Efforts to selectively target and disrupt established tumor vasculature have largely failed to date. We hypothesized that a vaccinia virus engineered to target cells with activation of the ras/MAPK signaling pathway (JX-594) could specifically infect and express transgenes (hGM-CSF, β-galactosidase) in tumor-associated vascular endothelial cells in humans. Efficient replication and transgene expression in normal human endothelial cells in vitro required either VEGF or FGF-2 stimulation. Intravenous infusion in mice resulted in virus replication in tumor-associated endothelial cells, disruption of tumor blood flow, and hypoxia within 48 hours; massive tumor necrosis ensued within 5 days. Normal vessels were not affected. In patients treated with intravenous JX-594 in a phase I clinical trial, we showed dose-dependent endothelial cell infection and transgene expression in tumor biopsies of diverse histologies. Finally, patients with advanced hepatocellular carcinoma, a hypervascular and VEGF-rich tumor type, were treated with JX-594 on phase II clinical trials. JX-594 treatment caused disruption of tumor perfusion as early as 5 days in both VEGF receptor inhibitor-naïve and -refractory patients. Toxicities to normal blood vessels or to wound healing were not evident clinically or on MRI scans. This platform technology opens up the possibility of multifunctional engineered vaccinia products that selectively target and infect tumor-associated endothelial cells, as well as cancer cells, resulting in transgene expression, vasculature disruption, and tumor destruction in humans systemically.