Marrow stromal cells (MSCs) are postnatal progenitor cells that can be easily cultured ex vivo to large amounts. This feature is attractive for cell therapy applications where genetically engineered MSCs could serve as an autologous cellular vehicle for the delivery of therapeutic proteins. The usefulness of MSCs in transgenic cell therapy will rely upon their potential to engraft in nonmyeloablated, immunocompetent recipients. Further, the ability to deliver MSCs subcutaneously -as opposed to intravenous or intraperitoneal infusions -would enhance safety by providing an easily accessible, and retrievable, artificial subcutaneous implant in a clinical setting. To test this hypothesis, MSCs were retrovirally engineered to secrete mouse erythropoietin (Epo) and their effect was ascertained in nonmyeloablated syngeneic mice. Epo-secreting MSCs when administered as 'free' cells by subcutaneous or intraperitoneal injection, at the same cell dose, led to a significant -yet temporaryhematocrit increase to over 70% for 55713 days. In contrast, in mice implanted subcutaneously with Matrigeltembedded MSCs, the hematocrit persisted at levels 480% for over 110 days in four of six mice (Po0.05 logrank). Moreover, Epo-secreting MSCs mixed in Matrigel elicited and directly participated in blood vessel formation de novo reflecting their mesenchymal plasticity. MSCs embedded in human-compatible bovine collagen matrix also led to a hematocrit 470% for 7578.9 days. In conclusion, matrixembedded MSCs will spontaneously form a neovascularized organoid that supports the release of a soluble plasma protein directly into the bloodstream for a sustained pharmacological effect in nonmyeloablated recipients.