Deconvolution of spatiotemporal transcriptomic heterogeneity in the glioblastoma ecosystem

Abstract
Glioblastoma (GBM) is the most common and lethal primary brain malignancy in adults, characterized by therapeutic resistance and inevitable relapse. Major clinical challenges are imposed by extensive intra-tumoral heterogeneity, diffuse infiltrative growth of the tumor, and bi-directional interactions with diverse non-malignant cell-types within the brain tumor microenvironment (TME). A better understanding of how the tumor cells organize spatially and interact with the TME to promote growth and invasion may reveal opportunities for improved therapeutic strategies. In this thesis, I explore transcriptional heterogeneity within the glioblastoma ecosystem using spatially profiled, temporal samples of GBM xenograft models. The species-specific distinction of the human tumor and mouse TME, coupled with spatial resolution, overcomes previous limitations in studying the invasive front and delineating co-existing non-malignant components within the tumor. By applying a novel computational framework based on unsupervised deconvolution, I characterize a compendium of 15 tumor cell gene expression programs set within the context of 90 mouse brain and TME cell types, cell activities, and anatomic structures. This approach reveals the spatial organization of tumor programs along an axis corresponding to tumor density and distinct colocalization patterns with spatiotemporally varying TME components. Notably, tumor-associated macrophages and reactive astrocytes colocalized with the tumor early in its growth, while an outward gradient of invasion programs, centered on hypoxia, was observed with tumor progression. Moreover, distinct tumor programs aligned with well-documented routes of GBM invasion including the white-matter tracts, perivasculature and parenchymal routes. Ligand-receptor analyses highlighted neuronal and extra-cellular matrix (ECM) signaling along these routes, and further analyses indicated that these routes could be distinguished by the expression of tumor and TME-derived ECM molecules. Lastly, using a network-graph of predicted protein-protein interactions, I identified sub-modules of genes serving as program network hubs that were highly prognostic in patient datasets. Taken together, spatial profiling of xenografts has revealed a granular repertoire of transcriptional programs and provides a basis for rational targeting of tumor and/or TME niches within the GBM ecosystem, paving the way for improved therapeutic interventions.
Description
Keywords
spatial transcriptomics, glioblastoma
Citation
Thoppey Manoharan, V. (2024). Deconvolution of spatiotemporal transcriptomic heterogeneity in the glioblastoma ecosystem (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.