The success of cancer gene therapies requiring in vivo gene transfer is severely hampered by the low efficacy of gene transfer, which has been difficult to improve. We therefore established a novel strategy to increase the share of transduced cells post gene transfer. We hypothesized that in vivo selection of tumor cells transduced with a suicide gene effectively enriches these cells within a tumor, thus allowing for an increased bystander effect after the prodrug is given, leading to enhanced eradication of tumor cells. We reasoned that in vivo enrichment should be achieved by exploiting the metabolism of the suicide gene product. For this 'enrichment-eradication' strategy we chose a fusion gene of cytosine deaminase and uracil phosphoribosyl transferase. Positive selection (enrichment) was to be achieved by concurrently giving N-(phosphonacetyl)-L-aspartate, an inhibitor of pyrimidine de novo synthesis, which leads to pyrimidine depletion-mediated death of non-transduced cells, and cytosine, to rescue fusion gene expressing cells via the pyrimidine salvage pathway. Negative selection (eradication) was to be induced by giving the prodrug 5-fluorocytosine. Indeed, murine NXS2 neuroblastoma cells transduced with the fusion gene were effectively enriched in vitro, leading to a near-complete bystander effect. In vivo enrichment-eradication of NXS2 cells led to decreased tumor growth. This proof-of-principle study shows that enrichment-eradication may compensate the effects of low in vivo gene transfer efficacy, a major obstacle in cancer gene therapy.