Abstract

Tumours are complex ecosystems comprised of tumour cells and various normal cell types. The interactions among these cells with diverse molecular characteristics lead to intratumoral heterogeneity and cancer evolution. However, little is known about the spatial organisation of the glioma tumour ecosystem, the cellular interactions therein, and their influence on patient outcomes. Spatially resolved transcriptomics, which measures the activity of hundreds of genes in situ, offers new avenues to address these long-standing questions.

We assembled a clinically annotated cohort of glioma patients spanning 278 specimens from low- and high-grade IDH-mutant astrocytoma, IDH-mutant and 1p/19q co-deleted oligodendroglioma, ependymoma, and glioblastoma. Using tissue microarrays from archival FFPE, we performed spatial transcriptomics profiling with the 10x Xenium platform. A custom panel of 344 genes, selected based on publicly available single-cell RNA sequencing datasets of gliomas, was interrogated.

The spatial transcriptomics data enabled annotation and spatial mapping of more than 20 tumour microenvironment (TME) cell types and various tumour cell types present across the 5 glioma entities. We observed broadly conserved transcriptional profiles of TME cell types across gliomas, but TME cell type composition and spatial organisation varied depending on tumour entity. In particular, lower amounts of macrophages and microglia as well as oligodendrocytes and oligodendrocyte progenitor cells were observed in oligodendroglioma tumours compared to other entities. High-grade astrocytoma samples contained more myeloid dendritic cells and monocytes in comparison to low-grade astrocytomas. A lower proportion of vascular cells correlated with better overall survival for glioblastomas and low- and high-grade astrocytomas. Tumour cells exhibited mostly entity-specific gene expression, but also distinct tumour cell states including AC-, OPC-, NPC-, and MES-like cell states.

Our work demonstrates the utility of the 10x Xenium platform for high-resolution spatial profiling of expression in archival FFPE samples. Preliminary mapping of the TME and tumour cell types unravelled a complex spatial architecture of the glioma ecosystems. Distinct compositions of the TME were found in each glioma subtype, which may help understand different mechanisms of malignancy. Associations between cellular spatial patterns and clinical data available for FFPE samples have the potential to provide better insights for improved patient stratification and treatment.