Transcriptome-Inferred metabolic subtypes define prognostic and immune ecosystems in osteosarcoma at single-cell...

Osteosarcoma exhibits substantial metabolic heterogeneity, yet existing classification schemes typically rely on narrow pathway windows and lack cellular-source attribution. Here, we integrated bulk transcriptomes from 157 osteosarcoma tumors (PKUPH-OS, n = 70; TARGET-OS, n = 87), quantified pathway-level activity via single-sample gene set enrichment analysis (ssGSEA) across curated metabolic gene sets, and performed consensus clustering to identify three stable metabolic subtypes: Cholesterogenic (C1), characterized by mTORC1-associated lipid biosynthesis and invasion-related signaling; Redox-Catabolic (C2), dominated by detoxification, glutathione/ROS buffering, and enhanced catabolic programs alongside immune engagement; and OXPHOS-Active (C3), featuring coupled mitochondrial oxidative phosphorylation and proliferative programs. Clinically, C2 displayed favorable survival (3-year overall survival: 90.8%), whereas C1 (60.0%) and C3 (61.9%) followed adverse trajectories (overall survival, P = 0.0037; progression-free survival, P = 0.011). Single-cell RNA sequencing of 16,276 cells with inferCNV-supported malignant-cell anchoring revealed distinct cellular origins: C1 signatures mapped predominantly to stromal/mesenchymal populations, C2 signatures to immune cells, and C3 signatures to malignant tumor cells. These observations apply within osteosarcoma the well-recognized principle that bulk transcriptomic signatures reflect cellular composition, providing an explicit cell-of-origin map of metabolic subtypes that supports niche-resolved risk stratification and cell-aware therapeutic hypothesis generation.