αvβ3 Integrin Limits the Contribution of Neuropilin-1 to Vascular Endothelial Growth Factor-induced Angiogenesis*

Both vascular endothelial growth factor receptors (VEGFR) and integrins are major regulators of VEGF-induced angiogenesis. Previous work has shown that β3 integrin can regulate negatively VEGFR2 expression. Here we show that β3 integrin can regulate negatively VEGF-mediated angiogenesis by limiting the interaction of the co-receptor NRP1 (neuropilin-1) with VEGFR2. In the presence of αvβ3 integrin, NRP1 contributed minimally to VEGF-induced angiogenic processes in vivo, ex vivo, and in vitro. Conversely, when β3 integrin expression is absent or low or its function is blocked with RGD-mimetic inhibitors, VEGF-mediated responses became NRP1-dependent. Indeed, combined inhibition of β3 integrin and NRP1 decreased VEGF-mediated angiogenic responses further than individual inhibition of these receptors. We also show that αvβ3 integrin can associate with NRP1 in a VEGF-dependent fashion. Our data suggest that β3 integrin may, in part, negatively regulate VEGF signaling by sequestering NRP1 and preventing it from interacting with VEGFR2.

Vascular endothelial growth factor (VEGF) 5 is a key regulator of physiological and pathological angiogenesis. The biological activities of VEGF are primarily mediated by two receptor tyrosine kinases, VEGF-receptor-1 (VEGFR1) and VEGF-receptor-2 (VEGFR2), although substantial evidence suggests that VEGFR2 is the major mediator of VEGF-induced endothelial cell (EC) responses (1). For these reasons, several drugs have been designed to inhibit VEGF binding to and/or signaling through VEGFR2 with the aim of inhibiting tumor angiogenesis (see Ref. 2). So far, these monotherapies have proven less successful than anticipated in a number of clinical trials. Some tumors, for example, become refractory to anti-VEGF therapies (3). This has led to the recent idea of simultaneously targeting convergent proangiogenic pathways.
VEGF-mediated signaling together with integrin-mediated signaling coordinates EC proliferation, migration, invasion, and tube formation (4). In particular, ␣v␤3 integrin is up-regulated in proliferating ECs during angiogenesis. Disrupting the interactions of this molecule with the ECM prevents blood vessel formation in a number of models in vivo (5). In apparent contrast, we have shown that genetic ablation of ␤3 integrin enhances VEGFR2-dependent pathological angiogenesis (6). The likely focal point of the coordination between the VEGFand integrin-mediated angiogenesis pathways is VEGFR2, whose signaling is thought to be, at least in part, modulated by interactions with ␣v␤3 integrin (7,8). Recent data have shown that blockade of both ␣v␤3 integrin and VEGFR2 increases the efficacy of antiangiogenic therapy in vivo (9). Our data suggest that the clinical efficacy of any therapeutic approach that targets these two interacting proteins also is likely to be dependent on other molecules that affect VEGFR2 or ␣v␤3 integrin activity. For example, VEGFR2 signaling is influenced additionally by associations with NRP1 (neuropilin-1), a co-receptor for some VEGF isoforms (10,11). Both peptides and antibodies that block the binding of VEGF to NRP1 can enhance anti-VEGF therapy and inhibit tumor growth (12)(13)(14). As such, NRP1 is an obvious point of possible convergence of two pathways (VEGF and ␤3 integrin) that control VEGFR2-dependent EC responses. We thus set out to determine what role NRP1 plays in ␤3 integrin-dependent, VEGF-induced angiogenesis.
Cell Cultures-Mouse lung ECs were isolated and cultured as described previously (16) from either WT (primary), ␤3-null (primary), WT temperature-sensitive SV40-large T-antigen immortalized mice (H2kb-tsA58, Charles River, Wilmington, MA), and ␤3-null/H2kb-tsA58 mice. SV40 ECs were cultured under non-permissive culture conditions for at least 48 h prior to any analysis, thus providing a primary cell equivalent. Polyoma-middle T-antigen immortalized lung endothelial cells (pMT) ECs were generated and cultured as described by May et al. (17). Human umbilical vein EC (HUVEC) were purchased from Lonza Biologics (Woking, UK) and cultured according to the manufacturer's instructions. Tissue culture plates for all experiments were coated with a mixture of collagen (30 g/ml), gelatin (0.1%), and fibronectin (10 g/ml).
Transfection of Endothelial Cells-Smart pool duplexes targeting Vegfr2, Nrp1, or ␤3 integrin were purchased from Dharmacon (Chicago, IL). A non-silencing "control" pool (cp) was used as a control. Mouse ECs at 40 -50% confluence were transfected in Opti-MEM TM medium (Invitrogen) using Oligofectamine TM (Invitrogen) and a final concentration of siRNA duplexes of 100 nM. After 4 h, Opti-MEM TM was replaced with EC medium, and cultures were maintained for up to 72 h before performing experiments. HUVEC were transiently transfected with ␤3⌬722 (18) using the appropriate Nucleofector TM kit (Amaxa Biosystems, Cologne, Germany).
Wound Closure Assay in Vitro-Due to their relatively slow growth in culture, we were unable to maintain siRNA knockdown in primary ECs long enough to achieve the confluent monolayer necessary to efficiently perform these studies. Therefore, these studies were performed in SV40 large T-antigen, temperature-sensitive ECs grown under primary-equivalent, non-immortalized conditions. Cells were grown to 50% confluence in 6-well plates, followed by transfection with siRNA duplexes for 72 h, after which time a confluent cell monolayer was achieved. The cells were serum-starved for 3 h in Opti-MEM TM medium, and then a wound was created using a pipette tip. Cultures were then stimulated with VEGF-A164 (20 ng/ml). To determine the rate of cell motility, the scratch width was measured at 0 and 8 h, using a grid inserted in the eyepiece of a Zeiss Telaval TM microscope. For each well at 8 h, the average distance between leading cells was measured at three points along the scratch (left edge, middle, right edge).

125
I-VEGF Binding Assays-pMT ECs were grown to subconfluence in 24-well dishes (1 ϫ 10 5 cells/well). Cells were transferred to 4°C, and all subsequent steps were performed on ice. Cells were washed twice with PBS and once with binding buffer (Opti-MEM TM , 2 mM HEPES pH 7.5, 0.2% gelatin, 1 g/ml heparin). Increasing concentrations of 125 I-VEGF-A165 (PerkinElmer Life Sciences) ranging from 2 to 40 ng/ml were bound to the cells for 2 h at 4°C. At the end of the binding reaction, cells were washed three times with PBS and solubilized with 1% Triton X-100, and bound counts were deter-mined. Nonspecific binding was determined in the presence of 1 g of unlabeled VEGF.
Primers for Vegfr2 were purchased from R&D Systems, and primers for Acta1 were from Invitrogen. Cycle conditions for all primers (Vegfr2 and Nrp1, 35 cycles; Acta1, 25 cycles) were those set out by R&D Systems for Vegfr2.
Western Blot Analysis-Cells were lysed in 1% Nonidet P-40buffer (see "Cell Signaling Assays") and processed as described (19). Densitometric readings of band intensities were obtained using the Image J TM software available at the National Institutes of Health web site.
Ex Vivo Aortic Ring Assays-Thoracic aortas were isolated and prepared for culture as described previously (6). One-mmthick rings were then transfected in 1 ml of Opti-MEM TM medium with the indicated siRNA "smart" pool duplexes (at a final concentration of 100 nM) using Oligofectamine TM . Transfections were performed in 24-well plates with Յ24 rings/well. After 18 h at 37°C, aortic rings were embedded in growth factor-reduced Matrigel TM (BD Biosciences) and cultured as described (6). Where indicated, culture medium was supplemented with VEGF-A164 at 25 ng/ml. Rings were fed every 2 days with fresh medium with or without VEGF. Sprouting microvessels were counted after 5 days. For RNA or protein isolation, a length of thoracic aorta equivalent to 24 rings was sliced down the midline with a scalpel and transfected as described above. 24 -48 h later, two treated aortas were pooled for subsequent RNA or protein extraction. Peptide inhibitors directed against NRP1 or ␤3 integrin were administered to embedded rings as described by Reynolds et al. (20).
Cell Signaling Assays-ECs were transfected with siRNA 48 h prior to analysis. Cells were starved for 3 h in Opti-MEM TM medium and then stimulated with VEGF (20 ng/ml) for 7 min (except where noted). Cells were lysed in 150 mM NaCl, 20 mM Hepes, 2.5 mM EDTA, 10% glycerol, 1% Nonidet P-40 plus freshly added Protease Inhibitor Mixture Set III TM and Phosphatase Inhibitor Mixture Set I TM (Merck). Western blot analyses were performed as described above.
Immunoprecipitation-Mouse ECs were transfected with siRNA 48 h prior to analysis. Cells were starved for 3 h in Opti-MEM TM medium and then stimulated with VEGF (20 ng/ml) for 7 min and processed as for cell signaling assays. 700 g of total protein was immunoprecipitated with either 5 l of the appropriate antibody or IgG control using the Catch and Release System TM (Millipore) overnight at 4°C. Column elution was in the supplied sample buffers. For ␤3⌬722 studies, transiently transfected HUVEC or pMT ECs were starved for 3 h in Opti-MEM TM medium and then transferred to 4°C for 30 min. VEGF (20 ng/ml) was added to the appropriate plates, and cells were incubated for an additional 30 min at 4°C. Cells were then transferred to 37°C for 7 min and processed as follows. For ␣v␤3 integrin immunoprecipitations, cells were lysed in 25 mM Tris-HCl, pH 7.6, 0.15% Tween 20, 5% glycerol, 0.5 mM EGTA, and protease/phosphatase inhibitors. For VEGFR2 immunoprecipitations, cells were lysed in 50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% Triton X-100, and protease/phosphatase inhibitors. Immunoprecipitations were performed using the Dynabeads TM protein-G immunoprecipitation kit (Invitrogen).
Subcutaneous Growth Factor-induced Angiogenesis Assay-Briefly, two autoclaved sponges (a kind gift from Daryl Harmon, Caligen Foam Ltd., Accrington, UK) of ϳ1 cm 3 in size were implanted subcutaneously in the flanks of anesthetized WT or ␤3-null mice. Every 2 days, sponges were injected in situ with 100 l of PBS containing 10 ng/ml VEGF-A164 and 100 M scr (LWRPTPA) or nrp1 (ATWLPPR) peptide (see Ref. 13). PBS alone was used as a negative control. After 14 days, sponges were excised and fixed for paraffin embedding. Vessel number density was measured (as described by Thurston et al. (22)) after immunostaining.
Histological Analysis of Sponge Sections-Infiltrating blood vessels were detected by immunohistochemical analysis using anti-endomucin antibody and the Vectastain Elite TM ABC staining kit from Vector Laboratories (Orton Southgate, UK). Images were acquired on a Zeiss Axiophot TM microscope equipped with a ϫ40/0.75 numerical aperture objective and a Nikon digital DXM1200 TM camera.
Duolink TM in Situ Proximity Ligation Assay (PLA)-Assay components were purchased from Cambridge Bioscience (Cambridge, UK) and were as follows: anti-rabbit PLA plus probe, anti-goat PLA minus probe, and detection kit-563. Cells were grown on 13-mm coverslips and treated as described below for confocal microscopy. The assay was performed following the manufacturer's protocol. Duolink TM controls were performed with anti-␣v␤3 integrin (clone LM609) and antitransferrin receptor (see above). Results were quantified using BlobFinder, available on the World Wide Web.
Confocal Fluorescence Microscopy-Cells were grown on 13-mm glass coverslips and treated as described above. Cells were washed in PBS, fixed in 4% paraformaldehyde, permeabilized in 0.01% saponin for 10 min on ice, and then incubated with primary antibodies overnight at 4°C. Cells were washed with PBS and incubated with Alexa-Fluor TM -conjugated secondary antibodies (Invitrogen) for 1 h at 37°C. Cells were analyzed by using a Zeiss LSM 710 Confocal TM microscope equipped with a ϫ63/1.4 numerical aperture oil immersion objective. Linear processing of images was performed with Adobe Photoshop CS3 TM software (Adobe, San Jose, CA).
Statistical Analysis-Except where noted, data are presented as means Ϯ S.E. Significant differences between means for ex vivo and in vivo studies were evaluated by the Mann-Whitney non-parametric -test. All other significant differences were evaluated by Student's t test. p Յ 0.05 was considered statistically significant.

NRP1 Inhibition Reduces VEGF-induced Angiogenesis in
␤3-null Mice to a Greater Extent than in WT Mice-We were interested in the contribution that NRP1 makes, in vivo, to pathological angiogenesis. Most tumor cells, however, express NRP1, which itself can promote angiogenesis (23), making it difficult to discriminate between the contribution of endothelial cell-derived and tumor cell-derived-NRP1 to any given result. Therefore, to determine whether NRP1 contributed differently to VEGF-mediated angiogenesis in wild-type (WT) versus ␤3 integrin-deficient (␤3-null) mice, we performed growth factor-dependent subcutaneous sponge implant angiogenesis assays (24). Briefly, synthetic sponges were implanted subcutaneously into WT or ␤3-null mice and then injected either with VEGF-A164 or VEGF in combination with an NRP1-directed peptide (ATWLPPR, nrp1) that was designed to inhibit VEGF-NRP1 binding (13,25). A scrambled peptide (LWRPTPA, scr) was used as a control. Vessel infiltration in sponges was assessed after 14 days by immunohistochemical detection of endomucin in blood vessels (Fig. 1A).
Quantification of blood vessel infiltration into the sponges showed that blood vessel infiltration in WT mice was induced by VEGF. This VEGF-induced response was enhanced in ␤3-null mice (Fig. 1B). Although treatment with the nrp1 peptide reduced vessel density in VEGF-treated sponges in both WT and ␤3-null mice, this effect was exacerbated in ␤3-null animals (60% in ␤3-null compared with 30% in WT), reducing their response to approximately WT levels (Fig. 1B).
Although we cannot rule out off-target effects of the nrp1 peptide, its efficacy in a VEGF-dependent pathway was illustrated by inhibiting VEGF-induced phosphorylation of the mitogen-activated protein kinases Erk1/2 (ERK), known down-stream targets of VEGF stimulation (26,27) (Fig. 1C). Furthermore, treatment of both WT and ␤3-null endothelial cells (ECs) with nrp1 peptide inhibited long term ERK phosphorylation when compared with scr peptide-treated cells (supplemental Fig. S1). These results imply that the enhanced ␤3-null response to VEGF in vivo is dependent on NRP1.
Ex Vivo Microvessel Sprouting of ␤3-null Aortic Rings, but Not WT Aortic Rings, Is Dependent on NRP1-Microvessel sprouting from aortic ring segments is a good model for examining neovascularization ex vivo. When grown in a three-dimensional matrix, aortic rings produce vessels with lumens and many of the salient features associated with angiogenesis, including tip cells and stalk cells involving endothelial and supporting cells (28). Aortic ring assays, therefore, provided a system for continuing our analyses of the functions of VEGFR2, NRP1, and ␤3 integrin during VEGF-A164-induced angiogenesis. We took, initially, an siRNA approach to knock down the expression of these proteins in WT and ␤3-null rings (Figs. 2, A and B). Reverse transcription-PCR (supplemental Fig. S2, upper panels) and Western blot (supplemental Fig. S2, lower panels) analyses confirmed that the expression of VEGFR2, NRP1, and ␤3 integrin were reduced substantially by siRNA treatment of aortic rings from either genotype.
As expected, the cp of siRNA had no affect on microvessel sprouting induced by VEGF, but Vegfr2 depletion inhibited sprouting to background levels in both genotypes when compared with either VEGF-only-treated or cp siRNA-transfected controls. In contrast, Nrp1 knockdown did not alter sprouting from WT rings but reduced significantly the degree of microvessel sprouting in ␤3-null rings, suggesting that ␤3 integrin limits the participation of NRP1 in VEGF-driven microvessel sprouting ( Fig. 2A). Importantly, ␤3 integrin depletion in WT rings increased VEGF-induced microvessel sprouting to levels similar in ␤3-null rings (Figs. 2, A and B). This indicates that the enhanced angiogenic response observed in ␤3-null aortic rings does not occur simply as a result of the long term absence of ␤3 integrin expression (through developmental compensation) because it can be replicated by ␤3 integrin knockdown. Moreover, a combination of nrp1 and ␤3 integrin siRNA treatment inhibited sprouting from WT aortic rings ( Fig. 2A). FIGURE 1. NRP1 antagonist treatment inhibits VEGF-mediated angiogenesis more substantially in ␤3-null mice than WT mice. A, immunohistochemistry of endomucin to detect infiltrating blood vessels in VEGF-induced angiogenesis in subcutaneously implanted sponges in the flanks of WT or ␤3-null animals. Sponges were injected every other day with either PBS (ϪVEGF) or VEGF-A164 (ϩVEGF) and either no peptide (none), scrambled peptide (LWRPTPA; scr), or neuropilin-1-inhibiting peptide (ATWLPPR; nrp1) and harvested after a total of 14 days. The arrows point to examples of endomucin-positive blood vessels. Scale bar, 50 m. B, quantification of VEGF-induced angiogenesis in vivo. The bar graph shows the relative number of blood vessels/mm 2 (means Ϯ S.E.) infiltrating subcutaneously implanted sponges after 14 days of treatment (ϩ) (or not (Ϫ)) with VEGF-A164 plus the indicated peptides. n ϭ 3-5 individual animals/genotype/treatment. *, p Յ 0.03; **, p Յ 0.01. C, representative Western blot analysis of phosphorylated ERK (pERK) levels in ␤3-null pMT ECs in the presence of scr or nrp1 peptide. Cells were treated with the indicated peptides for 30 min and then stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were Western blotted for pERK. Blots were stripped and reprobed for total ERK. The accompanying bar graph represents densitometric results (means Ϯ S.E.) showing relative VEGF-A164-induced pERK/total ERK levels from three independent experiments. *, p Յ 0.02.

␤3 Integrin Regulates Neuropilin-1 Function
To further extend the relevance of these observations beyond that of ␤3 integrin deficiency, we studied the effects of blocking ␤3 integrin function using ␣v␤3/␣v␤5 RGD-mimetic inhibitors in WT aortic rings. As expected, VEGF-stimulated sprouting from WT aortic rings in the presence of either cp siRNA alone or with scr peptide alone (Fig. 2C). The degree of this response was not altered by treatment with Nrp1 siRNA or nrp1 peptide. In the presence of cp siRNA or scr peptide, sprouting was inhibited, however, in a dose-dependent manner with the ␣v␤3/ ␣v␤5 RGD-mimetic inhibitor S 36578 (20,29,30). This inhibition was augmented by the addition of Nrp1 siRNA or nrp1 peptide (Fig. 2C). These studies provide evidence that our findings are not restricted to the ␤3-knock-out model but relevant to wild-type systems.
Taken together, our findings indicate that in the presence of ␤3 integrin, NRP1 normally plays a small role in VEGFdriven angiogenesis both in vivo and ex vivo but that when ␤3 integrin expression is reduced, angiogenic processes are more NRP1-dependent.
In addition to VEGF, a number of proangiogenic heparinbinding growth factors interact with NRPs (31)(32)(33)(34). The interaction between FGF2 and NRP1 in particular is thought to be of physiological relevance because NRP1 was found to enhance the growth-stimulatory activity of FGF2 on endothelial cells (33). We showed that WT aortic rings exhibited a significant increase in fibroblast growth factor-stimulated microvessel sprouting. ␤3-null aortic rings, on the other hand, failed to respond to fibroblast growth factor stimulation (supplemental Fig. S3). This lack of response in ␤3-null microvessel sprouting is probably due to the dependence of fibroblast growth factor responses on ␣v␤3 integrin (35,36). Thus, ␣v␤3 integrin is unlikely to regulate negatively the signaling between FGF2 and NRP1 as it appears to do between VEGF and NRP1.
The aortic ring ex vivo angiogenic assay preserves many of the complex interactions between various cell types that might influence VEGF-A164 responses. Although VEGFR2 expression largely is EC-specific, the expression of NRP1 and ␤3 integrin is more widespread. Therefore, the observed effects in ␤3-null rings might be attributable to cell types other than ECs. Thus, we directed ␤3 integrin deficiency to endothelial cells in this system by infecting WT rings from EC-specific Tie1-Crepositive transgenic animals (37) with a lentivirus carrying a Cre-inducible shRNA construct (21) directed against ␤3 integrin. Infection of ␤3 integrin-specific shRNA constructs into Tie1-Cre-positive, but not Tie1-Cre-negative, rings showed a significant increase in VEGF-A164-induced microvessel sprouting (Fig. 3A). This indicated that the reduction of ␤3 integrin specifically in ECs is sufficient to enhance VEGF-induced angiogenic responses.
We then infected both WT/Tie1-Cre-positive and ␤3-null/ Tie1-Cre-positive rings with Cre-inducible shRNA constructs against Vegfr2 and Nrp1 (Fig. 3B). Infection with scr shRNA constructs did not affect the VEGF-induced microvessel sprouting in WT or ␤3-null rings. Microvessel sprouting in the presence of Vegfr2 shRNA was inhibited in both genotypes. However, microvessel sprouting in ␤3-null, but not WT, rings  ␤3 Integrin Regulates Neuropilin-1 Function DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49 was inhibited by Nrp1 shRNA. In addition, infection with ␤3-specific shRNA enhanced VEGF-mediated sprouting of WT rings to levels similar to those found in VEGF-stimulated ␤3-null rings. Together, these results establish that knocking down Nrp1 specifically in Tie1-Creexpressing endothelial cells is sufficient to demonstrate that VEGFmediated angiogenic responses are more dependent on NRP1 when ␤3 integrin expression is reduced in endothelial cells.
VEGF-induced Motility of ␤3-Null ECs in Vitro Is Dependent on NRP1-Confident that we were studying an EC-specific process, we next investigated the contributions made by VEGFR2, NRP1, and ␤3 integrin to VEGF-A164-mediated events in vitro. In particular, we tested the effect of NRP1 depletion on EC migration, an essential feature of EC behavior during angiogenesis. Due to their relatively slow growth in culture, we were unable to maintain siRNA knockdown long enough in primary ECs to achieve the confluent monolayer necessary to efficiently perform migration studies in primary ECs. Therefore, these studies were performed in SV40 large T-antigen, temperature-sensitive ECs grown under primary-equivalent, non-immortalized conditions. We examined the effects of Vegfr2, Nrp1, and ␤3 integrin knockdown on EC motility in in vitro wound closure assays. Our results showed that transfection of endothelial cells with cp siRNA did not affect the expected increase in VEGF-mediated ␤3-null wound closure when compared with WT controls (19) but that knocking down Vegfr2 in both WT and ␤3-null ECs suppressed significantly EC motility (Fig. 4A). Although Nrp1 knockdown in WT ECs did not affect VEGF-stimulated motility, it did suppress the otherwise enhanced motility of ␤3-null ECs. Once again, ␤3 integrin knockdown in WT ECs gave rise to an enhancement in VEGF-stimulated wound closure similar to that observed for ␤3-null ECs (when compared with WT con- trols) (Fig. 4A). Western blot analyses confirmed the specificity of the vegfr2 and nrp1 siRNA smart pools for their respective targets (Fig. 4B). Our data indicate, therefore, that VEGFdriven EC motility is dependent on NRP1 when ␤3 integrin expression is absent.
Enhanced VEGFR2-mediated Signaling in ␤3-Null Endothelial Cells Is NRP1-dependent-To understand better the molecular mechanisms underlying our results so far, we next investigated the levels of VEGFR2 and NRP1 expressed by WT and ␤3-null primary lung ECs. As we have shown previously (6), VEGFR2 levels were elevated ϳ2-fold in ␤3-null ECs when compared with WT controls. Moreover, NRP1 expression levels were increased ϳ3-fold in ␤3-null cell lysates (Fig. 5A). Similar increases in VEGFR2 and NRP1 expression were observed in SV40 immortalized ECs. We recognized, therefore, that using these cells, we would not be able to distinguish between increased VEGF-induced responses due simply to the elevated levels of VEGFR2 and NRP1 expression or altered interactions between the two molecules.
To overcome this problem, we isolated populations of pMT lung ECs from both WT and ␤3-null mice that expressed roughly equivalent levels of VEGFR2 and NRP1. We also showed that transduction of human ␤3 integrin into ␤3-null pMT ECs did not affect VEGFR2 or NRP1 levels (Fig. 5B). In addition, the saturation kinetics of 125 I-VEGF binding to each of the cell lines was indistinguishable (supplemental Fig. S4A). Furthermore, although ␤3-null pMT ECs showed an enhanced response to VEGF-A164 stimulation as measured by ERKphosphorylation, all three cell lines maintained a dose responsiveness to increasing concentrations of VEGF-A164 (supplemental Fig. S4B).
The results from these immortalized ECs indicate, therefore, that the affinity of VEGF binding to ECs is unchanged in the absence of ␤3 integrin and that the enhanced response of ECs to VEGF-A164 in the absence of ␤3 integrin (as seen in primary ECs) is not due solely to increased expression of VEGFR2 and/or NRP1. Thus, these pMT EC lines provided a good model system for studying the effects of ␤3 integrin expression on the functions of and interactions between VEGFR2 and NRP1 independent of changes in their expression levels.
We first examined the effect of Nrp1 depletion on VEGFinduced autophosphorylation of VEGFR2 (38) in WT versus ␤3-null pMT ECs (Fig. 5C). Nrp1 knockdown reduced VEGFinduced VEGFR2 phosphorylation in ECs of both genotypes, showing that NRP1 contributes to VEGFR2 phosphorylation in both genotypes.
We then tested, in these pMT ECs, the influence of NRP1 expression on the activation of ERK after VEGF-A164 stimulation (26,27). In WT cells, VEGF induced a significant increase in ERK phosphorylation; Nrp1 depletion had a small, but significant, effect on this response (Fig. 5D). The enhanced level of VEGF-stimulated ERK phosphorylation in ␤3-null cells was reduced to approximate WT levels by Nrp1 depletion, suggesting that the enhanced VEGF response in ␤3-null cells is dependent on NRP1. Human ␤3 integrin expression in ␤3-null ECs restored a WT-like response (Fig. 5D). Taken together, our data demonstrate that ␤3 integrin expression limits the contribution of NRP1 to VEGF-A164-dependent signaling even when the expression levels of VEGFR2 and NRP1 are unaltered after ␤3 integrin deletion.
Enhanced VEGF Responses in the Absence of ␤3 Integrin Are Dependent on Elevated Interactions between VEGFR2 and NRP1-We next tested the influence of ␤3 integrin expression on VEGFR2/NRP1 association in pMT ECs. We demonstrated that the co-immunoprecipitation of VEGFR2 and NRP1 was elevated in ␤3-null pMT EC lysates when compared with WT controls, even in the absence of VEGF stimulation and that VEGF-A164 stimulation further enhanced this association (Fig.  6A). Nrp1 knockdown in replicate samples served as a specificity control for the co-immunoprecipitation reactions.
These studies suggest that in the absence of ␤3 integrin, VEGFR2/NRP1 associations are elevated. We wished to test, however, whether a VEGF-induced coupling between VEGFR2 and NRP1 was required for elevated angiogenic responses in the absence of ␤3 integrin. One way to do this is to use the differential ability of VEGF-A164 and VEGF-A120 (the mouse isoform of human VEGF-A121) to couple VEGFR2 and NRP1. VEGF-A164 is known to interact with VEGFR2 and NRP1 and act as a bridge between these two molecules (10). In contrast, VEGF-A120 can bind to either VEGFR2 or NRP1, but no evidence for a trimeric complex of VEGF-A120⅐VEGFR2⅐NRP1 has been demonstrated (1,11,39). Therefore, by testing the effects of VEGF-A120 on aortic ring sprouting and ERK phosphorylation, we tested VEGF responses independent of VEGFR2/NRP1 coupling. As expected, VEGF-A120 stimulation did not increase the association between VEGFR2 and NRP1 in either WT or ␤3-null ECs (supplemental Fig. S5). Furthermore, WT and ␤3-null aortic rings exhibited similar angiogenic responses to VEGF-A120 that were inhibited by Vegfr2 knockdown but not by Nrp1 knockdown (Fig. 6B). In addition, VEGF-A120-induced ERK phosphorylation was not affected by Nrp1 knockdown in either genotype (Fig. 6C). Thus, in contrast FIGURE 5. Enhanced VEGF-mediated endothelial cell signaling in ␤3-null ECs is highly dependent on NRP1. A, representative Western blot analyses of VEGFR2 and NRP1 levels in primary lung WT and ␤3-null ECs. Blotting for HSC70 provides a loading control. The bar graphs below represent densitometry results (means Ϯ S.E.) from three or four independent experiments. B, representative Western blot analyses of total VEGFR2 and NRP1 levels in pMT WT, ␤3-null, and ␤3-null-expressing human ␤3 integrin (ϩh␤3) ECs show that in these cells, VEGFR2 and NRP1 levels remain unchanged between genotypes. Relative ␤3 integrin levels are also shown. Blotting for HSC70 provides a loading control. The underlying bar graph represents relative levels of the indicated proteins compared with HSC70 for the shown blots. C, representative Western blot analysis of phosphorylated VEGFR2 levels in pMT ECs. Cells were treated for 48 h with the indicated siRNA duplexes and then stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were Western blotted for phosphorylated VEGFR2 (pVEGFR2). These blots were stripped and reprobed for total VEGFR2 and NRP1 (to assess its knockdown). HSC70 was used as a loading control. The accompanying bar graph represents densitometric results (means Ϯ S.E.) showing relative VEGF-A164-induced phosphorylated VEGFR2/total VEGFR2 levels from three independent experiments. *, p Յ 0.05; **, p Յ 0.005. D, representative analysis of pERK levels in pMT ECs. Cells were treated for 48 h with the indicated siRNA duplexes and then stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were Western blotted for pERK. These blots were stripped and reprobed for total ERK and NRP1 (to assess its knockdown). HSC70 was used as a loading control. The accompanying bar graph represents densitometric results (means Ϯ S.E.) showing relative VEGF-A164-induced pERK/total ERK levels from three or four independent experiments. *, p Յ 0.01; **, p Յ 0.005. FIGURE 6. VEGFR2/NRP1 associations are enhanced in ␤3-null pMT ECs. A, representative analysis of the co-immunoprecipitation of NRP1 with VEGFR2 in WT and ␤3-null pMT ECs. Cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were immunoprecipitated (IP) with anti-VEGFR2 and Western blotted (WB) for NRP1. Blots were stripped and reprobed for VEGFR2. 48 h pretreatment of replicate samples with Nrp1 siRNA duplexes (as indicated) served as a control for NRP1 blotting. The bar graph below represents densitometric results showing the relative VEGF-A164-induced levels of NRP1/VEGFR2 co-association (means Ϯ S.E.) from four independent experiments. *, p Յ 0.05. B, quantification of VEGF-A120 mediated microvessel sprouting displayed by WT or ␤3-null aortic rings after depletion of Vegfr2 (r2) or Nrp1 using siRNA duplexes. Bars show total number of microvessel sprouts/aortic ring (means Ϯ S.E.) after 5 days of stimulation (ϩ) (or not (Ϫ)) with VEGF-A120. n Ͼ 25 individual rings/genotype/treatment. *, p Յ 0.01; *, p Յ 0.001. nsd, no significant difference. C, representative analysis of pERK levels in pMT ECs. Cells were treated for 48 h with the indicated siRNA duplexes and then stimulated (ϩ) (or not (Ϫ)) with VEGF-A120 for 7 min. Protein lysates were Western blotted for pERK. Blots were stripped and reprobed for total ERK. The accompanying bar graph represents densitometric results (means Ϯ S.E.) showing relative VEGF-A120-induced pERK/total ERK levels from four independent experiments. *, p Յ 0.01; **, p Յ 0.005. nsd, no statistical difference.
to the enhanced ␤3-null angiogenic responses to VEGF-A164, where VEGFR2 and NRP1 are coupled together, the lack of VEGFR2/NRP1 coupling after stimulation with VEGF-A120 did not increase angiogenic responses in the ␤3-nulls. This suggests that the enhanced VEGFR2/NRP1 association in ␤3-null ECs is functional and required for the elevated angiogenic responses to VEGF-A164.
p38 MAPK is another downstream target of VEGF-A164 whose state of activation is indicative of a functional coupling between VEGFR2 and NRP1 (40). Our conclusions are supported further, therefore, by the observation that VEGFA-164induced p38 MAPK phosphorylation is more sensitive to Nrp1 depletion in ␤3-null pMT ECs when compared with either WT ECs or ␤3-null ECs that are expressing human ␤3 integrin (supplemental Fig. S6).
The Cytoplasmic Tail of ␤3 Integrin Regulates Interactions between VEGFR2 and NRP1-Point mutations in the cytoplasmic tail of ␤3 integrin (where tyrosines 747 and 759 have been replaced with phenylalanines (DiYF) (41)) have indicated that this motif of ␤3 integrin is required for angiogenesis and that interactions between ␤3 integrin and VEGFR2 can positively regulate VEGFR2 signaling (8). We therefore wished to determine whether the cytoplasmic tail of ␤3 integrin regulates the interaction between VEGFR2 and NRP1. We took advantage of a human ␤3 integrin mutant (␤3⌬722), which lacks most of its cytoplasmic tail but is still able to dimerize with endogenous ␣v integrin and engage ligand (18,42).
Using an antibody directed specifically against ␣v␤3 integrin, we confirmed the expression of the ␤3⌬722 mutant in transiently transfected ␤3-null pMT ECs by flow cytometry. Fulllength human ␤3 integrin was used as a control; both isoforms of ␤3 integrin coupled with ␣v integrin and were expressed on the surface of ␤3-null pMT ECs (Fig. 7A).
We examined then the ability of the two different ␤3 integrin isoforms to regulate VEGF-induced ERK phosphorylation. As expected, VEGF-induced ERK phosphorylation in ␤3-null pMT ECs was elevated compared with WT pMT ECs. As shown previously, full-length human ␤3 integrin restored VEGF-induced ERK phosphorylation to WT-like levels. The ␤3⌬722 mutant, however, was not able to "rescue" the ␤3-null response to the same degree as full-length ␤3 integrin (Fig. 7B).  ␤3⌬722. B, representative analysis of ERK phosphorylation in WT pMT ECs, mock-transfected ␤3-null pMT ECs, or ␤3-null pMT ECs transfected with either ␤3⌬722 or full-length human ␤3 integrin (ϩh␤3). Cells were transfected with the indicated constructs. 48 h later, cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were Western blotted for pERK. These blots were stripped and reprobed for total ERK. The accompanying bar graph represents densitometric results indicating relative VEGF-A164-induced pERK/total ERK levels from the shown blot. C, representative analysis of the co-immunoprecipitation of NRP1 with VEGFR2 in mock-transfected WT pMT ECs, mock-transfected ␤3-null pMT ECs, or ␤3-null pMT ECs transfected with either ␤3⌬722 or full-length human ␤3 integrin (ϩh␤3). 48 h after transfection, cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A164 for 7 min. Protein lysates were immunoprecipitated (IP) with anti-VEGFR2 and Western blotted for NRP1. Blots were stripped and reblotted for VEGFR2. The accompanying bar graph represents relative VEGFR2/NRP1 associations for the shown blot; the base-line association between VEGFR2 and NRP1 in WT non-stimulated cells has been arbitrarily set to 1.
More importantly, ␤3⌬722-transfected cells maintained the enhanced VEGF-induced association between VEGFR2 and NRP1 that was observed in mock-transfected ␤3-null pMT ECs; indeed, as a result of a decreased expression of VEGFR2 in the presence of the ␤3⌬722, the relative association of VEGFR2 with NRP1 was more dramatic. Transfection with full-length human ␤3 integrin led to a reduction in the association of these two molecules to WT-like levels (Fig. 7C). These studies highlight two important points: 1) the association between VEGFR2 and NRP1 is regulated by the expression of ␤3 integrin, and 2) the absence of the ␤3 integrin cytoplasmic tail is sufficient to increase the association of VEGFR2 and NRP1.
VEGF-A164 Induces an Interaction between ␣v␤3 Integrin and NRP1 That Can Negatively Regulate VEGFR2 Signaling-It has been demonstrated that NRP1 can associate with ␤1 integrin (43). Furthermore, Veldembri et al. (44) have shown that, via interactions with ␣5␤1 integrin, NRP1 regulates adhesion to fibronectin. This raised the possibility that interactions between ␣v␤3 integrin and NRP1 may regulate the processes we have been examining.
Unfortunately, transfection of pMT ECs with either fulllength human ␤3 integrin or the ␤3⌬722 mutant did not produce high enough levels of ␣v␤3 integrin to efficiently detect its expression by immunoprecipitation. Therefore, to study the interactions between ␤3 integrin, NRP1, and VEGFR2, we chose to use HUVEC, which are routinely used as a model system for studying both VEGF-and integrin-mediated angiogenic events.
We showed that ␣v␤3 integrin and NRP1 associate in a coimmunoprecipitable complex (Fig. 8A, left). This association could be demonstrated by immunoprecipitating ␣v␤3 integrin and Western blotting for NRP1 or vice versa. IgG isotype controls and total cell lysates were used to demonstrate the specificity of the immunoprecipitations. As a further control, we immunoprecipitated ␣v␤3 integrin and Western blotted for the transferrin receptor (which is expressed in HUVEC at levels comparable with NRP1) and showed no co-association between the two molecules. Stripping the same blot and reprobing for NRP1 recapitulated the co-immunoprecipitation of ␣v␤3 integrin with NRP1. In addition, fluorescent confocal mircroscopy confirmed close spatial association of ␣v␤3 integrin and NRP1 in whole cells (Fig. 8A, right).
In order to analyze in more detail the interaction between ␣v␤3 integrin and NRP1, we made further use of the ␤3⌬722 mutant; there is evidence to suggest that this mutation has an impact on lateral associations with other transmembrane proteins (18). We postulated that by transfecting HUVEC with the ␤3⌬722 mutant, we might, therefore, be able to alter the interactions of ␤3 integrin with other transmembrane proteins, such as NRP1.
We transiently transfected HUVEC with the ␤3⌬722 mutant and demonstrated, by Western blot analysis, expression of the ␤3⌬722 mutant (albeit at lower levels when compared with endogenous ␤3 integrin; Fig. 8B, inset). Stimulation with VEGF-A165 (the human equivalent of mouse VEGF-A164) induced a significant increase in the co-association between ␣v␤3 integrin and NRP1 in mock-transfected cells (Fig. 8B). In contrast, the association of ␣v␤3 integrin with NRP1 was reduced significantly by expression of the ␤3⌬722 mutant. Total cell lysates acted as loading and specificity controls.
The reduced co-immunoprecipitation and relative association of ␣v␤3 integrin and NRP1 in the presence of the ␤3⌬722 mutant was enticing. Although total ␤3 integrin levels did not change in the presence of the mutant, because the total level of immunoprecipitated ␣v␤3 integrin appeared slightly reduced in the presence of the mutant, the interactions between ␣v␤3 integrin and NRP1 were substantiated using the Duolink TM in situ PLA, which measures "direct" interactions between molecules in whole cells by producing a fluorescent amplification product between two primer-ligated secondary antibodies. We showed that treatment with VEGF-A165 induced co-association of ␣v␤3 integrin and NRP1 in mock-transfected cells but FIGURE 8. VEGF-A induces an interaction between ␣v␤3 integrin and NRP1. A, Western blots. Representative analysis of the co-immunoprecipitation of NRP1 with ␤3 integrin in VEGF-stimulated human ECs. 1, protein lysates were immunoprecipitated (IP) with anti-␣v␤3 integrin (or an isotype IgG control) and Western blotted for NRP1. Blots were stripped and reprobed for ␤3 integrin. 2, protein lysates were immunoprecipitated with anti-NRP1 (or an isotype IgG control) and Western blotted for ␤3 integrin. Blots were stripped and reprobed for NRP1. 3, blotting in total cell lysate (TCL) served as a control for levels of NRP1 and ␤3 integrin in the starting sample. As a further control for the specificity of the immunoprecipitation (4), protein lysates were immunoprecipitated with anti-␣v␤3 integrin and Western blotted for transferrin receptor. The same blot was stripped and reprobed for NRP1 and for ␤3 integrin. Blotting in total cell lysate served as a control for levels of NRP1, TFR, and ␤3 integrin in the starting sample. Micrographs, fluorescent confocal microscopy analysis of ␣v␤3 integrin (red) and NRP1 (green) shows that the two molecules can be spatially co-localized in VEGF-stimulated human ECs (white arrows). Scale bar, 2 m. B, representative analysis of the co-immunoprecipitation of NRP1 with ␣v␤3 integrin in mock-transfected and ␤3⌬722 (⌬722)-transfected HUVEC. Cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A165 for 7 min. Protein lysates were immunoprecipitated with anti-␣v␤3 integrin and Western blotted for NRP1. Blots were stripped and reprobed for ␤3 integrin. Blotting in total cell lysates served as a control for starting levels of ␤3 integrin and NRP1. HSC70 served as a loading control in total cell lysates. The accompanying bar graph represents densitometric results showing the relative VEGF-A165-induced levels of ␤3 integrin/NRP1 co-association (means Ϯ S.E.) from three independent experiments. *, p Յ 0.05. The inset is a Western blot showing that both endogenous and truncated ␤3 integrin are expressed in ␤3⌬722-transfected HUVEC (C). Upper micrographs, representative staining of closely associated endogenous ␤3 integrin and NRP1 (red staining) in mock-transfected and ␤3⌬722-transfected HUVEC as detected by the Duolink TM in situ PLA. The system elicits a visible signal only when the two antibodies are in close proximity. Examples of intense staining in VEGF-A165-treated, mock-transfected cells are demarcated by white arrows. Nuclei are stained blue. Scale bar, 10 m. Lower micrographs, representative controls for the PLA in VEGF-A165-stimulated HUVEC demonstrating the signals generated between ␤3 integrin and an IgG-matched control for NRP1 (␤3/IgG), between ␤3 integrin and NRP1 (␤3/NRP), or between ␤3 integrin and transferrin receptor (␤3/TFR). Examples of intense staining between ␤3 integrin and NRP1 are demarcated by white arrows. Scale bar, 10 m. The bar graph below represents quantification of the signal between ␤3 integrin and NRP1 in VEGF-stimulated (ϩ) and -unstimulated (Ϫ) mock-transfected or ␤3⌬722-transfected HUVEC. Also shown, as a negative control, is the quantification of the signal between ␤3 integrin and the transferrin receptor in untransfected HUVEC cells. Bars, the average number of signals/cell taken through a midcell confocal z-section and measured using BlobFinder TM . *, p Յ 0.05. nsd, no statistical difference. D, representative analysis of pERK levels in mock-transfected and ␤3⌬722-transfected HUVEC. Cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A165 for 7 min. Protein lysates were Western blotted for pERK. Blots were stripped and reblotted for total ERK. The accompanying bar graph represents densitometric results (means Ϯ S.E.) showing relative VEGF-A165-induced pERK/total ERK levels from four independent experiments. *, p Յ 0.05; **, p Յ 0.01. E, representative analysis of the co-immunoprecipitation of NRP1 with VEGFR2 in mock-transfected and ␤3⌬722-transfected HUVEC. Cells were stimulated (ϩ) (or not (Ϫ)) with VEGF-A165 for 7 min. Protein lysates were immunoprecipitated with anti-VEGFR2 and Western blotted for NRP1. Blots were stripped and reblotted for VEGFR2. not VEGF-treated ␤3⌬722-transfected cells (Fig. 8C, upper  micrographs). As controls for the specificity of the PLA reaction, we saw no VEGF-induced interaction between ␣v␤3 integrin and an IgG control antibody or between ␤3 integrin and the transferrin receptor, which is expressed in HUVEC at levels comparable with NRP1 (Fig. 8C, lower micrographs).
Finally, we showed that even low level expression of the ␤3⌬722 mutant was sufficient to increase significantly VEGF-A165-stimulated ERK phosphorylation when compared with mock-transfected controls (Fig. 8D). Furthermore, ␤3⌬722 expression also induced a significant increase in VEGF-A165stimulated VEGFR2/NRP1 association when compared with similarly treated mock controls (Fig. 8E).
Overall, our data establish not only that the presence of the ␤3 integrin cytoplasmic tail can limit the interaction of VEGFR2 and NRP1 but that ␣v␤3 integrin can associate with NRP1 in a VEGF-dependent manner. Our data suggest a model whereby ␣v␤3 integrin may sequester at least a proportion of NRP1, thus limiting its interaction with VEGFR2. In contrast, in the absence of ␣v␤3 integrin, both VEGFR2 and NRP1 levels are elevated (at least in primary ECs; see Fig. 5), but, more importantly, base-line associations between the two are enhanced. This leads to a more readily formed close association between VEGFR2 and NRP1 that, upon VEGF stimulation, equates with an enhanced response to VEGF stimulation.

DISCUSSION
Both ␣v␤3 integrin and NRP1 can be expressed by endothelial cells and tumor cells (10,45,46). As such, a precise dissection of their individual roles in tumor angiogenesis is complicated. Nevertheless, it is generally accepted that both are up-regulated by ECs during neovascularization.
Complete ablation of NRP1 is lethal and leads to vascular remodeling and branching defects (47). These defects are further enhanced by the additional loss of NRP2 (48). Although the two proteins may have overlapping functions early in vascular development, the expression of each becomes restricted later with NRP1 expressed in arteries and NRP2 in veins and lymphatic vessels (49,50). Furthermore, Nrp1 deletion specifically in ECs leads to vascular defects similar to those seen in the total knock-out (51), and defects have been observed in the guidance of endothelial tip cells in the developing hindbrain in Nrp1 KO embryos (52).
We demonstrate that EC-specific expression of ␤3 integrin limits VEGF-induced angiogenesis by serving as a negative regulator of VEGFR2/NRP1 co-association in ECs. VEGF-A164mediated events in vivo, ex vivo, and in vitro were enhanced and dependent on NRP1 expression when ␤3 integrin was absent. In striking contrast, NRP1 contributed less substantially to these processes when ␤3 integrin was expressed.
We have shown that when ␤3 integrin is absent or depleted, blocking NRP1 expression significantly reduces VEGF-mediated aortic ring microvessel sprouting, EC wound closure in vitro, and VEGFR2-mediated signaling. These results correlate with an enhanced association between VEGFR2 and NRP1 in ␤3-null ECs, as detected by co-immunoprecipitation, even under non-VEGF-stimulated conditions. Although two groups have shown that the association between NRP1 and VEGFR2 is dependent on VEGF treatment (55, 56), Whitaker et al. (38) have shown that co-immunoprecipitation of VEGFR2 and NRP1 in COS-1 cells co-expressing the two receptors and in HUVEC can occur independently of VEGF-A stimulation. Our data suggest that the interaction between VEGFR2 and NRP1 is more readily achieved in the absence of ␤3 integrin.
Mac Gabhann and Popel (57,58) have comprehensively modeled the interactions between VEGFR2, NRP1, and VEGFs, and their predictions are in agreement with in vitro studies. Their models demonstrate that the coupling of NRP1-bound VEGF-A165 drives the formation of VEGFR2⅐NRP1 complexes. Although the influence of the ECM on sequestering VEGF has been considered, what is missing from the computational models is the role of other surface receptors that may positively or negatively regulate this association.
By expressing a ␤3 integrin cytoplasmic tail mutant in ␤3-null pMT ECs, we show that the ␤3 integrin cytoplasmic domain is involved in the regulation of VEGFR2/NRP1 associations and downstream signaling. In HUVEC, expression of the ␤3⌬722 mutant altered interactions between ␣v␤3 integrin and NRP1 and between VEGFR2 and NRP1. At present, we cannot distinguish between the following two possibilities of how the ␤3⌬722 mutant elicits these changes in HUVEC: 1) the presence of ␤3⌬722 cytoplasmic tail mutation by itself affects coassociation between ␤3 integrin and NRP1, or 2) the presence of the ␤3⌬722 mutant also changes the confirmation of endogenous ␤3 integrin, and this alters associations with NRP1. Nevertheless, our studies indicate that the presence of other surface molecules needs to be factored into both mathematical and biological studies that examine how VEGFR2 and NRP1 interact.
It is apparent that ␤3 integrin regulates VEGFR2 function, and therefore VEGF-induced angiogenesis, in both positive and negative ways and that the two are not mutually exclusive. For example, ␤3 integrin deficiency can enhance VEGFR2 expression and function (6,19). In addition, we demonstrate here that ␤3 integrin (and importantly the cytoplasmic domain of ␤3 integrin) negatively regulates VEGFR2 signaling by inhibiting the interaction between VEGFR2 and NRP1. However, studies with animals carrying ␤3 integrin with cytoplasmic tail point mutations (where tyrosines 747 and 759 have been replaced with phenylalanines (DiYF) (41)) have implied that this motif of ␤3 integrin is required for angiogenesis and that interactions between ␤3 integrin and VEGFR2 can positively regulate VEGFR2 signaling (8). Although beyond the scope of this study, it is tempting to speculate that the DiYF cytoplasmic tail mutant form of ␤3 integrin might inhibit VEGFR2/NRP1 interactions more than wild-type ␤3 integrin and thus reduce angiogenic responses.
Classically, anti-␣v␤3 integrin therapy failure is attributed to there being multiple proangiogenic pathways within ECs that ␤3 Integrin Regulates Neuropilin-1 Function DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49 compensate for the loss of ␣v␤3 integrin function. Recent data have shown that low doses of some of these drugs can enhance VEGF-mediated angiogenesis by affecting VEGFR2 recycling (20). Our data strongly imply that this therapy might be unsuccessful as a consequence of removing an inhibitory angiogenic regulator. This raises the possibility that the efficacy of antiangiogenic therapy depends on integrin expression and function. Since the absence of ␤3 integrin expression in ECs enhances significantly the role played by NRP1 in VEGF-mediated events, targeting both VEGFR2 and NRP1, when ␤3 integrin expression is low, may prove highly effective. A related observation is that in some tumors, such as angiosarcomas, expression of ␤3 integrin subunits on ECs decreases during the course of malignant transformation (59). In such situations, targeting NRP1 may prove more effective in preventing disease progression.
If ␣v␤3 integrin expression is up-regulated on proliferating ECs during angiogenesis and during vascular remodeling, our data would suggest that targeting NRP1 expressed by ECs to control angiogenesis may be of limited use. However, our data indicate that simultaneously targeting ␤3 integrin and NRP1 may significantly increase the efficacy of antiangiogenic therapy, a hypothesis we are currently testing.