Abstract / Summary
Metastatic progression is the principal driver of mortality in prostate cancer (PCa); however, existing clinicopathologic tools cannot reliably identify which ostensibly localized tumors will later metastasize. To address this gap, we developed Met-Score, a biologically interpretable transcriptomic signature derived from primary tumor gene expression to quantify metastatic competence. Using a random-effects meta-analysis across six independent primary tumor cohorts ( n = 1,000 patients; 306 metastatic events), we identified a 45-gene metastatic progression program comprising 27 up-regulated and 18 down-regulated genes. For clinical validation, we trained and locked an L2-regularized logistic regression model using these genes to generate a Met-Score probability for metastatic progression. In the development cohorts, Met‑Score showed consistent gene-level directional effects across studies (meta-analytic pooled AUC = 0.81). Under a single locked deployment with no per-cohort refitting, competing-risk time-dependent AUCs were 0.75 and 0.71 at 5 and 10 years in JHU ( n = 239; 93 metastatic events) and 0.79 at both horizons in Durham VA ( n = 555; 40 events). Met-Score remained associated with metastasis-free survival after adjustment for pathological Gleason grade in both validation cohorts. Exploratory analyses within Gleason 7 disease suggested additional risk resolution beyond Grade Group, warranting prospective confirmation. In cross-sectional diagnostic biopsy RNA-seq cohorts, Met-Score distinguished de novo metastatic hormone-naïve from localized disease (AUC 0.93–0.98), supporting biological detection of the Met-Score program in primary biopsy tissue. Single-cell RNA-seq analyses of tumors biopsies across localized, metastatic hormone-sensitive, and metastatic castration-resistant disease demonstrated significantly higher tumor cell Met-Score in metastatic disease, with enrichment concentrated in proliferative tumor programs and accompanied by marked cell type-restricted expression patterns among component genes. In the bone metastatic niche, Met-Score showed cell type-specific enrichment in tumor tissue relative to distal marrow, particularly among lymphoid populations. Finally, the directional Met-Score was lower after batiraxcept treatment in a metastatic bone patient-derived xenograft model and showed concordant shifts across independent genetic and pharmacologic perturbation datasets, supporting perturbation responsiveness of the underlying transcriptional program. Met-Score is a reproducible transcriptomic signature that captures a conserved metastatic-competence program detectable in primary tumor profiles, reflects tumor and microenvironmental transcriptional states, and is responsive to therapeutic and genetic perturbation, supporting prospective evaluation for improved risk stratification, including within Gleason 7 disease.