P01.06.B THERAPEUTIC REPURPOSING REVEALS NOVEL INTERVENTIONS FOR HETEROGENEOUS GLIOBLASTOMA

Abstract BACKGROUND The clinical management of glioblastoma (GBM) has remained unchanged since 2005, where the one-size-fits-all approach of surgical resection, radiotherapy, and chemotherapy was established. A critical factor in this therapeutic failure is the profound intratumoral heterogeneity of GBM linked to treatment resistance. There is an urgent need to explore novel treatment options, including drug candidates beyond the classical repertoire of antineoplastic agents. MATERIAL AND METHODS We developed a pipeline to identify drugs suitable for repurposing in GBM consisting of drug prediction via graph-based approaches, sophisticated screening and examination of mechanism of action. Selected candidates were screened in two patient-derived neurosphere (PDN) lines that replicate levels of intratumoural heterogeneity seen in the patients’ tumour tissue. We screened 8 compounds across a range of therapeutic classes via a live-cell imaging system (IncuCyte) and generated single cell RNA-seq (Parse Biosciences) of PDNs treated at concentrations inhibiting 25% and 50% of growth. RESULTS Our identification pipeline successfully revealed a diverse range of therapeutics, including classical antineoplastics as well as unexpected compounds such as anti-epileptics and anti-inflammatories, highlighting the power of graph-based methods to uncover hidden biological relationships and novel therapeutic opportunities. 80% of identified compounds inhibited growth of PDNs, with potencies varying between models, reflecting their heterogeneous nature. Single-cell sequencing revealed unique and shared gene expression programs across different compounds and PDN models, enabling characterisation of individualised responses in a dose-dependent manner and examination of how tumour state composition affects overall responses. Leveraging the broad gene-body coverage of our dataset, we interrogated alternative splicing events and uncovered compound-specific differential splicing patterns, further characterising responses. Although PDNs were exposed to each compound as a monotherapy, our computational framework was then used to predict personalised drug combinations for each PDN model, which were further validated in vitro. CONCLUSION There is an urgent unmet need for novel and effective therapeutics for GBM patients. Our work demonstrates that identification and validation of compounds outside common antineoplastic agents can yield promising results, and that these therapies can synergise. Our end-to-end pipeline provides a comprehensive framework for the discovery and development of new personalised therapeutic approaches for GBM.