Human mesenchymal stem cells (MSCs) are capable of differentiating into multiple mesenchymal lineages including chondrocytes, osteocytes, adipocytes, and marrow stromal cells. Using a nonhuman primate model, we evaluated nonhuman primate MSCs as targets for gene therapy. Baboon MSCs (bMSCs) cultured from bone marrow aspirates appeared as a homogeneous population of spindle-shaped cells. bMSCs were capable of differentiating into adipocytes and osteocytes in vitro and chondrocytes in vivo. bMSCs were genetically modified with a bicistronic vector encoding the human erythropoietin (hEPO) gene and the green fluorescent protein (GFP) gene. Transduction efficiencies ranged from 72 to 99% after incubation of MSCs with retroviral supernatant. Transduced cells produced from 1.83 x 10(5) to 7.12 x 10(5) mIU of hEPO per 10(6) cells per 24 hr in vitro before implantation. To determine the capacity of bMSCs to express hEPO in vivo, transduced bMSCs were injected intramuscularly in NOD/SCID mice. In a separate experiment, transduced bMSCs were loaded into immunoisolatory devices (IIDs) and surgically implanted into either autologous or allogeneic baboon recipients. Human EPO was detected in the serum of NOD/SCID mice for up to 28 days and in the serum of five baboons for between 9 and 137 days. NOD/SCID mice experienced sharp rises in hematocrit after intramuscular injection of hEPO-transduced bMSCs. The baboon that expressed hEPO for 137 days experienced a statistically significant (p < 0.04) rise in its hematocrit. These data demonstrate that nonhuman primate MSCs can be engineered to deliver a secreted and biologically active gene product. Therefore, human MSCs may be an effective target for future human gene therapy trials.