Stimulation of beta 1 integrin induces tyrosine phosphorylation of vascular endothelial growth factor receptor-3 and modulates cell migration.

Interactions between integrins and tyrosine kinase receptors can modulate a variety of cell functions. We observed a cooperative interaction between the beta(1) integrin and vascular endothelial growth factor receptor-3 (VEGFR-3 or Flt4) that appeared to be required for cell migration. By using VEGFR-3-transfected 293 cells (293/VEGFR-3) or primary dermal microvascular endothelial cells (DMEC), we found that stimulation with either soluble or immobilized extracellular matrix (ECM) proteins, collagen or fibronectin (FN), resulted in the increased tyrosine phosphorylation of VEGFR-3 in the absence of a cognate ligand. This increased tyrosine phosphorylation of VEGFR-3 was diminished by pretreatment with a blocking antibody against the beta(1) integrin. Cross-linking with anti-beta(1) integrin antibody induced a similar degree of tyrosine phosphorylation of VEGFR-3. Stimulation with collagen or FN induced an association between beta(1) integrin and VEGFR-3 in both 293/VEGFR-3 and primary DMEC cells. Collagen or FN-induced tyrosine phosphorylation of VEGFR-3 was inhibited by treatment with cytochalasin D, an inhibitor of actin polymerization. Collagen or FN was able to induce the migration of 293/VEGFR-3 or DMEC cells to a limited extent. However, migration was dramatically enhanced when a gradient of the cognate ligand, VEGF-D, was added. VEGF-D failed to induce cell migration in the absence of ECM proteins. Introducing a mutation at the kinase domain of VEGFR-3 or treatment with blocking antibody against either VEGFR-3 or beta(1) integrin inhibited cell migration induced by ECM and VEGF-D, indicating that signals from both beta(1) integrin and VEGFR-3 are required for this cell function.


Interactions between integrins and tyrosine kinase receptors can modulate a variety of cell functions. We observed a cooperative interaction between the ␤ 1 integrin and vascular endothelial growth factor receptor-3 (VEGFR-3 or Flt4) that appeared to be required for cell migration. By using VEGFR-3-transfected 293 cells (293/ VEGFR-3) or primary dermal microvascular endothelial cells (DMEC), we found that stimulation with either soluble or immobilized extracellular matrix (ECM) proteins, collagen or fibronectin (FN), resulted in the increased tyrosine phosphorylation of VEGFR-3 in the absence of a cognate ligand. This increased tyrosine phosphorylation of VEGFR-3 was diminished by pretreatment with a blocking antibody against the ␤ 1 integrin. Cross-linking with anti-␤ 1 integrin antibody induced a similar degree of tyrosine phosphorylation of VEGFR-3. Stimulation with collagen or FN induced an association between ␤ 1 integrin and VEGFR-3 in both 293/VEGFR-3 and primary DMEC cells. Collagen or FNinduced tyrosine phosphorylation of VEGFR-3 was in-
hibited by treatment with cytochalasin D, an inhibitor of actin polymerization. Collagen or FN was able to induce the migration of 293/VEGFR-3 or DMEC cells to a limited extent. However, migration was dramatically enhanced when a gradient of the cognate ligand, VEGF-D, was added. VEGF-D failed to induce cell migration in the absence of ECM proteins. Introducing a mutation at the kinase domain of VEGFR-3 or treatment with blocking antibody against either VEGFR-3 or ␤ 1 integrin inhibited cell migration induced by ECM and VEGF-D, indicating that signals from both ␤ 1 integrin and VEGFR-3 are required for this cell function.
The interaction of cells with extracellular matrix (ECM) 1 proteins generates a series of signaling events that can regulate growth, differentiation, adhesion, and migration (1,2). The adhesive interactions between cells and ECM are mediated by integrins, a family of cell surface receptors that bind to ECM proteins including collagens, vitronectin, laminins, and fibronectin. Integrins are heterodimeric molecules composed of ␣and ␤-subunits. At least 17 different ␣-subunits and 8 ␤-subunits have been identified to date. The various combinations of the ␣and ␤-subunits yield integrin dimers with diverse ligand specificity and biological activities (3).
Integrins not only function as ECM adhesion molecules but also transduce biochemical signals into the cell to affect various proteins and second messengers, such as tyrosine kinases, serine/threonine kinases, lipid mediators, low molecular weight GTPases, and intracellular calcium fluxes (4,5). Integrins can synergistically act with growth factors to modulate cell functions and share signaling pathways initiated by receptor tyrosine kinases (RTKs) (6 -9). In certain types of cells, integrinmediated cell adhesion appears to be required for optimal activation of growth factor receptors (2). Engagement of integrins by ECM proteins can modulate the activation of RTKs, even in the absence of their ligands (9 -11). The possible molecular basis of such cross-talk between integrins and RTKs may involve the physical interaction of these two classes of proteins, forming macromolecular complexes on the cell membrane. Thus, the ␣ v ␤ 3 integrin can be immunoprecipitated in complexes with the receptor for insulin (12,13), PDGF (13)(14)(15), or VEGF (15,16); ␣ 5 ␤ 1 and perhaps other ␤ 1 integrins can associate with the EGF receptor (10,11), whereas ␣ 6 ␤ 1 or ␣ 6 ␤ 4 can associate with the ErbB-2 receptor (17). ECM-induced aggregation of integrins can trigger co-aggregation and autophosphorylation of the PDGF (13)(14)(15) or EGF (10,11) receptor. However, how such interactions occur and their functional consequences have not been well delineated.
Integrin-mediated interaction of endothelial cells with ECM proteins is essential for angiogenesis, a complex process that involves extracellular matrix remodeling, endothelial cell migration and proliferation, and maturation of new endothelial cells into mature blood vessels. Angiogenesis is regulated by various growth factors such as VEGF, placenta growth factor, fibroblast growth factor, and PDGF. Synergistic effects of these growth factors and certain ECM proteins have been described (15, 18 -24). VEGF receptor-3 (VEGFR-3 or Flt4) is a member of a subfamily of class III receptor tyrosine kinases characterized by seven extracellular immunoglobulin-like domains and a split, intracellular domain containing kinase activity (25,26). Two glycoproteins, VEGF-C and VEGF-D, have been identified as the ligands for VEGFR-3 (27)(28)(29). Tissue expression of the VEGFR-3 receptor is relatively restricted, being found predominantly in lymphatic and microvascular endothelium (30 -32 It is likely that VEGFR-3 is important for lymphangiogenesis (33)(34)(35).
In this report, we present evidence that the ECM proteins, collagen and fibronectin, through their binding to ␤ 1 integrin, can stimulate the tyrosine phosphorylation of the VEGFR-3 receptor and induce the formation of a ␤ 1 integrin⅐VEGFR-3 receptor complex. Furthermore, endothelial cell migration required signals from both the ligand-activated VEGFR-3 and the ECM-activated ␤ 1 integrin.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Recombinant VEGF-D was purchased from R & D Systems (Minneapolis, MN). Rabbit polyclonal antibody for VEGFR-3 (Flt4) was obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The mouse anti-phosphotyrosine monoclonal antibody (mAb) 4G10 was a generous gift of Dr. Brian Druker (University of Oregon). Mouse anti-human ␤ 1 integrin-blocking mAb and anti-␤ 1 integrin mAb for immunoprecipitation and Western blotting were purchased from Chemicon International Inc. (Temecula, CA). Normal rabbit serum and purified normal rabbit IgG were purchased from Organon Teknika Corp. (West Chester, PA). The recombinant human VEGFR-3 (Flt4)/Fc chimera (VEGFR-3/Ig) containing the extracellular domain of human VEGFR-3 and the Fc portion of human immunoglobulin were purchased from R & D Systems (Minneapolis, MN). Cytochalasin D was obtained from Calbiochem. Poly-L-lysine was purchased from Sigma. Electrophoresis reagents and nitrocellulose membrane were obtained from Bio-Rad. Protein A-Sepharose CL-4B and Gamma-Bind ® plus Sepharose ® were obtained from Amersham Pharmacia Biotech. The protease inhibitors leupeptin, aprotinin, and ␣ 1 -antitrypsin and all other reagents were obtained from Sigma. Calcium and potassium-free phosphate-buffered saline (PBS) was obtained from Life Technologies, Inc.
Cells and Cell Culture-293 cells stably transfected with VEGFR-3 (293/VEGFR-3) (28) were kindly provided by Genentech Inc. (South San Francisco, CA) and shown to be mycoplasma-free prior to their expansion in culture. 293/VEGFR-3 cells were cultured in DMEM with 10% fetal calf serum. Dermal microvascular endothelial cells (DMEC) were purchased from Clonetics Inc. (Palo Alto, CA) and expanded in EGM-2 medium. 5-8 passages were used in the described experiments.
Stimulation of Cells by ECM Proteins-293/VEGFR-3 cells were starved in serum-free DMEM for 20 h or DMEC were starved in EGM-2 medium without serum and without supplements for 5 h. After starvation, cells were washed and incubated with PBS containing 0.1 mM Na 3 VO 4 for 20 min. Cells were then stimulated with collagen (8 g/ml), fibronectin (10 g/ml), or VEGF-D (100 ng/ml) in PBS for different times as indicated at 37°C. For stimulation with immobilized ECM proteins, cells were detached by PBS containing 0.2 mM EDTA, washed twice with PBS, and resuspended at 5 ϫ 10 6 /ml in PBS containing 0.1% bovine serum albumin (BSA). Cells were then stimulated by loading onto tissue culture dishes coated with collagen or fibronectin for different times as indicated at 37°C. After stimulation, cells were washed with ice-cold PBS containing 0.1 mM Na 3 VO 4 , and cell lysates were prepared in a lysis buffer containing 50 mM Hepes, pH 7.0, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 10 g/ml each of aprotinin, leupeptin, and pepstatin. Because receptor complexes can be dissociated in Triton X-100 buffer (36), 15 mM CHAPS was used to replace Triton X-100 in the above lysis buffer for studying the co-immunoprecipitation of VEGFR-3 with ␤ 1 integrin. Total cell lysates were clarified by centrifugation at 10,000 ϫ g for 10 min. Protein concentrations were determined by protein assay (Bio-Rad).
Immunoprecipitation and Western Blot Analysis-For the immunoprecipitation studies, identical amounts of protein from each sample were clarified by incubation with protein A-Sepharose for 1 h at 4°C. Following the removal of protein A-Sepharose by brief centrifugation, the solution was incubated with different primary antibodies as detailed below for each experiment for 4 h or overnight at 4°C. Immunoprecipitation of the antibody-antigen complexes was performed by incubation for 2 h at 4°C with 75 l of protein A-Sepharose (10% suspension) or 50 l of GammaBind ® plus Sepharose ® . Nonspecific bound proteins were removed by washing the Sepharose beads three times with lysis buffer and one time with PBS. Bound proteins were solubilized in 40 l of 2ϫ Laemmli sample buffer and further analyzed by immunoblotting. Samples were separated on SDS-PAGE and then transferred to nitrocellulose membranes. The membranes were blocked with 5% non-fat milk protein and probed with primary antibody for 2 h at room temperature or 4°C overnight. Immunoreactive bands were visualized using horseradish peroxidase-conjugated secondary antibody and the enhanced chemiluminescent (ECL) system (Amersham Pharmacia Biotech). Each experiment was repeated at least 3 times, and the presented blots are representative of these.
Cross-linking Stimulation with Anti-␤ 1 Integrin Antibody-293/ VEGFR-3 cells were starved in serum-free DMEM, washed with PBS, and incubated with anti-␤ 1 integrin antibody (10 g/ml in PBS) at 4°C for 45 min. After washing with cold PBS, cells were incubated with rabbit anti-mouse IgG (10 g/ml in PBS) for 15 min at 4°C and then 10 min at 37°C. Total cell lysates were prepared as described above. Tyrosine phosphorylation of VEGFR-3 was examined by immunoprecipitation with anti-VEGFR-3 antibody followed by Western blot analysis with anti-phosphotyrosine antibody.
DNA Constructs, Site-directed Mutagenesis, and Transfection-The wild type VEGFR-3 expression construct was generated by cloning the long form of VEGFR-3 cDNA into the vector, PRK5, as described previously (28). The expression construct of the mutant form of VEGFR-3 was made in PRK5, in which a Gly to Arg mutation at the 857 site was generated by an in vitro site-directed mutagenesis kit, QuickChange TM XL (Stratagene, La Jolla, CA), using oligonucleotide primers as follows: P1, 5Ј-CGGCGCCTTCAGGAAGGTGGT-3Ј; P2, 5Ј-ACCACCT-TCCTGAAGGCGCCG-3Ј. The constructs were verified by DNA sequencing using the ABI PRISM 3777 System. For transient expression of wild type or mutant VEGFR-3 in 293 or 293/VEGFR-3 cells, Effectene TM Transfection Reagent (Qiagen, Valencia, CA) was used according to the manufacturer's instructions.
Migration Assays-Migration assays were performed in triplicate using 8-m pore filters (Transwell, 24-well cell clusters; Costar, Boston, MA). The lower side of each filter was coated with type I collagen isolated from rat tail (50 g/ml in PBS), FN (25 g/ml in PBS), poly-Llysine (1 mg/ml), or BSA (50 g/ml) as a control for 2 h at room temperature. The filters were blocked with PBS containing 1% BSA for 1 h at 37°C and then rinsed with migration medium (DMEM with 0.5% BSA for 293 or 293/VEGFR-3 cells, RPMI 1640 with 0.5% BSA for DMEC), and the supernatant was aspirated immediately before loading the cells. Cells (about 80% of confluence) were harvested by releasing them from flasks with 2 mM EDTA for 293 or 293/VEGFR-3 cells or with trypsin/EDTA for DMEC. Cells were washed two times with migration medium and then loaded into the upper chamber of the inserts (2 ϫ 10 5 cells/0.1 ml migration medium). The inserts were then carefully transferred to another well containing 650 l of migration medium with various concentrations of VEGF-D (R & D Systems). The plates were incubated at 37°C in 5% CO 2 for 3.5 h. Non-migrated cells were removed from the upper chambers by wiping the upper surface with a cotton-tip applicator, and the migrated cells on the under surface were fixed and stained with Diff-Quik ® stain Kit (Baxter Healthcare Corp., McGaw Park, IL). The number of migrated cells was counted in 10 randomly selected high power (ϫ 200) fields per insert. No cells were found in the lower chambers where the inserts were placed. Each determination represents the average of three individual inserts, and error bars represent mean Ϯ S.D. Statistical significance was determined using the Student's t test.

Collagen or Fibronectin Induces Tyrosine Phosphorylation of VEGFR-3-
To assess if collagen activates VEGFR-3, we first examined the tyrosine phosphorylation of VEGFR-3 in 293/ VEGFR-3 cells after stimulation with a soluble form of collagen. Serum-starved 293/VEGFR-3 cells were stimulated with human type I collagen (8 g/ml) or VEGF-D (50 ng/ml) for various times. VEGFR-3 was immunoprecipitated by anti-VEGFR-3 antibody and probed with anti-phosphotyrosine monoclonal antibody (4G10). As shown in Fig. 1A, collagen stimulation induced a significant increase in the tyrosine phosphorylation of VEGFR-3. This increased tyrosine phosphorylation was rapid and transient, detected as early as 1 min, reached a maximum at 5 min, and returned to a basal level at 30 min after stimulation (Fig. 1A, upper panel). Equal amounts of VEGFR-3 were present in each lane (Fig. 1A, lower panel). Used as a positive control, VEGF-D also stimulated the tyrosine phosphorylation of VEGFR-3. We also observed a similar induction of VEGFR-3 phosphorylation by type I collagen isolated from rat tail (data not shown). In addition to type I collagen, we demonstrated a similar induction of VEGFR-3 tyrosine phosphorylation following treatment with other forms of human collagen including types III-V (data not shown). These results indicated that the soluble form of various types of collagen can activate VEGFR-3.
Because collagen encountered in vivo will usually be immobilized as part of the extracellular matrix, we next examined the tyrosine phosphorylation of VEGFR-3 in 293/VEGFR-3 cells stimulated by immobilized collagen. Serum-starved 293/ VEGFR-3 cells were detached with PBS containing 1 mM EDTA, washed twice, and resuspended in PBS with 0.1% BSA at 5 ϫ 10 6 /ml. The cells were then loaded onto collagen I-coated dishes. The dishes were centrifuged briefly and incubated at 37°C for different times. The tyrosine phosphorylation of VEGFR-3 was then examined as described above. As shown in Fig. 1B, tyrosine phosphorylation of VEGFR-3 was stimulated by immobilized collagen with similar kinetics to that induced by soluble collagen (Fig. 1B, upper panel). We verified similar protein loading in each lane of the immunoblot (Fig. 1B, lower  panel).
We also found that stimulation with either soluble (Fig. 1C) or immobilized forms (data not shown) of FN induced the tyrosine phosphorylation of VEGFR-3 in 293/VEGFR-3 cells.
To exclude the possibility that VEGFR-3 phosphorylation induced by collagen or FN was mediated through autocrine stimulation by its natural ligands, 293/VEGFR-3 cells were stimulated with collagen, FN, or VEGF-D in the absence or presence of the recombinant human VEGFR-3 (Flt4)/Fc chimera protein (VEGFR-3/Ig). VEGFR-3/Ig has been shown to bind to both VEGF-D and VEGF-C and to block ligand/receptor binding (37,38). As shown in Fig. 1D, the tyrosine phosphorylation of VEGFR-3 induced by VEGF-D was blocked by VEGFR-3/Ig, whereas the collagen-or FN-stimulated phosphorylation of VEGFR-3 was not inhibited in the presence of VEGFR-3/Ig. These results indicated that matrix protein-induced phosphorylation of VEGFR-3 was not mediated by the release of receptor ligands.
To investigate the effects of collagen or FN on normal primary cells expressing endogenous VEGFR-3, we utilized DMEC which express robust levels of VEGFR-3. 2 We showed that both collagen and FN induced the tyrosine phosphorylation of VEGFR-3 in DMEC (Fig. 2, A and B).
These results indicate that the extracellular matrix protein, collagen or FN, can activate VEGFR-3 by inducing its tyrosine phosphorylation.
Collagen or FN-induced Tyrosine Phosphorylation of VEGFR-3 Is Mediated by ␤ 1 Integrin-Receptors for collagen include ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrins, whereas the major receptors for FN are ␣ 4 ␤ 1 and ␣ 5 ␤ 1 , which have been found to be expressed in many cell types (40,41). The functional roles of collagen or FN in regulating cell proliferation, survival, adhesion, and migration are mediated through such cell surface expressing molecules. Thus, we tested whether the collagen-or FN-stimulated tyrosine phosphorylation of VEGFR-3 was mediated by ␤ 1 integrin. To that end, we used blocking antibody against ␤ 1 and treated the 293/VEGFR-3 cells before collagen or fibronectin stimulation. As shown in Fig. 3, treatment of 293/VEGFR-3 cells with anti-␤ 1 blocking antibody, but not control isotype IgG antibody, inhibited the collagen (Fig. 3A) or FN (Fig. 3B)-induced tyrosine phosphorylation of VEGFR-3. These results indicated that the collagen-or FN-induced tyrosine phosphorylation of VEGFR-3 was mediated through ␤ 1 integrin.
To investigate further the role of ␤ 1 integrin in the activation of VEGFR-3, experiments were performed to assess the tyrosine phosphorylation of VEGFR-3 in 293/VEGFR-3 cells after cross-linking with anti-␤ 1 integrin antibody. Serum-starved 293/VEGFR-3 cells were incubated with anti-␤ 1 or control isotype IgG for 45 min at 4°C, followed by incubation with sheep anti-mouse IgG for 30 min at 4°C. The variously treated dishes were incubated at 37°C for 10 min, and the cells were then treated with a lysis buffer. The tyrosine phosphorylation of VEGFR-3 was analyzed by immunoprecipitation with anti-VEGFR-3 antibody and Western blotting with anti-phosphotyrosine monoclonal antibody (4G10). As shown in Fig. 4, cross- linking of anti-␤ 1 antibody, but not control normal mouse IgG, with sheep anti-mouse IgG, resulted in an increase in VEGFR-3 tyrosine phosphorylation. These results indicated that ␤ 1 integrin-mediated enhancement of the tyrosine phosphorylation of VEGFR-3 was dependent on the clustering of these cell surface integrin molecules.

VEGFR-3 Associates with ␤ 1 Integrin and Is Enhanced by Collagen or FN Stimulation-To investigate whether VEGFR-3
associates with ␤ 1 integrin, we performed co-immunoprecipitation and Western blot analysis. Serum-starved 293/VEGFR-3 cells were stimulated by soluble collagen or FN for different times. Cell lysates were then immunoprecipitated with anti-VEGFR-3 antibody and subjected to Western blot analysis with anti-␤ 1 integrin antibody. As shown in Fig. 5A, collagen or fibronectin stimulation induced an association of VEGFR-3 with ␤ 1 integrin. A similar association of VEGFR-3 with ␤ 1 integrin was shown in DMEC (Fig.  5B). This enhanced association was confirmed by immunoprecipitating with anti-␤ 1 integrin antibody and Western blotting with anti-VEGFR-3 antibody (data not shown).
Disruption of actin cytoskeleton integrity decreased the collagen-induced phosphorylation of VEGFR-3.
The integrity of the actin cytoskeleton has been shown to be necessary for many integrin-mediated signaling events (6). We investigated whether actin cytoskeleton integrity was required for the collagen-induced tyrosine phosphorylation of VEGFR-3. Before stimulation with collagen, 293/VEGFR-3 cells were incubated with 0.4 M cytochalasin D, a known inhibitor of actin polymerization. We demonstrated that the collagen-induced phosphorylation of VEGFR-3 was inhibited by cytochalasin D (Fig. 6).
Cooperation between ␤ 1 Integrin and VEGFR-3 Modulates Cell Migration-Our data suggested cooperation between ␤ 1 integrin and VEGFR-3 signaling. We investigated the possible functional significance of this observation in regulating cell migration. In a Transwell migration assay, 293/VEGFR-3 cells or parent 293 cells were loaded in the migration insert filters coated with collagen or BSA. The transmigration was examined as described under "Experimental Procedures." As shown in Fig. 7, coating with collagen increased the number of migrating 293/VEGFR-3 and 293 cells compared with coating with BSA. However, the migration of 293/VEGFR-3 cells was higher (2.5-fold) than that of the 293 cells. To test the specific role of VEGFR-3 or ␤ 1 integrin in collagen-induced cell migration, cells were pretreated with anti-VEGFR-3 or anti-␤ 1 integrin antibody before performing the migration assays. We observed that the migration of 293/VEGFR-3 cells through collagen-coated inserts was reduced by either blocking antibody treatment. To explore further the possibility that activation of VEGFR-3 by ␤ 1 integrin stimulation contributed to the enhanced cell migration on collagen, we generated a VEGFR-3 mutant in which Gly-857 was replaced by Arg using site-directed mutagenesis. This missense mutation was initially identified as one of the mutations found in patients with hereditary lymphedema (42)(43)(44). These mutant VEGFR-3 proteins lack tyrosine kinase activity and inhibit autophosphorylation of the wild type receptor (44). We observed that this mutation resulted in the loss of tyrosine phosphorylation of this receptor in response to either collagen or VEGF-D stimulation, when transiently expressed in 293 cells (Fig. 8A). We then examined the migration through collagen-coated filters of 293 cells transfected with wild type or mutant VEGFR-3 or control vector. As shown in Fig. 8B, transfection with wild type but not the mutant form of VEGFR-3 increased the number of migrated cells in comparison with the control vector-transfected cells. These results suggested that collagen-enhanced migration was mediated via the activation of both VEGFR-3 and ␤ 1 integrin. We next examined the migration of 293/VEGFR-3 cells in response to VEGF-D, the ligand for VEGFR-3. We showed that VEGF-D, in a concentration-dependent manner, induced cell migration in collagen-coated but not in BSA-or poly-L-lysinecoated migration inserts (Fig. 9A). Treatment with blocking antibody against ␤ 1 integrin as described above diminished the VEGF-D-increased migration on collagen (Fig. 9B). We also demonstrated that transfection with the mutant form of VEGFR-3 inhibited the 293/VEGFR-3 migration through collagen-coated filters (Fig. 9C).
We further investigated whether the observed cooperative effects of collagen and VEGF-D were also involved in modulating the migration of primary endothelial cells. Similar to that demonstrated in 293/VEGFR-3 cells, VEGF-D-induced DMEC migration was dependent on the presence of collagen. VEGF-D alone failed to induce cell migration through BSA-coated filters. The chemotactic effect of VEGF-D was partially blocked by anti-VEGFR-3 antibody and was completely blocked by anti-␤ 1 integrin antibody (Fig. 10). As observed previously in 293/VEGFR-3 cells, VEGF-D at a concentration range of 1-100 ng/ml did not increase the migration of DMEC through poly-Llysine-coated filters (data not shown), even though they were able to adhere to this substrate. Thus, optimal migration of DMEC appears to involve the cooperative action of both the VEGFR-3 and ␤ 1 integrin signaling pathways.

DISCUSSION
The signals that regulate endothelial migration are not well understood. Such migration is an important part of angiogenesis and thus is critical to physiological development as well as to tumor growth and spread (23,45). Prior data on fibroblasts, epithelial cells, or endothelial cells indicate that cooperativity between integrin receptors and RTKs, specifically PDGF and EGF, can have important functional outcomes in modulating cell migration (13,14,46,47). We have focused on endothelial cells and how the ECM proteins, collagen and FN, by stimulating ␤ 1 integrins, can modulate the effects of VEGFR-3. In model 293 cells transfected with VEGFR-3, as well as in primary microvascular endothelial cells derived from the dermis, we demonstrated cross-talk between this angiogenic receptor and the ␤ 1 integrin. This apparent cooperativity correlated with an important functional outcome, that of cell migration.
ECM proteins can act synergistically with growth factors in modulating cell proliferation, differentiation, survival, and migration. Analysis of signaling events triggered by engagement of integrins by ECM proteins has demonstrated a remarkable similarity to those activated by growth factor receptors (6), indicating that there exists a cross-talk mechanism between integrins and growth factor receptors. Although the mechanism of such interactions is not clear, it can be explained at least in part by the findings in some recent studies. These studies have provided evidence that ECM proteins can activate certain tyrosine kinase receptors such as the PDGF receptor-␤, EGF receptor, fibroblast growth factor receptor, insulin receptor, and the RON receptor in the absence of their ligands (9 -11, 36, 48, 49). In the present work, we demonstrated a similar induction of VEGFR-3 tyrosine phosphorylation by the ECM protein, collagen or FN. Our data provide another example of the transactivation of RTKs by ECM proteins, suggesting that the ligand-independent activation of growth factor receptors can be a broadly used mechanism in cell adhesion-mediated signaling and in the synergistic effects of the ECM and growth factors.
ECM-triggered signaling is generally considered to be mediated by integrins. However, it has been documented recently (50 -52) that the ECM protein, collagen, can directly bind to the discoidin domain receptor (DDR) tyrosine kinases, DDR1 and 293 or 293/VEGFR-3 cells were preincubated without or with anti-VEGFR-3 blocking antibody or control normal goat IgG (NGIgG) or with blocking antibody against ␤ 1 integrin or its control normal mouse IgG (NMIgG). The variously treated cells were allowed to migrate through collagen-or BSA-coated filters, and the migrated cells were stained and counted as described under "Experimental Procedures." Each determination represents the average of three individual inserts, and error bars represent the mean Ϯ S.D. DDR2, and activate these receptors independently of integrins. In the present study, we showed that both the collagen-and FN-induced tyrosine phosphorylation of VEGFR-3 could be blocked by anti-␤ 1 integrin antibody (Fig. 3), indicating that this transactivation was mediated by the integrin rather than through direct binding of VEGFR-3 by the ECM proteins. Although ECM proteins may bind to different ␣-chains of the ␤ 1 integrin complexes, both of the proteins tested in our study, collagen and FN, were found to be able to induce the tyrosine phosphorylation of VEGFR-3. These results indicate that the various integrin ␣-subunits may not confer specificity in the ECM protein-induced tyrosine phosphorylation of VEGFR-3. In fact, ligation of ␤ 1 integrin by cross-linking with anti-␤-chain specific antibody was shown to be sufficient to activate VEGFR-3 (Fig. 4).
The molecular mechanisms underlying VEGFR-3 phosphorylation by ␤ 1 integrin remain to be defined. Previous studies (12)(13)(14)(15)(16)(17) on the transactivation of other RTKs have demonstrated that engagement of integrins by ECM proteins can induce the physical association of integrins and RTKs. Formation of macromolecular complexes may result in the oligomerization, transphosphorylation, and activation of RTKs. Our results showed that collagen or fibronectin stimulation induced an association between ␤ 1 integrin and VEGFR-3 in 293/ VEGFR-3 cells and DMEC (Fig. 5). However, we were able to visualize this association consistently only when low stringency detergents such as CHAPS were used in the lysis buffer, indicating that this association is weak. The weak association suggests that the binding between these two molecules may be short lived or have low affinity and thus could be easily dissociated during isolation and incubation. Interestingly, VEGF receptor-2, an orphan receptor for VEGFR-3, has been reported to associate with ␣ v ␤ 3 , but not ␤ 1 integrin, in endothelial cells upon VEGF stimulation (15,16). Whether VEGFR-3 is also able to associate with ␣ v ␤ 3 needs to be examined. Considering the co-expression of both receptors in microvascular endothelial cells, they may play different roles through their interaction with different classes of integrins. This hypothesis will be addressed in the future. Intracellular signaling triggered by ECM proteins may also be required for the transactivation of RTKs by integrins. It has been reported that ECM protein-induced activation of the RON receptor is dependent on the kinase activity of both c-Src and RON itself (49). Other studies have demonstrated that integrin ligation can lead to the recruitment of RTKs to focal adhesions (53)(54)(55) and that actin cytoskeleton integrity appears to be required for the integrin-induced activation of RTKs (56,57). In support of this notion, we demonstrated that disruption of the actin cytoskeleton with cytochalasin D inhibited the collagen- (Fig. 6) or fibronectin (data not shown)-induced tyrosine phosphorylation of VEGFR-3. The cytoplasmic tails of most integrins are generally short and lack intrinsic enzymatic activity. Therefore, integrins transduce signals by associating with adaptor proteins that connect the integrins to the cytoskeleton and cytoplasmic kinases, leading to the formation of focal adhesion complexes. Upon integrin activation, cytoskeletal proteins, such as paxillin, talin, vinculin, and ␣-actinin and signaling molecules such as the focal adhesion kinase, related adhesion focal tyrosine kinase (Pyk2), c-Src, phospholipase C, protein kinase C, phosphatidylinositol 3-kinase, small GTPbinding proteins, GRB2, etc., are recruited to the focal adhesion complexes (6,39,58). Whether these proteins participate in the transactivation of VEGFR-3 by ␤ 1 integrin remains to be investigated.
To assess the functional significance of the observed transactivation of VEGFR-3 by ECM proteins, we analyzed the effects of the ECM protein, collagen, and VEGF-D, the ligand for VEGFR-3, in modulating cell migration. We observed that there was no cell migration when cells were plated on BSA-or poly-L-lysine-coated filters (Figs. 7, 9, and 10). However, when cells were plated on collagen-coated filters, cells were able to migrate. Compared with the 293 cells, 293/VEGFR-3 cell migration was greater, and this increased migration could be blocked by either anti-VEGFR-3 or anti-␤ 1 integrin blocking antibody. Furthermore, we found that introducing a point mutation (Gly-857 3 Arg) at the first kinase domain of VEGFR-3 interrupted its tyrosine phosphorylation in response to either collagen or VEGF-D stimulation (Fig. 8A). When this mutant was transfected into 293 cells, cell migration on collagen was lower as compared with that transfected with the wild type receptor (Fig. 8B). This indicated that activation of VEGFR-3 by the ECM contributed to the enhanced cell migration on collagen. VEGF-D failed to induce cell migration on BSA-or poly-L-lysine-coated filters in either 293/VEGFR-3 cells or DMEC. However, VEGF-D concentration gradients induced significant cell migration when the filters were coated with collagen ( Figs. 9 and 10). The VEGF-D-induced cell migration on collagen-coated filters was inhibited by treatment with anti-VEGFR-3 antibody or by transfection with a mutant form of the receptor (Fig. 9). In addition to collagen, other ␤ 1 integrin-binding matrix proteins, such as FN, vitronectin, and laminin, were also found to support VEGF-D-mediated cell migration (data not shown). Poly-L-lysine is a non-integrin substrate that is known to mediate cell adhesion in an integrin-independent manner. We demonstrated that 293 cells, 293/VEGFR-3, or DMEC were able to adhere on this substrate. However, plating cells on poly-L-lysine did not enhance the tyrosine phosphorylation of VEGFR-3, and no significant change in cell migration FIG. 9. Migration of 293/VEGFR-3 cells on collagen-or BSA-coated filters in response to various concentrations of VEGF-D. A, 293/VEGFR-3 cells (2 ϫ 10 5 cells/0.1 ml migration medium) were loaded onto the upper chambers of the inserts. The undersides of the filters in these inserts were coated with BSA, poly-L-lysine, or collagen. Cells were allowed to migrate, in the presence of various concentrations of VEGF-D, into the lower chamber at 37°C in 5% CO 2 for 3.5 h. B, 293/VEGFR-3 cells were first incubated with blocking antibody against ␤ 1 integrin or with control normal mouse IgG. Cell migration was detected as described above. C, 293/VEGFR-3 cells were transfected with an expression plasmid containing a mutant form of VEGFR-3 or with control vector (PRK5). Forty eight h following transfection, cell migration through collagen-coated filters in response to VEGF-D (100 ng/ml) was examined. The migrated cells on the undersurface were stained and counted as described under "Experimental Procedures." Each determination represents the average of three individual inserts, and error bars represent the mean Ϯ S.D. was observed in response to VEGF-D concentration gradients when the filters were coated with poly-L-lysine (Fig. 9). 3 These studies suggest that signals from both integrins and RTKs are required for optimal induction of dermal microvascular endothelial cell migration.
Taken together, our findings suggest that microvascular endothelial cells, which express VEGFR-3 and respond to VEGF-D, can be modulated in their biology by ECM proteins like collagen and fibronectin. Further study will elucidate the downstream mediators of the ␤ 1 integrin and VEGFR-3 receptor with regard to their signaling to the cytoskeleton, because the integrity of the actin cytoskeleton was required for the cooperative effect. Similarly, such cooperative interactions are likely to signal to the cell nucleus and initiate a program of gene expression relevant to cell migration. Improved understanding of such events in the endothelium should provide insights into both physiological angiogenesis as well as tumordriven angiogenesis.