The identification of appropriate mouse models could be useful in carefully evaluating the actual role of the in vivo development of antigen-loss variants during antigen-specific vaccine therapy of human tumors. In this study we investigated the level of efficacy of a MART-1/Melan-A-specific CD8+ T cell clone against its autologous melanoma in a severe combined immunodeficiency (SCID) mouse model, in which the tumor cells expressed in vivo heterogeneous and suboptimal levels of MART-1. The subcutaneous co-injection of the MART-1/Melan-A-reactive T cell clone A42 with MART-1/Melan-A+ autologous human melanoma cells into SCID mice caused a total inhibition of tumor growth. However, the systemic treatment with A42 clone lymphocytes resulted in only 50-60% inhibition of tumor growth, although the T cell clone targeted the tumors and the MART-1+ cells virtually disappeared from the tumors. This study suggests that an immunotherapy based on the expansion of an antigen-specific T cell clone generated in vitro is highly efficient in abolishing tumor growth when the target antigen is fully expressed, but leads to in vivo immunoselection of antigen-loss variants in the presence of suboptimal levels of antigen expression. Furthermore, this work shows that human tumors/SCID mouse models may be useful in evaluating the in vivo efficacy of adoptive immunotherapies.