EPCO-18. DECIPHERING SPATIAL DIVERSITY OF GLIOBLASTOMA GENETIC ARCHITECTURE

Abstract Glioblastoma (GBM) is hallmarked by profound inter- and intratumoral heterogeneity. Current research has mapped key genetic drivers to distinct transcriptional profiles within GBM. Nevertheless, the intricate interactions between genetic alterations and spatial-transcriptional heterogeneity remain elusive. We utilized nestin-Tva C57BL/6 mice to induce de novo GBM via oncogene overexpression alongside shRNA-mediated silencing or deletion of tumor suppressors. We modeled critical driver mutations: PDGFB, NF1, and EGFR. Employing 10X Genomics platform for single-cell RNA sequencing, combined with spatially resolved transcriptomics and proteomics (Nanostring GeoMx), we integrated data using both horizontal (mutual nearest neighbors) and vertical (weighted nearest neighbors) approaches, analyzed through the SPATA2 software. Advanced computational techniques, including graph neural networks and single-cell deconvolution, were leveraged to unravel the genotypic and spatial variability within tumors and their microenvironments. Our findings uncover persistent cellular heterogeneity across diverse genetic contexts. However, specific genetic mutations precipitated distinct dominant transcriptional states: NF1 and EGFR mutations drove mesenchymal and astrocytic-like states, respectively, whereas PDGFB fostered a neural progenitor/cell-like state. These transcriptional shifts were closely linked to unique microenvironmental modifications, such as elevated neuronal signaling in PDGFB-driven GBMs and significant immune infiltration in NF1 and EGFR models. Through weighted co-expression network analysis (WGCNA) and gene set enrichment analysis (GSEA) paired with spatial-niche deconvolution, we identified a mesenchymal transition leading to an immunosuppressive microenvironment in NF1-loss driven GBMs and increased lymphocyte presence around blood vessels in EGFR-amplified GBMs. These patterns were consistent with human spatially resolved transcriptomic data, emphasizing the genetic complexity and heterogeneity of human GBMs. Our genetically engineered mouse models serve as a robust platform for dissecting GBM with defined genetic mutations. Leveraging sophisticated analytical techniques and computational modeling, this study underscores the profound interplay between genetic drivers and the tumor microenvironment, paving the way for targeted therapeutic strategies and precision medicine in GBM.