression or representing low-confidence populations were removed. The analysis pipeline (PCA, Harmony integration, and clustering) was then repeated on the filtered dataset, resulting in the identification of nine major cell populations, which were annotated based on canonical protein marker expression.

Spatial correlation gene expression analysis

To identify gene expression patterns within immune aggregates and their surrounding 50-micrometer border regions, we employed the InSituCor⁷¹ R package. For each cell, we generated an environment expression profile by averaging gene expression across its 50 nearest neighbors. To control for potential confounding factors, such as cell type abundance and transcript count, we calculated a conditional correlation matrix from residuals after regressing the environment expression against these confounders.

Quantification of endothelial cell density in standardized tissue regions

To define standardized regions for endothelial cell density analysis, we developed a computational pipeline that creates consistent measurement zones extending from the epidermal border into the dermis. Using basal keratinocyte coordinates, we applied a nearest-neighbor algorithm to create an ordered path through these points, approximating the epidermal-dermal junction. A cubic spline smoothing function generated a smooth reference curve representing the tissue boundary. From this curve, we constructed analysis polygons by creating a parallel boundary shifted 600 μm into the dermis, with shift direction determined by the vector from the basal keratinocyte centroid toward the tissue centroid. Within each polygon, endothelial cell density was calculated by normalizing cell counts by polygon area using the concave hull method.

Pathway and gene signature analysis

We used the decoupleR⁷² R package to estimate pathway activities across studied conditions. Pathway scores were generated using the multilevel model with the top 500 target genes per pathway. For interferon response analysis, we quantified type I and II interferon signatures from the Reactome database using the UCell⁷³ R package, which calculates signature enrichment in individual cells through a rank-based approach.

Fibroblast сytokines and hypoxia treatment

Human dermal fibroblasts (gift from Dr. Erin Theisen, MD, PhD, Brigham and Women’s Hospital, Boston, MA) were seeded at a density of 2 × 10⁴ cells/well in quadruplicate in two sets of 96-well plates. Cells were cultured in complete media (5% fetal bovine serum, HEPES, MEM amino acids, L-glutamine, penicillin-streptomycin, nonessential MEM amino acids, 2-mercaptoethanol, and gentamicin). Plates were incubated overnight at 37 °C in a 5% CO₂ incubator. The next day, the media was removed, and cells were treated as follows: complete condition media (control), complete media with IL1-β (1 ng/mL), complete media with IFN-β (5 ng/mL), or complete media with IFN-γ (5 ng/mL). One set of plates was transferred to a hypoxia chamber maintained at 0.5% O₂, while the other set was maintained under normoxic conditions. After a 24-h incubation period, RNA was extracted using TRIzol reagent (Invitrogen, cat. # 15596026) according to the manufacturer’s protocol. cDNA synthesis was performed using the QuantiTect Reverse Transcription kit (Qiagen, #205314). Quantitative PCR (qPCR) was carried out using the Brilliant III Ultra-Fast SYBR Green qPCR master mix (Agilent Technologies, #600883) on an Agilent AriaMx Real-Time PCR system. Human primer sequences were obtained from Origene Technology, Inc (Rockville, MD) and purchased from Integrated DNA Technologies, Inc (Coralville, IA).

Flow сytometry apoptosis assay

Dermal fibroblasts at passage 6 were seeded at 5 × 10⁴ cells/well in 48-well plates in 5% serum medium (normoxia: n = 6 wells; hypoxia: n = 3 wells) and incubated overnight under respective conditions. The following day, cells were prepared for the Annexin V assay. Dead cells were generated by heating trypsinized normoxic cells at 56 °C for 20 min and subsequently mixed 1:1 with normoxic live cells. Four experimental groups were analyzed: unstained controls, dead/live cell mixture, normoxic cells, and hypoxic cells. Staining was performed according to the manufacturer’s protocol. Briefly, cells were trypsinized, transferred to V-bottom 96-well plates, washed with cell staining buffer, and resuspended in Annexin V-APC staining solution for 15 min at room temperature in the dark. Following a wash with a binding buffer, cells were incubated with SYTOX Green for 15 min at room temperature in the dark. Samples were then transferred to FACS tubes containing 200 μL binding buffer and analyzed immediately by flow cytometry. Unstained controls received binding buffer alone in place of dyes.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(17.8MB, pdf)}

41467_2026_74871_MOESM2_ESM.pdf^{(63.5KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(66.2KB, xlsx)}

Reporting Summary^{(92.7KB, pdf)}

Transparent Peer Review File^{(1.8MB, pdf)}

Source data

Source Data^{(1.7MB, zip)}

Acknowledgements

We thank members of the Wei lab, Rashighi lab, and Korsunsky lab for helpful discussions. We thank members of the 10x Genomics team, including Rachel Gate, Steve Kujawa, Navid Farahani, Anne Monahan, and Adam Semple, for their time and generous support of this project.

Author contributions

Conceptualization: K.S.A., N.S., M.R., R.A.V., I.K., and K.W. Funding acquisition: K.W. and N.S. Single-cell and spatial single-cell transcriptomics data generation: C.G., J.L., S.A.P., and K.W. CODEX imaging data generation: S.P.T.Y., L.P., and S.J. Skin biopsy acquisition and patient management: N.S., R.L.C., K.A., E.T., T.B., A.L., W.J.C., and K.H. Data analysis: K.S.A. Validation: S.K., K.B. Figure and schematic generation: K.S.A. and A.N.K. Writing-original draft: K.S.A. and K.W. Writing-review and editing: K.S.A., N.S., M.R., R.A.V., K.W., and all other authors. Supervision: K.W. All authors have read and agreed to the published version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Ana-Maria Lennon-Dumenil and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by a Brigham and Women’s Hospital Department of Medicine-Broad Institution collaborative research award and a Dermatology Foundation Patient-Directed Investigation Grant (N.S.). K.W. is supported by NIH-NIAMS K08AR077037, a Doris Duke Foundation Clinical Scientist Development Award, and a Burroughs Wellcome Fund Career Award for Medical Scientists. M.R. is supported by NIH-NIAID 5U01AI176310 and NIH-NIAID 5U01AI168640. R.L.C. is supported by the Dermatology Foundation Women’s Health Career Development Award. S.P.T.Y. is a MacMillan Family Foundation Awardee of the Life Sciences Research Foundation. We thank the Harvard Skin Disease Research Center, supported by NIAMS P30AR069625, for providing normal skin samples used in this study.

Data availability

Raw single-cell RNA-seq data have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE334588. The single-cell RNA-sequencing and spatial transcriptomics objects generated and analyzed during the current study are also available in the Zenodo repository at https://zenodo.org/records/20386623. Raw output files from Xenium Prime 5k spatial transcriptomics are available at https://zenodo.org/records/20388043. The multiplexed CODEX imaging dataset is available in the Zenodo repository at https://zenodo.org/records/18246178. Source data are provided with this paper.

Competing interests

Reagents for single-cell and spatial transcriptomes were provided by 10x Genomics as part of a sponsored research agreement with K.W. The remaining authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Ksenia S. Anufrieva, Neda Shahriari.

These authors jointly supervised this work: Mehdi Rashighi, Ruth Ann Vleugels, Kevin Wei.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-74871-7.

References

1.Lundberg, I. E. et al. Idiopathic inflammatory myopathies. Nat. Rev. Dis. Primers7, 86 (2021). [DOI] [PubMed] [Google Scholar]
2.Mainetti, C., Terziroli Beretta-Piccoli, B. & Selmi, C. Cutaneous manifestations of dermatomyositis: a comprehensive review. Clin. Rev. Allergy Immunol.53, 337–356 (2017). [DOI] [PubMed] [Google Scholar]
3.Sigurgeirsson, B., Lindelöf, B., Edhag, O. & Allander, E. Risk of cancer in patients with dermatomyositis or polymyositis. a population-based study. New Engl. J. Med.326, 363–367 (1992). [DOI] [PubMed] [Google Scholar]
4.Airio, A., Pukkala, E. & Isomäki, H. Elevated cancer incidence in patients with dermatomyositis: a population based study. J. Rheumatol.22, 1300–1303 (1995). [PubMed] [Google Scholar]
5.Qiang, J. K., Kim, W. B., Baibergenova, A. & Alhusayen, R. Risk of malignancy in dermatomyositis and polymyositis. J. Cutan. Med. Surg.21, 131–136 (2017). [DOI] [PubMed] [Google Scholar]
6.Hu, T. & Vinik, O. Dermatomyositis and malignancy. Can. Fam. Physician65, 409 (2019). [PMC free article] [PubMed] [Google Scholar]
7.Kang, E. H. et al. Temporal relationship between cancer and myositis identifies two distinctive subgroups of cancers: impact on cancer risk and survival in patients with myositis. Rheumatology55, 1631–1641 (2016). [DOI] [PubMed] [Google Scholar]
8.Ogawa-Momohara, M. et al. Strong correlation between cancer progression and anti-transcription intermediary factor 1γ antibodies in dermatomyositis patients. Clin. Exp. Rheumatol.36, 990–995 (2018). [PubMed] [Google Scholar]
9.Aussy, A., Boyer, O. & Cordel, N. Dermatomyositis and immune-mediated necrotizing myopathies: a window on autoimmunity and cancer. Front. Immunol.8, 992 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.He, S. et al. High-plex multiomic analysis in FFPE at subcellular level by spatial molecular imaging. Nat. Biotechnol.40, 1794–1806 (2022). [DOI] [PubMed] [Google Scholar]
11.Ma, F. et al. Systems-based identification of the Hippo pathway for promoting fibrotic mesenchymal differentiation in systemic sclerosis. Nat. Commun.15, 210 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Mangiola, S. et al. sccomp: robust differential composition and variability analysis for single-cell data. Proc. Natl. Acad. Sci. USA120, e2203828120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Liu, N. et al. hoodscanR: profiling single-cell neighborhoods in spatial transcriptomics data. Preprint at bioRxiv10.1101/2024.03.26.586902 (2024). [DOI] [PMC free article] [PubMed]
14.Magro, C. M., Segal, J. P., Crowson, A. N. & Chadwick, P. The phenotypic profile of dermatomyositis and lupus erythematosus: a comparative analysis. J. Cutan. Pathol.37, 659–671 (2010). [DOI] [PubMed] [Google Scholar]
15.Billi, A. C. et al. Nonlesional lupus skin contributes to inflammatory education of myeloid cells and primes for cutaneous inflammation. Sci. Transl. Med.14, eabn2263 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Ramirez Flores, R. O., Lanzer, J. D., Dimitrov, D., Velten, B. & Saez-Rodriguez, J. Multicellular factor analysis of single-cell data for a tissue-centric understanding of disease. Elife12, e93161 (2023). [DOI] [PMC free article] [PubMed]
17.Shoffner-Beck, S. K. et al. Lupus dermal fibroblasts are proinflammatory and exhibit a profibrotic phenotype in scarring skin disease. JCI Insight9, e173437 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Chen, R.-Y. et al. The role of PD-1 signaling in health and immune-related diseases. Front. Immunol.14, 1163633 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Huard, C. et al. Correlation of cutaneous disease activity with type 1 interferon gene signature and interferon β in dermatomyositis. Br. J. Dermatol.176, 1224–1230 (2017). [DOI] [PubMed] [Google Scholar]
20.Fiorentino, D. et al. Efficacy, safety, and target engagement of dazukibart, an IFNβ specific monoclonal antibody, in adults with dermatomyositis: a multicentre, double-blind, randomised, placebo-controlled, phase 2 trial. Lancet405, 137–146 (2025). [DOI] [PubMed] [Google Scholar]
21.Kim, T. W. et al. A critical role for IRAK4 kinase activity in toll-like receptor-mediated innate immunity. J. Exp. Med.204, 1025–1036 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Brown, G. J. et al. TLR7 gain-of-function genetic variation causes human lupus. Nature605, 349–356 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Zhang, N., Lei, T., Xu, T., Zou, X. & Wang, Z. Long noncoding RNA SNHG15: a promising target in human cancers. Front. Oncol.13, 1108564 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Chen, C., Feng, Y., Wang, J., Liang, Y. & Zou, W. Long non-coding RNA SNHG15 in various cancers: a meta and bioinformatic analysis. BMC Cancer20, 1156 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Dunlap, G. S. et al. Single-cell transcriptomics reveals distinct effector profiles of infiltrating T cells in lupus skin and kidney. JCI Insight7, e156341 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Tsoi, L. C. et al. IL18-containing 5-gene signature distinguishes histologically identical dermatomyositis and lupus erythematosus skin lesions. JCI Insight5, e139558 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Turnier, J. L. et al. Imaging mass cytometry reveals predominant innate immune signature and endothelial-immune cell interaction in juvenile myositis compared to lupus skin. Arthritis Rheumatol.74, 2024–2031 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Patel, J., Maddukuri, S., Li, Y., Bax, C. & Werth, V. P. Highly multiplexed mass cytometry identifies the immunophenotype in the skin of dermatomyositis. J. Investig. Dermatol.141, 2151–2160 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Patel, J. et al. Multidimensional immune profiling of cutaneous lupus erythematosus in vivo stratified by patient response to antimalarials. Arthritis Rheumatol.74, 1687–1698 (2022). [DOI] [PubMed] [Google Scholar]
30.Ogawa-Momohara, M. et al. Multiplexed mass cytometry of cutaneous lupus erythematosus and dermatomyositis skin: an in-depth, B-cell-directed immunoprofile. J. Investig. Dermatol.145, 190–193 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Turnier, J. L. et al. Comparison of lesional juvenile myositis and lupus skin reveals overlapping yet unique disease pathophysiology. Arthritis Rheumatol.73, 1062–1072 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Yokogawa, M. et al. Epicutaneous application of toll-like receptor 7 agonists leads to systemic autoimmunity in wild-type mice: a new model of systemic lupus erythematosus. Arthritis Rheumatol.66, 694–706 (2014). [DOI] [PubMed] [Google Scholar]
33.Xu, H. & Qian, J. Vasculopathy in dermatomyositis. Chin. Med. J.137, 247–249 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Ibrahim, S. I., Mansour, H. E., Azrag, G. A. & Hussein, S. A. E. Nailfold capillaroscopic changes in dermatomyositis and polymyositis Egyptian patients: relation to disease activity. Egypt. J. Hosp. Med.81, 1292–1298 (2020). [Google Scholar]
35.Mugii, N. et al. Association between nail-fold capillary findings and disease activity in dermatomyositis. Rheumatology50, 1091–1098 (2011). [DOI] [PubMed]
36.Johnson, D. et al. Nailfold capillaroscopy abnormalities correlate with disease activity in adult dermatomyositis. Front. Med.8, 708432 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Mumtaz, S. et al. Microvascular abnormalities between anti-TIF1-γ-associated dermatomyositis with and without malignancy. BMC Rheumatol.9, 50 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Peng, Y., Zhang, S., Zhao, Y., Liu, Y. & Yan, B. Neutrophil extracellular traps may contribute to interstitial lung disease associated with anti-MDA5 autoantibody positive dermatomyositis. Clin. Rheumatol.37, 107–115 (2018). [DOI] [PubMed] [Google Scholar]
39.Seto, N. et al. Neutrophil dysregulation is pathogenic in idiopathic inflammatory myopathies. JCI Insight5, e134189 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Tsioumpekou, M., Krijgsman, D., Leusen, J. H. W. & Olofsen, P. A. The role of cytokines in neutrophil development, tissue homing, function and plasticity in health and disease. Cells12, 1981 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Li, T. M. et al. The interferon-rich skin environment regulates Langerhans cell ADAM17 to promote photosensitivity in lupus. Elife13, e85914 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Girard-Madoux, M. J. H., Kel, J. M., Reizis, B. & Clausen, B. E. IL-10 controls dendritic cell-induced T-cell reactivation in the skin to limit contact hypersensitivity. J. Allergy Clin. Immunol.129, 143–50.e1–10 (2012). [DOI] [PubMed] [Google Scholar]
43.Chow, W. H. et al. Cancer risk following polymyositis and dermatomyositis: a nationwide cohort study in Denmark. Cancer Causes Control6, 9–13 (1995). [DOI] [PubMed] [Google Scholar]
44.Oldroyd, A. G. S. et al. International guideline for idiopathic inflammatory myopathy-associated cancer screening: an International Myositis Assessment and Clinical Studies Group (IMACS) initiative. Nat. Rev. Rheumatol.19, 805–817 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Fiorentino, D. F. & Casciola-Rosen, L. Autoantibodies and cancer association: the case of systemic sclerosis and dermatomyositis. Clin. Rev. Allergy Immunol.63, 330–341 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Ma, J. et al. A blueprint for tumor-infiltrating B cells across human cancers. Science384, eadj4857 (2024). [DOI] [PubMed] [Google Scholar]
47.Kroeger, D. R., Milne, K. & Nelson, B. H. Tumor-infiltrating plasma cells are associated with tertiary lymphoid structures, cytolytic T-cell responses, and superior prognosis in ovarian cancer. Clin. Cancer Res.22, 3005–3015 (2016). [DOI] [PubMed] [Google Scholar]
48.Fiorentino, D. & Casciola-Rosen, L. Autoantibodies to transcription intermediary factor 1 in dermatomyositis shed insight into the cancer-myositis connection. Arthritis Rheumatol.64, 346–349 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Fiorentino, D. F. et al. Immune responses to CCAR1 and other dermatomyositis autoantigens are associated with attenuated cancer emergence. J. Clin. Investig.132, e150201 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.You, S. et al. Lymphatic-localized Treg-mregDC crosstalk limits antigen trafficking and restrains anti-tumor immunity. Cancer Cell42, 1415–1433 (2024). [DOI] [PubMed] [Google Scholar]
51.Poschel, D. B. et al. PD-L1 restrains PD-1Nrp1 Treg cells to suppress inflammation-driven colorectal tumorigenesis. Cell Rep.43, 114819 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Strauss, L. et al. Targeted deletion of PD-1 in myeloid cells induces antitumor immunity. Sci. Immunol.5, eaay1863 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Gordon, S. R. et al. PD-1 expression by tumour-associated macrophages inhibits phagocytosis and tumour immunity. Nature545, 495–499 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Perry, J. A. et al. PD-L1-PD-1 interactions limit effector regulatory T cell populations at homeostasis and during infection. Nat. Immunol.23, 743–756 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol.42, 293–304 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Germain, P.-L., Lun, A., Garcia Meixide, C., Macnair, W. & Robinson, M. D. Doublet identification in single-cell sequencing data using. F1000Research10, 979 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Young, M. D. & Behjati, S. SoupX removes ambient RNA contamination from droplet-based single-cell RNA sequencing data. Gigascience9, giaa151 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods16, 1289–1296 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Almet, A. A. et al. A roadmap for a consensus human skin cell atlas and single-cell data standardization. J. Investig. Dermatol.143, 1667–1677 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Fiskin, E. et al. Multi-modal skin atlas identifies a multicellular immune-stromal community associated with altered cornification and specific T cell expansion in atopic dermatitis. Nat. Commun.17, 3194 (2026). [DOI] [PMC free article] [PubMed]
61.Joost, S. et al. Single-cell transcriptomics reveals that differentiation and spatial signatures shape epidermal and hair follicle heterogeneity. Cell Syst.3, 221–237.e9 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Korsunsky, I., Nathan, A., Millard, N. & Raychaudhuri, S. Presto scales Wilcoxon and auROC analyses to millions of observations. Preprint at bioRxiv10.1101/653253 (2019).
63.Chen, Y., Chen, L., Lun, A. T. L., Baldoni, P. L. & Smyth, G. K. edgeR v4: powerful differential analysis of sequencing data with expanded functionality and improved support for small counts and larger datasets. Nucleic Acids Res.53, gkaf018 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol.16, 1–13 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Feng, Y. et al. Spatial analysis with SPIAT and spaSim to characterize and simulate tissue microenvironments. Nat. Commun.14, 2697 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Dos Santos Peixoto, R. et al. Characterizing cell-type spatial relationships across length scales in spatially resolved omics data. Nat. Commun.16, 350 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Black, S. et al. CODEX multiplexed tissue imaging with DNA-conjugated antibodies. Nat. Protoc.16, 3802–3835 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Yiu, S. P. T. et al. Same-Slide Spatial Multi-Omics Integration with IN-DEPTH Reveals Tumor Virus-Linked Spatial Reorganization of the Tumor Microenvironment. Cancer Discov. 10.1158/2159-8290.CD-25-0775 (2026). [DOI] [PMC free article] [PubMed]
69.Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm for cellular segmentation. Nat. Methods18, 100–106 (2021). [DOI] [PubMed] [Google Scholar]
70.Tan, Y. et al. SPACEc: a streamlined, interactive Python workflow for multiplexed image processing and analysis. Nat. Commun.16, 10652 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Danaher, P. et al. InSituCor: exploring spatially correlated genes conditional on the cell type landscape. Genome Biol. 26, 105 (2025). [DOI] [PMC free article] [PubMed]
72.Badia-I-Mompel, P. et al. decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinform. Adv.2, vbac016 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Andreatta, M. & Carmona, S. J. UCell: robust and scalable single-cell gene signature scoring. Comput. Struct. Biotechnol. J.19, 3796–3798 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(17.8MB, pdf)}

41467_2026_74871_MOESM2_ESM.pdf^{(63.5KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(66.2KB, xlsx)}

Reporting Summary^{(92.7KB, pdf)}

Transparent Peer Review File^{(1.8MB, pdf)}

Source Data^{(1.7MB, zip)}

Data Availability Statement

[CR1] 1.Lundberg, I. E. et al. Idiopathic inflammatory myopathies. Nat. Rev. Dis. Primers7, 86 (2021). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Mainetti, C., Terziroli Beretta-Piccoli, B. & Selmi, C. Cutaneous manifestations of dermatomyositis: a comprehensive review. Clin. Rev. Allergy Immunol.53, 337–356 (2017). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Sigurgeirsson, B., Lindelöf, B., Edhag, O. & Allander, E. Risk of cancer in patients with dermatomyositis or polymyositis. a population-based study. New Engl. J. Med.326, 363–367 (1992). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Airio, A., Pukkala, E. & Isomäki, H. Elevated cancer incidence in patients with dermatomyositis: a population based study. J. Rheumatol.22, 1300–1303 (1995). [PubMed] [Google Scholar]

[CR5] 5.Qiang, J. K., Kim, W. B., Baibergenova, A. & Alhusayen, R. Risk of malignancy in dermatomyositis and polymyositis. J. Cutan. Med. Surg.21, 131–136 (2017). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Hu, T. & Vinik, O. Dermatomyositis and malignancy. Can. Fam. Physician65, 409 (2019). [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Kang, E. H. et al. Temporal relationship between cancer and myositis identifies two distinctive subgroups of cancers: impact on cancer risk and survival in patients with myositis. Rheumatology55, 1631–1641 (2016). [DOI] [PubMed] [Google Scholar]

[CR8] 8.Ogawa-Momohara, M. et al. Strong correlation between cancer progression and anti-transcription intermediary factor 1γ antibodies in dermatomyositis patients. Clin. Exp. Rheumatol.36, 990–995 (2018). [PubMed] [Google Scholar]

[CR9] 9.Aussy, A., Boyer, O. & Cordel, N. Dermatomyositis and immune-mediated necrotizing myopathies: a window on autoimmunity and cancer. Front. Immunol.8, 992 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.He, S. et al. High-plex multiomic analysis in FFPE at subcellular level by spatial molecular imaging. Nat. Biotechnol.40, 1794–1806 (2022). [DOI] [PubMed] [Google Scholar]

[CR11] 11.Ma, F. et al. Systems-based identification of the Hippo pathway for promoting fibrotic mesenchymal differentiation in systemic sclerosis. Nat. Commun.15, 210 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Mangiola, S. et al. sccomp: robust differential composition and variability analysis for single-cell data. Proc. Natl. Acad. Sci. USA120, e2203828120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Liu, N. et al. hoodscanR: profiling single-cell neighborhoods in spatial transcriptomics data. Preprint at bioRxiv10.1101/2024.03.26.586902 (2024). [DOI] [PMC free article] [PubMed]

[CR14] 14.Magro, C. M., Segal, J. P., Crowson, A. N. & Chadwick, P. The phenotypic profile of dermatomyositis and lupus erythematosus: a comparative analysis. J. Cutan. Pathol.37, 659–671 (2010). [DOI] [PubMed] [Google Scholar]

[CR15] 15.Billi, A. C. et al. Nonlesional lupus skin contributes to inflammatory education of myeloid cells and primes for cutaneous inflammation. Sci. Transl. Med.14, eabn2263 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Ramirez Flores, R. O., Lanzer, J. D., Dimitrov, D., Velten, B. & Saez-Rodriguez, J. Multicellular factor analysis of single-cell data for a tissue-centric understanding of disease. Elife12, e93161 (2023). [DOI] [PMC free article] [PubMed]

[CR17] 17.Shoffner-Beck, S. K. et al. Lupus dermal fibroblasts are proinflammatory and exhibit a profibrotic phenotype in scarring skin disease. JCI Insight9, e173437 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Chen, R.-Y. et al. The role of PD-1 signaling in health and immune-related diseases. Front. Immunol.14, 1163633 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Huard, C. et al. Correlation of cutaneous disease activity with type 1 interferon gene signature and interferon β in dermatomyositis. Br. J. Dermatol.176, 1224–1230 (2017). [DOI] [PubMed] [Google Scholar]

[CR20] 20.Fiorentino, D. et al. Efficacy, safety, and target engagement of dazukibart, an IFNβ specific monoclonal antibody, in adults with dermatomyositis: a multicentre, double-blind, randomised, placebo-controlled, phase 2 trial. Lancet405, 137–146 (2025). [DOI] [PubMed] [Google Scholar]

[CR21] 21.Kim, T. W. et al. A critical role for IRAK4 kinase activity in toll-like receptor-mediated innate immunity. J. Exp. Med.204, 1025–1036 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Brown, G. J. et al. TLR7 gain-of-function genetic variation causes human lupus. Nature605, 349–356 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Zhang, N., Lei, T., Xu, T., Zou, X. & Wang, Z. Long noncoding RNA SNHG15: a promising target in human cancers. Front. Oncol.13, 1108564 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Chen, C., Feng, Y., Wang, J., Liang, Y. & Zou, W. Long non-coding RNA SNHG15 in various cancers: a meta and bioinformatic analysis. BMC Cancer20, 1156 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Dunlap, G. S. et al. Single-cell transcriptomics reveals distinct effector profiles of infiltrating T cells in lupus skin and kidney. JCI Insight7, e156341 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Tsoi, L. C. et al. IL18-containing 5-gene signature distinguishes histologically identical dermatomyositis and lupus erythematosus skin lesions. JCI Insight5, e139558 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Turnier, J. L. et al. Imaging mass cytometry reveals predominant innate immune signature and endothelial-immune cell interaction in juvenile myositis compared to lupus skin. Arthritis Rheumatol.74, 2024–2031 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Patel, J., Maddukuri, S., Li, Y., Bax, C. & Werth, V. P. Highly multiplexed mass cytometry identifies the immunophenotype in the skin of dermatomyositis. J. Investig. Dermatol.141, 2151–2160 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Patel, J. et al. Multidimensional immune profiling of cutaneous lupus erythematosus in vivo stratified by patient response to antimalarials. Arthritis Rheumatol.74, 1687–1698 (2022). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Ogawa-Momohara, M. et al. Multiplexed mass cytometry of cutaneous lupus erythematosus and dermatomyositis skin: an in-depth, B-cell-directed immunoprofile. J. Investig. Dermatol.145, 190–193 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Turnier, J. L. et al. Comparison of lesional juvenile myositis and lupus skin reveals overlapping yet unique disease pathophysiology. Arthritis Rheumatol.73, 1062–1072 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR32] 32.Yokogawa, M. et al. Epicutaneous application of toll-like receptor 7 agonists leads to systemic autoimmunity in wild-type mice: a new model of systemic lupus erythematosus. Arthritis Rheumatol.66, 694–706 (2014). [DOI] [PubMed] [Google Scholar]

[CR33] 33.Xu, H. & Qian, J. Vasculopathy in dermatomyositis. Chin. Med. J.137, 247–249 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Ibrahim, S. I., Mansour, H. E., Azrag, G. A. & Hussein, S. A. E. Nailfold capillaroscopic changes in dermatomyositis and polymyositis Egyptian patients: relation to disease activity. Egypt. J. Hosp. Med.81, 1292–1298 (2020). [Google Scholar]

[CR35] 35.Mugii, N. et al. Association between nail-fold capillary findings and disease activity in dermatomyositis. Rheumatology50, 1091–1098 (2011). [DOI] [PubMed]

[CR36] 36.Johnson, D. et al. Nailfold capillaroscopy abnormalities correlate with disease activity in adult dermatomyositis. Front. Med.8, 708432 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Mumtaz, S. et al. Microvascular abnormalities between anti-TIF1-γ-associated dermatomyositis with and without malignancy. BMC Rheumatol.9, 50 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Peng, Y., Zhang, S., Zhao, Y., Liu, Y. & Yan, B. Neutrophil extracellular traps may contribute to interstitial lung disease associated with anti-MDA5 autoantibody positive dermatomyositis. Clin. Rheumatol.37, 107–115 (2018). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Seto, N. et al. Neutrophil dysregulation is pathogenic in idiopathic inflammatory myopathies. JCI Insight5, e134189 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Tsioumpekou, M., Krijgsman, D., Leusen, J. H. W. & Olofsen, P. A. The role of cytokines in neutrophil development, tissue homing, function and plasticity in health and disease. Cells12, 1981 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Li, T. M. et al. The interferon-rich skin environment regulates Langerhans cell ADAM17 to promote photosensitivity in lupus. Elife13, e85914 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Girard-Madoux, M. J. H., Kel, J. M., Reizis, B. & Clausen, B. E. IL-10 controls dendritic cell-induced T-cell reactivation in the skin to limit contact hypersensitivity. J. Allergy Clin. Immunol.129, 143–50.e1–10 (2012). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Chow, W. H. et al. Cancer risk following polymyositis and dermatomyositis: a nationwide cohort study in Denmark. Cancer Causes Control6, 9–13 (1995). [DOI] [PubMed] [Google Scholar]

[CR44] 44.Oldroyd, A. G. S. et al. International guideline for idiopathic inflammatory myopathy-associated cancer screening: an International Myositis Assessment and Clinical Studies Group (IMACS) initiative. Nat. Rev. Rheumatol.19, 805–817 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Fiorentino, D. F. & Casciola-Rosen, L. Autoantibodies and cancer association: the case of systemic sclerosis and dermatomyositis. Clin. Rev. Allergy Immunol.63, 330–341 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Ma, J. et al. A blueprint for tumor-infiltrating B cells across human cancers. Science384, eadj4857 (2024). [DOI] [PubMed] [Google Scholar]

[CR47] 47.Kroeger, D. R., Milne, K. & Nelson, B. H. Tumor-infiltrating plasma cells are associated with tertiary lymphoid structures, cytolytic T-cell responses, and superior prognosis in ovarian cancer. Clin. Cancer Res.22, 3005–3015 (2016). [DOI] [PubMed] [Google Scholar]

[CR48] 48.Fiorentino, D. & Casciola-Rosen, L. Autoantibodies to transcription intermediary factor 1 in dermatomyositis shed insight into the cancer-myositis connection. Arthritis Rheumatol.64, 346–349 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Fiorentino, D. F. et al. Immune responses to CCAR1 and other dermatomyositis autoantigens are associated with attenuated cancer emergence. J. Clin. Investig.132, e150201 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.You, S. et al. Lymphatic-localized Treg-mregDC crosstalk limits antigen trafficking and restrains anti-tumor immunity. Cancer Cell42, 1415–1433 (2024). [DOI] [PubMed] [Google Scholar]

[CR51] 51.Poschel, D. B. et al. PD-L1 restrains PD-1Nrp1 Treg cells to suppress inflammation-driven colorectal tumorigenesis. Cell Rep.43, 114819 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Strauss, L. et al. Targeted deletion of PD-1 in myeloid cells induces antitumor immunity. Sci. Immunol.5, eaay1863 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Gordon, S. R. et al. PD-1 expression by tumour-associated macrophages inhibits phagocytosis and tumour immunity. Nature545, 495–499 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.Perry, J. A. et al. PD-L1-PD-1 interactions limit effector regulatory T cell populations at homeostasis and during infection. Nat. Immunol.23, 743–756 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol.42, 293–304 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Germain, P.-L., Lun, A., Garcia Meixide, C., Macnair, W. & Robinson, M. D. Doublet identification in single-cell sequencing data using. F1000Research10, 979 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Young, M. D. & Behjati, S. SoupX removes ambient RNA contamination from droplet-based single-cell RNA sequencing data. Gigascience9, giaa151 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods16, 1289–1296 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Almet, A. A. et al. A roadmap for a consensus human skin cell atlas and single-cell data standardization. J. Investig. Dermatol.143, 1667–1677 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Fiskin, E. et al. Multi-modal skin atlas identifies a multicellular immune-stromal community associated with altered cornification and specific T cell expansion in atopic dermatitis. Nat. Commun.17, 3194 (2026). [DOI] [PMC free article] [PubMed]

[CR61] 61.Joost, S. et al. Single-cell transcriptomics reveals that differentiation and spatial signatures shape epidermal and hair follicle heterogeneity. Cell Syst.3, 221–237.e9 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Korsunsky, I., Nathan, A., Millard, N. & Raychaudhuri, S. Presto scales Wilcoxon and auROC analyses to millions of observations. Preprint at bioRxiv10.1101/653253 (2019).

[CR63] 63.Chen, Y., Chen, L., Lun, A. T. L., Baldoni, P. L. & Smyth, G. K. edgeR v4: powerful differential analysis of sequencing data with expanded functionality and improved support for small counts and larger datasets. Nucleic Acids Res.53, gkaf018 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol.16, 1–13 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR65] 65.Feng, Y. et al. Spatial analysis with SPIAT and spaSim to characterize and simulate tissue microenvironments. Nat. Commun.14, 2697 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Dos Santos Peixoto, R. et al. Characterizing cell-type spatial relationships across length scales in spatially resolved omics data. Nat. Commun.16, 350 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Black, S. et al. CODEX multiplexed tissue imaging with DNA-conjugated antibodies. Nat. Protoc.16, 3802–3835 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR68] 68.Yiu, S. P. T. et al. Same-Slide Spatial Multi-Omics Integration with IN-DEPTH Reveals Tumor Virus-Linked Spatial Reorganization of the Tumor Microenvironment. Cancer Discov. 10.1158/2159-8290.CD-25-0775 (2026). [DOI] [PMC free article] [PubMed]

[CR69] 69.Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm for cellular segmentation. Nat. Methods18, 100–106 (2021). [DOI] [PubMed] [Google Scholar]

[CR70] 70.Tan, Y. et al. SPACEc: a streamlined, interactive Python workflow for multiplexed image processing and analysis. Nat. Commun.16, 10652 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Danaher, P. et al. InSituCor: exploring spatially correlated genes conditional on the cell type landscape. Genome Biol. 26, 105 (2025). [DOI] [PMC free article] [PubMed]

[CR72] 72.Badia-I-Mompel, P. et al. decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinform. Adv.2, vbac016 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Andreatta, M. & Carmona, S. J. UCell: robust and scalable single-cell gene signature scoring. Comput. Struct. Biotechnol. J.19, 3796–3798 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Spatial transcriptomics identifies immune-stromal niches associated with cancer in adult dermatomyositis

Ksenia S Anufrieva

Neda Shahriari

Ce Gao

Stephanie Pei Tung Yiu

Shideh Kazerounian

Rochelle L Castillo

Jessica Liu

Sean A Prell

Kartik Bhamidipati

Khashayar Afshari

Lindsay Parmelee

Anastasia N Kazakova

Erin Theisen

Teresa Bowman

Avery LaChance

William J Crisler

Kimberly Hashemi

Ilya Korsunsky

Sizun Jiang

Mehdi Rashighi

Ruth Ann Vleugels

Kevin Wei

Abstract

Introduction

Results

High-resolution spatial and single-cell atlas reveals interferon-driven tissue reorganization in lesional skin biopsies from DM and CLE

Fig. 1. Spatial and single-cell analysis reveals similar patterns of skin remodeling under the action of interferon in DM and CLE.

Single-cell analysis reveals immune cell heterogeneity in DM

Fig. 2. Heterogeneity of immune cells determines differences in pathologic features of DM and CLE.

Multicellular factor analysis uncovers distinct stromal and immune cell signatures in non-cancer-associated DM skin

Unique spatial organization of immune cell niches reveals disease-specific archetypes in DM and CLE lesions

Fig. 3. Spatial organized immune niches define cancer- and non-cancer-associated DM.

Disease-associated transcriptional programs define distinct immune niches

Fig. 4. Activated Treg and PD-L1+ mregDC distinguish immune niches in non-cancer-associated DM.

Vascular remodeling and metabolic reprogramming underlie distinct immune niche phenotypes

Fig. 5. Disease-specific vascular and stromal niches are shaped by hypoxia and inflammation.

Fig. 6. Hypoxia-driven suppression of vascular fibroblast identity and its spatial correlation with inflammatory signals.

Discussion

Methods

Patient recruitment and tissue sample collection

FFPE single-cell RNA sequencing

Xenium slide preparation

Xenium analyzer setup and data acquisition

Single-cell and spatial single-cell data preprocessing

Major and granular cell states identification

Spatial analysis of cell states organization—compartment analysis

Cell composition analysis

Sample stratification using multicellular factor analysis

Identification and analysis of immune niches in lesional samples

Finding stroma associated with immune niches

Cell interactions analysis

Multiplexed CODEX imaging with antibody-based protein detection

Quantification and analysis of CODEX data

Spatial correlation gene expression analysis

Quantification of endothelial cell density in standardized tissue regions

Pathway and gene signature analysis

Fibroblast сytokines and hypoxia treatment

Flow сytometry apoptosis assay

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Competing interests

Footnotes

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles