Mapping of the sites for ligand binding and receptor dimerization at the extracellular domain of the vascular endothelial growth factor receptor FLT-1.

The vascular endothelial growth factor (VEGF) receptor FLT-1 has been shown to be involved in vasculogenesis and angiogenesis. The receptor is characterized by seven Ig-like loops within the extracellular domain. Upon VEGF binding FLT-1 becomes phosphorylated, which has been thought to be preceded by receptor dimerization. To further investigate high affinity binding of VEGF to FLT-1 and ligand-induced receptor dimerization, we expressed in Sf9 cells the entire extracellular domain comprising all seven Ig-like loops: sFLT-1(7) and several truncated mutants consisting of loop one, one and two, one to three, one to four, and one to five. The corresponding proteins, named sFLT-1(1), (2), (3), (4), and (5) were purified. Only mutants sFLT-1(3) to (7) were able to bind 125I-VEGF with high affinity. No binding of VEGF was observed with sFLT-1(1) and sFLT-1(2), indicating that the first three Ig-like loops are involved in high affinity binding of VEGF. The binding of VEGF to sFLT-1(3) could be competed with placenta growth factor (PlGF), a VEGF-related ligand, suggesting that high affinity binding of VEGF and PlGF is mediated by the same or closely related contact sites on sFLT-1. Deglycosylation of the sFLT-1(3), (4), (5), and (7) did not abolish VEGF binding. Furthermore, unglycosylated sFLT-1(3), expressed in Escherichia coli, was able to bind VEGF with similar affinity as sFLT-1(3) or sFLT-1(7), both expressed in Sf9 cells. This indicates that receptor glycosylation is not essential for high affinity binding. Dimerization of the extracellular domains of FLT-1 upon addition of VEGF was detected with all mutants containing the Ig-like loop four. Although sFLT-1(3) was able to bind VEGF, dimerization of this mutant was inefficient, indicating that sites on Ig-like loop four are essential to stabilize receptor dimers.

The vascular endothelial growth factor (VEGF), 1 a potent mitogen for endothelial cells, is an important angiogenic factor also involved in the differentiation of endothelial cells and the development of the vascular system (1,2). It has been shown to be implicated in human diseases such as diabetic retinopathy, rheumatic arthritis, and cancer (3). VEGF in particular appears to be the most important angiogenic factor of many solid tumors, promoting vascularization and formation of metastases (4).
Four different VEGF isoforms have been described so far, all encoded by a single gene: VEGF 121 , VEGF 165 , VEGF 189 , and VEGF 206 (5). All different isoforms are secreted dimeric proteins, sharing similarities with platelet-derived growth factor (PDGF) and belong to the family of growth factors containing a cysteine knot motif (6).
Two receptor tyrosine kinases, FLT-1 (7,8) and KDR/FLK-1 (9, 10), have been identified, which bind VEGF with high affinity. Both receptors belong to the type III tyrosine kinases and are characterized by seven Ig-like loops within their extracellular domain and a split kinase domain within the cytoplasmatic moiety (11). The Ig-like loop motif is a common feature of extracellular domains of membrane-anchored proteins. Members of the immunoglobulin superfamily are often involved in cell surface recognition (12). Both VEGF receptors contain several putative N-glycosylation sites and the apparent molecular weights of the mature proteins suggest that both receptors are extensively glycosylated (8,10).
The activation of growth factor receptors in general is preceded by the formation of receptor dimers and subsequent receptor phosphorylation. The resulting phosphotyrosine residues are docking sites for signal coupling components such as SH-2 proteins (13). The molecular structures that are responsible for ligand/receptor interaction and ligand-induced dimerization are poorly understood for most receptors. Since the dimeric structure of VEGF is a prerequisite of receptor activation, it can be speculated that one VEGF molecule bridges two receptors via two similar recognition sites, as has been suggested for PDGF (14). Characterization of VEGF binding to its receptors by mutational analysis of the ligand supports the assumption that VEGF has two contact sites for its receptors (15).
Nothing is known about which part of the receptor is involved in ligand binding and which part of the receptor's extracellular domain is involved in VEGF-dependent dimerization. To investigate these questions, we generated several soluble mutants of the extracellular domain of FLT-1, each consisting of a different stretch of Ig-like loops. The data presented here suggest that the recognition site for VEGF is located on the first three Ig-like loops, whereas dimerization is stabilized due to an additional domain located on Ig-like loop four. Glycosylation was found not to be a prerequisite of high affinity binding of VEGF to FLT-1.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-Sf9 cells were cultured routinely in 1-liter spinner flasks (Technomara, FRG) in ExCell 400 (JRH Bioscience, UK) without any further supplements. Human umbilical vein endothelial cells (HUVECs) were obtained from PromoCell (FRG) and cultured according to the provided protocol. VEGF 165 was expressed in Escherichia coli and purified as described previously (16). VEGF 165 was iodinated to a specific activity of 100,000 cpm/ng by using the chloramine T method (17) (Immundiagnostik, FRG). PlGF 152 was expressed with the baculovirus/insect cell system as described previously (18). N-Glycosidase F and disuccinimidyl suberate (DSS) were obtained from Boehringer Mannheim (FRG), N-hydroxysuccinimidyl-biotin was from Pierce (FRG).
Expression and Purification of Soluble FLT-1 Receptor Mutants-Soluble N-terminal flt-1 fragments were cloned by polymerase chain reaction from the full-length cDNA clone 3-7 (7) using the upstream primer: 5Ј-GGAATTCCGCGCTCACCATGGTCAGC-3Ј, containing an EcoRI site and various downstream primers, which all contain an artificial stop codon and a BamHI site. The coding sequences ended at bp 642 for sFLT-1(1), at bp 957 for sFLT-1(2), at bp 1243 for sFLT-1(3), at bp 1531 for sFLT-1(4), at bp 1915 for sFLT-1(5) and at bp 2497 for sFLT-1(7), respectively. The polymerase chain reaction products were purified with the Quiax DNA gel extraction kit (Quiagen, FRG) and subcloned into the baculovirus transfer vector pVL1392 as EcoRI/ BamHI fragments. Plasmids containing the cDNA were isolated from transfected bacteria and then used for transfection into Sf9 cells along with wild-type baculovirus DNA. Recombinant baculoviruses were obtained using the BaculoGold TM transfection kit following standard protocols (Pharmingen, San Diego, CA). For protein expression, Sf9 cells at a density of 2 ϫ 10 6 cells/ml were infected with a multiplicity of infection of 10. 72 h after infection, cell-free conditioned medium was filtered and applied to a heparin-Sepharose column (Pharmacia, FRG). The column was washed with either 10 ml of 20 mM Tris-Cl, pH 7.4, for sFLT-1(1) and (2), or 10 ml of 0.4 M NaCl for sFLT-1(3), or 10 ml of 0.6 M NaCl for sFLT-1(4), (5), and (7). Bound proteins were eluted by increasing the NaCl concentration by steps of 0.2 M. The monoclonal antibody 7A6 was used to identify all the various sFLT-1 proteins by Western blotting.
Bacterial Expression and Purification of sFLT-1(3)-A 0.9-kilobase pair NcoI/BamHI fragment encoding amino acids 31-338 of human FLT-1 was generated by polymerase chain reaction and ligated into the bacterial expression vector His-pET (16). Thus the expression plasmid encoded FLT-1 amino acids 31-338 fused to an N-terminal 6 ϫ His-tag and amino acids Met and Glu, which were derived from the artifical NcoI site. For bacterial expression, the plasmid construct was transduced into E. coli strain BL21(DE3) carrying an inducible T7 RNA polymerase gene (19). Bacterial cultures of 250 ml of LB medium containing 100 g/ml ampicillin were grown in shaking flasks at 37°C to an A 600 of 0.8. Isopropyl-␤-D-thiogalactoside was added to a final concentration of 0.4 mM, and the culture was grown for another 3 h. Cells were harvested and washed, and the pellet was frozen at Ϫ80°C. Upon use, cells were thawed at 37°C, resuspended in 25 ml of buffer A (50 mM Tris-HCl, 10 mM 2-mercaptoethanol, 2 mM EDTA, 5% (v/v) glycerol, 0.2 mg/ml lysozyme, 10 g/ml DNase I, pH 8.0) and incubated for 30 min at 22°C. The suspension was sheared by five high speed treatments of 20 s in an Ultra-Turrax dispersing apparatus and incubated for 10 min at 22°C. The mixture was cooled on ice and sonicated six times for 15 s with the microtip of a Bronson Sonifier 250. After the addition of sodium deoxycholate and Nonidet P-40 to a final concentration of 0.05% and 1% (w/v), respectively, the mixture was incubated for 10 min at 4°C and then centrifuged at 10,000 rpm at 4°C for 30 min. The pellet was resuspended in 25 ml of buffer A supplemented with 0.05% sodium deoxycholate and 1% Nonidet P-40 and recentrifuged. The inclusion body pellet was solubilized in 25 ml of buffer B (6 M guanidine HCl, 0.15 M NaCl, 0.1 M dithiothreitol, 50 mM NaPO 4 buffer at pH 6.5). The solubilized protein was dialyzed two times against 250 ml of buffer C (6 M urea, 0.1 M dithiothreitol, 50 mM MES at pH 5.5) at 4°C, and the volume of the dialysate was reduced 20-fold by ultrafiltration (Millipore Ultrafree-15). The concentrated protein solution was mixed with 2 volumes of buffer D (6 M urea, 0.5 M cystamine, 0.1 M glycine, 20 mM Hepes at pH 7.4) and incubated for 4 h at 4°C with gentle agitation. For refolding, the solution was sequentially diluted to final 0.3 M urea by the addition of PBS, 50 mM glycine. The refolded sFLT-1(3) protein was concentrated by ultrafiltration and referred to as Deglycosylation of Proteins-For deglycosylation of recombinant sFLT-1 mutants, N-glycosidase F was used. SDS (final concentration, 0.3%) was added to sFLT-1 proteins (250 -500 ng in 10 l, total volume), and the mixture was incubated at 95°C for 2 min. Then 10 l of 2 ϫ reaction buffer were added (100 mM sodium phosphate, pH 7.2, 20 mM EDTA, 1% Nonidet P-40) and again incubated at 95°C for 2 min. After cooling down, N-glycosidase F (0.2 unit/assay) was added, and the mixture was incubated at 37°C overnight. The reaction was stopped with 4 ϫ SDS sample buffer, and proteins were resolved by SDS-PAGE.
Antibodies and Western Blot Analysis-Approximately 50 g of sFLT-1(7) dissolved in PBS and emulsified with Freund's complete adjuvant were injected both intraperitoneally and subcutaneously into Lou/C rats and again as a booster 8 weeks after the first immunization. Fusion of the myeloma cell line P3X63-Ag8.653 with rat immune spleen cells was performed essentially as described previously (20). Supernatants were screened for anti-sFLT-1 antibodies using a solid phase enzyme-linked immunosorbent assay. The mAb 7A6 was able to detect the various sFLT-1 mutant proteins and was characterized as a rat IgG2a. For Western analysis, aliquots (20 l) of the elution fractions of the heparin-Sepharose columns were resuspended in SDS sample buffer and resolved by SDS-PAGE. The gels were electroblotted onto a polyvinylidene difluoride membrane, and blots were blocked for 30 min with 5% milk powder in TBST (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, 0.1% Tween-20) and incubated for 1 h in the same buffer with 1 g/ml mAb 7A6. The blots were washed twice in water and TBST and incubated for 45 min in TBST containing a 5000-fold diluted peroxidaseconjugated mouse anti-rat IgG/goat anti-rat IgG ϩ IgM (Jackson Im-munoResearch Labs., Inc., West Grove, PA). The blots were washed as described before, and peroxidase-coupled antibodies were visualized using the ECL chemiluminescent Western blotting detection system (Amersham, FRG).
Ligand Blotting-For ligand blotting, 0.5 g of purified sFLT-1 mutant proteins was mixed with nonreducing SDS sample buffer, and SDS-PAGE was performed. After blotting the gel onto a polyvinylidene difluoride membrane at 4°C, the filter was blocked in TBS containing 2% bovine serum albumin for 30 min. The filter was incubated with 20 ng/ml 125 I-VEGF 165 for 2 h. After extensive washing with TBS, the filter was exposed to a Kodak X-Omat x-ray film.
Solid Phase Binding Assays-sFLT-1 mutant proteins (0.5-1 g/ well) were absorbed to the surface of 96-well cluster plates (Nunc-Immuno Plate MaxiSorp TM ) for 18 h at 4°C. The plates were washed twice with TBS, nonspecific sites were blocked with 2% bovine serum albumin in TBS for 1 h, and the plates were washed again with TBS. In competition experiments, the binding of 125 I-VEGF 165 (10 ng/ml) was competed by increasing amounts of recombinant human VEGF 165 or PlGF 152 as indicated in the figure legends. The wells were washed four times with TBS, bound protein was solubilized in 100 l of 0.3 M NaOH, 1% SDS, and radioactivity was counted in a gamma-counter (Beckmann, FRG).
For saturation binding curves, biotinylated VEGF was used. Immobilized sFLT-1(3), (7), and sFLT-1(3) E. coli were washed three times with PBS containing 0.1% Tween-20, and nonspecific sites were blocked with 0.5% bovine serum albumin, 0.5% Tween-20 in PBS for 2 h at room temperature. Binding was carried out with increasing amounts of biotinylated VEGF 165 for 1 h. Unspecific binding was determined in parallel dishes by the addition of a 100-fold excess of unlabeled VEGF. Biotinylated VEGF was quantified using a streptavidin complexed to alkaline phosphatase (Calbiochem, FRG) and p-nitrophenyl phosphate disodium (Sigma, FRG) as a substrate.
Cell-based Binding Assay-Binding assays with endothelial cells were done as described previously (21). Briefly, HUVECs (passage 4) were seeded in growth medium at 5 ϫ 10 4 cells/well in 24-well cluster plates. After 72 h the cells were washed extensively with binding buffer (Dulbecco's modified Eagle's medium, 25 mM Hepes, 1 mg/ml bovine serum albumin, pH 7.4) and incubated for 2 h at 4°C with binding buffer containing 1 ng/ml 125 I-VEGF 165 . Increasing amounts of sFLT-1 were used for competition. The cells were washed three times with binding buffer, bound protein was solubilized with 250 l of 0.3 M NaOH, 1% SDS/well, and radioactivity present in the lysates was quantified.
Chemical Cross-linking of Protein Complexes-For cross-linking, 5 g of sFLT-1(3), (4), and (5) were mixed with 100 ng of VEGF 165 in PBS, 0.1% Tween 20 in a final volume of 100 l and incubated on ice for 2 h. DSS was added to a final concentration of 1 mM, and the reaction mixture was incubated for another 15 min. The protein was trichloroacetic acid-precipitated, washed twice with acetone, and resuspended in 30 l of SDS sample buffer. The proteins were resolved by SDS-PAGE and identified by Western blotting with the mAb 7A6. In a second approach, 125 I-VEGF 165 (20 ng/assay) was used for the cross-linking reaction with sFLT-1(3), (4), and (5). The gel was vacuum-dried and exposed to a Kodak X-Omat x-ray film.
[ 3 H]Thymidine Incorporation Assay-HUVECs (passage 3) were seeded at a density of 1 ϫ 10 4 cells/well into 48-well cluster plates in growth medium. Cells were allowed to attach overnight at 37°C. Growth medium was replaced by basal medium (1.7% fetal calf serum), and a constant concentration of VEGF 165 (3 ng/ml) together with increasing amounts of the different sFLT-1 proteins were added 24 h later. Incubation was continued for additional 18 h, then 1 Ci of [ 3 H]thymidine (56.7 Ci/mmol, DuPont NEN) was added. Cells were kept at 37°C for an additional 6 h. Cell monolayers were fixed with methanol, washed with 5% trichloroacetic acid, solubilized in 0.3 M NaOH, and counted by liquid scintillation.

Recombinant Expression of sFLT-1 Mutants-
To express the soluble extracellular domain of the VEGF receptor FLT-1 and its truncated mutants, constructs were made by introducing stop codons at various sites of the flt-1 cDNA as shown in Fig.  1. The resulting cDNAs were cloned into the baculovirus transfer vector pVL1392, and all proteins were expressed in Sf9 insect cells. Since all constructs contain the N-terminal leader sequence but lack the transmembrane domain, the expressed proteins were expected to be secreted from the infected Sf9 cells. To enable detection of the expressed proteins, monoclonal antibodies had been raised against the extracellular domain of FLT-1. The monoclonal antibody 7A6 was found to detect the soluble FLT-1 as well as all truncated derivatives. Western blot analysis of partially purified sFLT-1 proteins from conditioned media of infected Sf9 cells confirmed that all FLT-1 mutants were expressed and secreted (Fig. 2). The apparent molecular masses were estimated from SDS-PAGE to be 16 kDa for sFLT-1(1), 28 kDa for sFLT-1(2), 45 kDa for sFLT-1(3), 57 kDa for sFLT-1(4), 72 kDa for sFLT-1(5), and 105 kDa for sFLT-1(7). sFLT-1(1) and sFLT-1(2) matched more or less the calculated molecular masses deduced from the amino acid sequence, whereas the estimated size of sFLT-1(3) to sFLT-1(7) was increased by about 10 -20 kDa as compared with the calculated molecular mass. This is most likely due to N-linked sugars since the extracellular domain contains 12 putative glycosylation sites (see Fig. 1).
Binding of VEGF and PlGF to sFLT-1 Mutants-To investigate whether the expressed sFLT-1 mutant proteins are able to bind VEGF 165 we used a solid phase binding assay on microtiter plates. Specific binding of VEGF 165 could be observed to all mutants containing the first three N-terminal Ig-like loops: sFLT-1(3), (4), (5), and (7). No specific VEGF binding could be obtained with sFLT-1(1) and sFLT-1(2) containing the first or the first and the second Ig-like loops, respectively (Fig. 3, inset). In a competition experiment, using increasing amounts of unlabeled VEGF 165 , comparable high affinity binding of VEGF to all sFLT-1 mutants containing the first three Ig-like loop could be demonstrated (Fig. 3). To make sure that lack of ligand binding of the two shorter mutants sFLT-1(1) and sFLT-1(2) was not due to inefficient immobilization to the microplates, we performed a cellular binding assay with HUVECs that express both VEGF receptors KDR and FLT-1 (21). All sFLT-1 mutants containing the first three Ig-like loops were able to compete efficiently with VEGF 165 binding to HUVECs, the mutants sFLT-1(1) and (2), which failed to bind VEGF in the solid phase assay, also failed to compete with VEGF binding to endothelial cells (Fig. 4). The slight decrease of total binding observed with the sFLT-1(1) and sFLT-1(2) preparations is statistically not significant.
PlGF, similar to VEGF, is a disulfide-bridged homodimer and shares about 30% identity with VEGF (18,22). PlGF has been found to bind to FLT-1 and to displace VEGF from this  (7), have been expressed in Sf9 insect cells and purified as described. The mAb 7A6, which detects all sFLT-1 mutant proteins, was used for Western blotting.
FIG. 3. Solid phase VEGF binding assay. Solid phase binding assay with the sFLT-1 mutant proteins has been performed on microtiter plates. Binding of radiolabeled VEGF to immobilized sFLT-1 mutant proteins was competed by increasing amounts of unlabeled VEGF. Specific binding as shown in the insert for the mutants sFLT-1(1)- (7) was estimated as the difference between total binding of VEGF and binding of labeled VEGF in the presence of a 100-fold excess of unlabeled ligand.
receptor (23)(24)(25). To test whether this is also true for the soluble extracellular domain of FLT-1, we performed the solid phase binding assay (Fig. 5). PlGF 152 competed with VEGF 165 for binding to sFLT-1(3) and (7). No difference could be detected between the full-length extracellular domain and sFLT-1(3), the shortest sFLT-1 mutant still exhibiting VEGF binding. From this we conclude conclude that PlGF and VEGF share similar contact sites on the receptor. (Fig. 6). A dose-dependent inhibition of VEGF-stimulated DNA synthesis could be observed with sFLT-1(3), (4), (5), and (7), confirming the results from the ligand binding studies. Addition of sFLT-1(1) and sFLT-1(2) had only a minor VEGF-antagonizing effect. From these experiments we conclude that the presence of Ig-like loop three is a prerequisite for high affinity binding of VEGF to its receptor.

Inhibition of VEGF-mediated DNA Synthesis by sFLT-1 Mutants-Since HUVECs proliferate in response to VEGF, we investigated the ability of the sFLT-1 mutants to antagonize VEGF-mediated incorporation of [ 3 H]thymidine into HUVECs
Effect of Glycosylation on VEGF Binding-The presence of putative N-glycosylation sites (Fig. 1) and the obvious differences between the apparent and calculated molecular weights of the sFLT-1 mutants (3) to (7) suggest that those sFLT-1 mutants, which are able to bind VEGF, are released from Sf9 cells as glycosylated proteins. To investigate whether protein glycosylation has any influence on VEGF binding, we incubated the recombinant sFLT-1 mutants with N-glycosidase F. The N-glycosidase F-treated proteins and appropriate controls were subjected either to Western blot analysis or ligand blotting (Fig. 7). The Western blot analysis revealed a significant decrease of the apparent molecular weight due to the glycosidase treatment, demonstrating that the polypeptides contain N-linked sugar residues (Fig. 7A). Incubation with radiolabeled 125 I-VEGF identified both the glycosylated and deglycosylated polypeptides as VEGF binding proteins (Fig. 7B). In several preparations we detected high molecular weight complexes in the absence of VEGF with the mAb 7A6. The appearance of these complexes varied in individual preparations of different sFLT-1 mutants. An example is shown for sFLT-1(5) in Fig. 7,  A and B (arrowheads). The reason for this complex formation is unclear. Since this effect was most obvious on nonreducing SDS gels, we speculate that the additional bands might be due to incorrect formation of disulfide bridges during the recombinant protein expression and partial proteolytic degradation.
To confirm that VEGF binding does not depend on glycosylation, we used a different experimental approach, expressing the sFLT-1 sequence corresponding to amino acid Asp 31 to His 338 in E. coli to prevent N-glycosylation. The amino acid sequence expressed in E. coli thus resembles sFLT-1(3) lacking the leader peptide. This recombinant protein was found in the inclusion body fraction of E. coli. The apparent molecular mass as estimated from SDS-PAGE was approximately 37 kDa, which is very close to the calculated molecular mass of 35 kDa. For refolding, the inclusion bodies were first solubilized in the presence of guanidine HCl. Oxidation and reduction of SH groups was performed to achieve the formation of proper disulfide bridges. The refolded protein was referred to as sFLT-1(3) E. coli and subjected to a solid phase VEGF binding assay (Fig. 8). The sFLT-1(3) E. coli showed a similar dose dependence of specific binding as compared with sFLT-1(3) and sFLT-1(7), which had been expressed in Sf9 cells. The K d values calculated from the binding data were estimated to be 47 ng/ml for sFLT-1(7), 71 ng/ml for sFLT-1(3), and 62 ng/ml for sFLT-1(3) E. coli . From these experiments we conclude that N-glycosylation on the extracellular domain of sFLT-1 is not a prerequisite of high affinity VEGF binding.
Dimerization of sFLT-1 Mutants-To test whether the soluble extracellular domain of the receptor is sufficient for VEGFdependent receptor dimerization, we investigated the formation of dimers in the presence of VEGF. The mutants sFLT-1(3), (4), and (5) were cross-linked with DSS after incubation without and with VEGF followed by Western blotting using the mAb 7A6 (Fig. 9A). In the absence of VEGF the sFLT-1 mutants were all found to be in the monomeric form. After incubation with VEGF, additional high molecular mass complexes were observed. The sizes of these additional protein complexes were found to be about 66 kDa for sFLT-1(3), 140 kDa for sFLT-1(4), and 180 kDa for sFLT-1(5), respectively. In the case of sFLT-1(4) and (5) these molecular masses correspond to a complex of 1 molecule of VEGF and 2 molecules of sFLT-1 mutants. In contrast, the cross-linked sFLT-1(3) complex rather matches the size of a VEGF-containing monomer and not of a dimer. From these data we conclude that sFLT-1(3), although binding VEGF with high affinity, is impaired to form stable dimers.
Similar results can be obtained by cross-linking the sFLT-1 mutants upon incubation with radiolabeled VEGF (Fig. 9B). The autoradiograph of the corresponding SDS-gel shows radiolabeled high molecular mass complexes at about 180 kDa for dimeric, VEGF-containing sFLT-1(5) and at about 140 kDa for dimeric VEGF-containing sFLT-1(4). Under the same experimental conditions the mutant sFLT-1(3) does not show substantial dimerization in the presence of VEGF. The size of the major labeled complex at about 70 kDa corresponds to only 1 molecule of sFLT-1(3) and 1 molecule of VEGF. Only a very weak labeling was found at 120 kDa, the size of the sFLT-1(3) dimer complexed with VEGF. These results suggest that sFLT-1(3) has lost almost completely the potential to form stable VEGF-dependent dimers and that the fourth Ig-like loop harbors a domain that supports and/or stabilizes the ligand-induced formation of receptor dimers. DISCUSSION We expressed different truncated mutants of the extracellular domain of the VEGF receptor FLT-1 to get further insight into the structure/function relationship for VEGF binding and receptor dimerization. The baculovirus/insect cell expression system has already been used to express a soluble extracellular domain of FLT-1, consisting of the Ig-like loops one to six, which retained the full capacity of VEGF binding (26). We used the same expression system to express different sFLT-1 mutants that were constructed by introducing artificial stop codons into appropriate positions, resulting in cDNAs for Iglike loop one, Ig-like loop one and two, Ig-like loop one to three, Ig-like loop one to four, Ig-like loop one to five, and Ig-like loop one to seven. Monoclonal antibodies were raised against the extracellular domain of FLT-1 to confirm protein expression and to follow protein purification. mAb 7A6 was able to detect all different constructs, indicating that its binding epitope is located within the first N-terminal Ig-like loop (Fig. 2).
To investigate binding of VEGF to the various sFLT-1 mutants, we used either a solid phase binding assay or an endothelial cell-based receptor binding assay. High affinity binding of VEGF to sFLT-1(3), (4), (5), and (7) could be observed with the solid phase binding assay. The same sFLT-1 mutants were also able to compete with VEGF receptors on HUVECs for VEGF binding. This is in agreement with the results obtained with the solid phase binding assay (compare Figs. 3 and 4). No binding of VEGF to sFLT-1(1) or (2) containing only the first, or the first and second N-terminal Ig-like loops, respectively, could be detected in the solid phase assay, and only a very weak competition, if any at all, was observed in cell-based binding assay (Figs. 3 and 4). Furthermore, VEGF-stimulated DNA synthesis in HUVECs was inhibited in a dose-dependent fashion by all sFLT-1 mutants except sFLT-1(1) and (2). From this we conclude that the contact site(s) of VEGF is(are) located within the first three N-terminal Ig-like loops of sFLT-1. We further hypothesize that especially Ig-like loop three contributes to VEGF binding, since the deletion of this loop completely abolishes VEGF binding. Similar results have been reported by others for c-Kit and the PDGF receptor, two receptors related to the VEGF receptors and consisting of five Ig-like loops in their extracellular domain. In each case ligand binding could be mapped to the first three Ig-like loops (27)(28)(29). Further mutational analysis for the PDGF receptor type ␣ revealed that high affinity binding of PDGF is mediated by Ig-like loops two and three and that loop one defines specificity between PDGF AA or PDGF BB (27).
PlGF, a growth factor related to VEGF, has been described to FIG. 8. VEGF binding to sFLT-1(3) E. coli . Recombinant sFLT-1(3) and (7) expressed in Sf9 insect cells and sFLT-1(3) E. coli were immobilized on microtiter plates and incubated with increasing amounts of biotinylated VEGF 165 . Bound VEGF was detected by the use of streptavidin complexed to alkaline phosphatase and nitrophenol phosphate as a substrate. Unspecific binding was estimated in the presence of a 100-fold excess of unlabeled VEGF 165 , and specific binding was calculated as the difference between total and unspecific binding.  (7) mutants have been deglycosylated by treatment with Nglycosidase F as described in the text. Deglycosylated mutants and the appropriate glycosylated controls were separated by SDS-PAGE under nonreducing conditions. The proteins were transferred to a polyvinylidene difluoride membrane and either used for Western analysis with mAb 7A6 (A) or incubated with 20 ng/ml 125 I-VEGF 165 (B). VEGF binding proteins were identified by autoradiography. A complex of two sFLT-1(5) molecules can be detected on both Western and ligand blots (arrowheads). bind to FLT-1 (23)(24)(25), and heterodimers of VEGF/PlGF have been found to activate VEGF receptors (30). Competition of PlGF with VEGF binding to FLT-1 has been found by others, suggesting that PlGF and VEGF share similar receptor recognition sites (23)(24)(25). We confirmed these results by testing the influence of PlGF on VEGF binding to sFLT-1 proteins in the solid phase binding assay (Fig. 5). PlGF was able to displace VEGF from both the entire extracellular domain sFLT-1(7) and sFLT-1(3), the shortest deletion mutant which retained VEGF binding.
The presence of potential N-glycosylation sites within the extracellular domain of FLT-1 and the difference between the apparent molecular weights obtained from SDS-PAGE and the calculated molecular weight, both suggest that sFLT-1(3) and larger sFLT-1 mutants were secreted as glycosylated proteins from infected Sf9 cells (Figs. 1 and 2). We therefore addressed the question, whether glycosylation of FLT-1 participates in the recognition of VEGF. Two lines of evidence demonstrate that glycosylation is not a prerequisite for ligand binding. First, enzymatic deglycosylation of sFLT-1 mutants, using N-glycosidase F, did not abolish VEGF binding (Fig. 7), and second, sFLT-1(3) expressed in E. coli retained the ability to bind VEGF (Fig. 8).
It has been shown, for a variety of receptor tyrosine kinases, that receptor activation is preceded by receptor dimerization (for review, see Heldin (13)). The ability of recombinant-purified extracellular receptor domains to dimerize in the presence of the appropriate ligand has already been described for the PDGF receptor ␣ and ␤ (14). To investigate whether the soluble extracellular domain of FLT-1 can dimerize and whether dimerization is VEGF-dependent, we performed chemical cross-linking experiments of sFLT-1 and VEGF complexes. Western analysis identifying sFLT-1 mutants with the monoclonal antibody 7A6 or the use of radiolabeled VEGF revealed that soluble receptor mutants sFLT-1(4) and (5) are able to form dimers in the presence of VEGF, whereas sFLT-1(3) associated predominantly in its monomeric form with VEGF ( Fig.  9). This leads us to the conclusion that Ig-like loop four is of particular importance for the formation of VEGF-mediated receptor dimerization. Similar observations have been reported for c-Kit, a receptor consisting of five Ig-like loops on the extracellular part. Ig-like loop four of c-Kit has been identified as an intrinsic ligand-dependent dimerization site. Receptor mutants of c-Kit lacking Ig-like loop four retained their capacity of ligand binding but no longer formed receptor dimers (31). No significant differences have been observed with respect to the affinity for VEGF binding to all sFLT-1 mutants (Figs. 3  and 4). Thus it is unlikely that Ig-like loop four has a strong impact on ligand recognition. However, it might be responsible for a direct interaction of two sFLT-1 monomers. Since efficient dimerization requires the presence of VEGF, presentation of the active dimerization site seems to be the consequence of a VEGF-dependent conformational change. Unlike the other Iglike loops, loop four lacks the disulfide bridge, which is thought to stabilize the two ␤ sheets of the Ig-like loop structure. Thus Ig-like loop four might gain additional flexibility. However, participation of other domains involved in formation of receptor dimers cannot be excluded.
The results presented here clearly suggest that VEGF binding can be attributed to the first three N-terminal Ig-like loops