Replication-selective oncolytic adenoviruses hold promise, but novel mechanisms must be identified to maximize intratumoral virus persistence, spread and therapeutic transgene-carrying capacity while maintaining safety. One of the main approaches to engineering cancer-selectivity has been to delete a viral gene that is theoretically expendable in cancer cells. Results with this approach have been mixed, however, as evidenced by controversy over Onyx-015 (E1B-55kD(À)) selectivity. We hypothesized that the functional redundancy between viral gene products might limit selectivity and/or potency with this approach. Antiviral immune inducers of apoptosis (eg TNF-a) have not been thoroughly investigated in previous studies. We therefore explored whether deletion of functionally redundant viral genes, E1B-19kD and E3B, both independently antagonize TNF-a, could lead to enhanced oncolytic potency while maintaining selectivity. Since tumors have numerous blocks in apoptotic pathways, we hypothesized that deletion of one or both gene regions would result in cancer-selectivity in the presence of TNF-a. We have previously shown that the E1B-19kD deletion resulted in enhanced viral spread in vitro and in immunocompetent tumor models in vivo. In contrast, the impact of E3B deletion, especially its in vitro selectivity and potency, was not thoroughly characterized, although it resulted in rapid immune-mediated viral clearance in vivo. Furthermore, previous publications indicated that doubledeleted mutants have selectivity but unsatisfactory efficacy. We compared the selectivity and potency of E1B-19kD(À), E3B(À) and E1B-19kD(À)/E3B(À) mutants to wild-type adenovirus. In cancer cells, the E1B-19kD(À) mutant had superior replication, spread and cytolysis (+) or (À) TNF-a; deletion of both E1B-19kD and E3B was relatively deleterious. In normal cells without TNF-a, similar results were obtained. In contrast, all three mutants were significantly inhibited in the presence of TNF-a. In immunocompetent mice, all three mutants were significantly inhibited in normal tissue. In tumors, only the E1B-19kD(À) mutant demonstrated enhanced replication, spread and antitumoral efficacy. Therefore, E1B-19kD deletion and E3B retention should be incorporated in oncolytic adenoviruses for enhanced safety and efficacy. In addition, functional redundant viral genes and their biological mediators/targets need to be carefully examined for the next generation of gene-deleted oncolytic viruses.