COMP-05. A DRUG ASSOCIATION MAP OF GLIOBLASTOMA INFORMS PRECISION TARGETING OF P53-Dependent METABOLIC STATES

Abstract Glioblastoma (GBM) is a malignant brain tumor with few therapeutic options. Because GBMs are genetically and functionally diverse, there is little systematic understanding of how hundreds of approved drugs may be active against GBM subpopulations. To identify new pharmacological subgroups of GBM, we studied comprehensively the variation in drug response across 100 well-characterised patient-derived GBM cell cultures and used statistical models to explain patient specific cellular drug response in terms of molecular markers, pathway signatures and drug mechanism-of-action. We find that the response to proteasome inhibitors, currently under clinical development for GBM, shows extreme variation across cases, which is phenocopied \emph{in vivo} and predicted by p53 pathway activity. The resistance to proteasome inhibitor response is explained by a subset of p53 functional GBM cases, which exhibit more efficient scavenging of reactive oxygen species (ROS), thereby protecting against ROS-dependent apoptosis. By interference against this protective feedback loop using secondary inhibitors, synergistic killing of GBM cells is achieved. Our results define a sizable metabolic subclass of GBM cells, with implications for precision and combination targeting.