We and others have proposed mammalian cells as gene delivery vehicles with the potential for overcoming physiological barriers to viral vectors. To that end, we previously have shown the potential of CD34+ endothelial progenitors for systemic gene delivery in a primate angiogenesis model. Here we seek to explore the utility of CD34+ cells of human origin as vehicles for toxin genes and, in particular, to measure their capacity to effect a cytotoxic bystander effect in human endothelium and tumor cells. To this end, CD34+ cells were transduced with TOZ.1, a nonreplicative herpes simplex vector encoding thymidine kinase. To test the capacity of CD34+ cells to induce a cytotoxic bystander effect in target cells, we performed mixing experiments, whereby TOZ.1-transduced CD34+ cells were mixed with either human vascular endothelial cells or human ovarian tumor cells (SKOV3.ip1). Cell viability was measured by the MTS assay. Lastly, mixtures of TOZ.1-transduced CD34+ cells and SKOV3.ip1 tumor cells were injected s.c. to evaluate the bystander effect in vivo. After transduction of CD34+ cells with TOZ.1, treatment with ganciclovir induced the killing of 99% of cells. In cell-mixing experiments, a linear correlation was observed between the percentages of TOZ.1-transduced CD34+ cells and total cell killing. For example, when 50% of CD34+ transduced cells were mixed with nontransduced SKOV3.ip1, >70% of all cells died. Similarly, when the same percentage was mixed with human vascular endothelial cells, > 80% of the total number of cells died. In vivo studies showed an abrogation of tumor formation when TOZ.1-transduced CD34+ cells and ganciclovir were administered. Our observations establish the feasibility of a method for cell-based toxin gene delivery into disseminated areas of tumor angiogenesis.