Integrin α9β1 Directly Binds to Vascular Endothelial Growth Factor (VEGF)-A and Contributes to VEGF-A-induced Angiogenesis*

Vascular endothelial growth factor A (VEGF-A) is a potent inducer of angiogenesis. We now show that VEGF-A-induced adhesion and migration of human endothelial cells are dependent on the integrin α9β1 and that VEGF-A is a direct ligand for this integrin. Adhesion and migration of these cells on the 165 and 121 isoforms of VEGF-A depend on cooperative input from α9β1 and the cognate receptor for VEGF-A, VEGF receptor 2 (VEGF-R2). Unlike α3β1or αvβ3 integrins, α9β1 was also found to bind the 121 isoform of VEGF-A. This interaction appears to be biologically significant, because α9β1-blocking antibody dramatically and specifically inhibited angiogenesis induced by VEGF-A165 or -121. Together with our previous findings that α9β1 directly binds to VEGF-C and -D and contributes to lymphangiogenesis, these results identify the integrin α9β1 as a potential pharmacotherapeutic target for inhibition of pathogenic angiogenesis and lymphangiogenesis.

Integrins are heterodimeric transmembrane proteins that can mediate cell adhesion, migration, and proliferation. Following activation by their respective ligands, integrins can also modulate these cell functions through coordinated cross-talk with growth factor receptors, including VEGF receptors (14,15) often utilizing signaling proteins common to both receptor pathways (13,16,17). Indirect evidence suggests that inhibition of another ␤1 integrin, ␣3␤1, can inhibit cell adhesion to VEGF-A, suggesting that VEGF-A could serve as an ␣3 ligand (18). We have previously shown that the integrin ␣9␤1 directly binds to the growth factors, VEGF-C and -D (19), a finding that may help explain the abnormal lymphatic phenotype of mice expressing a null mutation of the ␣9 subunit. Because the VEGF homology domain of VEGF-C and -D shares 40% homology with VEGF-A, we hypothesized that VEGF-A might also be a ligand for ␣9␤1 and could potentially modulate VEGF-A-induced angiogenesis.
VEGF-A can be synthesized in a variety of forms based on alternative splicing. A previous study has shown that the integrins ␣3␤1 and ␣v␤3 specifically modulate cellular interactions with the 165-kDa form of VEGF-A but not the 121-kDa form (18). These results suggest that any association between these integrins and VEGF-A likely does not involve interaction with sequences encoded by exon 6. To determine whether interactions between ␣9␤1 and VEGF-A involved similar sites in VEGF-A, we incorporated both splice variants into our studies.
Here we report that ␣9␤1 does bind directly to VEGF-A and does cooperate with VEGF-R2 to modulate in vitro endothelial cell adhesion and migration on both VEGF-A165 and VEGF-A121. We also show that VEGF-A (but not bFGF)-induced angiogenesis in chick CAMs can be inhibited by antibody to ␣9␤1, suggesting that this interaction could have in vivo relevance.
Immunoprecipitation, SDS-PAGE, and Western Blot Analysis-For immunoprecipitation of VEGF-R2, HMVEC were grown in 6-well plates with full growth media until 70% confluent and subsequently in basal media with 0.1% BSA for 4 h. Cells were then added to 12-well dishes coated with VEGF-A (23,24). Cells were exposed to VEGF-A or medium alone for 5-30 min, washed with phosphate-buffered saline/sodium orthovanadate (NaV 10 mM), and then lysed with buffer containing 20 mM Tris, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Triton X, 0.5% sodium deoxycholate, 10% glycerol, 25 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 10 mM NaV, and protease inhibitors (Complete Mini EDTA-free, Roche Applied Science). Pre-cleared lysates were immunoprecipitated with 2 g of antibody bound to protein A-Sepharose beads (Amersham Biosciences). The beads were washed with lysis buffer, resuspended in Laemmli sample buffer, boiled, resolved on 8% SDS-PAGE, and transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore, Billerica, MA). The membrane was blocked in 5% milk/Tris-buffered saline with 0.1% Tween (TBST) for 1 h at room temperature and then probed with anti-VEGF-R2 antibody.
For immunoblotting of paxillin pY 31 , ERKTpY 185/187 , 4G10, and ␤-actin, proteins were suspended in Laemmli sample buffer and resolved on 4 -15% gradient SDS-PAGE, before transfer to polyvinylidene difluoride membrane. The membrane was blocked in 5% BSA, 0.1% Tween, probed with the appropriate specific primary antibodies, washed three times with TBST, and subsequently probed with horseradish peroxidase-conjugated secondary antibodies and developed using chemiluminescence (ECL, Amersham Biosciences).
Flow Cytometry-Cultured cells were trypsinized, washed with phosphate-buffered saline, and incubated with the appropriate primary antibodies and then appropriate phycoerythrinconjugated secondary antibody. Phycoerythin-conjugated VEGF-R2 was used to detect VEGF-R2. Fluorescence of labeled cells was determined with a flow cytometer (BD Biosciences).
Adhesion Assay-Assays were performed as described previously (19) with some minor modifications. After coating 96-well microtiter plates (ICN, Linbro/Titertek, Aurora, OH) with VEGF-A at 4°C overnight, wells were blocked with 3% bovine serum albumin (BSA, Sigma) for 30 min at 37°C. After trypsinization, cells were incubated with or without relevant antibodies (20 g/ml) for 30 min on ice, and 5 ϫ 10 4 cells/well were seeded. Adherent cells were fixed and stained with 1% formaldehyde, 0.5% crystal violet, 20% methanol for 30 min, and the number of adherent cells was evaluated by measuring absorbance at 595 nm in a microplate reader (SpectraMax 190, Molecular Devices, Sunnyvale, CA).
Migration Assay-Assays were performed as described previously (19) using 8-m Transwell plates (Corning Costar, Cambridge, MA), either uncoated or coated, at 4°C overnight with relevant ligand (VEGF-A) or 1% BSA as a binding control. After trypsinization, 5 ϫ 10 4 cells were incubated with or without relevant inhibitors for 30 min on ice and seeded into the top chamber of the Transwell (22). 1% fetal calf serum or soluble VEGF-A was added to the bottom well, to serve as a chemoattractant, and the plates were incubated at 37°C for 3 h. Cells that migrated and adhered to the bottom surface of the Transwell membrane were fixed, stained (DiffQuik, Pierce), and counted in 10 high power (ϫ25) fields for each condition.
Binding Assay-Purified ␣9␤1 (19) and ␣v␤6 (20) integrins were used in solid phase binding assays with VEGF-A, as described previously (19). Recombinant VEGF-A (5 g/ml) was coated on 96-well microtiter plates (Nunc ImmunoPlate, Naperville, IL), and purified ␣9␤1 or ␣v␤6 at various concentrations was added for 2 h at room temperature in the presence or absence of ␣9 blocking antibody or 10 mM EDTA. The extent of ␣9␤1 binding was detected using A9A1 antibody (20 g/ml, 1 h at 37°C). Following labeling with horseradish peroxidase (BD Biosciences), binding was quantified by measuring absorbance at 450 nm with a microplate reader (Molecular Devices).
CAM Assay-Chick eggs were maintained in a humidified 39°C incubator (Lyon Electric, Chula Vista, CA). Pellets containing 0.5% methylcellulose plus recombinant human VEGF-A (50 ng) or bFGF (150 ng) were placed onto the CAM of 10-day-old chick pathogen-free embryos (SPAFAS; Charles River Breeding Laboratories, Wilmington, MA). The CAMs were exposed by cutting a small window in the egg shell to facilitate application of the pellet. Relevant antibodies or agonist/antagonist compounds were applied to the site 24 h after stimulation with VEGF protein. CAMs were imaged on day 13, both following fixation and excision or with real time live imaging, using a digital camera (Canon Supershot6) attached to a Zeiss stereomicroscope. Angiogenesis was quantified by counting branch points arising from tertiary vessels from a minimum of 10 specimens from three separate experiments.
Statistical Methods-Data are presented as mean values Ϯ S.D. from at least three separate experiments unless otherwise stated.

Integrin ␣9␤1 Mediates Cell Adhesion and Migration on VEGF-A-To determine whether cells use integrin ␣9␤1
to adhere to VEGF-A, adhesion assays were performed using two different cell lines, ␣9and mock-transfected MEF and SW-480 cells (Fig. 1). Flow cytometry with the anti-␣9␤1 antibody, Y9A2, showed robust expression of ␣9␤1 in the ␣9-transfected cell types and no expression in mock transfectants (Fig. 1, A and C). ␣9-Transfected MEF demonstrated concentrationdependent adhesion to VEGF-A ( Fig. 1B) that was abolished by the ␣9␤1 blocking antibody Y9A2. In contrast, mocktransfected MEF did not adhere to VEGF-A above background levels of attachment to BSA-coated wells. Because cells have also been shown to adhere to VEGF-A in an ␣3␤1 integrin-dependent manner, we performed assays in a second ␣9-transfected cell line (SW-480) known to express ␣3␤1 (Fig. 1C). ␣9␤1-dependent adhesion on VEGF-A was also demonstrated in these cells. In mock-transfected cells, blocking ␣3␤1 inhibited cell adhesion, whereas blockade of ␣9␤1 or VEGF-R2, neither of which are expressed in these cells, had no effect (Fig. 1D). In ␣9, compared with mocktransfected cells, adhesion to VEGF-A in the absence of blocking antibodies was substantially enhanced, as demonstrated by higher absorbance values. In these cells, blockade of ␣9␤1 caused substantial inhibition of adhesion, whereas blocking antibody to ␣3␤1 had no detectable effect (Fig. 1E), perhaps because ␣9␤1 was expressed at substantially higher levels than ␣3␤1 (Fig. 1C). Because these transfected cells do not express VEGF-R2, the cognate receptor for VEGF (25,26), these findings suggest that ␣9␤1 might interact directly with VEGF-A.

␣9␤1 Integrin Is a Receptor for VEGF-A
ited by antibody to ␣9␤1 (Fig. 2B). Because ␣3␤1 and ␣v␤3 also interact with VEGF-A (18), we performed adhesion assays using HMVEC expressing these two integrins and ␣9␤1 (Fig. 2, A and C) in the presence of blocking antibodies to each of the three integrins. Fig. 2D shows that inhibition of any single integrin inhibited adhesion of these cells to VEGF-A, with nearly complete inhibition caused by antibodies to ␣9␤1 or ␣v␤3 and partial inhibition with antibody to ␣3␤1. This was not because of artifactual effects of antibody binding to integrins on these cells, because a nonblocking antibody to ␣9␤1 had no effect.
The Integrin ␣9␤1 Directly Binds VEGF-A-To determine whether the findings from adhesion assays were a result of direct binding of VEGF-A165 to ␣9␤1, we performed solid phase protein-protein binding assays. ␣9␤1 bound to VEGF-A in a concentration-dependent fashion, and binding was inhibited by blocking antibody to ␣9␤1 or by chelating divalent cations (10 mM EDTA), as expected for authentic integrin-ligand interactions (Fig. 3A). In contrast, the irrelevant integrin, ␣v␤6, showed no binding to VEGF-A.
To investigate the relative role of ␣9␤1, ␣3␤1, and ␣v␤3 integrins in cell migration, we performed HMVEC migration assays on either immobilized VEGF-A165 (Fig. 3E, top panel) or VEGF-A121 (Fig. 3E, bottom panel) in the absence or presence of inhibitory antibodies to either of the three integrins. Consistent with the specialized role of ␣9␤1 in facilitating accelerated cell migration, inhibition of ␣9 resulted in the greatest reduction in cell migration on VEGF-A165. HMVEC migration on VEGF-A121 was also ␣9 integrin-dependent and, as expected, not dependent on integrins ␣3␤1 or ␣v␤3. Because VEGF-A121 and VEGF-A165 are physiologically active in a soluble form, we tested whether soluble VEGF-A-induced cell migration was also ␣9␤1-dependent. Fig. 3F (top and  bottom panels) shows that both soluble VEGF-A165-and -121-induced endothelial cell migration was inhibited by ␣9␤1 blocking antibody. As for responses to immobilized These findings taken together with our previous report showing that the homologous proteins VEGF-C and -D also bind to ␣9␤1 (19) suggest that the ␣9␤1 binding domain resides within the VEGF homology domain (exons 1-5). However, a binding site in exon 8, which is present in both VEGF165 and VEGF121, cannot be excluded.
␣9␤1 and VEGF-R2 Act Coordinately to Mediate Endothelial Cell Adhesion and Migration-To assess the relative contribution of VEGF-R2 and ␣9␤1 to HMVEC adhesion and migration on VEGF-A, cells were treated with ␣9␤1 blocking antibody and/or the VEGF-R2 tyrosine kinase inhibitor SU1498 (27). Inhibition of either ␣9␤1 or VEGF-R2 significantly decreased endothelial cell adhesion to VEGF-A165 (Fig. 4A), and the combination of both inhibitors caused maximal inhibition. Neither inhibitor affected adhesion to the irrelevant ligand, collagen (Fig. 4B). Inhibition of ␣9␤1 or VEGF-R2 also decreased endothelial cell migration on VEGF-A (Fig. 4C), and again inhibition was maximal by the combination of both inhibitors. Inhibition of either ␣9␤1 or VEGF-R2 also decreased endothelial cell migration in response to soluble VEGF-A165 or -121 (Fig. 4D), although in this case the effects of both inhibitors together were not different from the effects of each individual inhibitor. Taken together, these findings suggest that both ␣9␤1 and VEGF-R2 are required for maximal cell adhesion and migration on immobilized VEGF-A and in response to soluble VEGF-A.
In Response to Immobilized VEGF-A, ␣9␤1 and VEGF-R2 Become Physically Associated and Both Contribute to VEGF-A-induced Phosphorylation of the Signaling Intermediates ERK and Paxillin-To determine whether VEGF-R2 and ␣9␤1 coordinately signal in response to VEGF-A, we examined the effects of VEGF-A165 on phosphorylation of the downstream signaling intermediates ERK and paxillin, in the presence or absence of inhibitors of either receptor (Fig. 5). In the absence of VEGF-A, tyrosine phosphorylation of each protein was minimal. When these cells were plated on immobilized VEGF-A, both ERK1/2 and paxillin (Fig. 5A,  top panel) were phosphorylated maximally after 15 min. Phosphorylation of each protein was inhibited by ␣9␤1 blocking antibody at all time points. Following HMVEC binding to immobilized VEGF-A, ERK and paxillin phosphorylation was also inhibited by the VEGF-R2 kinase inhibitor (Fig. 5B). When cells were exposed to soluble VEGF-A (Fig. 5C), paxillin was maximally phosphorylated after 5 min, an effect that was inhibited by either blocking ␣9␤1 or VEGF-R2. However, ERK phosphorylation in response to soluble VEGF-A was not affected by inhibition of ␣9␤1 (data not shown).
We also wished to determine whether co-ligation of VEGF-R2 and ␣9␤1 by VEGF-A leads to physical association of these two receptors. By co-immunoprecipitation, we could detect minimal association of VEGF-R2 and ␣9␤1 in the absence of VEGF-A, and this association was substantially increased when plated on immobilized VEGF-A (Fig. 5D). Using irrelevant antibody, VEGF-R2 was unable to be immunoprecipitated in the absence or presence of VEGF-A. Also, we were not able to co-immunoprecipitate VEGF-R2 and ␣9␤1 from cells exposed to soluble VEGF-A (data not shown). Furthermore, using either VEGF-R2-phosphospecific antibodies (pY951, 1054/1059) or immunoprecipitation, we found that in the presence of VEGF-A or the ␣9␤1-specific ligand Tnfn3RAA (where Tnfn3RAA is ␣9-specific ligand, recombinant third fibronectin repeat of tenascin C in which arginine-glycine-aspartic acid is mutated to RAA), VEGF-R2 phosphorylation was not ␣9␤1-dependent (data not shown).
VEGF-A-induced Angiogenesis Is ␣9␤1-Dependent-To determine the in vivo relevance of our in vitro findings, we performed chick CAM assays to measure VEGF-A-induced angiogenesis, in the presence or absence of ␣9␤1 blocking or nonblocking antibody. VEGF-A induced significant angiogenesis that was inhibited by the ␣9␤1 blocking antibody, Y9A2, but not by the nonblocking antibody, A9A1 (Fig. 6, A and B), or by FIGURE 5. Following binding of VEGF-A, ERK and paxillin are activated by coordinate signaling through ␣9␤1 and VEGF-R2. A, protein lysates from HMVEC exposed to immobilized VEGF-A in the presence or absence of ␣9 inhibiting antibody after 5, 15, or 30 min were immunoblotted for phospho-ERK or phosphopaxillin. B, in separate experiments, protein lysates exposed to immobilized VEGF-A165 for 15 min in the presence or absence of ␣9 blocking antibody or VEGF-R2 tyrosine kinase inhibitor or both inhibitors were immunoblotted for phospho-ERK (top panel) or paxillin (bottom panel). C, immunoblot as in B above, but protein lysates were from cells exposed to soluble VEF-A165 (top panel) or VEF-A121 (bottom panel). D, HMVEC lysates from separate experiments were protein immunoprecipitated (IP) using ␣9␤1 antibody or the irrelevant antibody to ␣v␤6 and immunoblotted (IB) with anti-VEGF-R2.
isotype-matched control antibody (data not shown). Also, as expected, VEGF-induced angiogenesis was inhibited by blocking VEGF-R2 (SU1498). Further inhibition was achieved by blocking both VEGF-R2 and ␣9␤1. In separate experiments, similar results were found when the CAMs were stimulated with VEGF-A121 (Fig. 6C). Consistent with previous reports and our in vitro work, inhibition of ␣3␤1 and ␣v␤3 did not result in inhibition of VEGF-A121-induced angiogenesis. In contrast, induction of angiogenesis by bFGF (Fig. 6D) was not inhibited by the ␣9 blocking antibody. These results suggest that the interaction between ␣9␤1 and VEGF-A is relevant to in vivo angiogenesis, and the role of ␣9␤1 in this process may be specific for angiogenesis induced by VEGF-A.

DISCUSSION
In this study we have identified the angiogenic growth factor VEGF-A as a ligand for the integrin ␣9␤1. Experiments with ␣9-transfected cells and primary human endothelial cells demonstrated that they adhere to and migrate on VEGF-A using ␣9␤1. In response to immobilized VEGF-A, VEGF-R2 and ␣9␤1 assemble into a macromolecular complex to cooperatively signal in an additive manner through phosphorylation of the downstream signaling intermediates ERK and paxillin. In vivo, endothelial cells respond to both immobilized VEGF-A, bound to the extracellular matrix, and to soluble VEGF-A, which may either be secreted from cells or released from the matrix through the action of matrix-degrading proteases. It appears that endothelial cell responses to soluble VEGF-A also depend on signals from ␣9␤1. In contrast to immobilized VEGF-A, we could not demonstrate critical roles for ERK phosphorylation or the formation of a physical complex between VEGF-R2 and ␣9␤1 following soluble VEGF-A stimulation. Using an in vivo angiogenesis assay, we provide evidence that the interactions between both immobilized and/or soluble VEGF-A and ␣9␤1 are likely to be biologically significant.
␣v␤5, ␣5␤1, and ␣4␤1 (9 -11, 28 -30) are at least three other integrins that have been implicated in angiogenesis. One could thus argue that the findings reported here are not novel or unexpected. However, each of the integrins thus far shown to contribute to angiogenesis appears to do so by distinct mechanisms (10,11,28,29). One such novel mechanism is through direct binding of growth factor to the integrin. Hutchings et al. (18) showed in protein-protein binding assays that VEGF-A165 bound directly to ␣v␤3. However, there has been only indirect evidence that growth factors may directly bind to ␤1 integrins. In the same study by Hutchings et al. (18), human umbilical artery endothelial cells were reported to adhere to immobilized VEGF-A in an ␣3␤1-dependent fashion. The authors suggested this adhesion was independent of VEGF-R2 based on the results of presumably down-regulating the receptor following cell pretreatment with VEGF-A. However, the authors did not provide biochemical evidence of direct binding in cell-free experiments. In this study we show that purified ␣9␤1 directly binds to recombinant VEGF-A and that cell adhesion to VEGF-A is not ␣3-dependent in the presence of ␣9␤1, suggesting the interaction with ␣9␤1 is more robust than with ␣3␤1.
Although both ␣3␤1 and ␣v␤3 integrins were found to bind the 165 isoform of VEGF-A, they did not bind VEGF-A121, suggesting the binding site for these integrins is encoded within exons 6 -8. In contrast, ␣9␤1 does bind to VEGF-A121 and, as we have shown previously, binds the homologous VEGF-C and -D growth factors (19), suggesting that a unique ␣9␤1-binding site is encoded within the VEGF homology domain (exons 1-5). Not only is the interaction of the ␣9␤1 integrin with VEGF-A121 unique for ␤1 integrins but also biologically significant as demonstrated by inhibition of cell migration and in vivo angiogenesis following blockade of ␣9␤1 activity.
There have been numerous previous reports of "cross-talk" between tyrosine kinase growth factor receptors and integrins (9,13,16,17). In some cases, this cross-talk has been associated with either growth factor or integrin ligand-induced co-association of both receptors (31). In the case of cross-talk between integrin ␣5␤1and epidermal growth factor receptor, ligation of the integrin induces phosphorylation of the receptor that appears to be independent of binding of the epidermal growth factor receptor itself (32,33). In contrast, activation of ␣9␤1 by its specific ligand, Tnfn3RAA, does not appear to lead to phosphorylation of VEGF-R2, and blocking ␣9␤1 has no effect on VEGF-A-induced phosphorylation of VEGF-R2. It thus appears that the signaling pathways activated by these two receptors intersect downstream of activation of VEGF-R2.
The role of ␣9␤1 in signaling responses to immobilized and soluble VEGF-A appears to be dissimilar. In response to immobilized VEGF-A, ␣9␤1 seems to play an important role in downstream phosphorylation of both paxillin and ERK and leads to the formation of a macromolecular complex containing both the integrin and VEGFR-2. In response to soluble VEGF-A, ␣9␤1 also contributes to phosphorylation of paxillin, but it does not appear to enhance ERK phosphorylation or induce the formation of a dual receptor complex. These differences might reflect the influence of the extracellular matrix and its associated receptors such as heparin sulfate proteoglycans (34) that facilitate immobilization of a high density of VEGF-A, which in turn may stimulate an increase in both ␣9␤1 and VEGFR-2 clustering. Nonetheless, the findings suggest that although important for full signaling through VEGF-R2 the integrin is not necessary and that paxillin phosphorylation may be the most relevant end point leading to enhanced cell migration in response to VEGF-A.
Our findings that blocking ␣9␤1 specifically inhibited CAM angiogenesis induced by the 165 and 121 isoforms of VEGF-A and not bFGF suggest that the selective interaction of VEGF-A with this integrin might be the mechanism by which ␣9␤1 enhances in vivo angiogenesis. This is supported by our findings that blocking both ␣9␤1 integrin and VEGF-R2 resulted in greater inhibition of angiogenesis than blocking either receptor alone. Of course, in vivo angiogenesis is a complex process, and we cannot be certain that inhibition of ␣9␤1 only inhibits angiogenesis by interfering with binding to VEGF-A. For example, the closely related integrin, ␣4␤1, was recently shown to contribute to angiogenesis following binding to one of its ligands, VCAM-1 (10). We have previously shown that ␣9␤1 is also a receptor for VCAM-1 (35), so it is conceivable that a similar mechanism contributes to the in vivo role of ␣9␤1 in angiogenesis. However, in contrast to results for ␣9␤1, blockade of ␣4␤1 inhibits angiogenesis induced by bFGF, so this effect is clearly not specific for VEGF-A. Furthermore, mice lacking VCAM-1 or the integrin ␣4 subunit (36,37) develop similar defects in vascular development, whereas development of the nonlymphatic vasculature appears to be normal in ␣9 knock-out mice (38). These results are most consistent with the hypothesis that the mechanisms by which ␣9␤1 and ␣4␤1 contribute to angiogenesis are distinct.
The absence of major vascular defects in ␣9 knock-out mice might suggest that the ␣9␤1 integrin is not essential for normal vascular development. However, this should not be taken as evidence that ␣9␤1 would not contribute to pathologic angiogenesis (39) as has been clearly demonstrated in the case of the ␣v integrins. Antagonists of ␣v␤3 or ␣v␤5 inhibit pathologic angiogenesis, but mice lacking ␤3 and/or ␤5 integrins have normal vascular development and even demonstrate enhanced tumor and ischemic angiogenesis (12). Additional experiments will be required using inhibitors of ␣9␤1 that are effective in mammalian disease models of angiogenesis such as tumor growth and metastasis to more definitively address this question. The results presented here, in our previous reports (19,38) and by others (40), demonstrate the unique role of ␣9␤1 in not only lymphatic development but also VEGF-induced angiogenesis and lymphangiogenesis and suggest that inhibition of this integrin could affect pathologic vasculogenesis supporting cancer cell growth and metastasis.