- Fazzari, Elisa;
- Azizad, Daria J;
- Yu, Kwanha;
- Ge, Weihong;
- Li, Matthew X;
- Nano, Patricia R;
- Kan, Ryan L;
- Tum, Hong A;
- Tse, Christopher;
- Bayley, Nicholas A;
- Haka, Vjola;
- Cadet, Dimitri;
- Perryman, Travis;
- Soto, Jose A;
- Wick, Brittney;
- Raleigh, David R;
- Crouch, Elizabeth E;
- Patel, Kunal S;
- Liau, Linda M;
- Deneen, Benjamin;
- Nathanson, David A;
- Bhaduri, Aparna
Glioblastoma (GBM) is the deadliest form of primary brain tumor with limited treatment options. Recent studies have profiled GBM tumor heterogeneity, revealing numerous axes of variation that explain the molecular and spatial features of the tumor. Here, we seek to bridge descriptive characterization of GBM cell type heterogeneity with the functional role of individual populations within the tumor. Our lens leverages a gene program-centric meta-atlas of published transcriptomic studies to identify commonalities between diverse tumors and cell types in order to decipher the mechanisms that drive them. This approach led to the discovery of a tumor-derived stem cell population with mixed vascular and neural stem cell features, termed a neurovascular progenitor (NVP). Following in situ validation and molecular characterization of NVP cells in GBM patient samples, we characterized their function in vivo. Genetic depletion of NVP cells resulted in altered tumor cell composition, fewer cycling cells, and extended survival, underscoring their critical functional role. Clonal analysis of primary patient tumors in a human organoid tumor transplantation system demonstrated that the NVP has dual potency, generating both neuronal and vascular tumor cells. Although NVP cells comprise a small fraction of the tumor, these clonal analyses demonstrated that they strongly contribute to the total number of cycling cells in the tumor and generate a defined subset of the whole tumor. This study represents a paradigm by which cell type-specific interrogation of tumor populations can be used to study functional heterogeneity and therapeutically targetable vulnerabilities of GBM.