In vitro experiments from our laboratory and others have suggested that herpes simplex virus thymidine kinase (HSV-TK)/ganciclovir (GCV) gene therapy depends on gap junctional intercellular communication (GJIC) to produce a strong bystander effect. Furthermore, we have shown that cells transduced with HSV-TK can be protected from GCV-mediated toxicity by GJIC with bystander cells. We wished to determine whether GJIC affected either the bystander or protective effect of the cytosine deaminase (CD)/5-flucytosine (5-FC) gene therapy approach, in which CD converts 5-FC to 5-fluorouracil (5-FU). To test this, we designed a coculture system using communication-competent WB rat hepatocytes and a noncommunicating subclone (aB1), which were transduced with CD and with antibiotic resistance genes so that we could independently determine the survival of the CD-containing or bystander cells. We found that, compared to the HSV-TK/GCV strategy, bystander killing resulting from treatment with CD/5-FC does not depend on GJIC. However, our most striking finding was that both communication-competent and -incompetent CD-transduced cells were preferentially killed, by a factor of up to 500, compared to bystander cells. The lesser dependence of the CD/5-FC system on GJIC, combined with the finding that most cancer cells lack the capacity for GJIC, suggest that the CD/5-FC system may be superior to the HSV-TK/GCV approach for gene therapy. However, the premature death of the CD-transduced 5-FU "factory" suggests that other strategies may be necessary to produce a sufficient quantity of 5-FU for a duration long enough to produce permanent tumor regression.