Abstract Brain metastases (BrMs) represent a significant clinical challenge with increasing incidence and poor prognosis. Here, we present a comprehensive proteogenomic characterization of pan-cancer BrMs (n=923) with patient-matched primary tumors (n=82) and glioblastomas (n=20), representing the largest cohort to date. Through consensus clustering of transcriptomic data, we identified and validated four distinct molecular subtypes (CMS1-4) with unique biological features. Integrated analysis of the transcriptomic and proteomic revealed that CMS1 was activated with neural system-associated pathways, CMS2 was characterized by extracellular matrix remodeling and increased immune cell infiltration, CMS3 exhibited a more epithelial phenotype along with metabolic reprogramming and CMS4 showed upregulation of cell cycle-related genes with elevated stemness scores. These molecular features were further validated through immunohistochemistry and multiplex immunofluorescence. Genomic analysis revealed subtype-specific mutational signatures and copy number alterations, with CMS1 exhibiting lower tumor purity and CMS4 displaying the highest chromosomal instability. Survival analysis indicates that CMS4 was associated with the worst prognosis, while CMS1 and CMS2 demonstrated more favorable responsiveness to radiotherapy and immunotherapy, respectively. Using patient-derived organoid models, we identify subtype-specific therapeutic vulnerabilities, notably metabolic inhibition sensitivity in CMS3 and CDK4/6 inhibition response in CMS4. Single-nucleus RNA sequencing further elucidates the complex cellular ecosystems and intercellular communication patterns within each subtype. This study creates a multi-omics resource for understanding the metastatic adaptation to the brain microenvironment and exploring personalized treatment strategies in BrMs.
Gao Zhang (2025) studied this question.