The use of E. coli purine nucleoside phosphorylase (PNP) to activate prodrugs has demonstrated excellent activity in the treatment of various human tumor xenografts in mice. E. coli PNP cleaves purine nucleoside analogs to generate toxic adenine analogs, which are activated by adenine phosphoribosyl transferase (APRT) to metabolites that inhibit RNA and protein synthesis. We created tumor cell lines that encode both E. coli PNP and excess levels of human APRT, and have used these new cell models to test the hypothesis that treatment of otherwise refractory human tumors could be enhanced by overexpression of APRT. In vivo studies with 6-methylpurine-2'-deoxyriboside (MeP-dR), 2-F-2'-deoxyadenosine (F-dAdo) or 9-β-D-arabinofuranosyl-2-fluoroadenine 5'-monophosphate (F-araAMP) indicated that increased APRT in human tumor cells coexpressing E. coli PNP did not enhance either the activation or the anti-tumor activity of any of the three prodrugs. Interestingly, expression of excess APRT in bystander cells improved the activity of MeP-dR, but diminished the activity of F-araAMP. In vitro studies indicated that increasing the expression of APRT in the cells did not significantly increase the activation of MeP. These results provide insight into the mechanism of bystander killing of the E. coli PNP strategy, and suggest ways to enhance the approach that are independent of APRT.