Transcriptomic characterization of patient gliomas and derived model systems reveals environmental influence on neurodevelopmental cellular states

Abstract Bulk tumor and single-cell RNA sequencing have revealed the remarkable molecular heterogeneity and plasticity of gliomas, leading to the definition of tumor subtypes and cellular states describing inter- and intra-tumoral levels of heterogeneity respectively. While there has been great interest in creating and revising these classifications, the combination of selective pressures from molecular alterations within the tumor cell as well as extrinsic factors within the tumor microenvironment (TME) driving these levels of heterogeneity remain to be fully described. The brain TME is a complex, regionally heterogeneous ecosystem of communicating non-malignant and malignant cell types and scavenge-able nutrients and metabolites. We hypothesized that distinct environmental contexts within and across tumors may drive the intra- and inter-tumoral heterogeneity of gliomas. To identify tumorigenic programs impacted by environmental context, we performed bulk RNA sequencing of over 40 triplets of patient glioma samples and their matched derivative models established in direct orthotopic mouse xenografts and conventional gliomasphere cultures. Comparative analyses revealed environment-associated programs including in vivo immune and neuroglial signaling as well as altered in vitro lipid metabolism and upregulated cell migration. Next, to investigate associations between tumor cellular state and environment-associated programs we performed single-cell RNA sequencing of 4 patient and model system triplets. By annotating single cell profiles with previously defined single cell atlases of the developing and adult brain, 9 major states reflecting neurodevelopmental cell types were identified. Single cell profiling further revealed cell state-specific expression of environment-associated programs. We simultaneously observed the divergence of model cellular state compositions towards states expressing programs associated with their environment. We conclude that the transcriptional evolution of cellular states is connected to their ecological role within the TME. Consequently, we find grade- and diagnosis-associated composition differences between patient tumors predictive of their ability to establish a model in vitro.