Abstract 5072: Single cell-based elucidation of molecularly distinct glioblastoma states and drug sensitivity

Abstract Glioblastoma (GBM) heterogeneity and plasticity represent critical challenges to the development of successful treatment options, because molecularly distinct subpopulations present different drug sensitivities, leading to resistance [Yuan et al., 2018; Neftel et al., 2019; Ding et al., 2019]. In fact, clinical outcomes for GBM, which accounts for >80% of all primary gliomas [4], have not improved over the last 20 years. Unfortunately, due to the sparse nature of single cell RNA-seq (scRNA-seq) profiles, prediction of subpopulation-specific dependencies and drugs has been difficult. To overcome this problem, we leveraged network-based algorithms, single cell versions of the ARACNe [Margolin et al., 2006], VIPER [Alvarez et al., 2016], and OncoTreat algorithms [Alvarez et al., 2018] (NY and CA Dept of Health approved, CLIA compliant algorithm, currently used in several clinical trials), to analyze a collection of scRNA-seq profiles from 16 GBM patients to elucidate Master Regulator (MR) proteins, representing molecular determinants and thus non-oncogene dependencies, of distinct, co-existing GBM cell states and to prioritize compounds capable of inverting their activity. OncoTreat prioritizes drugs by assessing their ability to invert the activity of subpopulation-specific MR proteins, based on the analysis of large-scale drug perturbation assays, thus supporting the identification of small molecule inhibitors targeting distinct tumor compartments within the same patient for potential combination therapy. Specifically, we used a library of genome-wide RNASeq profiles representing the response of proneural (PN) (HF2597 neurosphere) and mesenchymal (MES) (U87 cell line) cells at 24h following treatment with >350 clinically relevant drugs generated by the Califano Lab. The top six drugs predicted to target MES and PN subpopulations were tested in acute slices from GBM resections allowing direct characterization of the drug-mediated depletion of subpopulations predicted to be sensitive in adjacent drug vs. vehicle control treated slices [Zhao et al., 2021].Confirming our predictions, we observed statistically significant depletion of the MES and PN compartments by 3 of 3 (Dasatinib, Bexarotene and Prednisone) and 2 of 3 (Givinostat and Ixazomib) OncoTreat-predicted drugs, respectively. Given its high validation rate, this study paves the road and provides significant insight into the future development of combination trials where these drugs may be administered either concomitantly or sequentially. Moreover, the computationally inferred differential activity of subtype-specific MR proteins (CD44, OLIG2, KDM2B, EPHB3, TOP2B and others) was experimentally validated by FACS in cultured patients-derived GBM neurospheres. Citation Format: Lorenzo Tomassoni, Wenting Zhao, Hongxu Ding, Kenneth Egbuji, Pasquale Laise, Mariano J. Alvarez, Julia L. Furnari, Michael G. Argenziano, Matei A. Banu, Jeffrey N. Bruce, Peter D. Canoll, Peter A. Sims, Andrea Califano. Single cell-based elucidation of molecularly distinct glioblastoma states and drug sensitivity [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5072.