MODL-08. ESTABLISHING IN VITRO MODELS TO RECAPITULATE PROTEIN-LEVEL PROGNOSTIC CELL PHENOTYPES FROM PRIMARY HUMAN GLIOBLASTOMA

Abstract

Abstract Glioblastoma (GBM) is a highly aggressive primary central nervous system tumor for which median survival continues to be low, despite multimodal treatment and surgical resection. Heterogeneity at many levels (genomic, cytoarchitectural, and metabolic) likely contributes to these poor outcomes, and single-cell approaches can help to identify prognostically meaningful cells and features within heterogeneous tumors at each of these levels. Recent studies using mass cytometry (CyTOF) and the accompanying analysis pipeline Risk Assessment Population Identification (RAPID) measured per-cell protein levels in > 2 million cells from 28 IDH1/2-wildtype primary glioblastoma tumors and identified 2 phenotypes within these tumors that are present in different abundances across patients and have contrasting prognostic value. Increased abundance of Glioblastoma Negative Prognostic (GNP) cells within these tumors is predictive of shorter overall survival (OS HR = 1.07 [95% CI 1.02–1.12], p = 0.003), while the converse is true for Glioblastoma Positive Prognostic (GPP) cells (OS HR = 0.93 [0.87–1.0], p = 0.05). To effectively study these phenotypes, reproducible in vitro models are needed that recapitulate the total protein expression and phosphorylation events that distinguish GNP and GPP cells. Using available RNA transcript data, a candidate subset of glioma stem cell models were identified and are now being profiled using standardized mass-tagged antibody panels. These culture models will be quantitatively compared against established GNP and GPP populations in patient tissue samples using the Marker Enrichment Modeling statistic as well as other machine learning tools. Development of these models will be critical to future experiments aimed at identifying possible mechanistic therapies that effectively eliminate each of these identified populations, as well as aligning these protein-level phenotypes to other genomic and functional data in glioblastoma samples and models.