Variation Among Intact Tissue Samples Reveals The Core Transcriptional Identities Of Cell Types In The Glioma Microenvironment

Abstract Adult low-grade gliomas generally progress to glioblastoma, a more aggressive CNS tumor with an extremely poor prognosis. Despite intensive efforts, numerous promising glioma therapies have failed to provide survival benefits. These failures reflect many factors, including intertumoral heterogeneity and immunosuppression by the tumor microenvironment (TME). We propose a novel approach to addresses these challenges through integrative deconvolution of bulk gene expression data generated from more than 5000 human gliomas and 7000 normal human brain samples. Inherent variation in the cellular composition and cellular activities of these samples allowed us to identify highly correlated modules of genes that represent specific cell types and cell states. By comparing gene coexpression modules in glioma vs. normal human brain, we have identified cell type-specific gene expression changes in the glioma TME that are highly reproducible. In contrast to single-cell methods, which sample only a fraction of the tumor tissue and fail to capture major nonmalignant cell-types, our results derive from billions of cells and thousands of individuals and are therefore highly robust. We find that a number of genes encoding cell-surface proteins are specifically up-regulated in immune and vascular cells of the glioma TME. Surprisingly, among those genes up-regulated in glioma vasculature are multiple members of the angiotensin pathway, suggesting non-canonical roles for these proteins in the glioma setting. We propose that these proteins may form a specific ‘zip code’ for glioma within the brain’s vasculature that can be targeted directly or by conjugation with existing drugs. More generally, our analytical approach has revealed reproducible gene expression changes in specific cell types of the glioma TME that provide more stable therapeutic targets than those that are expressed by genetically mutable malignant cells. We have also discovered novel, aberrantly coexpressed genes in microglia, oligodendrocytes, and astrocytes which we are testing in state-of-the-art human brain assembloid systems.