Tumor growth and metastasis are angiogenesis-dependent. Virtually all solid tumors are neovascularized by the time they are detected. However, there is a prevascular phase during early tumor development where few or no tumor cells are angiogenic and expansion of the tumor is restricted to a few mm3. When enough tumor cells become angiogenic, the tumor can expand progressively and shed metastatic cells. This angiogenic switch has recently been quantitated for human breast cancer, as well as for prostate cancer. We have studied the problem of how tumors switch to the angiogenic phenotype by using transgenic mice in which tumors develop at a predictable time and in discrete prevascular and vascular stages. When the transgene is the bovine papilloma virus (BPV) genome, angiogenic fibrosarcomas develop from non-angiogenic precursors called fibromatoses. The fibrosarcomas secrete growth factors for capillary endothelial cells. In contrast, the fibromatoses do not secrete endothelial cell growth factors. When the transgene consists of the large "T" antigen of SV40 under the control of the rat insulin promoter, 70% of pancreatic islets become hyperplastic and 4-10% of these become angiogenic at 6-7 weeks. Tumors arise from these neovascularized hyperplastic islets and reach > 1000 x the volume of the preangiogenic islets. The onset of angiogenic activity coincides with the secretion of acidic fibroblast growth factor (aFGF) and other growth factors not fully identified at this writing. These studies help to explain the switch to the angiogenic phenotype during tumorigenesis and provide models to discover antiangiogenic therapies directed at the source of angiogenic activity. Such therapy, when developed, may be co-administered with currently available angiogenesis inhibitors which are directed at the target of angiogenic activity, vascular endothelial cells.