Glioblastoma (GBM) is the most common and lethal primary brain tumor in adults and is driven by self-renewing glioblastoma stem cells (GSC) that persist after therapy and seed treatment-refractory recurrent tumors. GBM tumors display a high degree of intra- and intertumoral heterogeneity that is a prominent barrier to targeted treatment strategies. This heterogeneity extends to GSCs that exist on a gradient between two transcriptional states or subtypes termed developmental and injury response. Drug targets for each subtype are needed to effectively target GBM. To identify conserved and subtype-specific genetic dependencies across a large and heterogeneous panel of GSCs, we designed the GBM5K-targeted guide RNA library and performed fitness screens in a total of 30 patient-derived GSC cultures. The focused CRISPR screens identified the most conserved subtype-specific vulnerabilities in GSCs and elucidated the functional dependency gradient existing between the developmental and injury response states. Developmental-specific fitness genes were enriched for transcriptional regulators of neurodevelopment, whereas injury response-specific fitness genes were highlighted by several genes implicated in integrin and focal adhesion signaling. These context-specific vulnerabilities conferred differential sensitivity to inhibitors of β1 integrin, focal adhesion kinase, MEK, and OLIG2. Interestingly, the screens revealed that the subtype-specific signaling pathways drive differential cyclin D (CCND1 vs. CCND2) dependencies between subtypes. These data provide a biological insight and mechanistic understanding of GBM heterogeneity and point to opportunities for precision targeting of defined GBM and GSC subtypes to tackle heterogeneity. Significance: CRISPR-Cas9 screens in a panel of patient-derived glioblastoma stem cells reveal heterogeneity in genetic vulnerabilities across subtypes that have important implications for targeted and combination treatment strategies for glioblastoma.
©2024 American Association for Cancer Research.