Tmic-70. harmonized single-cell landscape, intercellular crosstalk and tumor architecture of glioblastoma

Abstract Glioblastoma (GBM) is the most aggressive and deadliest brain tumor in adults. Single-cell RNA sequencing (scRNA-seq) has helped to grasp the complexity of the cell states and dynamic changes within the tumor microenvironment (TME). To better capture the molecular underpinnings of transcriptomic variation across different individuals and reconstruct the intricate TME, it is necessary to gather a significant cohort that can empower our understanding of the complex GBM biology. In addition, little is known about the spatial disposition of the cells within the tumor, neighboring patterns, and intercellular crosstalk. Taking advantage of the recent developments in computational methods, we set out to build the GBM cellular landscape and created the ‘GBM-VERSE’, a curated reference that combines 16 datasets, and 113 patients spanning over 330 thousand cells. Our resource characterizes cells at different levels (coarse and detailed), identifying gene programs that define each cell type/state and providing a harmonized annotation. We demonstrated the applicability of the GBM-VERSE to map for newly generated data, providing a robust annotation tool for other GBM scRNA-seq datasets. We expanded the GBM-VERSE through transfer learning which increased the likelihood of identifying under-represented phenotypic states and incorporating cell types that were not present in the core reference. From the biological point of view, the GBM-VERSE facilitated the construction of an inferred global map of the cell-cell interactome. Our analysis highlighted the crucial role of MES-like cancer cells as the primary source of pro-angiogenic signaling. We employed the GBM-VERSE to chart the spatial architecture of GBM by deconvolving next-generation sequencing-based ST data. We found a well-defined spatiotemporal distribution of the neoplastic cells that were further inspected using an in situ sequencing-based platform. The GBM-VERSE represents a framework that allows the streamlined integration, and interpretation of new data and provides a platform for exploratory analysis, hypothesis generation, and testing.