Using pathogenic simian immunodeficiency virus (SIV) infection of macaques as a model, we explored the limits of the protective immunity elicited by recombinant subunit vaccines and examined factors that affect their efficacy. Envelope gp 160 vaccines, when used in a live recombinant virus-priming and subunit-protein-boosting regimen, protected macaques against a low-dose, intravenous infection by a cloned homologous virus SIVmne E11S. The same regimen was also effective against intrarectal challenge by the same virus and against intravenous challenge by E11S grown on primary macaque peripheral blood mononuclear cells (PBMC). However, only limited protection was observed against uncloned SIVmne. Priming with live recombinant virus was more effective than immunization with subunit gp 160 alone, indicating a potential advantage of native antigen presentation and the possible role of cell-mediated immunity in protection. Whole gp 160 was more effective than the surface antigen (gp 130), even though both antigens elicited similar levels of neutralizing antibodies. Animals immunized with the core (gag-pol) antigens failed to generate any neutralizing antibody and were all infected following challenge. However, their proviral load was 10-100-fold lower than that of the control animals, indicating that immune mechanisms such as cytotoxic T lymphocytes (CTL) may play a role. Finally, animals immunized with both the core and the envelope antigens generated significant protective immunity, even with relatively low neutralizing antibodies. Taken together, these results indicate that multiple mechanisms may contribute to protection. It may therefore be advantageous to incorporate multiple antigens in the design of recombinant subunit vaccines against acquired immunodeficiency syndrome (AIDS).