This report examines a major barrier to suicide gene therapy in cancer and other diseases: namely, bystander cell killing. Existing vectors for in vivo gene delivery are inefficient and often transduce or transfect less than 1% of target cells. The E. coli PNP gene brings about cellular necrosis under conditions when 1 in 100 to 1 in 1000 cells express the gene product in vitro. In vivo bystander killing at or near this magnitude has not been reported previously. In the present experiments, transfection of cells with the E. coli PNP gene controlled by a SV40 promoter resulted in 30 nmol 6-methyl purine deoxyriboside (MeP-dR) converted per milligram tumor cell extract per hour (or conversion units (CU)). This level of expression led to elimination of entire populations of tumor cells in vitro after treatment with MeP-dR. Much earlier killing was observed using a tat transactivated E. coli PNP vector (approximately seven-fold higher activity, 230 CU). In vivo effects on tumor growth were next examined. Human ovarian tumors transfected with E. coli PNP were excised 5 days after i.p. implantation from the peritoneal cavities of mice in order to determine both E. coli PNP enzymatic activity and the fraction of cells expressing the gene. PNP activity at 5 days after gene transfer was approximately 170 CU and was expressed in approximately 0.1% of the tumor cells as judged by in situ hybridization. The expression of E. coli PNP at this level produced a 30% increase in life span (P < 0.001) and 49% reduction in tumor size (P < 0.005) after MeP-dR treatment, as compared with control tumors. Our observations lead to the conclusion that pronounced bystander killing by E. coli PNP is conferred in vivo, and that vectors capable of transgene expression in as few as one in 1000 cells can produce substantial antitumor effects if expression on a per cell basis is very high.