DMG-26. Uncovering the molecular basis for distinct disease characteristics between subtypes of pediatric high-grade diffuse gliomas

Abstract Pediatric-type high-grade diffuse gliomas (pHGGs) are aggressive brain cancers that lack effective treatments and portend a poor prognosis. Many of these cancers are driven by mutations in histone H3-encoding genes, affecting either lysine at position 27 (K27) or glycine at position 34 (G34). Interestingly, specific combinations of genetic lesions are associated with distinct disease characteristics, including localization in the brain, age of onset, and disease aggressiveness. It is critical to uncover the molecular basis for these distinctions, as they could point to tumor subtype-specific treatments. Towards this goal, we developed several knock-in mouse models that carry genetic lesions corresponding to those seen in patients. Crucially, these lesions were engineered in the endogenous genes, and targeted to specific cell types in the brain. Thus, our models replicate key disease features, including physiological levels of oncogene expression, spontaneous tumor development, and cancer progression within an intact micro-environment. For example, we have generated models for two main subtypes of diffuse midline gliomas that both carry H3-K27M and ACVR1 mutations, but differ by mutually exclusive lesions in TP53 or PIK3CA. The resulting mice show distinct phenotypes and survival. Notably, we observed a strong synergy between the H3-K27M, ACVR1, and PIK3CA mutations, such that each individual mutation substantially increases the severity of the disease driven by the two other lesions. Finally, using functional genomic screens in human pHGG cells, we surprisingly observed that some tumors are sensitive to the loss of genes whose inactivation is predicted to have an opposite oncogenic effect in other pHGGs . Overall, our results contribute to identify critical distinctions between pHGG types, and could inform the development of patient-specific treatments.