The Mesenchymal State Drives Stemness and Limits Differentiation in Glioblastoma

Glioblastoma (GBM) is the most common and lethal primary brain tumor in adults, characterized by high heterogeneity and limited therapeutic options. Recent single cell omics studies have emphasized the importance of the proneural-mesenchymal axis and strengthened the link between GBM and neurodevelopment. Furthermore, GBM comprises a well-known stem-like cell compartment known as glioblastoma stem cells (GSCs), responsible for most of the tumor’s malignant features. Targeting GSCs by inducing their differentiation into less malignant cancer cells holds great therapeutic potential. Yet, despite recent advances in the field, the identity and regulatory drivers of GSCs remain poorly defined. In this work, I combined published single-cell dataset analysis with in vitro models to investigate GBM stemness and the tumor’s response to differentiation cues. First, through integrative analysis of single-cell RNA sequencing data, I refined the landscape of GBM cellular states, revealing a high abundance of mesenchymal-like and astrocyte-like states, and a correlation between developmental and inflammatory phenotypes. RNA velocity analysis showed that the hierarchical architecture of GBM is confined to the proneural compartments, while mesenchymal and astrocytic states are more heterogeneous, with phenotypes determined by stemness, stress, and immune interactions. Moreover, I observed promiscuous stemness marker expression across states, supporting the concept of GBM stemness as a plastic gradient rather than a hierarchy. Comparison with human subventricular zone profiles revealed a striking similarity between a mesenchymal-like cell subpopulation and the adult radial glia. By integrating single-nucleus ATAC sequencing data, I highlighted the AP-1 transcriptional network as a central driver of the mesenchymal, radial glia-like state of GBM cells. I further established a comprehensive differentiation assay incorporating phenotypic, transcriptomic, and proteomic readouts. Applying this system to a comprehensive panel of GSCs, I demonstrated that GBM cells exhibit heterogeneous responses to differentiation signals, with marked discrepancies in transcript and protein changes for both stemness and lineage-specific genes, particularly in GSCs that are insensitive to differentiation cues. Finally, RNA sequencing analysis uncovered that cells with strong mesenchymal phenotypes display pronounced resistance to lineage commitment. Together, these findings highlight the mesenchymal state as a key barrier to differentiation in glioblastoma, laying the groundwork for targeting its underlying drivers with the aim of restoring differentiation capacity and potentially improving therapeutic outcomes.