Successful retroviral gene transfer into murine hematopoietic stem cells indicates the potential for somatic gene therapy in the treatment of certain human hereditary diseases. We developed a canine model to test the applicability of these techniques to a preclinical model of human marrow transplantation. Previously we reported that canine CFU-GM could be infected with retroviral vectors carrying either the gene for a mutant dihydrofolate reductase (DHFR) or neomycin phosphotransferase (NEO). This study reports six lethally irradiated dogs transplanted with autologous marrow cocultivated with retroviral vector-producing cells. This procedure conferred drug resistance to 3% to 13% of the CFU-GM. Three dogs infected with either the NEO or DHFR virus engrafted, but we detected no drug-resistant CFU-GM. Three dogs were given marrow infected with a DHFR virus and received methotrexate (MTX) as in vivo selection; all three had evidence of engraftment. In the surviving dog, we detected 0.03% to 0.1% MTX-resistant CFU-GM at 3 to 5 weeks posttransplant during in vivo selection. These results indicate that we can reconstitute lethally irradiated dogs with autologous marrow exposed to retroviral vectors and suggest that gene transfer into hematopoietic cells is feasible on a large scale. However, the low-level transient gene expression indicates that considerable obstacles remain before human gene therapy can be considered.