Relating Glioblastoma Heterogeneity To Human Fetal Glial Development Through High Resolution Single Nuclei Transcriptomics

Abstract Glioblastoma (GBM) is thought to be driven by a therapy-resistant cancer stem cell population that recapitulates developmental phenotypes. Direct comparisons of GBM to glial states during human fetal development are limited due to paucity of data from late prenatal gestation, when gliogenesis is thought to occur. Here, we generated a comprehensive single nuclei RNA sequencing (snRNAseq) dataset of approximately 200,000 nuclei taken from the germinal matrix and the cortical plate of 16 fetal postmortem samples, ranging from 17 to 41 gestational weeks, enabling high spatiotemporal resolution of late neurogenesis and early-to-peak gliogenesis. We performed unbiased clustering to identify broad cell types within each sample and integrated all fetal samples to analyze evolving glial states and relationships across two regions and four developmental stages. Subclustering analysis of developing glia from the germinal matrix and cortical plate resolved developmental cell type signatures that are absent in the adult brain. Trajectory inference and pseudo-time analyses reconstructed relationships within these glial lineages and states, identifying a robust common glial progenitor population (GPC) with distinct signature, preceding both oligodendrocyte progenitor cell (OPC) and astrocyte lineage commitment during late prenatal development. We then performed snRNAseq on approximately 30,000 nuclei taken from the core and infiltrating edge of two surgically resected GBM samples with IDH-mutant and IDH-wildtype status and EGFR amplification. Uniform manifold approximation and projection (UMAP) dimensionality reduction revealed distinct neoplastic and non-neoplastic population clusters within each GBM sample. Projecting our previously defined neural stem cell / progenitor signatures onto each GBM UMAP identified notable predominance of the GPC-like developmental signature throughout both GBM tumors with focal minor contributions from the OPC-, transit amplifying-, and astrocyte-like signatures. The high spatial and temporal resolution of the generated roadmap dissolves GBM intratumoral heterogeneity into distinct developmental molecular states driven by potentially targetable regulatory networks.