Integrative spatial analysis reveals a multi-layered organization of glioblastoma

Abstract

Glioma contains malignant cells in diverse states. Hypoxic regions are associated with a unique histology of pseudopalisading cells, while other regions appear to have limited histological organization, reflecting the diffuse nature of glioma cells. Here, we combine spatial transcriptomics with spatial proteomics and novel computational approaches to define glioma cellular states at high granularity and uncover their organization. We find three prominent modes of cellular organization. First, cells in any given state tend to be spatially clustered, such that tumors are composed of small local environments that are each typically enriched with one major cellular state. Second, specific pairs of states preferentially reside in proximity across multiple scales. Despite the unique composition of each tumor, this pairing of states remains largely consistent across tumors. Third, the pairwise interactions that we detect collectively define a global architecture composed of five layers. Hypoxia appears to drive this 5-layered organization, as it is both associated with unique states of surrounding cells and with a long-range organization that extends from the hypoxic core to the infiltrative edge of the tumor. Accordingly, tumor regions distant from any hypoxic foci and tumors that lack hypoxia such as IDH-mutant glioma are less organized. In summary, we provide a conceptual framework for the organization of gliomas at the resolution of cellular states and highlight the role of hypoxia as a long-range tissue organizer.