Vascular endothelial growth factor receptor-1 (VEGFR-1) is a tyrosine kinase receptor for several growth factors of the VEGF family. Endothelial cells express a membrane-spanning form of VEGFR-1 and secrete a soluble variant of the receptor comprising only the extracellular region. The role of this variant has not yet been completely defined. In this study, we report that the secreted VEGFR-1 is present within the extracellular matrix deposited by endothelial cells in culture, suggesting a possible involvement in endothelial cell adhesion and migration. In adhesion assays, VEGFR-1 extracellular region specifically promoted endothelial cell attachment. VEGFR-1-mediated cell adhesion was divalent cation-dependent, and inhibited by antibodies directed against the α5β1 integrin. Moreover, VEGFR-1 promoted endothelial cell migration, and this effect was inhibited by anti-α5β1 antibodies. Direct binding of VEGFR-1 to theα 5β1 integrin was also detected. Finally, binding to VEGFR-1 initiated endothelial cell spreading. Altogether these results indicate that the soluble VEGFR-1 secreted by endothelial cells becomes a matrix-associated protein that is able to interact with the α5β1 integrin, suggesting a new role of VEGFR-1 in angiogenesis, in addition to growth factor binding.
The assembly of endothelial cells (EC) into vascular structures requires the establishment of cell-cell interactions as well as the activation of cell membrane transducing receptors by soluble ligands and by components of the extracellular matrix.
Vascular endothelial growth factor receptor-1 (VEGFR-1/Flt-1) is one of the tyrosine kinase transmembrane receptors for VEGF (De Vries et al., 1992), and the only known high affinity receptor for VEGF-B and placenta growth factor (PlGF) (Korpelainen and Alitalo, 1998). Despite being closely related to the type III tyrosine kinase-Fms/Kit/platelet derived growth factor (PDGF), VEGFR-1 is classified into a distinct class of receptors composed of seven immunoglobulin (Ig)-like domains in the extracellular region (Shibuya et al., 1990). The VEGFR-1 II Ig-like domain comprises the regions that are principally involved in VEGF and PlGF binding and in the activation of the signal transduction cascade (Davis-Smyth et al., 1996).
VEGFR-1 is mainly expressed on endothelial cells, but other cell types, such as monocytes, macrophages (Sawano et al., 2001) and tumour cells (Bellamy et al., 1999; Lacal et al., 2000; Masood et al., 2001), have been shown to express this receptor on the cell surface. In endothelial cells, VEGF-mediated activation of VEGFR-1 has been reported not to induce cell proliferation efficiently (Landgren et al., 1998; Rahimi et al., 2000; Seetharam et al., 1995), whereas it plays a prominent role in cell migration (Barleon et al., 1996; Clauss et al., 1996).
A differently spliced form of VEGFR-1 mRNA encoding a soluble receptor variant (sVEGFR-1/sFlt-1) has been isolated in cultured endothelial cells (Kendall and Thomas, 1993) and different cell lines (Inoue et al., 2000). sVEGFR-1 is thought to be a naturally produced VEGF antagonist that inhibits the mitogenic effects of this cytokine by functioning as a dominant-negative trapping protein (Inoue et al., 2000) or by forming non-signalling complexes with VEGFR-2 (Kendall et al., 1996), but the physiological role of this soluble variant has not been fully characterised yet.
Gene knockout studies have demonstrated that VEGFR-1 is essential for development and differentiation of the embryonic vasculature (Fong et al., 1995). Mouse embryos homozygous for a targeted mutation in the VEGFR-1 locus die in utero at day 8.5 to 9.0 (Fong et al., 1995). In these animals, EC develop in both embryonic and extraembryonic sites, but fail to organise in normal vascular channels, suggesting that VEGFR-1 is primarily involved in vascular morphogenesis. Moreover, embryos lacking VEGFR-1 display an increased outgrowth of EC and hemangioblast commitment (Fong et al., 1999). The excess of EC in these animals inhibits the proper organisation of vascular structures. In contrast, mice carrying a homozygous deletion limited to the intracellular kinase domain of VEGFR-1 show a correct development of blood vessels (Hiratsuka et al., 1998). This selective knockout is still able to produce the soluble form of the receptor and displays a truncated form comprising only the receptor extracellular domain on the cell surface. The phenotype of these animals suggests that VEGFR-1 has a role that is independent of its tyrosine kinase activity.
In this study, we localised the soluble VEGFR-1 within the extracellular matrix deposited by EC in culture, and demonstrated that it is able to support EC adhesion and migration through the interaction with the α5β1 integrin.
Materials and Methods
Antibodies and reagents
Two polyclonal sera were generated, one against a peptide mapping in the second Ig-like domain of the receptor extracellular region (anti-VEGFR-1), and the other against a peptide mapping in the unique C terminus of the soluble form of VEGFR-1 (anti-sVEGFR-1). Rabbit immunisation was carried out by PRIMM, Milan, Italy.
The anti-VEGFR-1 polyclonal antibody was tested for VEGFR-1 specificity in ELISA and western blotting. No cross reaction was observed against either VEGFR-2/Fc or VEGFR-3/Fc chimeras.
Rabbit polyclonal antibody (pAb) H-225, C-17 (anti-VEGFR-1), and C-20 (anti-VEGFR-2) were from Santa Cruz Biotechnology Inc. (Santa Cruz, CA), while the goat pAb recognising the N terminus of VEGFR-1 (AF321) was from the R&D Systems (Minneapolis, MN). The monoclonal antibody (mAb) anti-human fibronectin FN 15 was from ICN Biomedicals Inc. (Costa Mesa, CA), whereas the mAb anti-fibronectin cell attachment fragment 3E3 was from Chemicon (Temecula, CA). Function blocking antibodies against various integrins were as follows: mouse mAb JBS5 (anti-α5β1), mouse mAb Jβ1a (anti-β1), mouse mAb LM609 (anti-αvβ3), and goat pAb anti-α5β1 (all purchased from Chemicon). The rat mAb GoH3 (anti-α6) was kindly provided by Dr A. Sonnenberg (The Netherlands Cancer Institute, Amsterdam, The Netherlands) and the mAb NKI-SAM-1 (anti-α5) was from Immunotech (Marseille, France). Recombinant human VEGFR-1/Fc, PDGFRβ/Fc, VEGFR-3/Fc, VEGFR-2/Fc chimeras, and recombinant human VEGF and placenta growth factor were purchased from R&D Systems. Human IgG1 were from Calbiochem (La Jolla, CA). Vitronectin, fibronectin, cycloheximide and monensin were obtained from Sigma-Aldrich (St Louis, MO). Trypsin was from ICN Biomedicals Inc. The GRGDTP and GRGESP peptides were from Invitrogen-Life Technologies (Paisley, UK), and the purified α5β1 integrin, octyl-β-D-glucopyranoside formulation, was from Chemicon.
Human umbilical vein endothelial cells (HUVEC) were isolated from freshly delivered umbilical cords, as previously described (Gimbrone, 1976), and cultured in Endothelial Cell Growth Medium-2 Kit from Clonetics (BioWhittaker Inc, Walkersville, MD). The human microvascular endothelial cell line (HMEC-1) was a generous gift of Dr F. J. Candal (Center of Disease Control and Prevention, Atlanta, GA) (Ades et al., 1992) and was cultured in MCDB 131 medium (Sigma-Aldrich) supplemented with 10% foetal bovine serum (Hyclone Laboratories, Logan, UT) plus hydrocortisone (Sigma-Aldrich) and epidermal growth factor (Austral Biological, San Ramon, CA). Normal human fibroblasts were isolated from human skin biopsies and cultured as previously described (Wirtz et al., 1987).
Preparation and analysis of the extracellular matrix (ECM)
The analysis of the ECM components was carried out according to previously described protocols (Delwel et al., 1993; Gagnoux-Palacios et al., 2001; Owensby et al., 1989), with minor modifications. Briefly, HUVEC were grown to confluence on 96-multiwell culture plates. Cell monolayer was then incubated overnight at 4°C with PBS/20 mM EDTA and washed with PBS/1% Triton X-100. This treatment leaves the ECM intact, free of cell debris and firmly attached to the well surface.
For the detection of the soluble VEGFR-1, the matrix was blocked with 1% BSA/PBS for 3 hours and then incubated for 2 hours with 10 μg/ml mAb against VEGFR-1 (H-225) or fibronectin (FN-15). After five washes with PBS/0.1% Tween 20, plates were incubated with a secondary biotinylated antibody (Vector Laboratories Inc., Burlingame, CA) for 2 hours at room temperature. Streptavidin-alkaline phosphatase conjugate and the appropriated substrate (4-nitrophenylphosphat, Roche Diagnostic, Basel, Switzerland) were used for detection. Absorbance was determined at 405 nm using a Microplate reader 3550-UV (Bio-Rad, Hercules, CA).
Immunoblot analysis of the ECM was also performed. Cell, ECM or total (containing both ECM and cells) samples were collected in the same final volume of SDS sample buffer (50 mM Tris-HCl pH 7.5, 100 mM dithiothreitol, 2% SDS, 0.1% bromophenol blue, 10% glycerol) and boiled for 5 minutes. Polypeptides were separated on 6% SDS-polyacrylamide gels and then transferred to supported nitrocellulose membranes (Hybond-C; Amersham Biosciences, Buckinghamshire, UK), using a Transphor TE 50X unit (Hoefer Scientific Instruments, San Francisco, CA). Membranes were blocked in 2% blocking solution (Roche Diagnostics)/TBS pH 7.5 for 1 hour, and incubated overnight at 4°C either with the goat pAb AF321 recognising the N terminus of VEGFR-1 (0.5 μg/ml), or with the rabbit pAbs recognising the C terminus of VEGFR-1 (C-17) and VEGFR-2 (C-20), diluted 1:200. After two washes with 0.1% Tween 20/TBS pH 7.5, membranes were incubated with the appropriated secondary antibody, diluted 1:5000 in 1% blocking solution/TBS, for 1 hour at room temperature and then washed four times with 0.1% Tween 20/TBS. Detection was carried out using the ECLTM western blotting detection reagents from Amersham Biosciences.
For immunofluorescence experiments, EC were seeded on untreated glass coverslips and left to reach confluence. Matrix was prepared as described above, and fixed with 4% paraformaldehyde in PBS for 5 minutes at room temperature. The anti-sVEGFR-1 polyclonal antibody or the anti-FN (FN-15) monoclonal antibody (1:100 dilution) were layered on the fixed matrix for 1 hour at 37°C. After several washes, secondary anti-rabbit or anti-mouse biotinylated antibodies (Vector Laboratories Inc.) were added, and then a streptavidin-FITC conjugate (Amersham Biosciences) was used. Matrix, on coverslips, were then mounted on slides and observed with a fluorescence microscope (Zeiss-Axiophot, Oberkochen, Germany). To test the specificity of the signal obtained with the anti-sVEGFR-1 pAb, the antibody was incubated with the corresponding blocking peptide and then used in the assay as described above.
Cell adhesion assay
Solid support was prepared by incubating immunological 96-multiwell plates with various concentrations of the receptor/Fc chimeras, fibronectin or vitronectin solubilised in PBS. After 2 hours, the coating solution was removed, and the well surface was blocked with 3% BSA in PBS for 18 hours before plating EC in serum-free medium at 3104 cells per well. After incubation at 37°C for the indicated time, the wells were washed with PBS, attached cells were fixed with 3% formaldehyde and stained with 0.5% crystal violet. The attachment efficiency was determined by quantitative dye extraction and spectrophotometric measurement of the absorbance at 540 nm using a Microplate reader 3550-UV (Bio-Rad). Results represent the mean of triplicate samples ± s.d. All experiments were repeated at least three times. In competition experiments, cells were preincubated with 10 μg/ml antibodies, 20 μg/ml VEGFR-1/Fc chimera, 0.4 mM RGD or RGE peptides for 15 minutes at room temperature before plating and were then left to adhere for 30 minutes at 37°C in the presence of the compounds. VEGF or placenta growth factor (20 μg/ml), EGTA (10 mM), EDTA (10 mM), MgSO4 (5 mM), MnCl2 (5 mM), CaCl2 (5 mM) or trypsin (0,05% w/v) were added at the time of cell plating.
Solid-phase binding assay
The solid-phase binding assay was performed as previously described (Mould et al., 1998). Immunological 96-multiwell plates were coated overnight at 4°C with 1μ g/ml purified α5β1 integrin (Chemicon). Plates were then blocked for 2 hours at room temperature with 1% BSA/PBS. VEGFR-1/Fc was diluted at the indicated concentrations in 25 mM Tris-HCl, 150 mM NaCl, 1 mM MgCl2 (buffer A), and overlaid on plates for 1 hour at 37°C. After three washes with buffer A, plates were incubated with a 1:10000 dilution of an anti-human IgG (Fc specific) alkaline phosphatase-conjugated antibody (Sigma-Aldrich) for 1 hour at 37°C. The appropriate substrate (4-nitrophenylphosphat, Roche Diagnostic) was then used for detection. Absorbance was determined at 405 nm using a Microplate reader 3550-UV (Bio-Rad). Where indicated, 1 mM EDTA or 10 μg/ml anti-α5β1 blocking antibody (mAb JBS5) were added during the binding assay. When EDTA was used, buffer A was prepared without MgCl2. VEGFR-2/Fc (20μ g/ml) was also used as a negative control.
The migration assays were performed in Boyden chambers, as previously described (Mensing et al., 1984). Polycarbonate filters (8 μm pore diameter, Nuclepore, Whatman Incorporated, Clifton, NJ) were coated with 5 μg/ml gelatine solution. The stimuli for chemotaxis were added to the lower chamber at the indicated concentrations and HUVEC (1.5×105) were loaded into the upper chamber. Chemokinesis was tested by including the VEGFR-1/Fc chimera only in the upper chamber, together with the cells, and, in selected experiments, the chimera was included in both the upper and the lower chambers. For haptotactic assays, the under surface of membrane filters, precoated in the upper surface with gelatine, was coated with 10 μg/ml VEGFR-1/Fc or vitronectin, as described (Nasreen et al., 2000). Background migration was measured by using filters coated with 10 μg/ml BSA. Migration medium (1 μg/ml heparin/0.1% BSA in EBM-2) was always used to prepare the solutions with the stimuli and as a negative control. After 2 hours incubation (5% CO2, 37°C), the filter was removed and cells were fixed in 3% paraformaldehyde in PBS and stained in 0.5% crystal violet. Cells from the upper surface were removed by wiping with a cotton swab. The chemotactic and chemokinetic responses were determined by counting the migrating cells attached to the lower surface of the filter in 12 randomly selected microscopic fields (×200 magnification) per experimental condition. Blocking of the α5β1 integrin was performed by preincubating the cells for 45 minutes with antibodies specific for this integrin (mAb JBS5), or with unrelated antibodies as controls (mAb GoH3), at room temperature and under constant shaking. Afterwards, cells were loaded into the Boyden chambers in the presence of the antibodies, and the migration assay was carried out as described above.
Cell spreading assays
Polystyrene Petri dishes, 100 mm diameter, were coated with 10 μg/ml VEGFR-1/Fc chimera or 10 μg/ml fibronectin, washed and blocked, as described above. EC (2.5×106/dish) were plated and left to adhere for 1 to 12 hours. Cell spreading was monitored using an inverted microscope. To show the F-actin distribution and microfilament organisation, glass coverslips were coated with 10 μg/ml VEGFR-1/Fc chimera or 10μ g/ml fibronectin for 2 hours at room temperature, then saturated by further incubation with 3% BSA/PBS. Coverslips were washed with PBS and cells seeded at 2.5× 104cells/cm2. Cells were fixed at different times with 4% paraformaldehyde for 5 minutes, washed twice with PBS and stained with fluorescein-labelled phalloidin (Sigma-Aldrich) for 30 minutes at room temperature. To detect fibronectin fibrils deposited by EC, coverslips were coated with 10 μg/ml VEGFR-1/Fc or 10 μg/ml vitronectin and treated as described above. Three hours after cell seeding, matrix was prepared as described above and then immunostained with an anti-fibronectin antibody (FN-15). After several washes, secondary anti-mouse biotinylated antibody was added, and then the streptavidin-FITC conjugate was used. Coverslips were then mounted on slides, and the preparations observed with a fluorescence microscope (Zeiss-Axiophot). In selected experiments, cycloheximide was added at 100 μg/ml before plating and monensin was added at 1 μM, 18 hours before plating. Experiments with RGD or RGE peptides (0.4 mM) were carried out by adding them to confluent EC monolayers and evaluating cell detachment after 1 hour incubation at 37°C.
The sequence of the soluble VEGFR-1 was compared to those of the proteins of known structure from the PDB (Berman et al., 2000) and to those in non-redundant databases of protein sequences [http://www.ncbi.nlm.nih.gov/ and (Holm and Sander, 1998)] by using the BLAST (Altschul et al., 1997) and SUPERFAMILY (Gough et al., 2001) servers available through the World Wide Web. Sequence homologues collected with BLAST were aligned using CLUSTALW (Thompson et al., 1994), whereas SUPERFAMILY provides multiple sequence alignments. Domain analysis was performed using the Pfam (Bateman et al., 2002) and SMART (Schultz et al., 1998) servers available via the World Wide Web.
VEGFR-1 is a component of the endothelial cell extracellular matrix
Previously reported data have shown that the soluble variant of VEGFR-1 is secreted by endothelial cells in culture (Kendall and Thomas, 1993). From a structural point of view, this receptor form is predicted to be formed by six domains belonging to the immunoglobulin superfamily, which produces a fold very commonly found in proteins involved in ECM-cell adhesion (Clothia and Jones, 1997; Hohenester and Engel, 2002), indicating that the structure of soluble VEGFR-1 is compatible with a role as an ECM protein. We therefore investigated whether the soluble VEGFR-1 could become a component of the extracellular matrix (ECM).
EC and human fibroblasts were cultured for 72 hours and then detached from the plates using a method that leaves the extracellular matrix intact (Delwel et al., 1993; Gagnoux-Palacios et al., 2001; Owensby et al., 1989). Human fibroblasts do not express either the membrane-bound or the soluble form of VEGFR-1, and were thus used as a negative control. The presence of VEGFR-1 within the matrix was detected by using an anti-VEGFR-1 antibody (H-225) against the extracellular region of the receptor. VEGFR-1 could be detected in the matrix produced by both HMEC-1 and HUVEC whereas no signal was obtained in the matrix deposited by human fibroblasts (Fig. 1A). The EC matrix was also analysed using a rabbit polyclonal antibody raised against a peptide mapping at the C terminus of soluble VEGFR-1 (anti-sVEGFR-1). The C-terminal region of the soluble receptor differs from that of the transmembrane VEGFR-1 since it is encoded by an alternatively-spliced RNA. As shown in Fig. 1B, soluble VEGFR-1 was detected in the EC matrix. As a positive control, the matrix was labelled with an anti-fibronectin antibody (Fig. 1B). Consistent with the ELISA and immunofluorescence findings, a western blotting analysis showed that the VEGFR-1 variant present within the ECM has a molecular weight that corresponds to the soluble form and is specifically recognised by an antibody directed towards the extracellular region of the receptor (Fig. 1C). Moreover, this polypeptide was not revealed by using an antibody directed towards the intracellular domain of the receptor (Fig. 1D). Both antibodies detected the presence of the transmembrane variant of VEGFR-1 in the cell extract, but not in the ECM fraction (Fig. 1C,D). Transmembrane VEGFR-1 appeared as a doublet, probably because of different glycosylation forms (Seetharam et al., 1995). VEGFR-2, analysed as a negative control, was only detected in the cell extract sample (data not shown). As a positive control and to further validate the ECM fraction obtained, the presence of fibronectin within this fraction was confirmed by immunoblotting (Fig. 1C and D). Altogether, these data demonstrate the presence of polypeptides from the extracellular domain of VEGFR-1 in the matrix produced by human endothelial cells.
VEGFR-1 is directly involved in human EC adhesion
The possible role of the VEGFR-1 extracellular region as a direct mediator of EC/matrix interactions was then tested using an in vitro cell adhesion assay. Ninety-six-multiwell plates were coated with increasing amounts of a chimeric protein comprising the extracellular region of VEGFR-1 fused to the human immunoglobulin Fc domain (VEGFR-1/Fc) or with fibronectin as a positive control. Wells were also coated with BSA as a negative control. The relative number of adherent HMEC-1 or HUVEC was quantified 1 hour after plating. As expected, EC effectively adhered to fibronectin, but also to VEGFR-1/Fc-coated surfaces at similar levels (Fig. 2A). A dose-dependent increase in cell adhesion was observed for both substrates as early as 30 minutes after plating. Comparable data were obtained using the HMEC-1 cell line or HUVEC primary cultures. To exclude non-specific interactions involving the Fc domain present in the fusion protein, or contaminants in the commercial preparation, PDGFRβ/Fc and VEGFR-2/Fc chimeric proteins or human IgG1 were used as controls. None of these substrates supported EC adhesion (Fig. 2B). Addition of VEGFR-1/Fc chimera to the cell suspensions abrogated cell attachment (Fig. 2C). Adhesion of HMEC-1 and HUVEC to VEGFR-1/Fc-coated surfaces was also specifically inhibited by a rabbit polyclonal anti-VEGFR-1 antibody, but not by the preimmune serum (Fig. 2C). The same antibody did not affect cell attachment on fibronectin-coated surfaces.
Integrin subunits mediate cell attachment to VEGFR-1
The capability of VEGFR-1/Fc to support EC adhesion was almost completely abolished by pretreatment of the cells with trypsin, indicating that the cell interaction with the matrix-associated VEGFR-1 is mediated by a protein (Fig. 3A). To assess whether divalent cations affect EC/VEGFR-1 interactions, HMEC-1 and HUVEC adhesion on VEGFR-1/Fc was assayed in the presence of different cations at a concentration of 5 mM. VEGFR-1/Fc-mediated EC attachment was supported by the addition of Mg2+ to the medium. Mn2+ further enhanced EC attachment, while Ca2+ inhibited it (Fig. 3A). Addition of 10 mM EGTA or EDTA reduced adhesion levels to those obtained in the presence of Ca2+. These data indicate that soluble VEGFR-1 binds to a divalent cation-dependent protein on the cell surface. Since cells interact with the extracellular matrix in a divalent cation-dependent way mainly through receptors of the integrin family (Mould et al., 1995; Smith et al., 1994), we investigated whether integrin subunits could be involved in the binding of the EC to VEGFR-1. The effect of anti-integrin antibodies on this interaction was therefore analysed. Considering the critical role played by theα 5β1 and αvβ3 integrins in angiogenesis both in vitro and in vivo (Brooks et al., 1994; Kim et al., 2000), we used blocking antibodies directed against these two integrins. Adherence of EC to VEGFR-1/Fc was greatly reduced in the presence of blocking antibodies directed either towards the α5β1 integrin or the β1 or α5 integrin subunits. An anti-αvβ3 antibody did not affect EC adhesion on VEGFR-1/Fc (Fig. 3B). As expected, EC binding to fibronectin was decreased by the anti-α5β1 blocking antibody and, to a lesser extent, by the anti-αvβ3 antibody (Fig. 3B). Simultaneous incubation of EC with the anti-α5β1 and anti-αvβ3 antibodies resulted in a greater inhibitory effect on adhesion to fibronectin, whereas no further inhibition could be observed on VEGFR-1/Fc (Fig. 3B). Indeed, the combination of the anti-α5β1 and anti-αvβ3 antibodies appeared slightly less effective in inhibiting cell adhesion on VEGFR-1/Fc than the anti-α5β1 antibody alone. This attenuated blocking might be due to steric hindrance between the two antibodies, as previously reported (Leong et al., 2002). As expected, the anti-αvβ3 antibody alone could almost completely block EC adhesion on vitronectin (Fig. 3B).
VEGFR-1 induces EC migration through the interaction with theα 5β1 integrin
The role of soluble VEGFR-1 in supporting EC adhesion and its localisation in the ECM suggested a possible involvement of this form in inducing EC migration, similarly to that already shown for other components of the ECM (Clark et al., 1988; Mensing et al., 1984). We therefore evaluated the ability of the VEGFR-1/Fc chimera to induce EC migration. As shown in Fig. 4A, VEGFR-1/Fc-stimulated EC chemotaxis, and this effect was dose-dependent and detectable when the protein was present at 1 μg/ml, reaching maximal activity at 5 μg/ml. At this concentration, VEGFR-1/Fc showed an effect comparable to that of the epidermal growth factor (EGF), a known chemoattractant for EC (Fig. 4A) (Chen et al., 1993). To determine whether VEGFR-1/Fc also stimulated chemokinesis (random cell movement), VEGFR-1/Fc was placed in the lower and/or upper chambers of the Boyden chamber as indicated in Fig. 4A. The presence of VEGFR-1/Fc in the upper chamber, together with the cells, induced a slight increase in cell motility, that was not concentration dependent (Fig. 4A). Modest chemokinetic activity was observed at lower VEGFR-1/Fc concentrations (1 μg/ml). In this condition, the chemokinetic response was slightly more relevant than the chemotactic one. At VEGFR-1 concentrations in which chemotaxis was substantial (5 μg/ml), the simultaneous presence of the chimera in the upper chamber abrogated the capability of EC to migrate through the filter (Fig. 4A). This result shows that EC are sensitive to a concentration gradient of VEGFR-1/Fc between the two chambers, and indicates that the chemotactic response is preponderant with respect to chemokinesis. Since ECM components usually stimulate cell motility through haptotactic mechanisms, the capability of VEGFR-1/Fc molecules immobilised on polycarbonate filters to induce haptotaxis of EC was analysed (Fig. 4C). An increased migration of EC to the under surface of the filters was observed (70% increase with respect to the background controls, Fig. 4C). Control filters coated with vitronectin showed a 100% increase in EC migration with respect to the background control (Fig. 4C). The role of the VEGFR-1/α5β1 interaction in both chemotaxis and haptotaxis was investigated by incubation of the EC with anti-α5β1 integrin blocking antibodies before the assay. An inhibition of almost 70% of the VEGFR-1/Fc-induced chemotaxis and 100% of the VEGFR-1/Fc-induced haptotaxis was observed when compared to EC pretreated with an unrelated antibody (Fig. 4B,D). In contrast, neither EGF-induced chemotaxis nor vitronectin-induced haptotaxis were significantly affected by the anti-α5β1 antibodies (Fig. 4B,D).
Characterisation of VEGFR-1 binding to the α5β1 integrin
To confirm the capability of VEGFR-1 to interact with α5β1, we analysed direct VEGFR-1/Fc-integrin interaction in vitro by using a solid-phase binding assay (Rehn et al., 2001). VEGFR-1/Fc specifically bound to immobilisedα 5β1 integrin, whereas direct binding of related proteins, such as VEGFR-2/Fc, was not detected (Fig. 5B). No significant binding was observed in control BSA-coated wells. Binding was concentration dependent (Fig. 5A), and the specificity of the interaction was confirmed by binding inhibition using the anti-α5β1 blocking antibody or EDTA (Fig. 5B).
These findings strongly support the assumption that cell adhesion to VEGFR-1 and VEGFR-1-induced cell migration are mediated by the interaction of this receptor with α5β1.
Since VEGFR-1 and fibronectin seem to bind the same integrin receptor, it could be hypothesised that VEGFR-1 interaction with the integrin could be mediated by fibronectin. In order to exclude this possibility, cell adhesion assays were performed in the presence of RGD peptides that should compete with integrin for binding to fibronectin. When EC were treated with the RGD peptide prior to plating on VEGFR-1/Fc, or with the peptide containing the RGE-related sequence, used as a control, no inhibitory effect was observed (Fig. 5C). As expected, treatment with the same concentration of RGD peptides blocked cell binding to fibronectin (Fig. 5C). These data confirmed that the EC interaction with VEGFR-1 is independent of the presence of other matrix proteins that bind through the RGD sequence. As further evidence, an anti-fibronectin antibody, reported to significantly block cell adhesion on fibronectin (Pierschbacher et al., 1981), was also used. In this assay, no significant inhibition of cell adhesion on VEGFR-1 was seen (data not shown).
In addition, neither VEGF nor placenta growth factor, another ligand of VEGFR-1, competed with EC adhesion to VEGFR-1/Fc (Fig. 5C), suggesting that different receptor sites are involved in growth factor or integrin binding.
VEGFR-1 induces EC spreading
To investigate whether the interaction between VEGFR-1 and theα 5β1 integrin that supports cell adhesion could also activate EC spreading, cells were seeded on VEGFR-1/Fc or fibronectin and analysed at different times. Cells plated on VEGFR-1/Fc spread and organised actin microfilaments in a similar manner to cells plated on fibronectin. However, cell spreading on VEGFR-1/Fc was slower than on fibronectin. On the latter substratum, EC were fully spread after 1 hour (Fig. 6A), whereas on VEGFR-1/Fc spreading was not detectable 1 hour after plating (Fig. 6B), and was completed only after 6 hours (Fig. 6C). Extracellular matrix proteins, newly produced by EC attached to VEGFR-1, might be required to achieve complete cell spreading after the primary interaction with VEGFR-1. We therefore tested whether de novo protein synthesis and secretion were required. Cells treated with cycloheximide, an inhibitor of protein translation, or monensin, which blocks protein secretion, still adhered to VEGFR-1/Fc, but only a few cells spread (Fig. 7A). The same treatment did not significantly affect spreading on fibronectin (Fig. 7A). Since EC bind to different extracellular matrix proteins mainly by recognising the RGD sequence, we tested the effect of such a peptide on VEGFR-1-induced cell spreading. The exposure to the RGD-containing peptide (but not to the peptide containing the RGE-related sequence, used as a control) of confluent EC monolayers seeded on VEGFR-1/Fc induced a significant rounding of cells (Fig. 7B), suggesting that EC spreading on this substratum could be dependent upon the cell secretion of other extracellular matrix proteins. We finally tested the secretion and organisation of endogenous fibronectin by EC adherent on VEGFR-1/Fc and on vitronectin, as a control. After 3 hours, cells seeded on VEGFR-1 began to spread and showed organised fibronectin fibrils in the newly deposited matrix, whereas cells plated on vitronectin were completely spread on this substratum, and no fibronectin fibrils could be detected (Fig. 7C).
The identification and characterisation of molecules that mediate cell-matrix interactions is of great importance in the understanding of multicellular tissue development. These interactions are extremely relevant in the creation of vascular structures, and vessel formation may be facilitated by shifting the predominance of cell-matrix contacts towards cell-cell contacts. Our data show, for the first time, that the soluble form of VEGFR-1 in the matrix deposited by EC in culture, suggesting that this molecule could function as an extracellular matrix protein. Sequence analysis shows that the domain architecture of the secreted VEGFR-1 is compatible with a role as a matrix protein involved in cell adhesion. Tandem repeats of domains, often with very similar structure, which probably arose by gene-duplication events, seem to be a common feature of both cell membrane and extracellular proteins involved in adhesion (Clothia and Jones, 1997; Hohenester and Engel, 2002). Moreover, our finding is in keeping with a recent work of Witmer and colleagues (Witmer et al., 2002), who characterised the expression pattern of the three VEGFR in human retina and suggested that VEGFR-1 could be present in the endothelial cell extracellular matrix in vivo.
The presence of the soluble VEGFR-1 within the matrix has therefore directed our research towards further investigation of the role of this receptor variant, in addition to that of growth factor binder. In our assays a VEGFR-1 polypeptide, which corresponds to the soluble variant, has been attached on a solid surface, mimicking the ECM deposited by the cells. In such an assay, this protein directly supported EC adhesion. We have demonstrated that antibodies directed towards the α5 or the β1 integrin subunits specifically inhibit the adhesion of EC to VEGFR-1. In addition, we have shown a direct interaction between VEGFR-1 and the α5β1 integrin in vitro, which strongly suggests that the α5β1 integrin is an EC ligand for the matrix-associated VEGFR-1. The α5β1 integrin has already been implicated in the regulation of several aspects of EC growth and differentiation. α5β1 binding to fibronectin results in the accumulation of signalling molecules and cytoskeletal components at focal adhesion sites and in the stimulation of tyrosine phosphorylation of proteins associated with focal adhesions (Hocking et al., 1998). An important role for α5β1 integrin in angiogenesis has been clearly established, since antagonists of this integrin block tumour-induced angiogenesis (Kim et al., 2000). Moreover, α5-null teratocarcinomas are poorly vascularised and α5-null embryoid bodies show delayed and reduced formation of tubular endothelial structures (Taverna and Hynes, 2001). Although a functional role for α5β1 integrin in vasculogenesis and embryonic angiogenesis has never been directly confirmed, the loss of the gene encoding the α5 integrin subunit is lethal during embryogenesis and associated with vascular and cardiac defects (Yang et al., 1993).α 5β1 integrin interacts with more than one ligand during angiogenesis, and our results raise the possibility that the VEGFR-1/α5β1 interaction also could play a role in the correct development of the primary vasculature or in the modulation of angiogenesis in the adult. Further studies are required to characterise the biological relevance of such an interaction in vivo.
Neither RGD nor other sequence motifs commonly recognised by integrinα 5β1 on different ligands could be mapped on the VEGFR-1 sequence (Koivunen et al., 1993). However, several ligands bind integrins through motifs that include structural features as well as specific residues. For example, the integrin binding motif of the vascular cell adhesion molecule-1 (VCAM-1) is a single aspartate residue at the end of a relatively long loop, and integrin binding residues in the intercellular adhesion molecules (ICAMs) are involved in long range interactions (Clothia and Jones, 1997). We are currently analysing the presence of such structural determinants in the VEGFR-1 protein.
When administered in solution to the cells, VEGFR-1 was capable of modulating EC migration in a chemotactic manner. In addition, when VEGFR-1 was immobilised on the lower surface of the filter, it sustained a haptotactic response in the EC. Both chemotaxis and haptotaxis appear to be mediated by the α5β1 integrin, since they were impaired by blocking antibodies directed against this adhesion molecule. The induction of EC motility implies the activation of a cellular response, probably mediated by theα 5β1 integrin itself as a consequence of VEGFR-1 binding. We had further indication of the initiation of such a response when we analysed EC spreading on VEGFR-1-coated surfaces. In this case, VEGFR-1 alone was not sufficient to support EC spreading. However, cell adhesion on VEGFR-1 triggered molecular activation mechanisms that resulted into new ECM protein secretion, finally leading to cell spreading. Such a process was blocked by both monensin and cycloheximide, supporting the hypothesis that EC spreading on VEGFR-1 requires protein synthesis and secretion. Moreover, we have shown that EC, left to adhere on VEGFR-1, secrete and organise a fibronectin matrix on which they subsequently spread. Therefore, in promoting EC spreading, VEGFR-1 acts differently from vitronectin, which induces direct EC spreading, and shows a behaviour similar to other proteins, such as fibrinogen (Dejana et al., 1990) that promote EC spreading via the release of newly synthesised matrix proteins.
To date the role of the soluble VEGFR-1 in angiogenesis has not been thoroughly characterised. Soluble VEGFR-1 has been proposed to be a negative regulator of VEGF-mediated signalling, acting as a decoy molecule or a soluble competitor. In addition to this function, our findings indicate that the soluble VEGFR-1 might be actively involved in other aspects of angiogenesis by playing a role in cell-matrix interactions. This new property of VEGFR-1 and the previously implied negative regulation of VEGF signalling need not be mutually exclusive. Indeed, neither VEGF nor placenta growth factor competed EC adhesion to VEGFR-1. Such a dual mechanism of action has already been shown for other molecules implicated in angiogenesis. Endostatin, for example, functions as an angiogenesis inhibitor when present in a soluble form, and as a mediator of EC adhesion and migration when bound to a solid support (Rehn et al., 2001).
The finding that a VEGF receptor may be directly involved in cell adhesion is completely novel. Previous reports demonstrated only an indirect role in cell adhesion explicated through the up-regulation of integrin expression. VEGF treatment of EC, in fact, enhances the expression of αvβ3 (Senger et al., 1996),α 1β1 and α2β1 (Senger et al., 1997) integrins. Direct interactions between integrins and transmembrane growth factor receptors on the surface of the same cell have also been described. VEGFR-2 binds to the αvβ3 integrin after activation by VEGF (Borges et al., 2000; Soldi et al., 1999), and this interaction seems to contribute to enhancing VEGFR-2 phosphorylation. Similar data is not yet available for VEGFR-1.
In addition to EC, several tumour cells express VEGFR-1 (Bellamy et al., 1999; Lacal et al., 2000; Masood et al., 2001), both as transmembrane and soluble protein. The VEGFR-1/integrin interaction described here might therefore also play a role in tumour cell adhesion to the extracellular matrix and could thus represent a new target for the development of compounds aimed at limiting tumour-induced angiogenesis and tumour metastatization.
We wish to thank Prof. J. Jiricny, Prof. K. Ballmer-Hofer, and Prof. A. Sonnenberg for helpful scientific discussion, and Dr T. Odorisio for critical review of our data. We would also like to thank N. De Luca and M. Teson for skilful technical assistance and M. Inzillo for artwork. This work was supported by grants from CNR, Progetto Finalizzato Biotecnologie (P.n.115.23287), from Ministero della Sanità, Italy and from the TMR Marie Curie Research Training scheme for funding (contract n° B104CT985056).
- Accepted May 16, 2003.
- © The Company of Biologists Limited 2003