Retroviral infection of bone marrow cells in long-term marrow cultures (LTMCs) offers several theoretical advantages over other methods for gene transfer into hematopoietic stem cells. To investigate the feasibility of this approach in a large animal model system, we subjected LTMCs from nine dogs to multiple infections with retrovirus containing the neomycin phosphotransferase gene (neo) during 21 days of culture. Feeder layers, cocultivation, polycations, and selection were not used. The in vitro gene transfer efficiency was 70% as determined by polymerase chain reaction amplification of neo sequences in colony- forming unit granulocyte-macrophage (CFU-GM) obtained from day-21 LTMCs. Day-21 LTMC cells were infused into autologous recipients with (four dogs) and without (three dogs) marrow-ablative conditioning. At 3 months posttransplant, up to 10% of marrow cells contained the neo gene. This percentage declined to 0.1% to 1% at 10 to 21 months posttransplant. Neo was also detected in individual CFU-GM, burst- forming unit-erythroid (BFU-E), and CFU-Mix progenitors derived from marrow up to 21 months postinfusion and in cultures of peripheral blood- derived T cells up to 19 months postinfusion. There was no difference in the percentage of neo-marked cells present when dogs that received marrow ablative conditioning were compared with dogs receiving no conditioning. Detection of neo-marked marrow cells almost 2 years after autologous transplantation in a large mammalian species shows that retroviral infection of marrow cells in LTMCs is a potentially nontoxic and efficient protocol for gene transfer. Further, our results suggest that marrow conditioning and in vivo selection pressure to retain transplanted cells may not be absolute requirements for the retention of genetically marked cells in vivo.