DIPG-38. PATIENT-DERIVED THREE-DIMENSIONAL AVATARS REVEAL CELL-AUTONOMOUS MECHANISMS DRIVING INVASION AND METASTATIC PROGRESSION IN DIFFUSE MIDLINE GLIOMAS

Diffuse midline gliomas H3K27-altered are pediatric tumors associated with dismal survival after diagnosis that are characterized by their ability to infiltrate the entire central nervous system. We show that the extent of tumor invasion and metastatic progression represent critical adverse prognostic factors affecting the survival of DMG patients. Yet, no genetic alterations associate to the tumor progression patterns. We therefore used glioma stem cells collected at diagnosis to create patient-specific 3D avatars. We found that their invasive behaviour faithfully predicts metastatic relapse in patients, thus proving DMG spread is mainly an autonomous ability of tumor cells. This relevant modelling strategy revealed the existence of phenotypic entities associated with distinct migration modalities and capacities. Using transcriptomics of both organoids and primary tumors, we identified the biological programs associated with variable invasiveness. Importantly, we show that Bone Morphogenetic Protein 7 (BMP7) signalling is a crucial tumour-autonomous regulator of DMG invasiveness via the control of a phenotypic switch. We further deciphered effectors downstream of BMP7 stimulation as actionable targets to reduce invasion. These findings have two possible therapeutic applications. First, glioma stem cell 3D models represent a tool for tumor phenotypic stratification to rapidly predict patients with exacerbated invasiveness and increased risk of metastatic progression. This may allow to propose whole-brain or cranio-spinal irradiation within weeks after diagnosis to prevent the appearance of distant foci. Second, targeting autocrine and short-range BMP7-related signalling may become a useful adjunct to the current therapeutics, which are mainly directed to control the disease locally.