Retroviral replicating vectors (RRVs) are a nonlytic alternative to oncolytic replicating viruses as anticancer agents, being selective both for dividing cells and for cells that have defects in innate immunity and interferon responsiveness. Tumor cells fit both these descriptions. Previous publications have described a prototype based on an amphotropic murine leukemia virus (MLV), encoding yeast cytosine deaminase (CD) that converts the prodrug 5-fluorocytosine (5-FC) to the potent anticancer drug, 5-fluorouracil (5-FU) in an infected tumor. We report here the selection of one lead clinical candidate based on a general design goal to optimize the genetic stability of the virus and the CD activity produced by the delivered transgene. Vectors were tested for titer, genetic stability, CD protein and enzyme activity, ability to confer susceptibility to 5-FC, and preliminary in vivo antitumor activity and stability. One vector, Toca 511, (aka T5.0002) encoding an optimized CD, shows a threefold increased specific activity in infected cells over infection with the prototype RRV and shows markedly higher genetic stability. Animal testing demonstrated that Toca 511 replicates stably in human tumor xenografts and, after 5-FC administration, causes complete regression of such xenografts. Toca 511 (vocimagene amiretrorepvec) has been taken forward to preclinical and clinical trials.