Many tumor-associated antigens are nonmutated, poorly immunogenic tissue differentiation antigens. Their weak immunogenicity may be due to “self”-tolerance. To induce autoreactive T cells, we studied immune responses to gp100/pmel 17, an antigen naturally expressed by both normal melanocytes and melanoma cells. Although a recombinant vaccinia virus (rVV) encoding the mouse homologue of gp100 was nonimmunogenic, immunization of normal C57BL/6 mice with the rVV encoding the human gp100 elicited a specific CD8⁺ T cell response. These lymphocytes were cross-reactive with mgp100 in vitro and treated established B16 melanoma upon adoptive transfer. To understand the mechanism of the greater immunogenicity of the human version of gp100, we characterized a 9-amino acid (AA) epitope, restricted by H-2D^b, that was recognized by the T cells. The ability to induce specific T cells with human but not mouse gp100 resulted from differences within the major histocompatibility complex (MHC) class I–restricted epitope and not from differences elsewhere in the molecule, as was evidenced by experiments in which mice were immunized with rVV containing minigenes encoding these epitopes. Although the human (hgp100_25–33) and mouse (mgp100_25–33) epitopes were homologous, differences in the three NH₂-terminal AAs resulted in a 2-log increase in the ability of the human peptide to stabilize “empty” D^b on RMA-S cells and a 3-log increase in its ability to trigger interferon γ release by T cells. Thus, the fortuitous existence of a peptide homologue with significantly greater avidity for MHC class I resulted in the generation of self-reactive T cells. High-affinity, altered peptide ligands might be useful in the rational design of recombinant and synthetic vaccines that target tissue differentiation antigens expressed by tumors.