Angiogenesis inhibiting agents are currently integral component of anticancer therapy. However, tumors, initially responsive to anti-angiogenic drugs or vascular targeting agents, can acquire resistance. The limited clinical efficacy might result from the heterogeneous nature of tumors or alternatively from the unique phenotype of tumor vascular cells, widely diverse from so-called 'normal' endothelium. Hence, defining the molecular mechanisms driving this diversity might provide a rational basis to design combinatory therapies that should be more effective in avoiding resistance. Herein, we demonstrated that tumor-derived endothelial cells (TECs) isolated from breast and kidney carcinomas retained an endothelial phenotype, but outspread independently of growth factors. Applying small interfering RNA approach, we demonstrated that interleukin (IL)-3, but not vascular endothelial growth factor, released by TECs, supports their autocrine growth and promotes in vivo vessel formation and tumor angiogenesis. Meanwhile, we found that the expression of the membrane-bound kit ligand (mbKitL) depends on IL-3, and it is crucial for adhesion of endothelial progenitor cells (EPCs) and inflammatory cells to TECs. These events required Akt activation. Finally, the finding that depletion of the mbKitL prevented EPC and inflammatory cell trafficking into vascular microenvironment, indicates that, as in bone marrow, the mbKitL can act as a membrane/adhesion molecule for c-Kit-expressing cells. These data provide evidences that an IL-3 autocrine loop can drive a tumor endothelial switch and that targeting IL-3 might confer a significant therapeutic advantage to hamper tumor angiogenesis.