Most cancer gene therapies using viruses have focused on replication-defective adenovirus, adeno-associated virus, retrovirus, vaccinia virus, and herpes simplex virus 1 (HSV-1). When using such replication-defective viruses, antineoplastic effects are achieved by delivery and expression of therapeutic transgenes. Although replication-defective viruses are used in a majority of current cancer gene therapy trials, 1 this approach has important drawbacks. Therapeutic transgene delivery is nonselective: both normal and cancer cells are infected. Replication-defective viruses are also unable to spread progeny virion to cells that were not initially infected. An alternative strategy exploits viral replication for tumor destruction, whereby infection of tumor cells by virus leads to cell destruction and simultaneous release of progeny virion that can infect adjacent tumor cells. 2 In this strategy, antineoplastic efficacy is dependent on viral replication; accordingly, it is important to maintain robust viral replication in neoplastic cells and simultaneously attenuate viral replication in nonneoplastic cells. Replication-conditional viruses can also deliver therapeutic transgenes to improve antineoplastic efficacy beyond that achieved by replication alone. 3 Tumor infection with replication-conditional viruses leads to longer transgene expression and better distribution throughout the tumor compared to tumor infection with replication-defective viruses. 4 Several viruses have been examined for their utility as replication-conditional oncolytic viruses, including HSV-1, 5 vaccinia virus, 6 Newcastle disease virus, 7 adenovirus, 8 and reovirus. 9 We have previously demonstrated preferential replication of an HSV-1 mutant hrR3 in tumors rather than normal tissue following intravascular administration. 10,11 hrR3 replication and oncolysis are attenuated in normal or quiescent cells because it is deficient in expression of viral ribonucleotide reductase (infected cell protein 6 [ICP6]). 12 The selectivity of hrR3 for liver tumors is related to significantly higher expression of ribonucleotide reductase and higher intracellular nucleotide pools in most tumors compared to surrounding normal tissues. 13 However, there are disadvantages to developing ICP6-defective HSV-1 mutants for clinical studies. ICP6-defective HSV-1 mutants are most effective against replicating cells, and accordingly quiescent cancer cells may be less susceptible to viral oncolysis. In addition, some virus may reach sites outside the liver despite regional administration, 14 and normal tissues with high replicative activity may be susceptible to infection and cytolysis by an ICP6-defective HSV-1 mutant. The normal tissues that are at greatest risk are those with high replicative activity, such as gastrointestinal mucosa and bone marrow. We have looked to manipulate viral genes that are critical for robust HSV-1 replication but, unlike viral ribonucleotide reductase, have no cellular homologues and whose function is poorly complemented by normal cells. The immediate-early ICP4 gene product is required for HSV-1 replication in cell culture. 15 Cells do not efficiently complement HSV-1 deficiency of ICP4, and accordingly, ICP4-defective HSV-1 mutants display markedly attenuated replication. The ICP4 gene is therefore a good candidate gene to regulate to achieve replication preferentially in cancer cells. Another good candidate gene to regulate is γ134.5, which is also required for robust HSV-1 replication. The γ134.5 gene product interacts with a cellular protein phosphatase to dephosphorylate elongation initiation factor 2α (eIF-2α) and permit cellular (and viral) protein translation to proceed. 16,17 Regulation of γ134.5 expression has been demonstrated by others to be an effective strategy to regulate HSV-1 lytic replication. 18 We examined a strategy in which ICP4 expression and γ134.5 expression are regulated by heterologous transcriptional regulatory elements for tumor-associated antigens. We first examined transcriptional regulatory elements for carcinoembryonic antigen (CEA) and MUC1/DF3 because they are overexpressed in a wide variety of epithelial cancers. We isolated HSV-1 mutants in which either ICP4 or γ134.5 expression is regulated by transcriptional regulatory elements for CEA and MUC1.