ANGI-07. EPIGENETIC MODULATION OF TRANSCRIPTION FACTOR BINDING DRIVES INVASIVE GLIOBLASTOMA PHENOTYPES

Abstract Glioblastoma is characterized by highly invasive behavior that contributes to therapeutic failure and recurrence, yet the stability and regulatory mechanisms of the invasive phenotype remain poorly understood. We established a serial selection model in patient-derived glioblastoma stem-like cells to generate an isogenic subpopulation with stable invasive potential (GBinv⁺) and compared it to the non-invasive counterpart (GBnaïve). GBinv⁺ cells showed a ~10-fold increased invasion in transwell and 3D spheroid assays, a phenotype maintained across passages and independent of proliferation. Integrative RNA sequencing and DNA methylation profiling identified 325 differentially expressed genes and ~6,700 differentially methylated CpG sites, enriched in promoters and enhancers. Invasive cells upregulated a distinct secretome enriched in ECM-modulating factors, including COL7A1, ICAM1, and P4HA2. Transcription factor activity was inferred using ISMARA, revealing increased activity of stress-related regulators (ATF4/6, IRF7) and reduced activity of MYC, E2F, and MYCN. Notably, 24 invasion-associated TFs contained CpGs within their core motifs, suggesting methylation-sensitive regulation. Motif enrichment at differentially methylated sites highlighted E-box motifs bound by bHLH factors (MYC, MAX, MNT), with hypermethylated regions showing reduced co-activator marks, whereas hypomethylated enhancers were enriched for MYC/MAX/E2F1 occupancy. Public ChIP-seq data confirmed that methylation states correlate with chromatin accessibility and TF binding. Integrated analysis demonstrated that although the global correlation between methylation and expression was weak, specific genes such as TNC and FLNC exhibited coordinated methylation-expression changes, identifying a subset of epigenetically modulated invasion effectors. These results support a model in which glioblastoma invasion is sustained by epigenetically reprogrammed transcription factor networks rather than transient environmental cues. Future work will use CUT&Tag to map genome-wide methylation-sensitive TF binding, focusing on MYC/MAX/MLX complexes in patient-derived glioma samples.