The Wilms' tumor antigen (WT1) is an attractive target for immunotherapy of leukemia. In the past, we isolated and characterized the specificity and function of a WT1-specific T-cell receptor. The goal of this translational study was to develop a safe and efficient WT1-T-cell receptor retroviral vector for an adoptive immunotherapy trial with engineered T cells. We generated a panel of retroviral constructs containing unmodified or codon-optimized WT1-T-cell receptor alpha and beta genes, linked via internal ribosome entry sites or 2A sequences, with or without an additional inter-chain disulfide bond in the T-cell receptor constant domains. These constructs were functionally analyzed in vitro, and the best one was tested in an autologous primary leukemia model in vivo. We identified a WT1-T-cell receptor construct that showed optimal tetramer staining, antigen-specific cytokine production and killing activity when introduced into primary human T cells. Fresh CD34(+) cells purified from a patient with leukemia were engrafted into NOD/SCID mice, followed by adoptive immunotherapy with patient's autologous T cells transduced with the WT1-T-cell receptor. This therapeutic treatment evidently decreased leukemia engraftment in mice and resulted in a substantial improvement of leukemia-free survival. This is the first report that patient's T cells, engineered to express the WT1-T-cell receptor, can eliminate autologous leukemia progenitor cells in an in vivo model. This study provides a firm basis for the planned WT1-T-cell receptor gene therapy trial in leukemia patients.