Abstract Glioblastoma (GBM), the most common and aggressive primary brain tumor, remains invariably refractory to therapy. The cellular heterogeneity of GBM is described in terms of four dominant transcriptional cellular states, two stem-like cellular states – neural progenitor cell-like (NPC-like) and oligodendrocyte progenitor cell-like (OPC-like) – and two more differentiated cellular states – astrocyte-like (AC-like) and mesenchymal-like (MES-like). In addition, cellular barcoding experiments evidenced the ability of GBM cells to transition between cellular states. These two hallmarks of GBM – cellular state heterogeneity and plasticity – underlie its therapeutic resistance. Prior studies indicate that environmental signals from the tumor microenvironment – such as secreted ligands from immune cells and neuronal input – play a role in driving specific cell states. We hypothesize that, (i) there are also epigenetic determinants of cellular states in GBM as evidenced by preferentially accessible transcription factor (TF) binding motifs, and that (ii) overexpression of specific TFs can drive cell state transitions. To test our hypotheses, we concurrently profiled single-nucleus gene expression (via RNA-seq) and chromatin accessibility (via assay of transposase-accessible chromatin using sequencing, ATAC-seq) in 27 GBM patient samples. Cell type annotations show our data accurately capture the malignant and microenvironmental composition of the tumors profiled. Leveraging our ATAC data, we identify DNA accessible regions within cells and samples via peak calling, subsequently defining robust peak sets across samples, and defining cis-regulatory element (CRE) modules in our data set. Applying this approach to our malignant cells, we identify 6 malignant CRE (mCRE) modules in GBM, including 2 shared mCREs enriched for mRNA processing and cell cycle regulation pathways, as well as 4 mCREs enriched for pathways associated with gene regulation of NPC-like, OPC-like, MES-like and AC-like programs. By scoring malignant cells for both transcriptional cell state and mCRE, we show that intermediate glial progenitor-like cell states have greater chromatin accessibility overlap with both neural progenitor (NPC-like) and more differentiated (AC-like) cellular states, suggesting they represent nodes for cell state transitions. We then identified TF-binding motifs enriched in cell-state specific mCRE modules and leverage this information along with expression data to nominate TFs hypothesized to drive cell state transitions. In the mCREs corresponding to more differentiated cell states (MES-like/AC-like), nominated TFs include those of the AP-1 family (Jun, JunB, Fos, FosL1, etc. ), while in the mCREs corresponding to stem-like cell states (NPC-like/OPC-like), we nominate neurodevelopmental TFs such as POU3F2, OLIG2, and NHLH1. Through TF overexpression in patient-derived GBM models, we demonstrate the ability of nominated TFs to drive cell state transitions. Altogether, our work proposes and validates a model of the epigenetic determinants of cellular state in GBM. Citation Format: L. Nicolas Gonzalez Castro, Min Yang, Alexander Jucht, Kevin Johnson, Noam Hadary, Sydney Dumont, Lillian Bussema, Roel Verhaak, Itay Tirosh, Mario L. Suvà. Epigenetic determinants of transcriptional cellular states in glioblastoma abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (6Suppl): Abstract nr B033.
Castro et al. (Mon,) studied this question.