Single-cell RNA sequencing reveals vascular heterogeneity and cell–cell communication remodeling in the glioma blood–tumor barrier

Abstract

Background

The blood–brain barrier (BBB) restricts drug delivery in glioma, yet its pathologically remodeled counterpart, the blood–tumor barrier (BTB), is a promising therapeutic target. A comprehensive view of the cellular composition and transcriptional heterogeneity of the vascular compartment across the BBB-to-BTB transition is essential for improving glioma therapy.

Methods

We performed a secondary single-cell RNA sequencing (scRNA-seq) analysis of the public GSE242044 dataset (7 patients; 13 tumor and adjacent non-tumor surgical specimens, comprising IDH-mutant lower-grade gliomas and IDH-wild-type gliomas). After stringent quality control, doublet removal, Harmony batch correction, and annotation with SingleR plus canonical markers, we dissected vascular cell heterogeneity and performed differential expression, Gene Set Variation Analysis (GSVA), pseudotime (Monocle/Slingshot), and CellChat ligand–receptor inference, with sample-level statistical comparisons (Kruskal–Wallis and Wilcoxon rank-sum tests on per-specimen values). Findings were assessed in three independent glioma scRNA-seq cohorts (GSE162631, GSE117891, GSE174554) and in the CGGA bulk cohort.

Results

Vascular cell composition (relative representation among the profiled vascular-enriched cells) showed trends across conditions: endothelial cells declined from 60.6 % to 47.5 %, whereas pericytes (15.8 % to 18.0 %) and immune cells increased in IDH-wild-type tumors (macrophages 1.3 % to 7.0 %; T cells 3.0 % to 7.5 %); astrocytes increased from 1.1 % to 4.5 %; these specimen-level trends were mostly not statistically significant (Supplementary Table S1). Cell–cell communication was predicted to be rewired, shifting from pericyte–smooth muscle cell crosstalk in non-tumor and IDH-mutant tissues toward pericyte–astrocyte and predicted vascular–immune interactions in IDH-wild-type tumors. Differential expression and enrichment analyses identified an IDH-wild-type vascular program dominated by extracellular matrix remodeling (COL1A1, COL3A1, COL4A1, LAMB1, TIMP1) and angiogenic activation (ANGPT2), with loss of quiescent endothelial markers (CA4, KLF2). Candidate markers (COL1A1, COL3A1, F3, S100A4, CSPG4, ANGPT2) showed no consistent prognostic association across independent cohorts.

Conclusions

This secondary analysis delineates transcriptional heterogeneity of BBB/BTB vascular cells and their inferred communication remodeling in glioma. The identified subtype markers are candidate biomarkers that require independent validation; our findings provide a framework for future studies—including spatial and protein-level validation of the candidate markers, functional permeability and coculture assays of the predicted vascular–immune interactions, and preclinical testing of BTB-modulating delivery strategies such as focused-ultrasound-mediated opening in glioma models.