TAMI-48. ENGRAFTMENT PHENOTYPES IN PRECLINICAL MODEL SYSTEMS REVEAL A FUNCTIONAL SUBGROUP OF PATIENT TUMORS DEPENDENT ON THE BRAIN TUMOR MICROENVIRONMENT FOR GROWTH

Abstract

Abstract The derivation of model systems from patient tumors is a requisite for reproducible and high throughput translational cancer research. However, not all tumors can form a model and those that do often fail to capture the molecular diversity specific to their cancer. The potential tumor-intrinsic underpinnings remain largely unknown. In gliomas, the brain tumor microenvironment (TME) is increasingly acknowledged as a regulator of tumor proliferation, invasion, and therapy response. The dissimilar environment of in vitro and heterotopic xenograft models could potentially play a role in the limited fidelity of these model systems. Here we established a culture-free workflow and biobank of 144 glioma direct-from-patient orthotopic xenografts (DPDOX) and 51 parallel gliomasphere cultures (GS). Our direct-from-patient workflow enabled the exclusive in vivo establishment of several gliomas – hereafter termed TME-dependent tumors – including low and high grade mtIDH gliomas and histone H3.3 G34 glioblastomas notoriously difficult to culture in vitro. Through molecular profiling of over 75 patient tumors and their matched derivative models, we find that DPDOX tumors preserve a gene expression signature of neural and glial interactions not found in GS and enriched in brain TME-dependent patient tumors. While these patient tumors span a diversity of clinical diagnoses, network-based inferred transcription factor activity suggests that they converge on shared master regulators of self-renewal driving proneural and OPC/NPC-like cellular state enrichment. Integrating multi-omic profiling from TCGA and other publicly available datasets reveals that this expression signature corresponds to a shared DNA methylation signature across disparate epigenetic subgroups. These findings suggest a brain TME dependence in patient tumors across multiple molecular and clinical classifications of glioma which leads to a lack of representation in model systems failing to recapitulate tumor-promoting components of the TME. Further this work provides a resource to guide translational investigations accounting for influences of the model environment.