TARGETING EPIGENETIC DYSREGULATION IN MEDULLOBLASTOMA WITH POOR PROGNOSIS

Abstract Medulloblastoma (MB) is the most common paediatric malignant brain tumour and is classified into four distinct molecular subgroups (WNT, SHH, G3 and G4), each of them further subdivided into subtypes with different prognosis and responses to therapy. Deregulation of chromatin modifier genes play an essential role in MB, particularly in the G4 subgroup. A BMI1High;CHD7Low molecular signature identifies patients with poor survival within this subgroup. We show that BMI1High;CHD7Low sustains MB growth through regulation of MAPK/ERK signalling and via a novel epigenetic regulation of inositol metabolism in both G4 MB cells and patients. These tumours display over-activation of MAPK/ERK signalling, sustaining tumour proliferation, and of AKT/mTOR pathway which leads to energetic rewiring characterised by enhanced glycolytic capacity and reduced mitochondrial function. We demonstrate that inositol administration counteracts this metabolic alteration, impairs proliferation and significantly extends survival in a pre-clinical model. Moreover, inositol synergises with cisplatin, a chemotherapy agent currently used in MB treatment, enhancing its therapeutic effect in vivo. Additionally, we identify a synergistic vulnerability of BMI1High;CHD7Low MB to a combination treatment with BMI1 and MAPK/ERK inhibitors that overcomes acquired resistance to single-drug therapy. Mechanistically, we observe a CHD7-dependent binding of BMI1 to MAPK-regulated genes underpinning the CHD7-BMI1-MAPK regulatory axis that is critical for the anti-tumour effect of the inhibitors in vitro and in a pre-clinical model. Moreover, we demonstrate that the BMI1High;CHD7Low molecular signature defines G4 MB patients with an enhanced ERK1-ERK2 phosphorylation activity. Importantly, cerebellar neural stem cells modelling the BMI1High;CHD7Low signature are not affected by BMI1 and MAPK/ERK inhibitors and do not show metabolic adaptation hence are resistant to the proposed treatments. In summary, we have identified two actionable vulnerabilities in a pre-clinical setting modelling a molecularly defined group of MB patients, paving the way for the design of signature-matched clinical trials.