Integrative single-cell analysis reveals immunogenic cell death-associated heterogeneity and identifies a robust prognostic signature in osteosarcoma with experimental validation.

Background: Osteosarcoma is a highly aggressive primary malignant bone tumor characterized by pronounced intratumoral heterogeneity and an immunosuppressive tumor microenvironment. Immunogenic cell death (ICD), a regulated form of cell death capable of activating antitumor immunity, has been implicated in cancer progression and treatment response. However, the cell type-specific distribution of ICD features and their clinical relevance in osteosarcoma remain poorly defined. Methods: Single-cell RNA sequencing data, multiple bulk transcriptomic cohorts, and clinical follow-up information were integrated to characterize ICD-related transcriptional features in osteosarcoma. scRNA-seq analysis of 11 osteosarcoma samples from GSE152048 was used to define cellular composition and ICD activity at single-cell resolution. At the bulk level, the TCGA-TARGET cohort served as the training set, with GSE16091 and GSE21257 as validation cohorts. An ICD-related prognostic signature was constructed using WGCNA and ensemble machine learning. Pathway activity, cell-cell communication, and drug sensitivity were further analyzed. Functional assays in MG-63 and U-2OS cells were performed to validate the role of the key gene NAP1L1. Results: Single-cell analysis revealed pronounced cell type-dependent heterogeneity of ICD activity, with macrophages showing the highest ICD scores and epithelial-like tumor cells the lowest. The ICD-based prognostic model consistently stratified patient survival across all cohorts. High-risk tumors exhibited enrichment of inflammatory and stress-related pathways, including TNFA_SIGNALING_VIA_NFKB, COMPLEMENT, and COAGULATION, whereas low-risk tumors were associated with WNT_BETA_CATENIN, HEDGEHOG, and KRAS_SIGNALING_DN pathways. Cell-cell communication analysis demonstrated increased interaction frequency and signaling strength between tumor cells and immune or stromal cells in the high-risk group. Drug sensitivity prediction indicated lower IC50 values in high-risk patients for Camptothecin (p = 0.00031), Cytarabine (p = 0.013), Sorafenib (p = 0.000083), and SN-38 (p = 0.0042). Functional experiments showed that NAP1L1 overexpression promoted proliferation, migration, invasion, and clonogenicity in osteosarcoma cells. Conclusion: This study delineates ICD-related transcriptional heterogeneity in osteosarcoma at single-cell resolution and establishes a robust ICD-based prognostic signature. The findings suggest that ICD reflects an integrated tumor state shaped by cellular programs and microenvironmental interactions, providing preliminary evidence for its utility in risk stratification and therapeutic exploration, which warrants further prospective validation.