The cDNA fragment of the VEGF165 N-terminal domain (NTD) (APMAEGGGQNHHEVVKFMDVYQRSYCHPIETLVDIFQEYPDEIEYIFKPSCVPLMRCGGCCNDEGLECVPTEESNITMQIMRIKPHQGQHIGEMSFLQHNKCECRPKKDRARQEN) was amplified by PCR using primers 5′-GGG​GAT​CCG​CAC​CCA​TGG​CAG​AAG​GAG​G-3′ and 5′-GGA​ATT​CTC​AAT​CtT​TCT​TTG​GTC​TGC​ATT​C-3′ with human VEGF165 cDNA (Open Biosystems, Lafayette, CO, United States) as a template, and it was sub-cloned into the BamHI/EcoRI site of the PET28a expression vector. The VEGF165 heparin-binding domain (HBD) (PCGPCSERRKHLFVQDPQTCKCSCKNTDSRCKARQLELNERTCRCDKPRR) of VEGF165 was amplified by PCR using primers 5′-CGG​GAT​CCC​CCT​GTG​GGC​CTT​GCT​CAG​AG-3′ and 5′-CGG​AAT​TCT​CAC​CGC​CTC​GGC​TTG​TCA​CAT​C-3′ with human VEGF165 cDNA (Open Biosystems, Lafayette, CO, United States) as a template, and it was sub-cloned into the BamHI/EcoRI site of the PET28a expression vector. The protein was synthesized in BL21 induced by IPTG as an insoluble protein and solubilized in 8 M urea, purified by Ni-NTA affinity chromatography, and refolded as described by Saegusa et al. (2009).

The cDNA fragment of KDR D1 (NTTLQITCRGQRDLDWLWPNNQSGSEQRVEVTECSDGLFCKTLTIPKVIGNDTGAYKCFYRETDL) was amplified by PCR using primers 5′-CGG​GAT​CCG​ACA​TAC​TTA​CAA​TTA​AGG​C-3′.

5′-CGG​AAT​TCT​CAA​GAT​CTG​TAA​TCT​TGA​ACA​TAG-AC-3′ with full-length KDR cDNA and sub-cloned into the BamHI/EcoRI site of PET28a vector, and protein was synthesized in BL21 induced by IPTG as an insoluble protein. The insoluble protein was solubilized in 8 M urea, purified by Ni-NTA affinity chromatography, and refolded as described by Saegusa et al. (2009). The cDNA fragment was sub-cloned into the BamHI/EcoRI site of pGEX2T vector, and protein was expressed in BL21 and purified by Glutathione–Sepharose affinity chromatography.

Site-directed mutagenesis was carried out as described by Saegusa et al. (2009). The presence of the mutation was verified by DNA sequencing.

Soluble αvβ3 (His-tagged) was immobilized on the CM5 sensor chip using a standard amine coupling procedure in HBS-P buffer (0.01 M HEPES, pH 7.4, 0.15 M NaCl, and 0.0005% of surfactant P20) with 1 mM of Mn²⁺. The HBD was injected at 50 μL/min for 1.8 min. HBS-P buffer with 1 mM of Mn²⁺ was then injected at 50 μL/min for 3 min to allow bound VEGF to dissociate from αvβ3.

KDR D1 (His-tagged) was immobilized on the CM5 sensor chip using a standard amine coupling procedure in HBS-EP buffer (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, and 0.0005% of surfactant P20). The HBD was injected at 50 μL/min for 1.8 min. The HBS-EP buffer was then injected at 50 μL/min for 3 min to allow the bound VEGFs to dissociate from the KDR D1.

Wells of 96-well microtiter plates were coated with PBS-diluted protein, as specified in the experiments, and incubated at 37°C for 1 h. The wells were then blocked with 0.1%BSA/PBS boiled at 80°C for 20 min to reduce background and then cooled to room temperature before application of 300 μL/well. After a 30-min incubation at room temperature, soluble αvβ3 (1 μg/mL) in HEPES-Tyrode buffer + 1 mM MnCl₂ (incubation buffer) was added and plates were further incubated at room temperature for 1 h. Unbound αvβ3 was removed by washing with incubation buffer, and bound αvβ3 was quantified using anti-β3 mAb (AV10), followed by HRP-conjugated anti-mouse IgG and peroxidase substrate 3,3′,5,5′-tetramethylbenzidine (TMB) solution, 100 μL/well. The reaction was stopped by the application of 2N HCl, 50 μL/well, and measurements were taken at OD = 450 nm with a plate reader.

Wells of 96-well microtiter plates were coated with HBD and incubated at 37°C for 1 hour. The wells were then blocked with 0.1% BSA/PBS boiled at 80°C for 20 min and then cooled before application of 300 μL/well. After 1 h of blocking, wells were incubated with KDR D1 fused to GST for 1 h at room temperature. After washing with 0.05% Tween 20/PBS, an anti-GST conjugated with HRP was applied for 1 hour at room temperature. Three washes were performed with 0.05% Tween-20/PBS before detection with the TMB solution, 100 μL/well. The reaction was stopped by the application of 2N HCl, 50 μL/well, and measurements were taken at OD = 450 nm with a plate reader.

Docking simulation of the interaction between the VEGF165 HBD (2VGH.pdb) and integrin αvβ3 was performed using AutoDock3, as described by Saegusa et al. (2009). In the present study, we used the headpiece (residues 1–438 of αv and residues 55–432 of β3) of αvβ3 (open-headpiece form, 1L5G.pdb). Cations were not present in αvβ3 during docking simulation (Mori et al., 2008; Saegusa et al., 2008; Fujita et al., 2015). We also performed a docking simulation without removing cations (Supplementary Figure S1).

In mutant VEGF165, amino acids R123, R124, K125, K140, R145, and R149 within the HBD were identified as required for integrin αvβ3 binding based on changing these residues to Ala (R123A/R124A/K125A/K140A/R145A/R149A) either in sterically relevant clusters or in combination. Molecular size values are shown in kDa.

His-tagged KDR was immobilized on Ni-NTR beads and incubated with full-length VEGF165 wild-type (WT) vs. VEGF165 mutant (mut) protein in the binding buffer (10 mM Tris, 0.15 M NaCl, pH 7.5) for 2 h at 4°C before elution and analysis of the retained protein by SDS-PAGE. Both wild-type and mutant VEGF165 bind to KDR.

Human umbilical vein endothelial cells (HUVECs) were starved for 4 h in M200 basal medium without low serum growth supplement and then treated with VEGF165 wild-type or mutant protein (10 ng/mL) for 10 min before lysis and Western blot analysis. Cells were lysed in lysis buffer 20 mM HEPES (pH 7.4), 100 mM NaCl, 10% glycerol, 1% Nonidet P-40, 1 mM MgCl₂, 1 mM PMSF, 20 mM NaF, 1 mM Na₃VO₄, and protease inhibitor mixture (Sigma-Aldrich). Cell lysates were analyzed by Western blotting using specific antibodies directed to ERK1/2 (p44/42) or phospho ERK1/2 p44/42 (Thr202/Tyr204). Bound IgG was detected using the HRP-conjugated second antibody and SuperSignal West Pico (Thermo Fisher Scientific). Images were evaluated using a Fuji LAS 4000 mini luminescent image analyzer and Multi Gauge v.3.0 software (Fujifilm, Tokyo, Japan).

Starved HUVECs, as detailed above, were treated with VEGF165 wild-type vs. mutant (10 ng/mL) for 10 min before lysis and Western blot analysis of β3 integrin subunit phosphorylation using anti-phospho β3 antibodies.

Starved HUVECs were treated for Western blot analysis. We stimulated the starved endothelial cells with WT and/or mutant VEGF165 for 10–20 min. We solubilized cells in lysis buffer (20 mM HEPES (pH 7.4), 100 mM NaCl, 10% glycerol, 1% Nonidet P-40, 1 mM MgCl₂, 1 mM PMSF, 20 mM NaF, 1 mM Na₃VO₄, and protease inhibitor mixture (Sigma-Aldrich)). We analyzed cell lysates by Western blotting using specific antibodies. Bound IgG was detected using HRP-conjugated second antibody and SuperSignal West Pico (Thermo Fisher Scientific). We analyzed the images using a Fuji LAS 4000 mini luminescence image analyzer and Multi Gauge v.3.0 software (Fujifilm, Tokyo, Japan).

Results are reported as the mean ± standard error of the mean and calculations performed using Prism 10, GraphPad. Where indicated, statistical analysis was performed using ANOVA.

It has been proposed that extracellular matrix-bound VEGF165 interacts with integrin αvβ3 expressed by endothelial cells, and evidence indicates that VEGF121, which lacks the HBD, does not have this property (Hutchings et al., 2003). This implies that the HBD of VEGF165 likely contains a binding site for αvβ3. However, the specifics of an HBD–αvβ3 interaction and the possible cellular consequences are unclear. In this study, we directly addressed the question of whether integrin αvβ3 recognizes and binds the VEGF165 HBD, and if so, whether this interaction can initiate αvβ3 activation at the molecular level and trigger ERK1/2 activation in endothelial cells as a first indication of endothelial cell responses known to be involved in VEGF165-stimulated functions. We found that soluble αvβ3 bound to immobilized VEGF165-HBD protein in a dose-dependent manner but not to the VEGF165 N-terminal domain (NTD) in an ELISA-type binding assay (Figures 1A,B). This binding was suppressed by heat treatment of the HBD, suggesting that the interaction requires native folding of this domain. αvβ3-HBD binding was inhibited by anti-integrin β3 function blocking mAb 7E3 (Figure 1C), whose epitope has been mapped to the classical ligand-binding site within the β3 integrin subunit (Puzon-McLaughlin et al., 2000; Artoni et al., 2004). The inhibitory effects of integrin antagonists suggest that the VEGF165 HBD binding to αvβ3 is specific. Furthermore, we found that binding of soluble αvβ3 to the VEGF165 HBD is cation-dependent (Mn²⁺>Mg²⁺ = Ca²⁺>EDTA) (Figure 1D) in a manner consistent with the reported αvβ3 ligand binding. In agreement with this observation, soluble αvβ3 is known to require 1 mM Mn²⁺ for full activation, as we previously documented (Mori et al., 2008). Moreover, we previously showed that several cytokines directly bind to soluble integrins, prominently including αvβ3 and its interaction with FGF1 and IGF1 (Mori et al., 2008; Saegusa et al., 2009). Thus, the present findings on the αvβ3-VEGF165 HBD interaction support the concept of a key role of integrin αvβ3 interaction with growth factors that have so far been thought to exert their biological properties primarily, if not exclusively, via their known cognate receptors (de Laat et al., 1999). Supporting this concept, our results documenting the cation requirement of VEGF165 HBD–αvβ3 binding and its inhibition by β3 function blocking mab 7E3, as well as the abrogation of this molecular interaction by heat denaturation of the HBD, measured by ELISA-type binding assays to investigate the properties of the purified binding partners reported here, are in line with our previous findings that the FGF1 and IGF1 interaction with integrin αvβ3 can modulate cell signaling responses to these growth factors. Using surface plasmon resonance analysis as an orthogonal approach to study the HBD-αvβ3 interaction and determine the binding affinity, K_D was calculated as 4.7 × 10⁻⁷ M (Figure 1E). Collectively, our results suggest that the VEGF165 HBD is a specific ligand for integrin αvβ3 (Figure 1F). These findings predict that VEGF165 binds to αvβ3 through the HBD and to KDR via the N-terminal domains (NTD), likely resulting in an integrin–VEGF165–KDR ternary complex (Figure 1F).

Although αvβ3 binds to extracellular matrix (ECM) ligands such as fibronectin and vitronectin, it is not straightforward to compare the αvβ3–VEGF165 HBD interaction to αvβ3–ECM interactions since αvβ3–ECM interactivity has mostly been measured based on cell adhesion assays. Cell adhesion assays measure cell interactions with their substrates mediated by multiple integrins and multiple ECM ligands, potentially including additional non-matrix molecules such as growth factors that bind to ECM, all of which may contribute to the interaction and cell adhesion process (Janus-Bell and Mangin, 2023; Sleeboom et al., 2024). Our present study, by design, directly addresses the molecular interaction between αvβ3 as a single integrin and the HBD of VEGF165 as a specific region within a single ligand whose effects on cell behavior and vascular responses have so far been attributed to VEGF165 recognition by its cognate receptor KDR. The present findings demonstrate that the VEGF165 HBD specifically binds to integrin αvβ3.

To predict which amino acid residues are involved in αvβ3 binding, we performed docking simulations between the HBD (2VGH.pdb) and αvβ3 (1L5G.pdb, with open headpiece) using AutoDock 3 (Figure 2A). We performed 50 cycles of docking simulation, and all docking poses were clustered (<0.5 root-mean-square deviation (RMSD)). Seventeen (34%) of the docking poses were in the first cluster. The pose in the first cluster with the lowest docking energy (−24.6 kcal/mol) was selected for further analysis. The simulation predicted that the HBD binds to αvβ3 at high affinity. Amino acid residues predicted to be involved in the HBD–αvβ3 interaction are Arg123, Arg124, Lys125, Lys140, Arg145, and Arg149 (Table 1). We selected these Arg and Lys residues of the VEGF165 HBD at the predicted binding interface for mutagenesis (Figure 2B). Binding of soluble αvβ3 to immobilized HBD wild-type vs. mutant protein was measured by ELISA-type binding assays. The Arg123/Arg124/Lys125 to Ala (R123A/R124A/K125A) mutation and the Lys140/Arg145/Arg149 to Ala mutation (K140A/R145A/R149A) partially suppressed the binding of αvβ3 to wild-type HBD (Figure 2C). The combined R123A/R124A/K125A/K140A/R145A/R149A mutation nearly completely blocked integrin binding (Figures 2D, E). These results indicate that these amino acid residues are critical for VEGF165 HBD binding to αvβ3, and they are consistent with the docking model.

The integrin αvβ3 structure (1L5G.pdb) contains eight Mn²⁺ cations. In our initial docking simulations, all these cations were removed from the integrin structure (Figure 2A). Since cations are critical for integrin binding, it is possible that an integrin structure without cations may not reflect ligand binding in biological conditions. We thus performed docking simulation using the αvβ3 integrin in which cations were not removed. We obtained an identical docking model (−22.5 kcal/mol) in the first cluster (Supplementary Figure S1). This indicates that the presence or absence of cations did not affect the prediction by docking simulation of the integrin–HBD interaction.

It has been proposed that integrin αvβ3 negatively regulates VEGF165 signaling since β3 KO mice had elevated levels of angiogenesis and tumorigenesis (Reynolds et al., 2002; Reynolds et al., 2005). Considering these results from β3 KO mouse models, the here proposed ternary complex between αvβ3, VEGF165, and KDR D2D3 may not be required for maximal mitogenicity of VEGF165. Instead, the presence of αvβ3 bound to the VEGF165 HBD might rather act as a negative modulator of VEGF165–KDR binding-induced mitogenic signaling. We, therefore, hypothesized that another protein might bind to the VEGF165 HBD and potentially sterically compete with αvβ3 to regulate the intensity of VEGF165-KDR induced mitogenicity. Since removal of the VEGF165 HBD markedly reduced KDR binding to VEGF165, as previously shown (Keyt et al., 1996) and consolidated in our binding experiments, we hypothesized that the VEGF165 HBD might also be recognized by KDR itself but involves a KDR domain distinct from its D2D3, which engage the VEGF165 NTD. The VEGF165 HBD is strongly positively charged (pI = 11). We, therefore, anticipated that KDR D1 might interact with the VEGF165 HBD as the KDR D1 domain is strongly negatively charged. To test this hypothesis, we performed binding experiments using orthogonal approaches. Applying the ELISA-type binding assay, we demonstrated that the VEGF165 HBD bound to KDR D1 (Figure 3A). This interaction was confirmed by surface plasmon resonance (SPR), which further added information on the binding dynamics (Figure 3B). Thus, we conclude that the VEGF165 HBD interacts with the KDR D1 region. These results indicate that KDR D1 can interact with the VEGF165 HBD and thus ligate VEGF165 via a previously unrecognized binding site.

Since our results identified a novel, potentially regulatory mechanism through which integrin αvβ3 binding to the VEGF165 HBD impacts cell signaling induced by VEGF165 binding to KDR, we explored the effect of VEGF165 HBD mutations on the newly discovered KDR D1 interaction. We studied the effect of mutations on the binding of KDR-D1 to the VEGF165 HBD. We were not able to use docking simulation between the HBD and the KDR D1 since KDR D1 is not folded and the 3D structure of KDR D1 is not available. We found that the combined mutations within the VEGF165 HBD blocked KDR D1 binding to the HBD of this ligand (Figure 3C). As mutations of these amino acids within the VEGF165 HBD also abrogated the integrin αvβ3 interaction with this ligand (Figure 3C), these findings indicate that the interaction sites of αvβ3 and KDR D1 within the VEGF165 HBD overlap (Figure 3D). To further locate the KDR D1 binding site within the VEGF165 HBD, we introduced point mutations of positively charged amino acid residues to Glu (charge reversal mutations) within the HBD to disturb the proposed charge-supported interaction between the VEGF165 HBD and KDR D1. The R123E/R124E/K125E mutation strongly suppressed the binding of HBD to KDR D1 (Figures 4A and B), and mutations of individual residues, R123E and R124E, but not K125E (Figure 4C), suppressed the binding. These results suggest that Arg123 and Arg124 are critical for KDR D1 binding and that αvβ3- and KDR D1-binding sites overlap in HBD (Table 1). Therefore, it is probable that αvβ3 and KDR D1 compete for binding to the HBD (Figure 4D) and that αvβ3 can interfere with KDR D1 binding to the VEGF165 HBD and, thereby, blunt VEGF165 signaling induced by the KDR–VEGF165 interaction. This concept is consistent with the observation that VEGF signaling is enhanced in β3 KO mice that lack integrin αvβ3 as a competitive modulator of VEGF165–KDR-driven signaling events. αvβ3 suppresses KDR D1 binding to the HBD. This is consistent with the observation that VEGF signaling is enhanced in β3 KO mice.

Full-length VEGF165 protein with the combined HBD R123A/R124A/K125A/K140A/R145A/R149A mutations (referred to as VEGF165-HBD mutant) (Figure 5A) was synthesized in E. coli as an insoluble protein, purified under denaturation, re-folded, and further purified by FPLC gel filtration. The protein migrated as a single band with the expected size (Figure 5A). The protein bound to immobilized KDR in pull-down assays (Figure 5B), indicating that KDR binding specificity was retained by the VEGF165-HBD mutant protein. However, importantly, the VEGF165-HBD mutant lacked integrin binding αvβ3, as expected. Used as a soluble ligand for human endothelial cell (HUVEC) cultures, VEGF165-HBD mutant protein failed to induce ERK1/2 activation and integrin β3 phosphorylation in these cells (Figures 5C–E). Furthermore, VEGF165-HBD mutant protein also failed to activate KDR Y-1175 phosphorylation in HUVECs (Figure 5F) despite the ability of the mutant protein to bind to KDR (Figure 5B). Notably, VEGF165-HBD mutant protein failed to induce proliferation of HUVECs, which is consistent with the defective signaling functions of the mutant protein (Figure 5G). In the case of VEGF165 signaling, integrin αvβ3 appears to negatively regulate VEGF165 signaling by competing with the cognate receptor KDR for VEGF165 binding. Intriguingly, our results reveal a novel interaction of KDR with VEGF165 based on a competing mechanism mediated through KDR D1 binding to the ligand’s HBD, where the interaction overlaps with integrin αvβ3 binding. VEGF165 is probably the first example where integrin αvβ3 negatively regulates signaling induced by the growth factor interaction with a cognate cell-surface receptor.