The Lymphangiogenic Vascular Endothelial Growth Factors VEGF-C and -D Are Ligands for the Integrin α9β1*

Mice homozygous for a null mutation of the integrin α9 subunit die 6–12 days after birth from bilateral chylothoraces suggesting an underlying defect in lymphatic development. However, until now the mechanisms by which the integrin α9β1 modulates lymphangiogenesis have not been described. In this study we show that adhesion to and migration on the lymphangiogenic vascular endothelial growth factors (VEGF-C and -D) are α9β1-dependent. Mouse embryonic fibroblasts and human colon carcinoma cells (SW-480) transfected to express α9β1 adhered and/or migrated on both growth factors in a concentration-dependent fashion, and both adhesion and migration were abrogated by anti-α9β1 function-blocking antibody. In SW-480 cells, which lack cognate receptors for VEGF-C and -D, both growth factors induced α9β1-dependent Erk and paxillin phosphorylation. Human microvascular endothelial cells, which express both α9β1 and VEGF-R3, also adhered to and migrated on both growth factors, and both responses were blocked by anti-α9β1 antibody. Furthermore, in a solid phase binding assay recombinant VEGF-C and -D bound to purified α9β1 integrin in a dose- and cation-dependent fashion showing that VEGF-C and VEGF-D are ligands for the integrin α9β1. The interaction between α9β1 and VEGF-C and/or -D may begin to explain the abnormal lymphatic phenotype of the α9 knock-out mice.

Integrins are heterodimeric transmembrane proteins, which serve as receptors for a variety of spatially fixed extracellular ligands (1). By virtue of their dual roles in adhesion and signaling and because of their close association with the actin cytoskeleton, integrins play important roles in regulating cell shape and cell migration (2,3). In vertebrates there are 8 identified integrin ␤ subunits and 18 ␣ subunits that form at least 25 different heterodimers (4). The integrin ␣9 subunit forms a single heterodimer with ␤1 and is expressed in epithelial cells, smooth and skeletal muscle, neutrophils, and a subset of endothelial cells (5,6). In vitro, the principal demonstrated function of ␣9␤1 is acceleration of cell migration, an effect that depends on unique sequences within the ␣9 cytoplasmic domain (7,8).
In a previous study we inactivated the ␣9 subunit in mice to better understand the function of ␣9␤1. In these mice lymph leaked into the pleural space (chylothorax), and the mice died 6 -12 days after birth. This phenotype was an unexpected finding which indicated that lymph vessel development and/or function was abnormal (9). On gross inspection, the thoracic duct and peripheral lymphatic vessels were present, but their integrity was compromised, as evidenced by chylothoraces and edema of the thoracic dermis, skeletal muscle, and pleural surface. To date the molecular mechanisms underlying the role of ␣9␤1 in lymphatic development and/or function remain unexplained.
Therefore, in this study, we hypothesized that the lymphatic abnormality in ␣9 knock-out mice could be explained by an interaction between ␣9␤1 and VEGF-C and/or -D. To address this question we used ␣9-transfected cell lines and primary microvascular endothelial cells to assess in vitro cell adhesion, migration, and receptor signaling and purified ␣9␤1 and VEGF-C or -D protein for solid phase binding assays.
We found that ␣9-transfected cells and primary microvascular endothelial cells, which endogenously express ␣9␤1, utilize ␣9␤1 to adhere to and migrate on VEGF-C and -D. This effect was inhibited by the specific ␣9␤1-blocking antibody Y9A2 and siRNA silencing of ␣9 protein expression. Furthermore, VEGF-C and -D directly bound to ␣9␤1 in a solid phase protein binding assay and activated ␣9␤1 signaling, as evidenced by Erk 1/2 and paxillin phosphorylation that was inhibited by anti-␣9␤1 antibody. These novel findings therefore identify the growth factors VEGF-C and -D as ligands for ␣9␤1 and provide a potential explanation for the abnormal lymphatic phenotype of the ␣9 knock-out mouse.
Transfected S2 cells were initially grown in 15-cm culture plates (Corning) and then transferred to 1-liter spinner flasks (Corning) for 5 days in reduced serum conditions. Protein secretion was induced with 500 M CuSO 4 (Sigma) for 5 days. The supernatant was collected following centrifugation (3000 ϫ g for 10 min at 4°C) and then filtered, dialyzed, and concentrated (Vivaflow 200, VivaScience Inc., Carlsbad, CA) in binding buffer (0.5 M NaCl, 50 mM Na 2 PO 4 , pH 8.0). Concentrated supernatant containing VEGF-C or -D was then applied to a Ni 2ϩ column (ProBond, Invitrogen), and the column was washed with binding buffer containing 50 M imidazole to remove contaminating proteins; the bound VEGF was eluted with buffer containing 300 M imidazole. The eluents were then dialyzed against 150 mM NaCl and 20 mM Tris-HCl, pH 7.5. Purity was assessed by silver staining (GelCode SilverSNAP, Pierce).
Immunoprecipitation, SDS-PAGE, and Western Blot Analysis-For immunoprecipitation of VEGF-R3, HMVEC were grown in 6-well plates with full growth medium until 70% confluent and subsequently in basal medium with 0.1% fetal bovine serum for 4 h. VEGF-C or -D was then added to the cells for 5 min, and the cells were washed with PBS/sodium orthovanadate (NaV i 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, 1 mM NaV i , and protease inhibitors (Complete Mini EDTAfree, Roche Applied Science). Lysates were centrifuged at 14,000 ϫ g for 10 min at 4°C and the precleared lysates immunoprecipitated with 2 g of rabbit anti-VEGF-R3 antibody bound to 30 l of protein A-Sepharose beads (Amersham Biosciences). The beads were washed five times with lysis buffer, resuspended in Laemmli sample buffer, boiled at 95°C for 6 min, resolved on 8% SDS-PAGE under reducing conditions, and transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore, Billerica, MA). The membrane was blocked in 5% milk/Trisbuffered saline with 0.1% Tween (TBST) for 1 h at room temperature and then probed with anti-phosphotyrosine antibody.
For immunoblotting of VEGF-C, VEGF-D, ␣9␤1, Erk 1/2, and paxillin, proteins were resuspended in Laemmli sample buffer and resolved on 15, 10, or 8% SDS-PAGE before transfer to a polyvinylidene difluoride membrane. The membrane was 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).
For analysis of Erk 1/2 and paxillin phosphorylation, SW-480 cells were grown to 70 -80% confluence in full growth media and then treated with 20 g/ml of cycloheximide. Cells were trypsinized, kept in suspension for 2 h in 6-well plates coated with 1% BSA (Sigma). Subsequently, 1 ϫ 10 6 cells/ml untreated or pretreated with relevant blocking antibodies were seeded into separate 6-well plates coated with either 1% BSA, VEGF-C or -D (3 g/ml), or Tnfn3RAA (2.5 g/ml). Tnfn3RAA is an ␣9␤1-specific ligand, which is a recombinant form of the third fibronectin type 3 repeat of tenascin C in which the arginineglycine-aspartic acid sequence is mutated to RAA, as described previously (25,26). After 60 min, cells were washed with PBS/NaV i (1 mM) and then lysed with a buffer containing 20 mM Tris, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Triton X, 25 mM NaF, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1 mM NaV i , and protease inhibitors (complete mini EDTA-free, Roche Applied Science). Lysates were centrifuged at 14,000 g at 4°C for 10 min, and Western blotting was performed as outlined above.
Flow Cytometry-Cultured cells were trypsinized, washed with PBS, blocked with normal goat serum at 4°C for 10 min, incubated with primary antibody for 20 min at 4°C and then with phycoerythrinconjugated goat anti-mouse antibody. Labeled cells were suspended in PBS, and fluorescence was determined for 5,000 cells with a flow cytometer (FACSort, BD Biosciences). Primary mouse monoclonal antibody Y9A2 (Wang, 1996 (10 g/ml)), was used to detect ␣9␤1 and 9D9f9 (5 g/ml, gift from Dr. K. Alitalo) to detect VEGF-R3.
Adhesion Assay-96-well non-tissue culture flat-bottomed microtiter plates (ICN, Linbro/Titertek, Aurora, OH) were coated with VEGF-C or -D at 4°C overnight and then blocked with 1% BSA (Sigma) in DMEM for 30 min at 37°C. After trypsinization cells were incubated with or without the ␣9␤1-blocking antibody Y9A2 (20 g/ml) for 30 min on ice, and 5 ϫ 10 4 cells/well were seeded (8). Control wells were treated with 1% BSA in DMEM. The plates were centrifuged (top side up) at 10 ϫ g for 5 min and incubated for 1 h (MEF) or 2 h (HMVEC) at 37°C. Non-adherent cells were removed by centrifugation (top side down) for 5 min at 48 ϫ g, and adherent cells were fixed and stained with 1% formaldehyde, 0.5% crystal violet, and 20% methanol for 30 min after which the wells were washed three times with PBS. Following solubilization in 2% Triton X-100, the number of adherent cells was evaluated by measuring absorbance at 595 nm in a microplate reader (Spectra-Max 190, Molecular Devices, Sunnyvale, CA).
Migration Assay-The undersurfaces of 12-well, 8-M Transwell plates (Corning Costar, Cambridge, MA) were coated with the relevant ligand (VEGF-C or -D) or 1% BSA as a binding control at 4°C overnight. Wells were then washed with PBS and blocked with 1% BSA in DMEM or PBS for 60 min at 37°C. After trypsinization, 5 ϫ 10 4 cells in DMEM were incubated with or without Y9A2 and/or MAZ51 for 30 min on ice and then seeded into the top chamber of the Transwell (8). 1% fetal calf serum 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 with DiffQuik fixative (Pierce), and the non-adherent cells were gently removed from the top with a Q-tip. After air drying, the membranes were stained with DiffQuik (Pierce), washed in water, and air-dried, and then the membranes were cut from the Transwell and mounted onto glass slides. Cells were counted in 10 high power (ϫ25) fields for each condition.
RNA Silencing of ␣9␤1-HMVEC were grown in 10-cm dishes with full growth media (Cambrex) until 70% confluent. Cells were then washed twice with serum-free media and transfected with siRNA (20 M, Ambion, catalog no. 86946) targeted to exon 4 of the ␣9 integrin using siPort Amine (Ambion) in Opti-MEM media. Cell media were replaced with full growth media 6 h after transfection and daily thereafter. Transfection efficiency was assessed by flow cytometry analysis of ␣9␤1 expression 48 -72 h after transfection and compared with mock and untransfected cells of the same passage. Cell adhesion assays using VEGF-C or VEGF-D as substrate were performed using transfected cells in which ␣9␤1 was successfully silenced and compared with untransfected and transfected cells in which ␣9␤1 was not silenced.
Production of Non-blocking Monoclonal Antibody to ␣9␤1-A non-blocking antibody to ␣9␤1 was produced for detection of integrin bound to the VEGF protein in binding assays. Murine L2 cells (ATCC) transfected with human ␣9 were injected into mice every 14 days for a total of three injections. 3 days after the final injection, lymphocytes were harvested and spin-combined with SP2/0 myeloma cells as described previously (27). After addition of selection media (RPMI 1640, 10% fetal calf serum, 1% penicillin/streptomycin, 1% sodium pyruvate, L-glutamine), single clones were selected, and the supernatants were screened by differential flow cytometry of mock-and ␣9-transfected cells. Antibody was purified by gradient anion exchange. Metabolic Labeling of Cells and Immunoprecipitation-To determine whether A9A1 was able to immunoprecipitate the ␣9␤1 integrin efficiently, ␣9-transfected SW-480 cells were grown for 4 h in methioninefree DMEM supplemented with 1% penicillin/streptomycin and then incubated with 0.5 mCi of 35 S for 24 h. Lysates of cells labeled with [ 35 S]methionine were then precleared with protein A-Sepharose for 1 h at 4°C. The precleared supernatant was incubated with 20 g of the primary antibody A9A1 at 4°C overnight. The beads and bound protein were washed four times with buffer containing 100 mM Tris, 300 mM NaCl, 1% Triton X-100, 0.1% SDS and then boiled in reducing Laemmli's buffer at 95°C for 6 min. Protein samples were separated by 8% SDS-PAGE, and autoradiography was performed after 1 week at Ϫ80°C.
Immunoaffinity Purification of Human Integrin ␣v␤6 -Secreted ␣v␤6 integrin was collected from conditioned media of Chinese hamster ovary cells stably transfected with truncated forms of each subunit of the integrin as described previously (28). The integrin was purified using immunoaffinity chromatography on columns with the ␣v␤6 antibody R6G9 covalently cross-linked to protein A-Sepharose beads. Bound proteins were eluted with 100 mM glycine, pH 3, into 0.05 volume of 1 M Na 2 PO 4 , pH 8 (29).
Binding Assay-Direct binding of VEGF-C and -D to ␣9␤1 was assessed by a solid phase binding assay in non-tissue-coated 96-well microtiter plates (Nunc ImmunoPlate, Naperville, IL). Recombinant VEGF-C or -D (5 g/ml) was attached to the plates, and purified ␣9␤1 was added for 2 h at room temperature in the presence or absence of 10 mM EDTA. Following five washes with PBS/1% BSA/0.05% Tween, the extent of ␣9␤1 binding was detected using the A9A1 antibody (20 g/ml, 1 h at 37°C). After incubation with biotinylated rat anti-mouse antibody (1:250, 1 h, room temperature), streptavidin-horseradish peroxidase was added for 20 min at room temperature followed by 3,3Ј,5,5Јtetramethylbenzidine substrate solution (Pharmingen). Absorbance was then measured at 450 nm with a microplate reader (Molecular Devices). The functional integrity of the purified ␣9␤1 integrin was assessed in a similar assay using the ␣9␤1-specific ligand Tnfn3RAA (5 g/ml) as described previously (25,26).
Statistical Methods-For statistical analysis of adhesion and migration assays comparisons were made by analysis of variance with a post-hoc Tukey's test. Statistical significance was assumed at p Յ 0.05 with respect to a two-tailed probability distribution. Data are presented as mean values Ϯ S.E. unless otherwise stated from at least three separate experiments.

RESULTS
Production and Purification of VEGF-C and -D-Fully processed VEGF-C and -D proteins were purified from transfected Drosophila S2 cells. Western blots (Fig. 1A) of secreted and purified VEGF-C (top panel) and VEGF-D (bottom panel) with anti-V5 antibody (left panels) demonstrate proteins with the expected molecular mass for VEGF-C (or -D) tagged with V5 and His (ϳ23-24 kDa). In addition, immunoblotting with VEGF-C-and -D-specific antibodies (right panels) showed that the V5-tagged proteins were indeed VEGF-C and -D. The purity of each recombinant protein was determined by silver staining of 15% SDS-polyacrylamide gels as shown in Fig. 1B. VEGF-C and -D-V5 induced phosphorylation of their cognate receptor, VEGF-R3. Fig. 1C shows the expected VEGF-R3 isoforms and confirms the activity of the purified VEGF proteins.
Binding of VEGF-C or -D Activates ␣9␤1 Integrin-We have shown previously that in SW-480 cells, activation of the ␣9␤1 integrin by established ligands results in phosphorylation of Erk 1/2 and paxillin. To assess whether binding of VEGF-C or -D similarly results in ␣9␤1-induced signaling, phosphorylation of these proteins was compared in mock-and ␣9-transfected SW-480 cells in the presence or absence of the ␣9␤1blocking antibody. Fig. 3A shows that after 15 min of ␣9 transfectants binding to adherent Tnfn3RAA (lanes 3 and 4), VEGF-C (lanes 5 and 6) or VEGF-D (lanes 7 and 8), Erk 1/2 phosphorylation was increased in an ␣9-dependent fashion. Similar to 1% BSA (lane 1), this response to VEGF was not seen in mock-transfected cells (lane 2). Paxillin phosphorylation was also induced after 15 min in ␣9 transfectants and was inhibited by the ␣9␤1-blocking antibody (Fig. 3B).

␣9␤1-Dependent Cell Adhesion Is Demonstrable in Primary
Endothelial Cells-To verify the biological significance of the results obtained with the transfected cell lines and to understand the role of ␣9␤1 in adhesion to VEGF-C and -D in the presence of their cognate receptor VEGF-R3, HMVEC were also studied. HMVEC were found to express both VEGF-R3 and ␣9␤1 as measured by flow cytometry (Fig. 4, A and B). Essentially, all of the cells expressed VEGF-R3 (Fig. 4A), belying their lymphatic origin, whereas the expression of ␣9␤1 varied from 40 -55% (Fig. 4B) depending on the lot purchased and decreased with increasing passage number. As a result, all experiments were performed between passages 3 and 7. Like ␣9-transfected cell lines, HMVEC demonstrated concentrationdependent adhesion to both VEGF-C (Fig. 4C) and VEGF-D (Fig. 4D). Adhesion to both substrates was inhibited by the ␣9␤1-blocking antibody Y9A2 (p Ͻ 0.05) to the levels measured in wells coated with 1% BSA, demonstrating that the presence of VEGF-R3 is not sufficient to mediate cell adhesion to VEGF-C or -D.
Similar results were obtained when ␣9␤1 was silenced using siRNA transfection (Fig. 5). The effectiveness of siRNA directed against exon 4 of ␣9 (␣9 siRNA) was assessed by measuring ␣9 expression on the cell surface by flow cytometry (Fig. 5A). The expression of ␣9 in these cells was compared against cells transfected with mock siRNA and cells not treated with siRNA (No siRNA). Adhesion assays were then performed using these cell populations. Fig. 5B shows that knock-down of ␣9 in HM-VEC cells significantly inhibits cell adhesion on both VEGF-C and VEGF-D. Mock siRNA cells adhered to VEGF-C or VEGF-D substrate to the same degree as non-transfected HMVEC.
Cell Migration on VEGF-C and -D Is also ␣9␤1-Dependent-We have shown previously that one of the principal func- tions of ␣9␤1 is enhancement of cell migration. Because migration of lymphatic endothelial cells into mesenchyme enriched for VEGF-C (48) and/or VEGF-D is potentially an important step in lymphangiogenesis, we next sought to determine the role of ␣9␤1 in migration on VEGF-C and -D. ␣9-Transfected MEF migrated on VEGF-C (Fig. 6A) and -D (Fig. 6B) in a concentration-dependent manner, whereas mock-transfected cells did not migrate on any concentration above values seen for migration on BSA. This effect depended on ligation of ␣9␤1, because migration on VEGF-C or -D was inhibited by the ␣9␤1-blocking antibody. Similar results were obtained for HM-VEC, which also demonstrated concentration-dependent migration on VEGF-C (Fig. 6C) and VEGF-D (Fig. 6D) that was inhibited by the ␣9␤1-blocking antibody.
To assess the relative contribution of VEGF-R3 and ␣9␤1 to HMVEC cell migration, cells were treated with the ␣9␤1-blocking antibody and/or the VEGF-R3-blocking chemical MAZ51 (49). Fig. 6E shows that blocking either ␣9␤1 or VEGF-R3 significantly inhibits cell migration. These results suggest that both ␣9␤1 and VEGF-R3 are required for maximal cell migration on immobilized VEGF.
Production of Non-blocking ␣9␤1 Antibody-A mouse antihuman monoclonal antibody (A9A1) was produced to serve as a detection antibody for ␣9␤1 in VEGF-C or -D solid phase binding assays. Flow cytometry using the antibody A9A1 shows detection of the ␣9␤1 integrin on ␣9-transfected MEF to a similar level as that of Y9A2, an established ␣9␤1 integrin antibody (Fig. 7A). To further characterize its specificity, A9A1 was used to immunoprecipitate lysates of [ 35 S]methionine-labeled ␣9-transfected SW-480 cells. Fig. 7B shows an autoradiograph of immunoprecipitated protein samples separated by 8% SDS-PAGE where the ␣9 and ␤1 bands of the heterodimer are clearly seen. Fig. 7C shows that in contrast to Y9A2, a blocking antibody to ␣9␤1, A9A1 did not inhibit adhesion of ␣9-transfected MEF to the ␣9␤1-specific ligand Tnfn3RAA. This antibody was thus suitable for use in ␣9␤1-VEGF binding assays.
VEGF-C and -D Bind the Integrin ␣9␤1-Solid phase binding assays demonstrated robust concentration-dependent adhesion of ␣9␤1 to VEGF-C and -D, both at 5 g/ml (Fig. 9, A and  B). In both cases, binding was completely inhibited by chelating divalent cations (EDTA, 10 mM) as expected for authentic integrin-ligand interactions. The irrelevant integrin ␣v␤6 (28) showed no binding to either VEGF-C or -D. DISCUSSION In this study we found that ␣9-transfected cells (lacking VEGF receptors) adhere and migrate on fully processed VEGF-C and -D. These cellular responses require functional ␣9␤1 because they are inhibited by a blocking antibody. These results alone suggest that an interaction between ␣9␤1 and the VEGF-C and -D proteins can affect cell behavior in the absence of VEGF-R3, the cognate receptor of the growth factors. Importantly, ␣9␤1was also shown to mediate adhesion and migration on VEGF-C and -D in primary human endothelial cells that co-express ␣9␤1 and VEGF-R3. Inhibition of migration by the ␣9␤1 antibody and/or by blocking VEGF-R3 suggests that VEGF-R3 is not sufficient to mediate maximal migration on VEGF-C or -D. Solid phase binding assays of purified VEGF-C or -D and ␣9␤1 verified that these cell functions are a result of VEGF-C and -D binding ␣9␤1 in an integrin-specific and cationdependent manner. Moreover, binding of VEGF-C or -D to ␣9␤1 leads to induction of downstream signals as demonstrated by ␣9␤1-dependent Erk 1/2 and paxillin phosphorylation. The finding that growth factors (VEGF-C and -D) bind to an integrin (␣9␤1) highlights a novel and important mechanism of integrin-growth factor interaction.
A number of previous studies have described cooperative interactions between integrins and growth factors (reviewed in Refs. 31 and 32). Growth factor ligation of its cognate receptor has been shown to lead to close physical association with integrins. For example, ligation of the platelet-derived growth factor receptor not only allowed its co-immunoprecipitation with the ␣v␤3 integrin but also potentiated ␣v␤3-dependent cell migration (33). In addition, input from VEGF-R2, stimulated by VEGF-A, has been shown to activate the ␣v integrins ␣v␤3 and ␣v␤5 and the ␤1 integrins ␣5␤1 and ␣2␤1 (34). Activated integrins have also been shown to modulate growth factor protein and receptor expression and signaling. For example, integrin ␣6␤4 increases VEGF-A translation in breast carcinoma cells through inactivation of the translational repressor 4E-BP1 (35). In addition, integrin-matrix interactions modulate expression of VEGF and fibroblast growth factor receptor 1 and 2 expression on HMVEC (36). The functional importance of growth factor/integrin interactions has also been demonstrated in vivo where blocking antibodies to ␣v␤3 and ␣v␤5 integrins inhibit basic fibroblast growth factor-and VEGF-induced angiogenesis, respectively (37). Finally, mice lacking the integrin ␣v␤5 are specifically protected from the increases in vascular permeability induced by VEGF-A, again presumably through cross-talk with VEGF-R2 (38).
In the current study, we describe a novel mechanism of growth factor-integrin interaction, direct ligation of an integrin by a growth factor. This interaction thus provides a mechanism for VEGF-C and -D to affect directly cell behavior (i.e. cell adhesion and migration) even in the absence of their cognate receptor, VEGF-R3. Furthermore, even in cells expressing the cognate receptor, ligation of the integrin ␣9␤1 is required for stable cell adhesion to these growth factors and substantially enhances cell migration across them. Our results do not exclude the possibility that ␣9␤1 may modulate cell function by activating another receptor capable of binding VEGF or inhibiting a VEGF-R3 suppressor. However, taken together these results strongly suggest that VEGF-C and -D directly bind to ␣9␤1. We speculate that simultaneous binding of both ␣9␤1 and VEGF-R3 by VEGF-C and/or -D, which is present at high concentration in the extracellular matrix, results in co-clustering of both receptors, facilitating cooperative interactions between the two receptors as demonstrated by our cell migration data. The specific binding sites on VEGF-C and -D and their relationship to the binding sites on other ␣9␤1 ligands, such as tenascin-C and osteopontin, are as yet undetermined. Because of the lack of a conserved binding sequence for known ␣9␤1 ligands, determination of this site for VEGF-C or -D would be difficult. Because the VEGF family members share a VEGF homology domain our results suggests that other VEGF family member proteins may also interact with ␣9␤1. It is also clear that in addition to integrins, other cell surface proteins can modulate VEGF responses, such as the other VEGF receptors (39) and the neuropilins (40). The potential interactions between these modulating proteins and ␣9␤1 remain to be determined.
Normal vessel development requires the coordinated function of both cellular and extracellular regulatory and effector proteins to ensure an optimal milieu for correct vessel morphogenesis and function (41). The molecular regulators of angiogenesis include the VEGF family of proteins and receptors FIG. 4. Primary endothelial cell adhesion to VEGF-C and -D is ␣9␤1-dependent. Flow cytometry analysis of HMVEC labeled with 9D9f9 (A) and Y9A2 (B) to measure expression of VEGF-R3 and ␣9␤1, respectively. Shaded areas represent cells labeled with no primary antibody, and the unshaded areas represent cells labeled with the primary antibody as indicated. Purified VEGF-C (C) and VEGF-D (D) were used as substrates for adhesion assays with HMVEC in the absence (dark bars) or presence (stippled bars) of the ␣9␤1-blocking antibody Y9A2. Cells were processed as described in the legend to Fig. 2. *, p Ͻ 0.05 compared with cells treated with Y9A2; #, p Ͻ 0.05 compared with 0 g/ml VEGF. Ab, antibody. (VEGF-R1, -R2, -R3), acidic and basic fibroblast growth factors, angiopoietins and Tie receptors, ephrins, integrins, and matrix metalloproteases (10,42). The molecules that regulate lymphangiogenesis have not been as well characterized. However, it is clear that VEGF-C and -D are key modulators of lymphatic development and function (21,22), although their relative roles remain to be determined (19). These growth factors induce lymphatic endothelial cell growth, survival, and migration in vitro (16) and lymphangiogenesis in vivo (43), mediated at least in part through their activation of VEGF-R3. In addition, loss of function mutations in VEGF-R3 have been described in a significant subgroup of patients with congenital human lymphedema (44).
The ␣9␤1 integrin has now been described to interact with a relatively large number of ligands including tenascin C, osteopontin, vascular cell adhesion molecule-1, coagulation factor XIII, and several members of the ADAMs family of transmembrane metalloproteinases (7,24,(45)(46)(47). The biological significance of ␣9␤1 interactions with most of these ligands remains to be determined. In contrast, we now show that the ligands VEGF-C and -D play a role in endothelial cell adhesion and migration, key cellular functions required for lymphangiogen- The observed bands represent the ␣9 and ␤1 subunits of the integrin. C, Tnfn3RAA, an ␣9-specific ligand, was used as substrate for adhesion assays in ␣9-transfected MEF in the absence (dark bars) or presence (diagonal bars) of the A9A1 antibody and compared with the blocking antibody Y9A2 (stippled bars). *, p Ͻ 0.05 compared with cells treated with Y9A2. Ab, antibody. esis. The most dramatic phenotypic feature of ␣9 null mice is the presence of bilateral chylothoraces, suggesting a functional defect in the collecting lymphatics of the thorax. Subtle edema and the accumulation of lymphocytes around some peripheral lymphatics provide further support for a role of this integrin in lymphangiogenesis (9). The phenotype of VEGF-D null mice is yet to be published; however, it is clear from E10.5 VEGF-C heterozygote mice that VEGF-C is expressed in mesenchymal cells surrounding the jugular vein toward which endothelial cells that already have a lymphatic differentiation (Prox-1positive) must migrate (48). Given the specialized role that ␣9␤1 plays in accelerated cell migration (8), we speculate that abnormal migration on VEGF-C and/or -D might be responsible for the functional defects that occur in ␣9 null mice. However, the fact that lymphatic vessels are formed in these mice suggests that either the presence of VEGF-R3 or some other unidentified receptor(s) can partially compensate for the loss of ␣9␤1.
In summary, the findings of this study describe a novel integrin-growth factor interaction whereby the lymphangiogenic proteins VEGF-C and -D bind the integrin ␣9␤1. Although the precise mechanisms by which binding of VEGF-C and/or -D to ␣9␤1 contributes to lymphatic development remain to be fully elucidated, these data strongly support a role for this interaction in explaining the lymphatic defects in ␣9 knock-out mice. FIG. 8. Purification of active ␣9␤1. ␣9␤1 integrin was purified from cell lysates of ␣9-transfected SW-480 cells by affinity chromatography using an ␣9␤1 antibody (A9A1) column. Proteins under reducing conditions were separated by 8% SDS-PAGE and analyzed by silver staining (A) and Western blot analysis with 1057, rabbit polyclonal antibody to ␣9␤1 (B). Molecular mass markers (kDa) are indicated. C, Tnfn3RAA, an ␣9-specific ligand, was used as substrate for binding assays with purified ␣9␤1 in the absence (diamonds) or presence (squares) of 10 mM EDTA.
FIG. 9. VEGF-C and -D bind directly to the integrin ␣9␤1. Purified VEGF-C (A) or VEGF-D (B) was used for solid phase binding assays with purified ␣9␤1 at various concentrations in the absence (diamonds) or presence (squares) of 10 mM EDTA. Similar assays were performed using purified ␣v␤6, an irrelevant integrin (triangles).