Exploring glioma organization and heterogeneity through spatial transcriptomics

Gliomas are well-known for their high degree of intratumor heterogeneity. Previously, we identified the cellular gene expression states and cellular hierarchies of glioblastoma and IDH-mutant gliomas by scRNA-seq. It is unclear to what extent gliomas are spatially heterogenous and the spatial relationships between the cellular gene expression states. We performed spatial transcriptomics on 20 glioma samples using 10X Visium and integrated our data with 13 samples from an external dataset (Ravi et al. 2022).

Using a multilayer clustering approach, we recovered spatial metaprograms representing conserved gene expression programs across all samples. While most spatial metaprograms were largely consistent with those derived from single-cell data, we identified several states unique to the spatial data including MES-RGL, inflammatory response, chromatin regulation, and metabolism. Proximity to hypoxia conferred greater spatial organization of malignant cells, with malignant cells further from hypoxia displaying increased spatial disorganization.

A systematic classification of samples into structured and disorganized regions revealed that both patterns can be found within the same tumor and even within the same tissue section. While structured areas reveal recurrent spatial preferences between malignant cell states (i.e. astrocyte-like with oligodendrocyte precursor-like), disorganized regions harbor random relationships on various levels, suggesting broader differences between cancer cells of the same expression state based on whether they occur in a structured or disorganized region.