Abstract
The BCL-X(L) anti-apoptotic protein is a clear cell Renal Cell Carcinoma (ccRCC) dependency; however, the mechanism of this dependence and its relevance in other aggressive kidney cancer contexts, including metastatic and/or rare RCC subtypes [e.g., Fumarate Hydratase (FH)-deficient and sarcomatoid RCCs], is unknown. Computational predictions, using a machine learning model trained on the human RCC TCGA dataset, and cell-based validations, confirmed BCL-X(L) dependence in all RCC subtypes. Remarkably, cell state changes, 'anoikis' programs, inflammatory state, and metabolic perturbations (e.g., fumarate production in FH-deficient RCCs) independently conferred increased BCL-X(L) dependence. Correlation studies revealed that increased AMPK isoform 2 (PRKAA2) expression is a kidney-specific biomarker of BCL-X(L) dependence. Indeed, pharmacological AMPK activation sensitized RCCs to BCL-X(L) blockade. Finally, using functional studies, we developed a multivariate model that accurately predicted BCL-X(L) dependence in RCC. Our studies offer biomarkers for patient stratification and credential BCL-X(L) as a subtype agnostic vulnerability in difficult-to-treat RCCs.