Die räumliche Vielfalt und genetische Komplexität des Glioblastoms entschlüsseln

Glioblastoma (GB) is marked by significant inter- and intratumoral heterogeneity, shaped by intricate interactions between genetic drivers and spatial transcriptional variations. This study aims to investigate how do spatial and genetic variations influence glioblastoma and its microenvironment.

To unravel these complexities, we developed nestin-Tva C57BL/6 mouse models of de novo GB through oncogene overexpression and targeted suppression or deletion of tumor suppressors, focusing on three pivotal mutations: PDGFB, NF1, and EGFR. We employed single-cell RNA sequencing (10X Genomics) and spatial multi-omics (Nanostring GeoMx, 10X Visium HD), integrating data through mutual nearest neighborhood (horizontal) and weighted nearest neighborhood (vertical) methods, analyzed using SPATA2 software.

Leveraging advanced computational tools, including graph neural networks and single-cell deconvolution, we delineated the genetic and spatial variability within tumors and their associated microenvironments. While cellular heterogeneity was pervasive across genetic backgrounds, specific mutations drove dominant transcriptional programs: NF1 loss favored mesenchymal-like states, EGFR amplification induced astrocytic-like phenotypes, and PDGFB overexpression promoted neural progenitor-like states. These genotype-specific shifts were accompanied by distinct microenvironmental features, such as enriched neuronal signaling in PDGFB-driven tumors and pronounced immune cell infiltration in NF1- and EGFR-mutant models.

Further analysis via weighted gene co-expression network analysis (WGCNA), gene set enrichment analysis (GSEA), and spatial niche deconvolution revealed key microenvironmental trends. NF1 loss promoted mesenchymal transitions alongside immunosuppressive niches, while EGFR amplification resulted in elevated perivascular lymphocyte recruitment. These findings were validated against human spatial transcriptomic datasets, underscoring their translational relevance.

Our work highlights the utility of genetically engineered mouse models in disentangling the relationships between genetic alterations, transcriptional states, and spatial heterogeneity in GB. By integrating single-cell and spatial transcriptomic approaches with sophisticated computational analysis, we provide new insights into the molecular and microenvironmental diversity of GB. These findings pave the way for more precise therapeutic strategies tailored to specific genetic and spatial tumor contexts.