Immunotherapy of cancer can lead to the selection of antigen loss variants, which provides strong rationale to target oncogenes that are essential for tumor growth or viability. To investigate this concept, we tagged the HER2/neu oncogene with epitopes from ovalbumin to confer recognition by T-cell receptor transgenic CD8(+) (OT-I) and CD4(+) (OT-II) T cells. Transgenic mice expressing neu(OT-I/OT-II) developed mammary adenocarcinomas at 6 to 10 months of age. Adoptively transferred naive OT-I cells (with or without OT-II cells) proliferated vigorously on encountering neu(OT-I/OT-II)-expressing tumors. This was followed by the complete regression of 37% of tumors, whereas others showed partial/stable responses (40%) or progressive disease (23%). Those tumors undergoing complete regression never recurred. In mice with multiple primary tumors, simultaneous regressions and nonregressions were often seen, indicating that immune evasion occurred at a local rather than systemic level. The majority of nonregressing tumors expressed Neu(OT-I/OT-II) and MHC class I, and many avoided rejection through a profound block to T-cell infiltration. Thus, T cells directed against an essential oncogene can permanently eradicate a subset of spontaneous, established mammary tumors. However, in other tumors, local barriers severely limit the therapeutic response. To maximize the efficacy of immunotherapy against spontaneous cancers, predictive strategies that take into account the heterogeneity of the tumor microenvironment will be required.