TMIC-26. PRECLINICAL MODEL SYSTEMS OF GLIOBLASTOMA REVEAL MICROENVIRONMENTAL PROGRAMS AND DEPENDENCIES IN PATIENT TUMORS

Abstract

Abstract Patient-derived model systems serve as a platform for translational research representing the heterogeneity of human cancers, and their success in recapitulating disease-driving genomic alterations is well-documented. While recent studies have demonstrated genomic and functional divergence in patient-derived models with passaging, the need for accurate preclinical models remains. Glioblastoma (GBM) is the most common and aggressive primary brain tumor, and thus far preclinical models have failed to consistently replicate the responses found in patients. We therefore aimed to evaluate the multi-omic fidelity of low-passage GBM model systems across in vitro and in vivo environments and to elucidate the molecular features in which they differ. To this end we established a biobank of glioma direct-from-patient orthotopic xenograft (GliomaPDOX) models and primary gliomasphere cultures (GSCs) and performed whole-exome and RNA sequencing of over 40 purified patient tumors and their matched GliomaPDOXs and GSCs to facilitate paired comparisons across a gradient of full tumor microenvironment (TME) presence. We observed global genomic and transcriptomic fidelity in both systems, but specific programmatic gene expression differences associated with cell-cell interactions in the brain TME, glial cell identity, and in vitro GSC-forming ability. GSCs and GSC-forming ability are strongly associated with an astrocytic gene expression signature, while more stem-like and oligodendrocytic patient tumors including IDH- and H3F3A-mutant GBMs more successfully engraft in GliomaPDOXs. This result implicates the brain TME as a support system for these more stem/oligo-like tumors. Transcription factor network analysis identified regulators of the NOTCH and MYC pathways as strongly enriched in this subgroup of patient tumors and their derivative xenografts, and provides potential targets for therapeutic intervention in near future experiments. Collectively, these findings underline the critical role of the TME in defining GBM cell state, reveal the heterogeneity of TME dependence across patient tumors, and link this dependency to therapeutically actionable molecular features.