Background Microvascular invasion (MVI) is a critical determinant of early recurrence and poor prognosis in hepatocellular carcinoma (HCC). The cellular and molecular mechanisms driving MVI, particularly the role of endothelial heterogeneity, remain incompletely understood. Methods We integrated single-cell and spatial transcriptomics from multiple HCC cohorts to identify MVI-associated endothelial subpopulations. GeneNMF and hdWGCNA were applied to define stable gene modules, while CellChat and NicheNet analyzed intercellular signaling. Prognostic relevance was assessed using machine learning–based risk models across TCGA-LIHC and external datasets. Functional assays including EdU proliferation, tube formation, and Western blot were conducted to validate the role of IGF2BP3 in tumor-associated endothelial cells. Results We identified a distinct MVI-associated endothelial subpopulation (MVI Endo) enriched in pro-angiogenic, EMT-like, and TGF-β-responsive pathways, predominantly localized near tumor vasculature. MVI Endo engaged in robust VEGF, ANGPT2, and TGF-β signaling with tumor and stromal cells. Machine learning–derived prognostic models highlighted IGF2BP3 as a central regulator, and functional assays demonstrated that IGF2BP3 enhances endothelial proliferation and tube formation via upregulation of VEGF-A and ANGPT2, supporting its involvement in the vascular phenotype associated with MVI. Conclusion This study characterizes the endothelial landscape associated with MVI, implicates IGF2BP3 as a critical regulator of vascular reprogramming, and provides a foundation for precision vascular-targeted therapies in HCC.
Zhang et al. (2026) studied this question.