Characterization of a Bicyclic Peptide Neuropilin-1 (NP-1) Antagonist (EG3287) Reveals Importance of Vascular Endothelial Growth Factor Exon 8 for NP-1 Binding and Role of NP-1 in KDR Signaling*

Neuropilin-1 (NP-1) is a receptor for vascular endothelial growth factor-A165 (VEGF-A165) in endothelial cells. To define the role of NP-1 in the biological functions of VEGF, we developed a specific peptide antagonist of VEGF binding to NP-1 based on the NP-1 binding site located in the exon 7- and 8-encoded VEGF-A165 domain. The bicyclic peptide, EG3287, potently (Ki 1.2μm) and effectively (>95% inhibition at 100 μm) inhibited VEGF-A165 binding to porcine aortic endothelial cells expressing NP-1 (PAE/NP-1) and breast carcinoma cells expressing only NP-1 receptors for VEGF-A, but had no effect on binding to PAE/KDR or PAE/Flt-1. Molecular dynamics calculations, a nuclear magnetic resonance structure of EG3287, and determination of stability in media, indicated that it constitutes a stable subdomain very similar to the corresponding region of native VEGF-A165. The C terminus encoded by exon 8 and the three-dimensional structure were both critical for EG3287 inhibition of NP-1 binding, whereas modifications at the N terminus had little effect. Although EG3287 had no direct effect on VEGF-A165 binding to KDR receptors, it inhibited cross-linking of VEGF-A165 to KDR in human umbilical vein endothelial cells co-expressing NP-1, and inhibited stimulation of KDR and PLC-γ tyrosine phosphorylation, activation of ERKs1/2 and prostanoid production. These findings characterize the first specific antagonist of VEGF-A165 binding to NP-1 and demonstrate that NP-1 is essential for optimum KDR activation and intracellular signaling. The results also identify a key role for the C-terminal exon 8 domain in VEGF-A165 binding to NP-1.


Vascular endothelial growth factor A (VEGF-A)
is an essential mediator of vasculogenesis and angiogenesis during embryonic development and plays a central role in pathophysiological neovascularization in human disease (1,2). Five isoforms of VEGF-A have been identified, VEGF-A 121 , VEGF-A 145 , VEGF-A 165 , VEGF-A 189 , and VEGF-A 206 , generated by alternative mRNA splicing of a single VEGF gene transcript containing 8 exons (3)(4)(5), of which VEGF-A 165 is the most abundant and biologically active.
VEGF-A 165 exerts its biological effects through high affinity binding to two tyrosine kinase receptors, Flt-1 (VEGFR1) and KDR (VEGFR2), which are expressed in most vascular endothelial cells (6). KDR binds VEGF-A 165 with lower affinity than Flt-1, and is also recognized by VEGF-C, VEGF-D, and VEGF-E (7-11), whereas Flt-1 is also a receptor for PlGF and VEGF-B (12,13). After binding and activation of KDR, VEGF-A 165 stimulates ERK activation and an array of other early signaling events followed by short and long term cellular biological effects including production of prostacyclin (PGI 2 ) and nitric oxide, increased cell survival, cell migration, proliferation, and angiogenesis (14 -22). The function of Flt-1 in the endothelium is unclear, but it is thought to regulate the activity of VEGF-A partly by acting as a decoy receptor, and in part through direct regulatory effects on KDR (23). Neuropilin-1 (NP-1) has recently been identified as a non-tyrosine kinase receptor for VEGF-A 165 , the heparin-binding PlGF-2 isoform, VEGF-B, and VEGF-E (24 -26). NP-1 was first identified as a receptor for semaphorin 3A (Sema3A), a member of a family of proteins involved in axonal guidance (27,28), and is expressed in endothelial cells, several tumor cell types, and in certain classes of neuron including cells of the dorsal root ganglion (DRG), olfactory, and optic nerves (24,29). NP-2 has a similar domain structure to NP-1 with 44% amino acid identity, and exhibits a distinct expression pattern in the developing nervous system (28,30). Sema3A induces neuronal growth cone collapse specifically through NP-1, whereas sema 3B, 3C, 3E, and 3F recognize both neuropilins, acting as NP-1 antagonists and NP-2 agonists (30 -33). A growing body of evidence also indicates a role for neuropilins in angiogenesis. Overexpression of NP-1 in mice results in increased capillary formation, vasodilatation, and malformation of the heart (34), whereas mice deficient in NP-1 exhibit defects in embryonic axonal patterning and an array of vascular abnormalities including defective development of large vessels and impaired neural and yolk sac vascularization (35). Inactivation of both NP-1 and NP-2 causes a more severe failure of embryonic vascularization resulting in death at E8.5 (36). Despite the strong evidence that NP-1 is essential for normal vascular development, the underlying mechanisms remain obscure, and the role of NP-1 in the biological functions of VEGF-A are not fully understood.
We recently reported that a specific bicyclic peptide based on the C-terminal NP binding domain of VEGF-A 165 (37) is an antagonist of VEGF binding to NP-1 and inhibits the anti-chemorepulsive effect of VEGF-A 165 in DRG neuronal explants (38). Here, we have identified the key features of EG3287 responsible for its antagonistic properties through a detailed structure-function analysis, and investigated its biological effects in vascular endothelial cells. Our findings show that the C-terminal six amino acid domain encoded by exon 8 plays a crucial role in VEGF-A 165 binding to NP-1. Evaluation of EG3287 in endothelial cells demonstrates that VEGF-A 165 binding to NP-1 is required for stable binding to KDR, full activation of KDR and downstream signaling and biological responses. This antagonist should be a valuable tool for probing the biological role of NP-1 in diverse cell types, and will also be useful for designing improved neuropilin antagonists.
Linear peptides were synthesized by an automated multiple solid phase approach using the Fmoc-Arg(Pbf)-p-alkoxybenzyl alcohol resin (0.59 mmol/g loading) or Fmoc-Rink Amide MBHA resin (0.59 mmol/g or 0.68 mmol/g loading). Amino acids were attached by Fmoc strategy on a 25-or 50-mol scale with a basis coupling time of 30 min followed by a recoupling step. Each amino acid was sequentially coupled to the growing peptide chain from the C to the N termini applying benzotriazol-1-yloxy-Tris-pyrrolidino-phosphonium hexafluoro-phosphate and N-methyl morpholine as coupling reagents via the active ester method. Removal of the N-Fmoc protecting group was carried out with 20% piperidine in DMF followed by sequential washes with DMF and DCM. The coupling reagent, Pybop, NMM, and all amino acid derivatives were dissolved in DMF (0.7 M, 4-fold excess). All solvents used were of HPLC grade quality.
The peptides were cleaved from the resin with simultaneous deprotection using 90% trifluoroacetic acid at room temperature for 3 h, either in the presence of 5% thioanisole, 2.5% water, and 2.5% ethanedithiol or with 95% trifluoroacetic acid, 2.5% water, and 2.5% triisopropylsilane. The cleavage mixture was filtered and precipitated in ice-cold methyl tert-butyl ether. The remaining resin was washed once with the cleavage reagent, filtered, and combined with the previous fractions. The precipitates were collected after centrifugation, washed three times with ice-cold methyl tert-butyl ether, and allowed to dry overnight at room temperature. The crude peptides were dissolved in 15% aqueous acetic acid and lyophilized for 2 days (40°C, 6 mbar).
General Method for Bicyclic Peptide Synthesis-The crude linear precursor, prepared as above, was dissolved in the minimum trifluoroacetic acid and diluted to 2 liter/0.25 mmol with water. The first disulfide bridge was formed between unprotected Cys residues using K 3 Fe(CN) 6. The peptide solution was adjusted to pH 7.5 with aqueous ammonium hydroxide. To this solution, 0.01 M K 3 Fe(CN) 6 was added dropwise to excess, until a slight yellow color remained. The completion of the reaction was confirmed by HPLC sampling after acidification. The pH of the solution was adjusted to 4 using 50% aqueous acetic acid. The crude reaction mixture was stirred with Bio-Rex 70 weak cation-exchange resin (Bio-Rad) overnight and packed into a glass column. After thorough washing with water, the peptide was eluted using 50% aqueous acetic acid and detected by TLC using ninhydrin. Ninhydrin-positive fractions were pooled and lyophilized. Crude material was purified via reverse-phase HPLC (Gilson) using a preparative C-8 column (see conditions for linear peptides). The purified fractions were collected, combined, and lyophilized. The second disulfide bridge was formed via I 2 -oxidation between Cys(Acm) protected residues. A solution of the peptide (5 mg/ml) in 10% aqueous trifluoroacetic acid was mixed vigorously with 8 equivalents of iodine, and the resulting suspension stirred for 1.5 h.
The progress of the cyclization reactions was monitored by analytical reverse-phase LC-MS. At completion of the reaction the excess iodine was quenched using 1 M ascorbic acid. The reaction mixture was diluted ϫ2 with 0.1% trifluoroacetic acid/water, filtered through a 0.45-m disposable filter and purified directly via preparative reverse-phase HPLC (Gilson) using a preparative C-8 column (see conditions for linear peptides). The relevant fractions were collected, evaporated, lyophilized, and stored at 4°C. Confirmation of the structure was performed by analytical reverse-phase LC-MS and/or MALDI mass spectroscopy and amino acid analysis.
Alternative Cleavage Procedure for Bicyclic Peptides-The peptides were cleaved from the resin with simultaneous deprotection using 82.5% trifluoroacetic acid at room temperature for 3 h in the presence of 5% thioanisole, 5% water, 2.5% ethanedithiol, and 6% (w/v) phenol. In the case of Cys-containing peptides, the cleavage mixture was filtered and precipitated in ice-cold diethyl ether. The remaining resin was washed once with trifluoroacetic acid, filtered, and combined with the previous fractions. The precipitates were stored at 4°C overnight and were collected by filtration, washed with ice-cold diethyl ether, and allowed to dry at room temperature. For non-Cys-containing peptides the cleavage mixture was filtered, the resin washed with trifluoroacetic acid then dichloromethane, and the filtrates combined and concentrated under vacuum to oil. The peptide was precipitated in ice-cold diethyl ether, collected by filtration and allowed to dry at room temperature. The crude peptides were dissolved in trifluoroacetic acid/acetonitrile/water and lyophilized overnight (Ϫ50°C, 6 mbar). Crude peptides were characterized by reverse-phase HPLC (Gilson) using an analytical C-18 column (Vydac 218TP54, 250 ϫ 4.6 mm, 5-m particle size, and 300 Å pore size) and a linear AB gradient of 0 -100% for B over 40 min at a flow rate of 1 ml/min, where eluent A was 0.1% trifluoroacetic acid /water and eluent B was 0.1% trifluoroacetic acid in 60% CH 3 CN/water. Mass was confirmed using MALDI-MS, and Ellman's color test confirmed the presence of free sulfhydryl groups where applicable.
NMR Spectroscopy and Three-dimensional Structure Determination-For NMR studies, 15 mg of EG3287 (Ser 138 -Arg 165 , VEGF-A 165 ) was prepared in 90% H 2 O, 10% D 2 O, with the resulting solution at pH 2.3. NMR spectra were acquired at 293 K and 298 K on a Varian INOVA spectrometer (operating at nominal 1 H frequency of 600 MHz) equipped with triple resonance probe including Z-axis pulse field gradients. All two-dimensional data were recorded in phase-sensitive mode using the States-TPPI method for quadrature detection, and data were apodized using shifted cosine-bell squared functions, followed by zero-filling once in each dimension. Water suppression was achieved using WATERGATE (39). Sequence-specific resonance assignments were obtained from two-dimensional 1 H TOCSY (80-ms mixing time), 1 H NOESY (100-ms mixing time), [ 1 H, 15 N] HSQC, and [ 1 H, 13 C] HSQC spectra (at natural isotopic abundance). All spectra were processed using NMRpipe/NMRDraw (40) and analyzed using ANSIG for openGL v1.0.3 (41). 1 H, 13 C, and 15 N chemical shifts were referenced indirectly to DSS, using absolute frequency ratios (42). Interproton distance restraints were derived from two-dimensional 1 H NOESY spectra with a mixing time of 250 ms. Cross-peaks were grouped into four categories according to their relative peak intensities: strong, medium, weak, and very weak, and were designated with interproton distance restraint bounds of 1.8 -2.5 Å, 1.8 -3.0 Å, 1.8 -3.5 Å, and 1.8 -4.5 Å, respectively. 0.5 Å was added for distances that involved methyl groups. The structure calculations were carried out using the PARALLH-DGv5.1 parameter, with the PROLSQ non-bonded energy function (43) within the CNS program (44), modified to allow floating stereochemistry, and to include active swapping, of prochiral centers (45). 229 interproton distance restraints (138 intraresidue, 56 sequential, 9 short range, 17 long range, and 9 ambiguous) were applied in restrained molecular dynamics simulated-annealing calculations.
Measurement of EG3287 Stability-EG3287 was recovered from tissue culture medium by a 1:1 addition of 20% trifluoroacetic acid and centrifugation of the precipitate. This method resulted in ϳ90% recovery. Quantitation of EG3287 was performed by Inpharmatica Ltd (Cambridge, UK) using a liquid chromatography mass spectrometric (LC-MS/MS) assay on a Micromass Quattro Micro.
Computational Chemistry-Molecules were analyzed using SYBYL 7.0 (Tripos Inc., St. Louis, MO). The Biopolymer tools in Sybyl were used to build some of the peptides, check structures for errors, correct atom types, and add/remove hydrogens as necessary. For energy minimizations or molecular dynamics simulations we used starting conformations as described by Fairbrother et al. (46) (VEGF-A55, Protein Data Bank code: 2vgh, Ala 111 -Arg 155 , VEGF-A 165 ) or EG3287 VEGF-A55 (Protein Data Bank code: 1KMX, Ser 138 -Arg 155 VEGF-A 165 ). The terminal amino acids and polar side chains were charged, and the peptides were solvated in a box of water molecules (Tripos explicit box solvation algorithm). Electrostatic charges were calculated with the method of Gasteiger and Marsili (47). Structures were minimized using the Tripos force field with a steepest descent gradient of 100 iterations followed by a conjugate gradient of 0.01 kcal/mol or a maximum of 10,000K iterations as termination criteria. During minimizations, a 12-Å non-bonded cut-off was applied, and when solvent was not present and ionic strength was set to 4.00.
For molecular dynamics simulations, all starting structures were solvated in a cubic box of explicit water molecules (edge 40 Å) using the XFIT solvation algorithm. This led to approximately ϳ1700 water molecules depending on the solute starting conformation. All simulations were performed using periodic boundary conditions and the SHAKE algorithm. Constant temperature simulations (NTV, 300°K) were run at a time step of 0.1 ps for a total period of 2000 ps. The Lennard-Jones interactions were evaluated with an 8.0-Å cutoff value. The non-bonded pair list was updated every 100 steps. Generated structures were stored in trajectory files every 5 ps, providing 400 conformers for each run. The collected structural data were analyzed with the graphic tools of the SYBYL 7.0. All calculations were performed on the dual Hewlett-Packard work station XW6000, 2 ϫ 2.8 GHz CPUs running REDHAT Enterprise Linux WS4.
Radiolabeled Ligand Binding-Confluent cells in 24-well plates were washed twice with phosphate-buffered saline. At 4°C various concentrations of peptides diluted in binding medium (Dulbecco's modified Eagle's medium, 25 mM HEPES pH 7.3 containing 0.1% bovine serum albumin) were added, followed by addition of the indicated concentration of 125 I-VEGF-A 165 (1200 -1800 Ci/mmol, GE Healthcare Plc) or 125 I-EGF. After 2 h of incubation at 4°C (or the indicated time at 37°C), the medium was aspirated, and washed four times with cold phosphate-buffered saline. The cells were lysed with 0.25 M NaOH, 0.5% SDS solution, and the bound radioactivity of the lysates was measured. Nonspecific binding was determined in the presence of 100-fold excess unlabeled VEGF-A 165 or EGF (R & D Systems). Equilibrium dissociation constants (K i ) for peptides were calculated from IC 50 values and the K d of VEGF-A 165 for NP-1 (0.3 nM) from Soker et al. (24), using the formula Cross-linking-Confluent cells were bound with 125 I-VEGF-A 165 at 4°C for 2 h as described above and then washed three times with phosphatebuffered saline. The bound 125 I-VEGF-A 165 was cross-linked to the cells by incubation with 1.5 mM disuccinimidyl suberate (DSS) for 20 min at room temperature. After three washes with phosphate-buffered saline at 4°C, the cells were solubilized in lysis buffer (64 mM Tris-HCl, pH 6.8, 0.2 mM Na 3 VO 4 , 2% SDS, 10% glycerol, 0.1 mM AEBSF, 5 g/ml leupeptin) and scraped off the plates. After centrifugation at 16,000 ϫ g for 20 min at 4°C, cross-linked 125 I-VEGF-A 165 -receptor complexes were subjected to 7.5% SDS-PAGE. Gels were dried and exposed to x-ray film.
Apoptosis-Subconfluent HUVECs in 6-well plates were washed twice with serum-free M199 medium (Invitrogen), and incubated with the indicated additions for 24 h. The cells were then trypsinized, collected by centrifugation, and stained with fluorescein-conjugated annexin V and propidium iodide (Roche Applied Science). After staining, the cells were analyzed by flow cytometry using a FACScan (Becton Dickinson). Annexin V-positive staining cells were counted as apoptotic cells. Propidium iodide-positive and annexin V-negative cells were not counted as apoptotic cells.
Cell Proliferation-HUVECs were seeded at a density of 1 ϫ 10 4 cells per well in 24-well plates. 24 h after plating, the medium was replaced with fresh EBM containing 0.5% FBS and 25 ng/ml VEGF-A 165 in the absence or presence of EG3287. After 3 days, the cell numbers were determined using a Sysmex CDA-500 cell counter.
Data Analysis-Data were analyzed using Prism (version 3.0) statistical packages. Differences in prostanoid production among four treatment groups were evaluated using the one-way analysis of variance with Bonferroni's multiple comparison tests. Differences in apoptotic frequencies between two groups were assessed by the chi-squared test. A value of p Ͻ 0.05 was taken as statistically significant.

Design and Synthesis of Bicyclic Peptides Mimicking the VEGF-A 111-165
Region-The NMR-derived three-dimensional NMR solution structure of the VEGF-A 111-165 region comprises two distinct protein subdomains, VEGF-A 111-137 and VEGF-A 138 -165 , each constrained and stabilized by two disulfide bonds and a network of hydrogen bonds (46). The C-terminal subdomain (Ser 138 -Arg 165 ) features a typical ␤-hairpin turn combined with a short (2-step) ␣-helix. Four positively charged amino acids are present in the helical region and three in the C-terminal exon 8 peptide. To elucidate the contribution of each subdomain to the binding and biological activity of VEGF-A 165 mediated through neuropilin, we evaluated the biological effects of peptides corresponding to the subdomains in endothelial cells.
The formation of one or two specific disulfide bonds in peptides corresponding to each subdomain was achieved by pairwise selective deprotection of cysteine residues. The first disulfide bridge was formed by oxidation using either Me 2 SO (48), or potassium ferricyanide (III) (49), the second by a simultaneous Acm deprotection/oxidation step mediated by iodine, under acidic conditions (Fig. 1). The route was optimized such that only a single purification step, using reverse-phase HPLC-MS, was required to isolate the final compound.
Structure and Stability of EG3287-The structure of each subdomain in solution was determined by a series of molecular dynamic simulations. The root mean-squared values of the C␣ chain for the two subdomains, between the average molecular dynamics structure and the starting conformation, were 3.5 and 0.45 Å for the N-and C-terminal subdomains, respectively. These values suggested that the C-terminal domain would be more likely to retain its structure in solution than the N-terminal domain (Fig. 2, A and B).
The resolved solution NMR structure for the C-terminal subdomain, described as EG3287, is shown in Fig. 2, C and D. The structure showed strong similarity both to the solution NMR structure of a recombinant protein comprising the C-terminal 55 amino acid residues of VEGF-A 165 (46,50), and to the molecular dynamics simulations for this region derived from the NMR structure (Fig. 2B). A 2.0-Å root mean-squared deviation value between the backbone carbons of EG3287 and the corresponding C-terminal domain of VEGF-A 111-165 indicates the close similarity of the three-dimensional polypeptide fold in the excised subdomain and correlates well with the molecular dynamics simulations.
A liquid chromatography mass spectrometric (LC-MS/MS) assay demonstrated no significant degradation when the peptide was incubated at 37°C for up to 24 h in water, EBM, or EBM supplemented with 10% fetal bovine serum, 10 ng/ml human epidermal growth factor, 12 g/ml bovine brain extract, 50 g/ml gentamicin sulfate, and 50 ng/ml FIGURE 1. Synthesis of EG3287. The peptide chain was synthesized on Wang resin using Fmoc chemistry. Connectivity of the disulfides was achieved using an unambiguous cyclization protocol. The first cyclization was achieved using Fe(CN) 6. Simultaneous deprotection of the Acmprotected Cys residues and cyclization using iodine provided, after HPLC purification, the bicyclic peptide EG3287. amphotericin-8. Measurement of EG3287 after incubation with HUVECs at 37°C indicated that whereas EG3287 was also largely unaffected after 24 h in the presence of serum-free EBM medium, it gradually decreased, with a half-life of ϳ39 h.
Structure-Function Analysis of the Requirements for VEGF-A 165 Binding to Neuropilin-Initially the two bicyclic peptides corresponding to the subdomains VEGF-A 111-137 (peptide 1) and VEGF-A 138 -165 (EG3287) were examined for binding to PAE/NP-1 cells. Only EG3287 exhibited significant activity (Table 1), localizing the main binding motifs to the C-terminal subdomain. EG3287 selectively inhibited 125 I-VEGF-A 165 binding to PAE cells expressing NP-1 with an IC 50 of 2.8 M (K i 1.2 M) and ϳ95% inhibition at 100 M, but had little effect on 125 I-VEGF-A 165 binding to PAE/KDR or PAE/Flt-1 cells (Table 1 and Fig. 3A). EG3287 also inhibited 125 I-VEGF-A 165 binding to MDA-MB-231 breast carcinoma cells, naturally expressing NP-1 but not KDR or Flt-1, with a very similar IC 50 of 3 M and complete inhibition of specific binding at 30 M (Fig. 3A). In HUVECs, naturally expressing KDR, Flt-1 and NP-1 receptors, EG3287 inhibited 125 I-VEGF-A 165 binding with an IC 50 of 20 M, and caused a maximum 72% inhibition at 100 M. When radiolabeled ligand binding experiments were performed at 37°C for different times, EG3287 also effectively inhibited 125 I-VEGF-A 165 binding to PAE/NP-1 cells and HUVECs, but had no effect on binding to PAE/KDR cells (Fig. 3B). EG3287 did not affect binding of 125 I-EGF to high affinity binding sites in MDA-MB-231 breast carcinoma cells expressing both NP-1 and EGFR, and EGF had little effect on 125 I-VEGF-A 165 binding to PAE/NP-1 at concentrations up to 1 g/ml (Fig.  3C), indicating that EGF does not compete with VEGF-A 165 for binding to NP-1.
Prevention of cyclization by substituting all four cysteines with the cysteine isostere aminobutyric acid (peptide 2), reduced inhibition of 125 I-VEGF-A 165 binding to PAE/NP-1 cells as compared with EG3287 (Table 1). To examine whether the reduced activity of the non-cyclized peptide was caused by loss of the necessary three-dimensional structure, we investigated the effect of altering the connectivity of the disulfide bonds by testing the two possible isomers with disulfide bonds between Cys 139 -Cys 146 and Cys 158 -Cys 160 (peptide 3) and between Cys 139 -Cys 160 and Cys 146 -Cys 158 (peptide 4). Peptide 4 showed a marked reduction in potency and affinity whereas the other isomer, peptide 3 displayed modestly reduced activity compared with EG3287 with a K i 3.8 M and 94% reduction in binding ( Table 1).
The basis for these differences was addressed by molecular dynamics simulations utilizing models of the two isomeric peptides. For peptide 4 there are two possible S-S bond networks and both analogues were constructed; however, one was of markedly higher energy than the other and was rejected. Molecular dynamics simulations for both structures in water stabilized fairly rapidly and the predicted structures are shown in Fig. 4. Peptide 4 adopted a stable conformation, but the ␣-helical region was distorted and adopted the shape of a poorly defined turn. In contrast, peptide 3 retained the shape of the ␣-helical region and the overall fold much better than peptide 4 (Fig. 4). This result is in accordance with the greater potency of peptide 3 and suggests a potentially important role for the ␣-helical region in binding.
We further investigated the conformational role of the ␣-helix by introduction of the ␣-aminoisobutyric acid (Aib) residue (peptide 5), which is widely used to stabilize helices (51,52). This had no effect on the binding affinity perhaps indicating that the helix was already stabilized by the constrained disulfide bond arrangement ( Table 1). The contribution of individual amino acids from the ␣-helix was examined by substitution of alanine for Lys 147 , Arg 149 , and Gln 150 . The K147A (peptide 6) mutation had no effect on binding, whereas the R149A (peptide 7) mutation decreased affinity for NP-1 (K i 6.8 M), indicating a contribution from this residue located toward the C-terminal portion of the ␣-helix (Table 1).

TABLE 1 Effects of peptides derived from VEGF-A 165 exons 7 and 8 on VEGF-A 165 binding to neuropilin
Sequences of exon7/8-derived peptides and positions of disulfide bonds are shown with IC 50 values Ϯ S.E.M. and K i values Ϯ S.E.M. (M) for inhibition of 125 I-VEGF-A 165 binding to PAE/NP-1 cells and (% inhibition at 100 M peptide). The structure of the VEGF-A 165 C-terminal domain is shown above. X denotes aminobutyric acid (isosteric non-cyclizing substitution for cysteine); Z denotes aminoisobutyric acid; residues highlighted in red are mutations (alanine substitutions); Ac denotes amino-terminal acetylation.

Role of the VEGF Exon 8-encoded Domain in Neuropilin Binding-
Whereas acetylation of the N terminus (peptide 9) had little effect on binding, transformation of the C terminus into the carboxamide (peptide 10) drastically reduced binding, indicating a key interaction with neuropilin at this point (Table 1). Surprisingly, an 8-residue linear peptide containing the 6 residues encoded by exon eight (peptide 11) demonstrated 73% inhibition of VEGF-A 165 binding at 100 M ( Table 1). The contribution of the residues located in the region encoded by exon 8 was determined by means of an alanine scan of a 7-residue exon 8 peptide (peptide 12) in which the cysteine was replaced by aminobutyric acid to prevent problems arising from peptide dimerization ( Table 2).  (Table 2). We also examined the effect of a peptide corresponding to the C terminus encoded by an alternative exon 8 found in VEGF-A 165 b (peptide 19), a recently identified isoform that is unable to stimulate endothelial cell proliferation, angiogenesis, or other biological activities (53). The peptide corresponding to the VEGF-A 165 b C terminus, Abu-RSLTRKD, had no effect on VEGF binding to PAE/NP-1 cells ( Table 2).
EG3287 Inhibits Cross-linking of VEGF with KDR and NP-1 in HUVECs-Though EG3287 had no effect on VEGF-A 165 binding to KDR in PAE/KDR cells in the absence of NP-1, we examined whether EG3287-mediated inhibition of VEGF binding to NP-1 in HUVECs could affect the ability of VEGF to bind to KDR. This question was addressed by investigating the effect of EG3287 on covalent cross-   4). B shows the least distortion from the EG3287 structure and has the highest biological activity.

TABLE 2 Effects of alanine substitutions in exon 8 on inhibition VEGF-A 165 binding to NP-1
Abbreviations and other details are as in legend to Table 1.

Characterization of a Neuropilin Antagonist
linking of 125 I-VEGF-A 165 to its receptors in HUVECs. As shown in Fig.  5, EG3287 strongly inhibited specific cross-linking of 125 I-VEGF-A 165 to two major bands of 250 kDa and 160 kDa corresponding to VEGF/ KDR and VEGF/NP-1 complexes, respectively. In parallel experiments, EG3287 had no effect on 125 I-VEGF-A 165 cross-linking to KDR in PAE/KDR cells, but abolished cross-linking to NP-1 in PAE/ NP-1 cells.
EG3287 Inhibits KDR Signaling-It was next determined whether inhibition of VEGF-A 165 binding to NP-1 by EG3287 had any effect on the ability of VEGF-A 165 to activate KDR. KDR phosphorylation at tyrosines 1054 and 1059 has been shown to be essential for kinase activity, whereas phosphorylation at Tyr 1175 is required for tyrosine phosphorylation of PLC-␥ and ERK activation (20,54). EG3287 markedly inhibited phosphorylation of KDR at Tyr 1054/1059 in a concentration-dependent manner with a significant effect at 10 M and almost complete inhibition at 100 M (Fig. 6A). EG3287 also reduced VEGF-A 165 -induced Tyr 1175 phosphorylation with a detectable effect at 30 M and maximum but not complete inhibition at 100 M.
A key early event following KDR activation is PLC-␥ tyrosine phosphorylation and association of PLC-␥ with KDR. EG3287 attenuated the increase in PLC-␥ phosphorylation at tyrosine 783 induced by treatment with VEGF-A 165 for 10 min, but did not cause a complete inhibition (Fig. 6A). Because VEGF-A 165 induces activation of ERKs1/2 via PLC-␥ and subsequent PKC activation, we next examined the effects of EG3287 on ERK activity. EG3287 inhibited VEGF-A 165 -induced activation of ERK1/2 in a concentration-dependent fashion, with a detectable decrease at 10 M, and a maximum effect at 30 -100 M, similar to the concentration dependence of the inhibition of high affinity VEGF-A 165 binding to NP-1 (Fig. 6A). In contrast to the inhibitory effects of the NP antagonist on VEGF-A 165 signaling, EG3287 had no effect on EGFinduced EGFR phosphorylation at tyrosines 845 or 1068 and EGF-induced ERK1/2 activation in MDA-MB-231 breast carcinoma cells, or on FGF-2-induced activation of ERKs 1 and 2 in HUVECs (Fig. 6B).
The PKC-dependent ERK activation pathway is essential for VEGF-A 165 stimulation of cPLA 2 activity leading to increased PGI 2 production (17, 21), a biological response that has been implicated in mediating VEGF-A-induced vascular permeability and vasculoprotective effects if VEGF-A in vivo (6). EG3287 caused a concentration-dependent decrease in VEGF-induced PGI 2 generation measured either 30 min or 2 h after addition of VEGF-A 165 . Inhibition was detectable at 10 M EG3287, with maximum inhibition at 100 M (Fig. 7). EG3287 also caused a similar concentration-dependent inhibition of VEGF-A 165 stimulation of PGE 2 production (Fig. 7).
We also examined the ability of EG3287 to inhibit activation of Akt. EG3287 caused a partial inhibition of VEGF-A 165 -induced Akt activation, and, similar to its effects on ERK activity, also attenuated basal Akt phosphorylation (Fig. 8A). However, treatment with EG3287 caused no increase in cell death and did not decrease the anti-apoptotic effect of VEGF-A 165 after 24 h, as determined by annexin V staining (Fig. 8B). EG3287 also had no significant effect on VEGF stimulation of endothelial cell proliferation (results not shown).

DISCUSSION
The conclusion that EG3287 is a specific inhibitor of binding to NP-1 is supported by the ability of the peptide to specifically compete with radiolabeled VEGF-A 165 for binding to PAE/NP-1 cells, and to MDA-MB-231 breast carcinoma cells endogenously expressing NP-1 but not KDR or Flt-1, with very similar potency, whereas VEGF-A 165 binding to PAE cells expressing either KDR or Flt-1 was unaffected. Furthermore, whereas the peptide completely abolished VEGF-A 165 binding to cells expressing only NP-1 receptors, it caused only a partial inhibition of binding to HUVECs expressing all three receptors, and inhibited binding to HUVECs with significantly reduced potency, consistent with a selective effect on binding to the NP-1 receptor population. In addition, the present study aimed to elucidate the features of EG3287 required for inhibition of VEGF binding to NP-1 and to characterize the effects of EG3287 on VEGF-A 165 receptor activation, intracellular signaling and downstream biological responses in endothelial cells.
The C-terminal 23 residues encoded by exon 7 in VEGF-A 165 were previously identified as the core NP-1 binding domain (37). EG3287 overlaps with this domain, except that it contains the C-terminal six residues of VEGF-A 165 encoded by exon 8 and lacks the first cysteine residue corresponding to position 22 of exon 7, previously identified as critical for the inhibitory activity of a glutathione S-transferase/exon 7 fusion protein (55). In addition, the structure of EG3287 is constrained by the introduction of two disulfide bonds, which were essential for optimal inhibition.
The strong similarity of the NMR-derived solution structures for EG3287 and the corresponding region of native VEGF-A 165 , obtained from the recombinant C-terminal 55 amino acids of VEGF-A 165 (46), indicates that the constrained, disulfide-bonded structure of EG3287 closely resembles the native structure of the VEGF-A 165 NP-1-binding domain. The EG3287 structure indicates several defined structural motifs, of which the most striking are a short ␣-helical region extending from residues 143-150, and a projecting region comprising the C-terminal six residues, encoded by exon 8. Similar stable small protein motifs have been observed in the Zn 2ϩ binding domain of carbonic anhydrase, the scorpion toxin scyllatoxin and a charybdotoxin analogue (56 -58), but EG3287 is unusual in constituting a stable protein domain with only 2 disulfide bridges between the ␣-helix and one of the ␤-sheet domains.
A surprising discovery was that peptides lacking the exon 7-encoded N-terminal residues and comprising largely the exon 8-derived sequence retained a significant degree of inhibitory activity, though with reduced potency compared with EG3287. Substitutions in this region indicated that the lysine, proline, and C-terminal arginine were all essential for inhibition of binding. The fact that amidation at the C terminus almost completely prevented inhibition of VEGF-A 165 binding to NP-1, further underscores the importance of the exon 8-encoded domain for NP-1 binding. Several substitutions designed to disrupt the predicted ␣-helical region reduced EG3287 activity, but did not abolish it, while alterations of the positions of the intramolecular disulfide bonds had significant though limited effects on the ability of EG3287 to inhibit NP-1 binding. Taken together, these results indicate an especially critical role for the C-terminal exon 8-derived region in VEGF-A 165 recognition of NP-1. This conclusion is further supported by the recent finding that the naturally occuring immunostimulatory peptide, Tuftsin (sequence TKPR), and its higher affinity antagonist (TKPPR), are both similar in sequence to the exon 8 region of VEGF-A, and block VEGF-A 165 binding to NP-1 (59).
The lack of a clearly defined role for the NP-1 intracellular domain suggests that this receptor is unlikely to play an independent role in VEGF-A 165 signal transduction, and NP-1 has been proposed to act principally as a docking molecule for VEGF-A 165 in endothelial cells. Given that EG3287 had no effect on VEGF-A 165 binding to KDR in  PAE/KDR cells, it was surprising that EG3287 markedly inhibited KDR phosphorylation at Tyr 1054 and Tyr 1059 induced by a maximally active VEGF-A 165 concentration. Tyr 1054 and Tyr 1059 are in the activation region of the tyrosine kinase domain and phosphorylation at these sites has been shown to be essential for maximal activation of the KDR kinase, mutation of both these residues reducing VEGF-A-stimulated receptor activation to ϳ10% of that for the native receptor (54). One explanation for the inhibitory effect of EG3287 on KDR activation is that NP-1 mediates a more stable physical association between VEGF-A 165 and KDR, consistent with the reduction of radiolabeled VEGF-A 165 cross-linking to KDR in HUVECs by EG3287, even though VEGF-A 165 cross-linking to KDR in PAE/KDR cells lacking NP-1 was unaffected. These findings are in agreement with a previous report showing that formation of complexes between KDR and NP-1 enhance VEGF-A 165 receptor binding (60), and support the argument that in cells co-expressing KDR and NP-1, cobinding to NP-1 is required for stable VEGF binding to KDR and full activation of KDR and distal signaling pathways.
Consistent with its ability to inhibit KDR activity, EG3287 also greatly reduced PLC-␥ tyrosine phosphorylation, ERK activation, and VEGF-A 165 -stimulated endothelial production of PGI 2 and PGE 2 . The fact that EG3287 inhibited all these early responses and VEGF-A 165 binding to NP-1 in a similar concentration range indicates strongly that these biological effects are mediated by specific antagonism of VEGF-A 165 binding to NP-1. It is noteworthy that a degree of KDR cross-linking, activation, and downstream signaling was resistant to inhibition of NP-1 binding by EG3287. Residual KDR phosphorylation at Tyr 1175 is likely to account for the partial effect of EG3287 on VEGF-A 165 -stimulated PLC-␥ tyrosine phosphorylation, which is critically dependent on KDR phosphorylation at this site (20). Despite inhibition of KDR activation and signaling, blockade of VEGF-A 165 binding to NP-1 by EG3287 partially inhibited Akt activation and had little or no significant effect on VEGF-A 165 -stimulated endothelial cell survival and proliferation. Interestingly, inhibition of NP-1 function by siRNA-mediated silencing was recently found to have only a modest inhibitory effect on VEGF-A 165induced proliferation (61). The lack of effect of EG3287 on mitogenic and survival responses to VEGF-A 165 , is probably attributed to the fact that prevention of NP-1 binding causes only a partial reduction in Akt activation and possibly other pathways that play key roles in these biological responses. Indeed, EG3287 pretreatment did not prevent rapid VEGF-A 165 -induced mobilization of intracellular Ca 2ϩ , 3 an essential early event in the mitogenic response to VEGF-A 165 (18). The lack of a stronger biological effect is presumably because maximum biological responsiveness can be attained, at least in some cases, through less than optimum signaling output. EG3287 also strongly inhibits VEGF-A 165 binding to tumor cells expressing predominantly NP-2 with similar efficacy and potency as in cells expressing only NP-1, 4 indicating that mediation of effects via NP-2 independently of NP-1 is unlikely to account for these findings.
Since the identification of NP-1 as a receptor for VEGF-A 165 , elucidation of the role of this receptor in endothelial functions of VEGF-A 165 has remained elusive. In conclusion, our studies with a specific antagonist of VEGF binding to NP-1, indicate that NP-1 is essential both for effective VEGF-A 165 binding to KDR and full activation of KDR, but plays a differential role in KDR signaling and biological response. The findings presented here further demonstrate that both the disulfidelinked three-dimensional structure of the C-terminal subdomain of exon 7 and the region encoded by exon 8 play key roles in VEGF-A 165 binding to NP-1. This discovery will be important for the future development of peptidomimetic and small molecule antagonists of VEGF-A 165 interactions with NP-1. The identification and characterization of a specific antagonist of NP-1 will be helpful in future studies aimed at delineating the functions of NP-1 and in developing novel therapeutic approaches targeted to neuropilins.