Processed Seurat dataset of endothelial cells from HCC mouse model from single cell RNA-seq (10x genomics) used in the manuscript.

<p dir="ltr">Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality, in part because advanced tumors often fail to respond to immunotherapy. Using <i>in vivo</i> imaging, single-cell transcriptomics, and lesion-level analyses in mouse models of HCC, we discovered that tumor lesions exceeding ~40 mm³ reprogram their endothelium from the characteristic fenestrated, sinusoidal phenotype to an immune-suppressive “capillarized” state, characterized by the loss of endothelial fenestrae. This size-dependent endothelial switch disrupts intravascular antigen recognition by CD8⁺ T cells, which in smaller lesions rely on cytoplasmic protrusions extending through fenestrations to sample tumor antigens. Analyses of clinical trial data at the individual-lesion level similarly reveal a loss of efficacy for immunotherapy beyond a critical tumor size threshold, whereas tyrosine kinase inhibitors do not show this limitation. To uncover the mechanisms underlying this endothelial reprogramming, we performed single-cell transcriptomic profiling and identified a distinct transcriptional program in non-responder lesions, including transcription factors likely responsible for loss of the canonical liver sinusoidal endothelial phenotype. Moreover, spatial transcriptomics in human HCC confirms that activated T cells cluster near vasculature displaying an immune-permissive signature analogous to that in smaller mouse lesions. Finally, we show genetic reprogramming of dysfunctional endothelial cells <i>in vitro</i> restores fenestrations and the canonical sinusoidal phenotype, and pharmacological restoration of endothelial fenestrations <i>in vivo</i> rescues intravascular T cell antigen recognition, together providing proof of principle for strategies aimed at reversing vascular barriers to immunity. These findings identify tumor size as a pivotal determinant of endothelial–immune interactions and underscore the potential of combination strategies—including approaches that normalize or re-engineer tumor vasculature in concert with adoptive T cell therapy (e.g., CAR T cells) or checkpoint blockade—to bolster CD8⁺ T cell–mediated control of advanced liver cancer.</p>