Many tumor-associated antigens are derived from nonmutated "self" proteins. T cells infiltrating tumor deposits recognize self-antigens presented by tumor cells and can be expanded in vivo with vaccination. These T cells exist in a functionally tolerant state, as they rarely result in tumor eradication. We found that tumor growth and lethality were unchanged in mice even after adoptive transfer of large numbers of T cells specific for an MHC class I-restricted epitope of the self/tumor antigen gp100. We sought to develop new strategies that would reverse the functionally tolerant state of self/tumor antigen-reactive T cells and enable the destruction of large (with products of perpendicular diameters of >50 mm2), subcutaneous, unmanipulated, poorly immunogenic B16 tumors that were established for up to 14 d before the start of treatment. We have defined three elements that are all strictly necessary to induce tumor regression in this model: (a) adoptive transfer of tumor-specific T cells; (b) T cell stimulation through antigen-specific vaccination with an altered peptide ligand, rather than the native self-peptide; and (c) coadministration of a T cell growth and activation factor. Cells, vaccination, or cyto-kine given alone or any two in combination were insufficient to induce tumor destruction. Autoimmune vitiligo was observed in mice cured of their disease. These findings illustrate that adoptive transfer of T cells and IL-2 can augment the function of a cancer vaccine. Furthermore, these data represent the first demonstration of complete cures of large, established, poorly immunogenic, unmanipulated solid tumors using T cells specific for a true self/tumor antigen and form the basis for a new approach to the treatment of patients with cancer.