In June 2021, a 28-year-old male presented to our hospital with a triad of diffuse facial erythema, progressive proximal limb weakness, and significant fatigue. He reported no systemic symptoms such as fever or weight loss. Physical examination revealed characteristic heliotrope rash and Gottron’s papules. Laboratory investigations were significant for markedly elevated muscle enzymes: creatine kinase (CK) 1428 U/L (ref: 50–310 for male), CK-MB 38 U/L (ref: 0-19), lactate dehydrogenase (LDH) 350 IU/L (ref: 120-250), aspartate aminotransferase (AST) 71 IU/L (ref: 13-35), and myoglobin 146.4 ng/mL (ref: 11.1-57.7). Electromyography demonstrated a myopathic pattern without denervation potential. Serological testing was positive for the anti-TIF1γ antibody (++; Euroline Autoimmune Inflammatory Myopathies 16 Ag (IgG) Profile, Euroimmun, Lübeck, Germany). A subsequent muscle biopsy confirmed the diagnosis of DM. Given the positive serum anti-TIF1γ antibody result, a whole-body CT scan, chosen for its lower cost compared to PET/CT, was performed for cancer screening. No tumor was detected. Additionally, the patient indicated that there is no family history of any specific health issues. The patient then received an 8-day hospitalization involved with daily intravenous methylprednisolone (60 mg). Following discharge, the rash had improved but persisted. Thus, an outpatient oral steroid taper was instituted, beginning at 40 mg daily and decreasing by 5 mg every two weeks until discontinuation. The patient was subsequently stable.

In April 2024, the patient presented to an outside oncology department with left cheek swelling and numbness. Concurrently, a rash was noted on the face and neck. A PET/CT scan revealed a hypermetabolic mass in the left parotid gland, suggestive of malignancy, along with concurrent enlargement of regional lymph nodes (Figures 1A–F). Subsequent MRI confirmed the presence of bilateral parotid space-occupying lesions. The patient subsequently underwent an extended parotidectomy with suprahyoid lymph node dissection and a right mandibular lymph node biopsy. Histopathological examination confirmed the diagnosis of lymphoepithelial carcinoma, supported by the following findings: H&E staining (Figure 1G) showed massive lymphocytic infiltration; immunohistochemistry was positive for CK5/6 and p63, with a Ki-67 proliferation index of 60% in tumor cells; and in situ hybridization (ISH) was positive for Epstein-Barr virus-encoded small RNAs (EBER). The diagnosis was further corroborated by the detection of high levels of EBV DNA in peripheral blood. Following surgery, the patient underwent three cycles of adjuvant therapy with gemcitabine (1.8 g/day), cisplatin (50 mg/day), and toripalimab (240 mg), all administered on a 3-week cycle. No severe adverse events were observed during this combination therapy. This was followed by consolidative radiotherapy with a total dose of 60 Gy in 30 fractions. By the completion of systemic therapy on September 15, 2024, the patient had achieved a near-complete response, corroborated by the clearance of EBV DNA to undetectable levels in peripheral blood (Figure 2).

Two months after completing systemic oncological treatment (December 2024), the patient developed a pronounced facial rash, recurrent proximal muscle weakness and myalgia. Respiratory symptoms included shortness of breath and a productive cough with yellow sputum. There was no reported fever or dysphagia. He was readmitted to our hospital for comprehensive evaluation. Physical examination revealed classic dermatomyositis findings: symmetrical, edematous, heliotrope erythema on the eyelids; a diffuse, V-shaped erythema on the chest and neck; and symmetrical, dark-purplish Gottron’s papules on the upper extremities. Proximal muscle strength was graded 4/5 with notable tenderness. SARS-CoV-2 nucleic acid was detected in a throat swab sample. Echocardiography revealed no evidence of cardiac disease. Laboratory studies showed a recurrence of elevated muscle enzymes: CK 335 U/L, CK-MB 26 U/L, LDH 357 IU/L, and AST 47 IU/L. Immunological testing was positive for antinuclear antibody (ANA) at a titer of 1:320 and anti-TIF1γ antibody at a titer of 1:30 (cell-based assay, RareDiagnostics, Xi’an, China). Furthermore, the absence of both rheumatoid arthritis (RA)-specific joint symptoms and systemic lupus erythematosus (SLE)-related mucocutaneous findings allowed for the exclusion of these diagnoses. Considering the temporal association with oncologic therapy and the classic clinical and serological profile, a diagnosis of recurrent DM was established (Figure 2).

In response to the DM flare, the immunosuppressive regimen was modified to intravenous cyclophosphamide (0.6 g every 2 weeks) for its dual autoimmune and oncological benefits, alongside a reduced dose of methylprednisolone (20 mg initially). During a 6-day hospitalization, the patient’s muscle weakness resolved and enzyme levels normalized. He was discharged on a maintenance methylprednisolone taper, which was later discontinued as his dysphagia and myalgia remained in remission; clinical focus shifted to the more critical paraneoplastic malignancy. A residual rash was present but non-bothersome (Figure 2).

Five months later, the rash worsened significantly, requiring re-initiation of cyclophosphamide. After six cycles reached the maximum cumulative dose when his BMI was 18.1, maintenance therapy was switched to subcutaneous methotrexate (10 mg/week), which was ineffective. Tofacitinib was then initiated for rash control, acknowledging the potential malignancy risk. Throughout this period, quarterly head/neck MRI surveillance confirmed the LEC remained stable (Figure 2).

Classic risk factors for CAM include older age and dysphagia (19, 28). In contrast, our 31-year-old patient developed CAM with a rare parotid LEC (18, 20, 29, 30), a malignancy seldom linked to DM. Therefore, to investigate the underlying immunologic features, we characterized the cellular atlas of PBMCs obtained at the time of first recurrent DM diagnosis. PBMCs from an age-matched (29-year-old) female healthy donor were used as a control (Figure 3A).

Our high-resolution dataset comprised approximately 140,654,478 transcripts across 23,600 cells, with a median detection of 2,290 genes per cell (Supplementary Figure S1A) and a minimum threshold of 550 genes per cell (Supplementary Figure S1B). Principal component analysis of the top 3,000 DEGs identified 21 distinct cell clusters (Supplementary Figure S2A), which were further annotated against established immune lineages using canonical marker genes (Supplementary Figure S2B). Following quality filtering and doublet removal, 22,665 high-quality cells were retained yielding representing all eight major immune cell populations (Figures 3B, C; Supplementary Figure S2C), with a ranked breakdown into 6,088 CD4⁺ T cells (26.9%), 5,587 CD8⁺ T cells (24.6%), 5,180 B cells (22.8%), 3,079 monocytes (13.6%), 2,264 NK cells (10.0%), 366 dendritic cells (DC,1.6%), 36 plasma cells (0.2%), and 65 platelets (0.3%).

A comparative analysis of overall immune cell frequencies identified a significant compositional shift in the case compared to HC. This shift was characterized by a relative expansion of B cells and CD4⁺ T cells, concurrent with a contraction of cytotoxic subsets (CD8⁺ T cells, NK cells) and monocytes (Figures 3D, E). To determine if these proportional changes reflected true changes in absolute cell numbers, we performed flow cytometry analysis of lymphocyte subsets, referencing clinical complete blood count (CBC) data and a 40-sample healthy cohort (Table 1). Flow cytometry analysis confirmed pronounced lymphopenia at both the initial diagnosis and a 10-month follow-up, with total lymphocyte counts below the healthy reference range (Table 1). The absolute numbers of T cells and NK cells were substantially reduced. While the absolute count of B cells is within the reference range. This indicates that the elevated B cell frequency in the scRNA-seq data was a relative increase, attributable to the more severe depletion of other lymphocyte populations (Table 1). The subsequent decline in absolute B cell counts during persistent disease suggests that the initial relative abundance may have been a maintenance effect of EBV infection, though EBV DNA was undetectable at the time of sample collection (31, 32).

While EBV can transform B cells through T cell-independent mechanisms (31), the role of B cells in this case of DM was unclear. Unsupervised clustering identified six distinct B cell sub-populations (Supplementary Figure S2E). Annotation with canonical markers revealed that the major population consisted of patient-derived naïve B cells (Figures 3F, G; Supplementary Figures S2E–G). These naïve B cells exhibited transcriptional levels comparable to other B cell subtypes (Figure 3H). Pathway enrichment analysis showed that although interferon (IFN) and antiviral response genes were active in the patient’s B cells, the inflammatory response was uniquely and predominantly elevated in the naïve B cell subset (Figures 3I–K). In contrast, cytokine and chemokine signaling showed a slight declining trend in the patient’s B cells (Supplementary Figure S2H) (33, 34). This suggests that the patient’s naïve B cells contribute to DM pathology may through a specific inflammatory program, rather than through broad antiviral, IFN, or cytokine-mediated activation.

To investigate the basis for naive B cell expansion, we performed gene set variation analysis (GSVA) using hallmark pathway gene signatures. This revealed significant enrichment in the case for proliferative pathways, including mitotic spindle, hedgehog signaling and WNT/β-catenin signaling (Figure 3L). Additionally, interferon pathways were more active in patient-derived naive B cells (Figures 3I, L). These proliferative and inflammatory characteristics were also observed in memory B cells (Figure 3M).

Given the established association between EBV and LEC (7, 35), we performed germline whole-exome sequencing (WES) on peripheral blood to identify potential driver genes or molecular mechanisms underlying DM recurrence. Sequencing generated 145,544,140 total reads, with a mean coverage depth exceeding 200×. We identified 98,002 single-nucleotide variants (SNVs) and 16,872 short insertions and deletions (indels). Of these, 24.04% of SNVs were exonic, and 47.24% of these SNVs were nonsynonymous (Supplementary Figures S3A, B). Among the indels, 4.12% were located in exonic regions, and 29.96% of these indels resulted in frameshifts (Supplementary Figure S3C, D). Putatively deleterious mutations are summarized in Supplementary Data S1. From the 338 genes harboring nonsynonymous exonic SNVs, we identified 29 genes implicated in immune regulation pathways (Supplementary Table S1). None of these genes have been previously reported as known susceptibility genes for DM. However, Complement C3d Receptor 2 (CR2, encode CD21 protein), which functions as a receptor for EBV binding on B and T lymphocytes, has been reported that rs1876453 in this gene are associated with susceptibility to systemic lupus erythematosus SLE (36). In our case, rs367567954 in CR2 is functional unknown but makes the amino acid changed and putatively deleterious.