Reduced Gene Module for Spatial Transcriptomics Recapitulates Single-Cell Molecular Classifications of Glioblastoma.

Abstract

Glioblastoma, IDH-wildtype, is the most common and malignant adult-type diffuse glioma. It is characterized by extensive heterogeneity and poor prognosis. Recent single-cell RNA-sequencing studies have revealed distinct transcriptional programs that define cellular states such as astrocyte-like, mesenchymal-like, neural-progenitor-cell-like, and oligodendrocyte-progenitor-cell-like. However, their spatial context remains unclear because of the limitations of single-cell and spot-based spatial transcriptomics. In the present study, we applied a single-cell molecular classification of glioblastoma to a spatial transcriptomic analysis using the Xenium in situ platform. A custom gene panel comprising 366 genes, including 100 additional genes beyond the predesigned central nervous system panel, was designed to enable high-resolution, cell-level analysis. We constructed reduced modules derived from the Neftel et al. classification and validated their reproducibility across public single-cell datasets, achieving approximately 80% concordance with the original modules. In three cases of glioblastoma, IDH-wildtype, spatial transcriptomic profiling revealed regional variations in transcriptional states; cellular tumor regions, regions of pseudopalisading necrosis were significantly enriched in reduced neural-progenitor-like-cells subtype 1, and reduced mesenchymal-like-cells subtype 2, respectively. Our data indicate that the novel reduced module provides a framework for investigating single-cell molecular states within a tissue context in glioblastoma.