Mycobacterium bovis BCG, the only vaccine available against tuberculosis (22), has also been used successfully as a vaccine against leprosy (29) and, for over 20 years, as a local immunotherapy against superficial bladder cancer (2). Mycobacterium smegmatis is a related organism which has potential advantages as a recombinant vaccine over M. bovis BCG, as it is nonpathogenic and frequently commensal in humans (36, 42). Indeed, emulsified M. smegmatis and M. bovis BCG have comparable antitumor activity in a tumor immunotherapy model (56). The recent realization that mycobacteria are promising vehicles for a new generation of recombinant vaccines lends urgency to investigating the relative value of different strains in that role.Despite their close relationship, there are several key differences between M. bovis BCG and M. smegmatis. M. bovis BCG is derived from the intracellular pathogen M. bovis and survives in host cells for months (31). In contrast, M. smegmatis is rapidly destroyed within infected cells and is nonpathogenic. Highly pathogenic mycobacteria such as M. tuberculosis can survive and replicate within cells because they inhibit phagosome maturation and hence their own degradation (11). Although the exact mechanism is not yet fully understood (47), it is known that such mycobacteria prevent acidification of the phagosome (13) and subsequent formation of a phagolysosome. M. bovis BCG shares many characteristics with M. tuberculosis and induces biochemical markers of maturation arrest (53, 54). On the other hand, M. smegmatis does not arrest phagosome maturation and is degraded rapidly by phagolysosomal proteases, which probably accounts for its lack of pathogenicity (28,54).Since the development of mycobacterium-Escherichia coli shuttle plasmids (23), there has been much interest in the generation of recombinant mycobacteria as vaccines against infectious disease, such as tuberculosis and human immunodeficiency virus. Furthermore, a recent study has also shown that recombinant M. bovis BCG can also protect against tumor challenge in model systems (15). The authors demonstrated that if they vaccinated mice with 10 2 or 10 4 M. bovis BCG secreting OVA and then challenged at day 30, 120, or 150 with B16-OVA, a significant delay in tumor growth was observed.It is generally accepted that the generation of effective antitumor and anti-infective immune responses requires antigenpresenting cells to prime cytotoxic T lymphocytes and T-helper cells by processing antigen and presenting it on MHC class I and class II molecules, respectively. Mice immunized with recombinant M. bovis BCG expressing human immunodeficiency virus antigens (1) and -galactosidase (49) generate specific cytotoxic T lymphocytes, but the mechanisms underlying anti-* Corresponding author. Present address: