To characterize the impact of increased production of TGF-b in a xenograft model of human breast cancer, TGF-b-responsive MDA-231 cells were genetically modi®ed by stable transfection so as to increase their production of active TGF-b1. Compared with control cells, cells that produced increased amounts of TGF-b proliferated in vitro more slowly. In vivo, however, tumors derived from these cells exhibited increased proliferation and grew at an accelerated pace. To evaluate the role of autocrine TGF-b signaling, cells were also transfected with a dominant-negative truncated type II TGF-b receptor (TbRII). Disruption of autocrine TGF-b signaling in the TGF-b-overexpressing cells reduced their in vivo growth rate. Co-inoculation of Matrigel with the TGF-b-overexpressing cells expressing the truncated TbRII compensated for their diminished in vivo growth capacity, compared with the TGF-b-overexpressing cells with an intact autocrine loop. Tissue invasion by the tumor was a distinctive feature of the TGF-b-overexpressing cells, whether or not the autocrine loop was intact. Furthermore, tumors derived from TGFb-overexpressing cells, irrespective of the status of the autocrine TGF-b-signaling pathway, had a higher incidence of lung metastasis. Consistent with the suggestion that TGF-b's enhancement of invasion and metastasis is paracrine-based, we observed no signi®cant di erences among the cell clones in an in vitro invasion assay. Thus, in this experimental model system in vitro assays of cell proliferation and invasion do not accurately re¯ect in vivo observations, perhaps due to autocrine and paracrine e ects of TGF-b that in¯uence the important in vivo-based phenomena of tumor growth, invasion, and metastasis.