TLSB-01 A PAN-CANCER MULTI-OMIC CLASSIFICATION OF BRAIN METASTASES REVEALS ACTIONABLE METASTATIC STATES AND A METABOLIC–IMMUNE BIOMARKER AXIS TO GUIDE RATIONAL COMBINATIONS

Abstract

Abstract Brain metastases (BrMs) remain therapeutically challenging because metastatic adaptations within the CNS are heterogeneous and often poorly captured by primary-tumor biomarkers, limiting rational patient stratification and trial design. We constructed a pan-cancer multi-omic atlas of BrMs (n = 1,032) integrating genomic and transcriptomic profiling with quantitative proteomics and targeted metabolomics, supported by orthogonal single-cell and spatial measurements, to derive a metastatic state–based classification that generalizes across tumor origins and across analytic platforms. Consensus analysis revealed four reproducible BrM states: a neural-like program (BrMS1), an immune-infiltrated state coupled to EMT/stromal activation (BrMS2), a metabolically reprogrammed state enriched for mitochondrial oxidative phosphorylation (OXPHOS) (BrMS3), and a highly proliferative, immune-excluded state (BrMS4). These states captured convergent cross-cancer programs, providing a unifying framework to interpret BrM biology beyond tissue of origin. To connect states to tractable liabilities, we performed targeted drug screening in patient-derived organoids (PDOs) and identified state-linked vulnerabilities, nominating mTOR-pathway dependence in BrMS3 and CDK4/6-axis dependence in BrMS4, thereby prioritizing rational combination strategies. Focusing on lung cancer BrMs, we analyzed an integrated multi-omic cohort (n = 154) with matched clinical annotation and tissue immune phenotyping, confirming a “mitochondria-high/immune-low” axis characterized by elevated OXPHOS signatures, suppressed immune signaling, and an immunosuppressive microenvironment that associated with inferior survival, supporting its utility as a biomarker for risk stratification. This biomarker-defined axis motivated a mechanism-informed therapeutic hypothesis: mitochondrial inhibition reduced OXPHOS-associated programs and impaired viability in LC-BrM PDOs, and in an orthotopic LC-BrM mouse model, combining mitochondrial inhibition with anti–PD-1 therapy produced the most durable survival benefit compared with either monotherapy. Together, these results establish a pan-cancer metastatic state framework, link it to actionable biomarkers and validated therapeutic hypotheses, and provide an evidence base for biomarker-guided clinical trials that pair metabolism-targeting with immunotherapy for metabolically reprogrammed, immune-suppressed BrMs, and inform companion diagnostic development for CNS-active regimens.