Recombinant human bone morphogenetic proteins (rhBMPs) can induce bone formation, but the inability to identify an ideal delivery system limits their clinical application. We used ex vivo adenoviral gene transfer to create BMP-2-producing bone-marrow cells, which allow delivery of the BMP-2 to a specific anatomical site. The autologous BMP-2-producing bone-marrow cells then were used to heal a critical-sized femoral segmental defect in syngeneic rats. Femoral defects in five groups of rats were filled with 5 x 10 6 BMP-2-producing bone-marrow cells, created through adenoviral gene transfer (twenty-four femora, Group I); twenty micrograms of rhBMP-2 (sixteen femora, Group II); 5 x 10 6 β-galactosidase-producing rat-bone-marrow cells, created through adenoviral gene transfer of the lacZ gene (twelve femora, Group III); 5 x 10 6 uninfected rat-bone-marrow cells (ten femora, Group IV); or guanidine hydrochloride-extracted demineralized bone matrix only (ten femora, Group V). Guanidine hydrochloride-extracted demineralized bone matrix served as a substrate in all experimental groups. Specimens that were removed two months postoperatively underwent histological and histomorphometric analysis as well as biomechanical testing. Twenty-two of the twenty-four defects in Group I (BMP-2-producing bone-marrow cells) and all sixteen defects in Group II (rhBMP-2) had healed radiographically at two months postoperatively compared with only one of the thirty-two defects in the three control groups (β-galactosidase-producing rat-bone-marrow cells, uninfected rat-bone-marrow cells, and guanidine hydrochloride-extracted demineralized bone matrix alone). Histological analysis of the specimens revealed that defects that had received BMP-2-producing bone-marrow cells (Group I) were filled with coarse trabecular bone at two months postoperatively, whereas in those that had received rhBMP-2 (Group II) the bone was thin and lace-like. Defects that had been treated with bone-marrow cells producing β-galactosidase (Group III), uninfected bone-marrow cells (Group IV), or guanidine hydrochloride-extracted demineralized bone matrix only (Group V) demonstrated little or no bone formation. Histomorphometric analysis revealed a significantly greater total area of bone formation in the defects treated with the BMP-2-producing bone-marrow cells than in those treated with the rhBMP-2 (p= 0.036). Biomechanical testing demonstrated no significant differences, with the numbers available, between the healed femora that had received BMP-2-producing bone-marrow cells and the untreated (control) femora with respect to ultimate torque to failure or energy to failure. This study demonstrated that BMP-2-producing bone-marrow cells created by means of adenoviral gene transfer produce sufficient protein to heal a segmental femoral defect. We also established the feasibility of ex vivo gene transfer with the use of biologically acute autologous short-term cultures of bone-marrow cells. Regional gene therapy is a novel approach to the treatment of bone defects. The limited duration of transgenic expression associated with first-generation adenoviral vectors is advantageous for this clinical application. This system of ex vivo gene transfer and the subsequent infection of bone-marrow cells with an adenovirus containing the BMP-2 cDNA could be adapted to enhance bone formation in humans.