We used a live attenuated murine cytomegalovirus (MCMV) mutant to analyze mechanisms of vaccination against acute and latent CMV infection. We selected MCMV mutant RV7 as a vaccine candidate since this virus grows well in tissue culture but is profoundly attenuated for growth in normal and severe combined immunodeficient (SCID) mice (V. J. Cavanaugh et al., J. Virol. 70:1365–1374, 1996). BALB/c mice were immunized twice (0 and 14 days) subcutaneously (s.c.) with tissue culture-passaged RV7 and then challenged with salivary gland-passaged wild-type MCMV (sgMCMV) intraperitoneally (i.p.) on day 28. RV7 vaccination protected mice against challenge with 10⁵ PFU of sgMCMV, a dose that killed 100% of mock-vaccinated mice. RV7 vaccination reduced MCMV replication 100- to 500-fold in the spleen between 1 and 8 days after challenge. We used the capacity to control replication of MCMV in the spleen 4 days after challenge as a surrogate for protection. Protection was antigen specific and required both live RV7 and antigen-specific lymphocytes. Interestingly, RV7 was effective when administered s.c., i.p., perorally, intranasally, and intragastrically, demonstrating that attenuated CMV applied to mucosal surfaces can elicit protection against parenteral virus challenge. B cells and immunoglobulin G were not essential for RV7-induced immunity since B-cell-deficient mice were effectively vaccinated by RV7. CD8 T cells, but not CD4 T cells, were critical for RV7-induced protection. Depletion of CD8 T cells by passive transfer of monoclonal anti-CD8 (but not anti-CD4) antibody abrogated RV7-mediated protection, and RV7 vaccination was less efficient in CD8 T-cell-deficient mice with a targeted mutation in the β2-microglobulin gene. Although gamma interferon is important for innate resistance to MCMV, it was not essential for RV7 vaccination since gamma interferon receptor-deficient mice were protected by RV7 vaccination. Establishment of and/or reactivation from latency by sgMCMV was decreased by RV7 vaccination, as measured by diminished reactivation of MCMV from splenic explants. We found no evidence for establishment of splenic latency by RV7 after s.c. vaccination. We conclude that RV7 administered through both systemic and mucosal routes is an effective vaccine against MCMV infection. It may be possible to design human CMV vaccines with similar properties.