Sessions

Glioblastoma, IDH-wildtype, is an aggressive central nervous system tumour associated with poor survival and frequent recurrence. Recurrence is characterised by molecular heterogeneity and therapeutic resistance, highlighting the importance of identifying molecular states and vulnerabilities associated with recurrence.

The aim of this study was to characterise transcriptional changes associated with recurrence of IDH-wildtype glioblastoma and to investigate associated pharmacological vulnerabilities using public databases. Paired primary and first-recurrence tumour samples from patients with IDH-wildtype glioblastoma (n=103) were analysed using data from the Glioma Longitudinal Analysis Consortium (GLASS). Following quality control, a cohort of paired primary and first-recurrence tumour samples was established, and differential gene expression was assessed. A consensus recurrence gene-expression profile comprising 265 genes was generated, including 168 genes with increased expression during recurrence (Recurrence-UP) and 97 with decreased expression (Recurrence-DOWN). The GLASS-derived Recurrence Score was defined as the difference between the mean percentile ranks of Recurrence-UP and Recurrence-DOWN genes. This profile was applied to glioma cell lines using Cancer Dependency Map (DepMap) RNA-seq data, and Recurrence Scores were calculated for the cell lines. The cell-line Recurrence Scores were integrated with dose-response data from the Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) database to identify recurrence-associated pharmacological vulnerabilities.

Inhibition of the 90 kDa heat shock protein (HSP90) emerged as a recurrence-associated vulnerability: in a class-level analysis of 12 HSP90 inhibitors, we identified a significant association between the Recurrence Score and sensitivity to HSP90 inhibitors (Spearman ρ=0.440; p=0.0032; 95% CI=0.122–0.678). An independent pharmacogenomic dataset provided support in the same direction. However, analysis based on DepMap CRISPR Gene Effect data did not identify a recurrence-associated genetic dependency involving either the gene encoding the HSP90 alpha isoform (HSP90AA1) or the gene encoding the HSP90 beta isoform (HSP90AB1). Pathway analyses showed associations with heat-shock regulation, protein folding and proteostasis. In contrast, the frozen Recurrence Score was strongly inversely correlated with the proliferation score (n=52; Spearman ρ=−0.596; p=3×10⁻⁶), but after adjustment for proliferation, the Recurrence Score was significantly associated with clinical benefit from selinexor (OR per 1 SD=0.266; 95% CI=0.102–0.695; p=0.0069).

Overall, these findings indicate that recurrence-associated transcriptional changes can be characterised by a reproducible molecular profile and linked to pharmacological vulnerabilities in glioma models. HSP90-related processes represent a recurrence-associated therapeutic vulnerability warranting further mechanistic investigation.