Recent experiments with genetically engineered tumors have generated renewed interest in active cellular immunotherapy as a cancer treatment modality. In order to consider the use of live tumor cells for immunotherapy in human cancer patients, it will be important to ensure that these cells do not themselves produce morbidity in the event the immune system fails to eliminate them. Toward this end, we have examined a strategy for eliminating genetically manipulated nonimmunogenic tumors in vivo. When B16F10 melanoma cells were transfected with the Herpes simplex virus 1 thymidine kinase (HSV-TK) gene, cells were rendered susceptible to killing by the nucleoside analogs acyclovir (ACV) and ganciclovir (GCV). B16-HSV-TK+ tumors established in C57BL6 mice were successfully "suicided" in vivo when GCV was administered by continuous infusion. However, late recurrences were observed even after 1 month of continuous GCV treatment. In vivo growth kinetics suggested that the recurrences resulted from a tiny number (< 20) of cells that had survived the GCV treatment. Interestingly, recurrent tumors were as sensitive to GCV as the parental B16-HSV-TK+ line. While these results demonstrate potential feasibility of the suicide gene strategy for active immunotherapy with live tumor cells, they also illustrate that approaches dependent on the intracellular generation of cell cycle-dependent toxins may fail to eliminate small numbers of cells that temporarily exit cell cycle or that are pharmacologically sequestered.