BACKGROUND: A histologic characteristic of glioblastoma tumors (GBM) is angiogenesis, which requires endothelial cell (EC) sprouting and migration. This is thought to be facilitated by signals in a specialized microenvironment, the perivascular niche, where glioma stem cells (GSC) reside in proximity to ECs. The goal of this study was to determine whether ECs and GSCs directly interact, the mechanism by which this occurs, and the effect on EC signaling and function. METHODS: We utilized multi-label immunofluorescence of GBM tissue and of co-cultured ECs and GSCs to evaluate the proximity of ECs to GSCs and effect of this interaction on signaling, respectively. Cell-cell binding assays, two-photon laser scanning microscopy of cells injected into brain slices in organotypic culture, live-video microscopy of 2D-migration analyzed by Cell Tracker and 3D-chemotactic migration assessed the mechanism of EC-GSC interaction and EC motility. In vivo effects of inhibiting EC interaction with GSCs were evaluated in an orthotopic mouse model of GBM. RESULTS: We found that GSCs were largely localized in very close proximity to ECs in GBM biopsies and bound to normal brain or GBM-derived ECs on Matrigel promoting EC network-formation. In multiple assays, we found that ECs and GSCs directly interact through a mechanism mediated by integrin αvβ3 on ECs binding the RGD-peptide in extracellular L1CAM on GSCs. Importantly, the direct interaction of ECs with GSCs increased the velocity and directional 2D motility of ECs, as well as the chemotactic migration of ECs. A cyclic RGD-peptide, neutralizing antibodies to integrin αvβ3 or L1CAM, and the downregulation of β3 on ECs or of L1CAM on GSCs all inhibited the EC-GSC direct interaction and the increased EC motility. Signaling studies showed that the EC-GSC interaction increased activation of integrin αvβ3, ERK and JNK in ECs and increased nuclear Akt in GSCs, suggesting bidirectional signaling. Soluble L1CAM also activated integrin αvβ3 on ECs. Cyclic RGD-peptide (Cilengitide) treatment of established mouse GBM tumors reduced the percent of Sox2-positive cells within 25-microns of vessels from 65%-to-29% and the percent Sox2-positive cells within 10-microns of vessels from 30%-to-17%. The ratio of Sox2-postiive cells/vessel was unaltered. CONCLUSIONS: These data suggest that (i) GSCs can promote EC motility and thereby angiogenesis through direct contact with ECs, and indicate a role for GSC L1CAM and EC αvβ3 in this interaction; and (ii) Soluble L1CAM activates integrin αvβ3 on ECs, thus circulating soluble L1CAM could interfere with the therapeutic effects of Cilengitide in patients. SECONDARY CATEGORY: Preclinical Experimental Therapeutics.