DYRK1A-mediated quiescence establishes a therapeutically-resistant reservoir for glioblastoma recurrence

Glioblastoma (GBM) is the most common and aggressive primary brain cancer in adults, owed to its extensive intratumoral heterogeneity and rapid disease progression. It is hypothesized that standard-of-care radio- and chemotherapy (temozolomide) treatments act as evolutionary pressures to select a minority subpopulation of temporally stable cells, which seed aggressive, therapeutically resistant recurrent tumors. This study aims to identify and target the therapeutic vulnerabilities of treatment-resistant GBM cells at diagnosis to proactively address disease progression. We developed and applied a deep learning model to analyze whole patient-matched primary and recurrent GBM specimens to nominate temporally stable cell populations by conserved morphology, successfully identifying these cells in 31 of 40 analyzed cases. Morphologic feature analysis of nominated populations revealed irregular nuclear shape metrics consistent with mesenchymal histology, including reduced nucleus circularity (-8.6%; p <0.0001; n =18), increased nuclear axis length (+8.2%; p <0.01; n =18) , and increased cell area (+13.0%; p <0.05; n =18). Features of stable cells further correlated with a 38.6% reduction in overall survival in an external glioma cohort ( p <0.05; n =51). To define molecular mechanisms underpinning stability, we performed mass spectrometry-based proteomic profiling of matched clinical specimens, which revealed the enrichment of DYRK1A-mediated cell quiescence programs via the DREAM complex as a potential mechanism for chemotherapy evasion ( p <0.01; n =36). Consistent with a quiescence-driven resistance mechanism, Ki67 immunohistochemistry staining demonstrated a 47.5% reduction in the proliferative index of stable subpopulations in the primary tumor compared to the tumor bulk ( p <0.05; n =9). To model these observations in vitro, we cultured CRISPR-Cas9 DYRK1A overexpressing GBM cells in an ex vivo mouse brain slice model and evaluated the effects of temozolomide administration. DYRK1A overexpression was found to ablate chemotherapy-induced GBM cell death in patient-derived cultures, compared to wildtype controls (p<0.05), confirming a functional role for DYRK1A in establishing therapeutic resistance in vitro. Pre-clinical studies of DYRK1A/DREAM complex inhibition have been shown to reverse tumour cell quiescence and sensitize resistant cells to chemotherapy in extra-CNS cancers. These findings extend this concept to GBM and support strategies to target drivers of therapeutic resistance at diagnosis. Current therapeutic approaches in GBM are limited to reducing disease burden; this study informs interventions to proactively anticipate and interrupt disease progression.