Many preclinical studies of cancer immunotherapy are based on the testing of a single vaccination strategy in several tumor models. Moreover, most of those studies used xenogeneic Ags, which, owing to their high immunogenicity, may not represent realistic models for the validation of cancer immunotherapies. To address these issues, we compared the vaccination efficacy of three well established strategies (i.e., naked DNA; peptide-pulsed dendritic cells (DC), or a mixture of peptide and the Escherichia coli toxin LTR72) using the xenogeneic OVA or the naturally expressed tyrosinase-related protein 2 (TRP-2) tumor Ag in the B16 melanoma model. C57BL/6 mice received one to three s.c. injections of peptide-pulsed DC or DNA, or one to four mucosal administrations of peptide-toxin mixture. One to 2 wk later, the animals were challenged s.c. with B16 or B16 cells expressing OVA (B16-OVA). Vaccination of mice with OVA induced in all cases melanoma-specific CTL and protection against B16-OVA. When TRP-2 was used, all three vaccines elicited B16-specific CTL, but only DC pulsed with the immunodominant T cell epitope TRP-2181-188 allowed protection against B16. Even more importantly, a vaccination regimen with TRP-2-pulsed DC, started 24 h after the injection of a lethal number of B16 cells, caused a therapeutic effect in 60% of the challenged animals. Our results strongly emphasize the relevance of the tumor Ag in the definition of immunotherapeutic strategies for cancer, and support the use of peptide-pulsed DC as cancer vaccine in humans.