Vascular Endothelial Growth Factor Receptor-2 and Neuropilin-1 Form a Receptor Complex That Is Responsible for the Differential Signaling Potency of VEGF165 and VEGF121 *

The two most abundant secreted isoforms of vascular endothelial growth factor A (VEGF165 and VEGF121) are formed as a result of differential splicing of the VEGF-A gene. VEGF165 and VEGF121 share similar affinities at the isolated VEGF receptor (VEGFR)-2 but have been previously demonstrated to have differential ability to activate VEGFR-2-mediated effects on endothelial cells. Herein we investigate whether the recently described VEGF165 isoform-specific receptor neuropilin-1 (Npn-1) is responsible for the difference in potency observed for these ligands. We demonstrate that although VEGFR-2 and Npn-1 form a complex, this complex does not result in an increase in VEGF165 binding affinity. Therefore, the differential activity of VEGF165 and VEGF121cannot be explained by a differential binding affinity for the complex. Using an antagonist that competes for VEGF165 binding at the VEGFR-2·Npn-1 complex, we observe specific antagonism of VEGF165-meditated phosphorylation of VEGFR-2 without affecting the VEGF121 response. These data indicate that the formation of the complex is responsible for the increased potency of VEGF165 versus VEGF121. Taken together, these data suggest a receptor-clustering role for Npn-1, as opposed to Npn-1 behaving as an affinity-converting subunit.

The product of the VEGF-A 1 gene is required for formation of the embryonic vasculature, because haploinsufficiency leads to embryonic lethality due to a failure in both angiogenesis and blood island formation (1). The effects of VEGF-A on development of the embryonic vasculature are mediated by an interaction with the VEGF receptor (VEGFR)-1 and -2 receptor tyrosine kinases (2)(3)(4), whereas effects on angiogenesis that occur in the adult animal appear to be mediated largely through VEGFR-2 (5)(6)(7)(8), with VEGFR-1 playing a modulatory role (9 -12).
VEGF-A exists in multiple protein isoforms with different heparin proteoglycan and extracellular matrix binding properties. These isoforms (VEGF 165 , VEGF 121 , VEGF 145 , VEGF 183 , VEGF 189 , and VEGF 206 ) arise from alternate splicing of the VEGF-A gene. VEGF 165 is the predominant isoform and has heparin binding activity, whereas the VEGF 121 isoform is freely soluble and is devoid of heparin binding activity (13). Similarly, another VEGF superfamily member, placenta growth factor (PlGF), exists in three different isoforms that also exhibit differential heparin binding ability (14). Although VEGF 165 and VEGF 121 bind to VEGFR-2 with equal affinity, their activity in biochemical assays that rely on activation of VEGFR-2 is not equivalent (15)(16)(17), implying that their ability to activate VEGFR-2 is not solely dependent on VEGFR-2 binding affinity.
Recently, neuropilin (Npn)-1 was identified as a receptor that binds VEGF 165 (18,19), VEGF-E (20), PlGF 152 (21), and VEGF-B (22), whereas Npn-2 has been identified as a receptor that binds PlGF 152 , VEGF 165 , and VEGF 145 (23). Neither Npn binds VEGF 121 , and Npn-1 will not bind the non-heparin-binding form of PlGF, PlGF 139 (21,23). Npn-1 and Npn-2 also bind various semaphorin ligands to mediate repulsive guidance activity in certain neuronal populations (24). The cytoplasmic domains of Npn-1 and Npn-2 are not required for semaphorin signaling (25), and they do not contain sequences predictive of enzymatic activity nor sequences predicted to be involved in coupling to intracellular signaling molecules. Nevertheless, the Npn-1 and Npn-2 proteins are obligate binding subunits of a semaphorin signaling receptor complex. Members of the plexin family appear to serve as the signaling subunits of the semaphorin receptor complex (26,27). Potential signals that lie downstream of the neuropilin-plexin receptor complex include activation of a pertussis toxin-sensitive G-protein (28) and activation of Rac1, a small molecular mass G-protein involved in cytoskeletal rearrangement (26,29). The carboxyl terminus of Npn-1 has been shown to bind to a PSD-95/Dlg/ZO domain containing a regulator of G-protein signaling protein (30,31), although the role of this protein in semaphorin signaling has not been established.
Although the role of Npn-1 in mediating the neuronal effects of the semaphorins is being elucidated, much less is known about the role of Npn-1 in VEGF 165 -mediated angiogenesis. Targeted disruption of the Npn-1 gene in mice results in embryonic lethality, with the embryos exhibiting defects in the formation of brachial arch-derived vessels, impaired neural vascularization, and defects in vascular development in the yolk sac that occur in addition to expected defects in neuronal patterning (32,33). Transgenic overexpression of Npn-1 results in embryonic lethality associated with excessive vascular formation, dilation, and hemorrhaging, as well as defects in skeletal morphogenesis and neuronal guidance (34,35). Because the semaphorin III knockout mouse also exhibits cardiovascular abnormalities (36), it is not currently known whether the * 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.
‡ To whom correspondence should be addressed: Dept. of Cardiovascular Research, Procter & Gamble Pharmaceuticals, Health Care Research Center, 8700 Mason-Montgomery Rd., Mason, OH 45040. Tel.: 513-622-2552; Fax: 513-622-1433; E-mail: rosenbaum.js@pg.com. 1 The abbreviations used are: VEGF, vascular endothelial growth factor; VEGFR, VEGF receptor; Npn, neuropilin; RTK, receptor tyrosine kinase; GFP, green fluorescent protein; PlGF, placenta growth factor; HUVEC, human umbilical vein endothelial cell; BSA, bovine serum albumin; PAGE, polyacrylimide gel electrophoresis. vascular effects attributed to the absence or the overexpression of Npn-1 are due to signaling alterations in the semaphorin or the VEGF systems. Because VEGF 121 does not bind to Npn-1 (18,23) and because mice engineered to express only the VEGF 121 isoform do not exhibit deficiencies in the development of the embryonic vasculature (37), it is possible that Npn-1 might not be required for the development of the vasculature in response to VEGF-A. However, these same mice exhibit defects in cardiac vascularization and suffer from ischemic cardiomyopathy, implying that VEGF 121 cannot substitute for VEGF 165 in development of the cardiac vasculature during the period of postnatal cardiac development (37). These data suggest that the inability of VEGF 121 to interact with Npn-1 may result in insufficient signaling to support the maturation of the cardiac vasculature.
The ability of VEGF 165 to bind to VEGFR-2 and Npn-1 through distinct regions of the VEGF 165 molecule led to the assertion that VEGFR-2 and Npn-1 may signal as a co-receptor complex (18,35). We demonstrate here that VEGFR-2 does indeed form a complex with Npn-1 in both heterologous systems as well as in cultured endothelial cells. We further demonstrate that, although the complex does not bind VEGF 165 with an affinity that is greater than that exhibited at either VEGFR-2 or Npn-1 alone, formation of this complex can explain the differential potencies that are observed in cultured endothelial cells for VEGF 165 -and VEGF 121 -mediated stimulation of VEGFR-2 activation.

EXPERIMENTAL PROCEDURES
Materials-Carrier-free recombinant VEGF 165 was purchased from R & D Systems Inc. (Minneapolis, MN). Na 125 I was purchased from Amersham Pharmacia Biotech. Heparin sulfate (catalog number 9399) and bovine serum albumin (BSA) (catalog number A3294) were purchased from Sigma. PD-10 columns and protein G-Sepharose beads were purchased from Amersham Pharmacia Biotech. Donor calf serum, LipofectAMINE-2000, and Optimem I were purchased from Life Technologies, Inc. Fetal bovine serum was purchased from Hyclone Laboratories (Logan, UT). pJFE14 vector and hVEGFR-2/pJFE14 construct were obtained from Regeneron Pharmaceuticals (Tarrytown, NY). pLTRMCSIRES-GFP and pLTR2hFLK1(full)IRES-GFP retroviral constructs were obtained from Regeneron Pharmaceuticals. The pBMN-Z-I-Neo vector used to construct hNpn-1 retroviral construct was obtained from Dr. Gary P. Nolan (Stanford University). The PlGF 152 Exon 6 peptide (amino acids 142-157) was synthesized by SynPep (Dublin, CA). All of the other reagents were obtained from commercial sources.
Plasmid Construction-For transient transfection in COS-1 cells, the hVEGFR-2 and hNpn-1 cDNAs were cloned into the mammalian expression plasmid pJFE14 containing the SR␣ promoter (38). For stable expression in Balb/c 3T3 A31 cells, hVEGFR-2, GFP, and hNeuropilin-1 cDNAs were generated and subcloned into the pBMN-Z family of retroviral vectors. The pLTRMCSIRES-GFP (referred to as the Mock vector) and pLTRhFLK1(full)IRES-GFP constructs are a bi-cistronic modification of the pBMN-Z vector, containing a GFP marker or containing GFP ϩ hVEGFR-2. The hNpn-1 cDNA was subcloned into the pBMN-Z-I-Neo vector.
Polyclonal Antibody Production-The Npn-1 rabbit polyclonal antibody (NP1ECD4) was raised against the purified peptide sequence (Ac-DLDKKNPEIKIDETGSTC-amide (amino acids 814 -830)) and affinity purified by QCB, a division of BIOSOURCE International (Hopkinton, MA). The sequence used for generation of the NP1ECD4 antibody is from the extracellular juxtamembrane region of Npn-1, and this antibody is therefore not predicted to recognize soluble Npn-1 (39), which has been demonstrated to bind VEGF 165 alone (40) or in Npn-1⅐VEGFR complexes (41). The VEGFR-2 rabbit polyclonal antibody (R2.2) was raised against the purified peptide sequence Ac-SKRKSR-PVSVKTFEDIPLEEPC-amide (amino acids 1225-1246), unique to the carboxyl-terminal domain of VEGFR-2, and affinity purified by QCB. The Npn-1 number 30 polyclonal antibody was the kind gift of Dr. David Ginty (Johns Hopkins University). This antibody was generated against amino acids 583-856 of rat neuropilin-1 (42). These residues correspond to the MAM domain and the extracellular juxtamembrane region and thus recognize only full-length Npn-1 protein. This antibody was used exclusively for Western blotting. The C-19 antibody (catalog number SC-7239) for neuropilin-1 was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) and will only recognize the fulllength Npn-1 protein. This antibody was sometimes substituted for the NP1ECD4 antibody for immunoprecipitations (as indicated in the figure legends).
Monoclonal Antibody Production-For generation of the immunogen for the J5F4A2 monoclonal antibody, the entire coding region of the VEGFR-2 extracellular domain (amino acids 1-760) was expressed and purified from Sf9 cells as an Fc fusion protein (Regeneron Pharmaceuticals). The amino-terminal sequence of the purified protein was determined to be a single sequence (ASVGL). This protein was designated FLK-1(Ig 1-7):Fc. Unique clones were tested by Western blot using hybridoma tissue culture supernatants and/or purified antibody. The VEGFR-2 monoclonal antibody (J5F4A2) showed the optimal reactivity to FLK-1(Ig 1-7):Fc and does not cross-react with VEGFR-1 (data not shown). Antibodies were purified from tissue culture supernatant by protein G column.
Iodination of VEGF 165 -Iodination was performed using a modification of methods described previously (15,43). Briefly, 5 g of carrierfree human recombinant VEGF 165 was suspended in 90 l of Dulbecco's phosphate-buffered saline. To the reaction tube, 1 mCi of Na 125 I was added, followed by 40 l of chloromine T (1 g/l in 0.5 M sodium phosphate buffer, pH 7.5), and incubated for 1 min. 50 l of sodium metabisulfite (2 g/l in sodium phosphate buffer, pH 7.5) was added to stop the reaction. 500 l of column elution buffer (0.5% BSA, 0.01% Tween 20 in Dulbecco's phosphate-buffered saline) was added to the reaction and transferred to pre-equilibrated PD-10 column for separation from unreacted iodine. The specific activity was corrected for column recovery and varied from 4,000 to 15,000 Ci/mmol.
Competition Binding Assay-2 ϫ 10 5 cells/well in a 12-well plate were plated 24 h prior to the experiment. Depending upon the equilibrium dissociation constant (K d ) of the receptor being studied, 100 -400 pM [ 125 I]VEGF 165 was incubated (4 h, 4°C) in the presence of increasing concentrations of unlabeled ligands in binding buffer containing 1 g/ml heparin sulfate and protease inhibitors. Unbound ligand was removed by washing with ice-cold BSA-free binding buffer. The cells were lysed with 250 l of RIP buffer containing 0.1% SDS and counted using a ␥ counter. Generation of competition curves and IC 50 values (concentration of unlabeled ligand that reduces radioligand binding to halfway between the upper and lower plateaus of a competition curve) were performed using the Prism software package (GraphPad Software Inc., San Diego, CA).
Saturation Analysis and Nonlinear Curve Fitting-For a direct measurement of receptor number and affinity, the saturation analysis was performed as described above except that increasing concentrations of 0.5-3000 pM [ 125 I]VEGF 165 were added in the presence or the absence of 30 nM unlabeled VEGF 165 to estimate nonspecific binding. This concentration range had previously been determined to discriminate a heterogeneous population of [ 125 I]VEGF 165 binding sites in some immortalized endothelial cell lines (data not shown). The maximum number of binding sites (B max ) and the K d values were obtained using the Prism software package, which performs a statistical assessment of goodness-of-fit to a one-binding site versus a two-binding site model. This methodology is preferred over the Scatchard analysis because it does not require a transformation and linearization of the data, which is known to distort the experimental errors associated with radioligand binding data (44). In all cases described herein, the optimal fit of the data was to a one-binding site model (Table I and  Western Blot Analysis-Cells were solubilized with ice-cold RIPA buffer (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 50 mM NaF, 1 mM sodium orthovanadate, 5 mM benzamidine, 0.5% IGEPAL CA-630, and protease inhibitors). The protein was separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes. The membranes were probed for VEGFR-2 (blocked with 5% BSA in Tris-buffered saline, 0.1% Tween 20 (0.1% TBST)) or Npn-1 (blocked in 0.5% Blotto in 0.1% TBST) using rabbit R2.2 (0.55 g/ml) or rabbit anti-Npn-1 number 30 (1:5,000) and detected by ECL.
Anti-phosphotyrosine Assay-HUVEC cells were plated (1 ϫ 10 6 cells/flask) in T-75 flasks 1 week prior to stimulation. The cells were rinsed with Dulbecco's modified Eagle's medium, serum-starved (1-2 h, 37°C), and then stimulated by incubation with increasing concentrations of VEGF in binding buffer containing 1 g/ml heparin sulfate for 5 min at 37°C. Cells were lysed with RIPA buffer containing protease inhibitors and passed through a 23-gauge needle. The samples were immunoprecipitated with R2.2 antibody (10 g) overnight at 4°C, washed with RIPA, separated by SDS-PAGE, and transferred onto polyvinylidene difluoride membranes. The membranes were probed with 0.5 g/ml 4G10 anti-phosphotyrosine antibody (Upstate Biotechnology Inc., Lake Placid, NY; catalog number 05-321) and detected by ECL. Membranes were stripped (200 mM glycine, pH 2.8) and processed for VEGFR-2 receptor normalization using R2.2 (0.55 g/ml). The bands were quantitated using ImageQuant (Molecular Dynamics), and the EC 50 values were determined from analysis of the dose-response curves generated with GraphPad Prism.

VEGF 165 and VEGF 121 Exhibit Differential Potency in Stimulating VEGFR-2 Phosphorylation
Despite Similar Affinity at VEGFR-2-Although VEGF 165 and VEGF 121 have the same affinity for the isolated purified VEGFR-2 receptor (15), these two ligands exhibit a differential ability to activate the VEGFR-2 receptor in vitro (15)(16)(17). In native HUVEC, this differential activity manifests as an increased potency of VEGF 165 relative to VEGF 121 . This difference in potency is characterized by an EC 50 (concentration of agonist that provokes a response halfway between the base-line and maximal responses) for VEGF 165 of 7.76 ϫ 10 Ϫ11 M (log EC 50 ϭ Ϫ10.11 Ϯ 0.113, n ϭ 14) and for VEGF 121 of 2.45 ϫ 10 Ϫ9 M (log EC 50 ϭ Ϫ8.61 Ϯ 0.091, n ϭ 14) in stimulating phosphorylation of VEGFR-2, with a representative experiment illustrated in Fig.  1. The reported similarity in affinity between VEGF 165 and VEGF 121 at the isolated VEGFR-2 receptor is not an artifact of producing the protein as a receptor body (15), because when the full-length receptor is expressed in cells similar results are observed ( Fig. 2A). In Fig. 2A, VEGFR-2 is overexpressed in COS-1 cells, and the ability of these two ligands to compete for [ 125 I]VEGF 165 binding is measured. In the COS-1 cells overexpressing the full-length VEGFR-2, the IC 50 values for VEGF 165 (IC 50 ϭ 6.34 ϫ 10 Ϫ10 M) and VEGF 121 (IC 50 ϭ 3.12 ϫ 10 Ϫ10 M) binding are nearly identical, indicating similar affinity to VEGFR-2 when it is expressed in isolation. Surprisingly, VEGF 121 has only limited ability to compete for [ 125 I]VEGF 165 binding in HUVEC (Fig. 2B), despite the presence of functional VEGFR-2 in these cells (Fig. 1). Because the HUVEC are reported to contain Npn-1 (18), we examined whether the binding observed in HUVEC could be reproduced in the COS-1 cell system. As expected, VEGF 121 does not compete in COS-1 cells where Npn-1 is expressed alone (Fig. 2C) but limited or no competition for binding is also observed in cells expressing Npn-1 in concert with VEGFR-2 (Fig. 2, B and D). The data from the COS-1 cells expressing defined receptor populations suggest that the relative inability of VEGF 121 to compete for binding in HUVEC may be attributed to excess Npn-1 expression relative to VEGFR-2. We therefore explored the possibility that VEGFR-2 and Npn-1 form a co-receptor complex and that the preferential ability of VEGF 165 to signal through this complex is responsible for the increased potency of VEGF 165 relative to VEGF 121 in the VEGFR-2 autophosphorylation assay.
VEGFR-2 and Npn-1 Form a Co-receptor Complex-To demonstrate the potential for the VEGFR-2⅐Npn-1 complex to form, we utilized an affinity labeling reciprocal immunoprecipitation experimental design. In this design, cells expressing each of the receptors in isolation or co-expressing both receptors are affinity-labeled with a high concentration of [ 125 I]VEGF 165 , chemically cross-linked, and immunoprecipitated with antibodies a Balb/c cells contain readily detectable levels of endogenous Npn-1 as measured in Western blot analysis (Fig. 7B). b Statistically greater then PGBMGH (p ϭ 0.050).
specific for one member of the suspected receptor complex. If the complex forms, the receptor-specific antibody should more effectively precipitate the receptor against which it was generated, and additionally precipitate a band of the appropriate size corresponding to the alternate receptor in the complex. Fig. 3A demonstrates that the complex can indeed form in the presence of ligand in the COS-1 overexpression system. In cells expressing only VEGFR-2 (lane 1) a single band of ϳ240 kDa is observed. This band represents [ 125 I]VEGF 165 cross-linked to VEGFR-2. Using our VEGFR-2 specific antibody, in COS-1 cells co-expressing VEGFR-2 and Npn-1 we are able to immunoprecipitate not only affinity-labeled VEGFR-2, but an additional doublet of bands that runs at the predicted size (ϳ140 kDa) for Npn-1 cross-linked to [ 125 I]VEGF 165 (Fig. 3A, left panel, lane 2). In cells overexpressing only Npn-1 (lane 3) the VEGFR-2 receptor is not present for the complex to form, and therefore we are unable to substantially detect the Npn-1 doublet. The data in lane 3 also indicate that the VEGFR-2 antibody does not cross-react with Npn-1, and the weak doublet detected in lane 3 may indicate formation of a complex between endogenous VEGFR-2 and the exogenous Npn-1. Interestingly, a weak doublet that corresponds to the size of the Npn-1 bands is also detected with the VEGFR-2 receptor-specific antibody in cells overexpressing VEGFR-2 (Fig. 3A, lane 1), which may indicate formation of a complex between endogenous Npn-1 (see Npn-1 Western data; Fig. 3B, right panel) in the presence of the exogenous VEGFR-2. The inability to detect these affinity-labeled bands in the COS-1 cells expressing empty vector (Fig. 3A, Mock, lane 4) provides a further indication that these bands are not nonspecific. The identity of the lower doublet is confirmed as being Npn-1 by the immunoprecipitation with the Npn-1-specific antibody (Fig. 3A, lane 7). The inability to detect bands in VEGFR-2 cells alone (lane 5) demonstrates that when Npn-1 is not overexpressed, the complex with endogenous Npn-1 and overexpressed VEGFR-2 is not detected because no affinity-labeled VEGFR-2 receptor is detected in cells immunoprecipitated by the Npn-1 specific receptor antibody. Lane 5 also demonstrates that the Npn-1 receptor-specific antibody does not cross-react with the VEGFR-2 receptor. When both receptors are co-expressed (Fig. 3A, lane 6) and immunoprecipitated using the Npn-1 receptor-specific antibody, a band corresponding to the size of affinity-labeled VEGFR-2 is apparent, in addition to the affinity-labeled Npn-1 doublet. COS-1 cells expressing Npn-1 alone immunoprecipitated with the Npn-1 antibody confirm the identity of the doublet as being due to [ 125 I]VEGF 165 binding to Npn-1 (lane 7). The inability to detect the doublet in COS-1 cells expressing empty vector (Fig.  3A, lane 8) again demonstrates the specificity of the immunoprecipitated bands. Finally, we observe high molecular mass bands either in cells expressing VEGFR-2 alone (Fig. 3A, lane  1), in cells expressing Npn-1 alone (lane 7), or in cells coexpressing these receptors (lanes 2 and 6). Because it is difficult to estimate the molecular masses of the bands in this region of the gel, and the individual receptor homodimers are of comparable molecular masses, we cannot identify the precise nature of these bands in cells co-expressing both VEGFR-2 and Npn-1.
It is noteworthy that more intense labeling of VEGFR-2 is observed upon co-expression with Npn-1 (Fig. 3A, lane 2 versus  lane 1) and that more intense labeling of Npn-1 is observed upon co-expression with VEGFR-2 (Fig. 3A, lane 6 versus lane  7). This is because co-expression results in increased expres- FIG. 1. VEGF 165 and VEGF 121 exhibit distinctly different EC 50 values in the VEGFR-2 phosphorylation assay. HUVEC cells were stimulated with ligand for 5 min at 37°C, lysed, and immunoprecipitated with an antibody specific to VEGFR-2 (R2.2). Following SDS-PAGE separation and transfer, the blots were probed using an anti-phosphotyrosine (pY) antibody (4G10; Upstate Biotechnology Inc.) and developed using standard ECL techniques (upper panel). These same blots were then stripped and reprobed using an antibody specific for VEGFR-2 (R2.2, lower panel). The films were scanned and quantitated using Image Quant (Molecular Dynamics). The nonlinear regression analysis (Prism) of the phosphotyrosine (pY)/R2 signal ratio yields EC 50 values of 34.9 pM and 2.09 nM for VEGF 165 and VEGF 121 , respectively. This experiment has been repeated 13 times with similar results.
sion of both VEGFR-2 and Npn-1 protein versus that obtained when either receptor is expressed alone (Fig. 3B) and is therefore not an indication of increased binding affinity at the receptor complex (see also  The left panel demonstrates that the Npn-1 doublet is co-precipitated with affinity-labeled VEGFR-2 when the two receptors are co-expressed. The right panel demonstrates that VEGFR-2 is co-precipitated in the complex when an Npn-1 specific antibody is used. In both panels, a high molecular mass band is observed that may represent the VEGFR-2⅐Npn-1 complex (see text). This experiment has been repeated three times with similar results. B, Western blot of matched lysates depicts the relative receptor expression level achieved (20 g of total protein was loaded per lane). Blotting antibodies are R2.2 for VEGFR-2 and the Npn-1 number 30 antibody for Npn- 1. in size to the doublet that is immunoprecipitated by the Npn-1 antibody in affinity-labeled COS-1 cells (Fig. 3A, lanes 6 and 7). The third band (Fig. 4, left panel, asterisk, and Fig. 7A, lane 1, asterisk) of the triplet observed in the HUVEC cells might represent an endogenous soluble form of Npn-1 because soluble Npn-1 is competent to bind [ 125 I]VEGF 165 (40) and may also form a complex with VEGFR-2 (41). This band is not observed in the COS-1 system (Fig. 3) because the Npn-1 cDNA used for these experiments would only produce the full-length receptor. In the anti-Npn-1 immunoprecipitates from HUVEC cells (Fig.  4, right panel, and Fig. 7A, lane 4), a doublet of the size seen in the COS-1 cells (Fig. 3A) is once again observed. The triplet seen in the anti-VEGFR-2 immunoprecipitates is not observed here because the Npn-1 antibody used for these experiments would not be predicted to immunoprecipitate the soluble form of Npn-1 (see "Experimental Procedures"). Similar to Fig. 3, high molecular mass bands are also observed with both antibodies, which might suggest the formation of a heteromeric receptor complex.
To confirm the identity of the bands in the affinity labeling experiments and to examine the possibility of ligand-independent complex formation, we utilized a reciprocal immunoprecipitation Western blot analysis experimental design, in either the presence or the absence of VEGF 165 , in both the HUVEC cells and COS-1 overexpression systems. As observed in the affinity labeling experiments (Fig. 3A), Fig. 5A demonstrates that we are able to immunoprecipitate Npn-1 with our VEGFR-2-specific antibody only in COS-1 cells that co-express VEGFR-2 in concert with Npn-1 (Fig. 5A, left panels). Conversely, we are also able to immunoprecipitate VEGFR-2 with the Npn-1-specific antibody only in COS-1 cells co-expressing VEGFR-2 with Npn-1 (Fig. 5A, right panels). The ability to detect the co-receptor complex in COS-1 cells appears to be independent of the presence of VEGF 165 ligand.
Western blot analysis of COS-1 and HUVEC cells demonstrates that the expression levels of VEGFR-2 and Npn-1 are substantially lower in the HUVEC cells relative to the COS-1 cells (Fig. 5B). The lower expression levels of the receptors, coupled with the limitation imposed by the efficiency of the receptor-specific antibodies, precludes detection of the receptor complex using the reciprocal immunoprecipitation Western technique in HUVEC cells. Detection of the co-receptor complex in the endogenous HUVEC system is only possible in the presence of ligand because of the sensitivity afforded by the use of radioiodinated ligand (compare Fig. 5A with Fig. 4). For this reason, we are unable to determine whether a ligand-independent complex exists in HUVEC cells. Although the COS-1 system indicates a potential for a ligand-independent complex to form in endogenous systems, we cannot exclude the possibility that the ligand-independent complex observed in the COS-1 cells is driven by the higher receptor expression levels obtained in the COS-1 cell system.

Formation of the VEGFR-2⅐Npn-1 Complex Does Not Result in an Increase in Ligand Binding
Affinity-With the formation of the co-receptor complex demonstrated, we investigated the mechanism by which the complex increased the signaling potency of VEGF 165 . The simplest explanation would be that Npn-1 acts as an affinity converter and that the affinity of VEGF 165 for VEGFR-2 is increased in the presence of Npn-1. Although the COS-1 system is an effective tool to determine whether the complex can form, a transient expression system does not ensure that co-expression of both receptors occurs in every cell. To avoid this caveat, we produced stable cell lines overexpressing VEGFR-2 on an Npn-1 background. We chose the Balb/c 3T3 A31 cell line in which to produce our stable cell lines because these cells are reported to have endothelial cell characteristics (45). Fig. 6A illustrates that the endogenous Balb/c 3T3 A31 cells contain detectable amounts of VEGF binding sites (vector control, PGBMGH) that can be explained by the expression of detectable amounts of Npn-1 (Fig. 6B). Not surprisingly, overexpression of human Npn-1 on this background increases the number of binding sites relative to vector control (p ϭ 0.050, B max ϭ 143,306 Ϯ 37,419, n ϭ 5, in Npn-1(2) versus 48,825 Ϯ 16,118, n ϭ 4 in PGBMGH) without a change in binding affinity (Table I). When VEGFR-2 is overexpressed on the Npn-1 background, there is a small but statistically insignificant increase in binding site number with no change in binding affinity (Fig. 6A and Table I, D7R2). The lack of a statistically significant increase in B max when VEGFR-2 is overexpressed on the Npn-1 background (Table I and Fig. 6A) can be attributed to the formation of the VEGFR-2⅐Npn-1 complex, which the VEGF 165 ligand perceives as indistinguishable from free VEGFR-2 or Npn-1. We attempted to further increase Npn-1 expression in the D7R2 cells but achieved only marginal success, as evidenced by the Western analysis in Fig. 6B and a lack of statistically significant increase in B max versus that in the parent D7R2 cells (Table I). Nevertheless, these data clearly indicate that binding affinity to Npn-1 is not increased in the presence of VEGFR-2, because we could not discern a subpopulation of higher affinity binding sites when the data were fit to a one-site versus two-site binding model (data not shown). Furthermore, the inability to detect a high affinity subpopulation is independent of the ratio of VEGFR-2 to Npn-1 because these ratios are quite different in the COS-1 and Balb/c cells (compare Western analyses in Figs. 3B, 5B, and 6B).
The limitation to this analysis is that the Balb/c 3T3 A31

FIG. 4. Affinity labeling and immunoprecipitation of the VEGFR-2⅐Npn-1 receptor complex in endogenous HUVEC cells.
HUVEC cells were affinity-labeled at 4°C for 4 h with a tracer concentration greater than the K d (700 pM [ 125 I]VEGF 165). The cells were lysed, immunoprecipitated (IP) using the receptor-specific antibody listed (for VEGFR-2, R2.2; for Npn-1, C-19), separated by SDS-PAGE, and developed using the Storm system (Molecular Dynamics). The left lane demonstrates that a triplet of affinity-labeled bands were co-precipitated with affinity-labeled VEGFR-2. The right lane confirms the identity of the top two bands of the triplet as being Npn-1 because an identical doublet is immunoprecipitated using an Npn-1 specific antibody (right panel). The asterisk represents the Npn-1 band not immunoprecipitated by the Npn-1 antibody (see text). Similar to what is observed in Fig. 3, a high molecular mass band is also present in the HUVEC immunoprecipitates, and this band may represent the VEGFR-2⅐Npn-1 complex (see text). This experiment has been repeated three times with similar results. cells do not allow us to compare the binding affinity at VEGFR-2 to that of Npn-1 when each are expressed alone. However, the binding to VEGFR-2 expressed alone in the COS-1 cells (3.39 ϫ 10 Ϫ10 M) is similar to that observed in the D7R2 cells (2.91 ϫ 10 Ϫ10 M), despite the much higher Npn-1 background in the D7R2 cells (Table I and Figs. 3B, 5B, and 6B). Furthermore, the affinity at Npn-1 in the Balb/c cells (4.17 ϫ 10 Ϫ10 M) is similar to that at VEGFR-2 expressed in the COS-1 cells (Table I). This indicates that the affinity of VEGF 165 is similar at either receptor, making it even less likely that we missed detection of a subpopulation of high affinity sites that is created upon co-expression of the two receptors.
Access to the VEGFR-2⅐Npn-1 Complex Can Explain the Increased Potency of VEGF 165 versus VEGF 121 for Activation of VEGFR-2-To demonstrate that the enhanced potency of VEGF 165 could be explained by enhanced signaling through the VEGFR-2⅐Npn-1 complex versus VEGFR-2 alone, we chose to specifically antagonize the binding of VEGF 165 to Npn-1 to see whether the potency of VEGF 165 in the anti-phosphotyrosine assay would be reduced to match that of VEGF 121 . A potential Npn-1 antagonist has been identified by Migdal et al. (21), who examined the effects of peptides generated from the Exon 6 and Exon 7 portions of PlGF 152 on the binding of [ 125 I]VEGF 165 in HUVEC cells. These authors concluded that a peptide consisting of the first 16 amino acid residues from Exon 6 of PlGF 152 was sufficient to block the binding of [ 125 I]VEGF 165 to a 120-kDa band whose characteristics were consistent with that of Npn-1. Fig. 7A   VEGFR-2 band and the Npn-1 triplet in HUVEC cells is detectable by immunoprecipitation with a VEGFR-2-specific antibody. An excess of cold VEGF 165 (30 nM) completely abrogates the labeling of both the VEGFR-2 band and the Npn-1 triplet (lane 2), and the PlGF 152 Exon 6 peptide competes for binding to the Npn-1 triplet with a proportional effect on the labeling of the VEGFR-2 band (lane 3). Furthermore, when HUVEC cells are affinity-labeled with [ 125 I]VEGF 165 and immunoprecipitated with an Npn-1-specific antibody, the PlGF 152 Exon 6 peptide completely abrogates the labeling of the Npn-1 triplet (lane 6). These data are consistent with the PlGF 152 Exon 6 peptide being able to compete for binding of VEGF 165 to Npn-1 and to the VEGFR-2⅐Npn-1 complex in the native HUVEC cells. The ability of the PlGF 152 Exon 6 peptide to bind preferentially to Npn-1 is demonstrated in Fig. 7B where the peptide is able to completely compete for [ 125 I]VEGF 165 binding in COS-1 cells expressing Npn-1 but does not compete effectively for binding to VEGFR-2 when it is expressed in the absence of Npn-1.
Because the PlGF 152 Exon 6 peptide is able to compete for VEGF 165 binding to Npn-1, we hypothesized that if Npn-1 is responsible for the increased potency of VEGF 165 , preventing the formation of the VEGFR-2⅐Npn-1 complex by blocking the binding of VEGF 165 to Npn-1 should reduce the signaling potency of VEGF 165 . Furthermore, because VEGF 121 does not have access to the Npn-1 receptor but activates VEGFR-2, the PlGF 152 Exon 6 peptide should have no effect on the signaling potency of VEGF 121 . Fig. 8 demonstrates that in the absence of the PlGF 152 Exon 6 peptide, VEGF 165 and VEGF 121 demonstrate significantly different signaling potencies in native HU-VEC (VEGF 165 EC 50 ϭ 283 pM; VEGF 121 EC 50 ϭ 1.84 nM). However, in the presence of 100 M PlGF 152 Exon 6 peptide, the signaling potency of VEGF 165 is dramatically reduced (VEGF 165 ϩPlGF 152 Exon 6 peptide EC 50 ϭ 1.06 nM), whereas the signaling potency of a ligand that does not have access to Npn-1 and that signals solely through VEGFR-2 (VEGF 121 ) is unaffected (VEGF 121 ϩPlGF 152 Exon 6 peptide EC 50 ϭ 2.38 nM). From these data, we conclude that the ability of VEGF 165 to bind to the VEGFR-2⅐Npn-1 complex is responsible for the differential potency of VEGF 165 relative to VEGF 121 in stimulating activation of VEGFR-2.

DISCUSSION
Neuropilin-1 has been identified as an isoform-specific receptor for VEGF 165 (18), the VEGF-E isoform VEGF orfNZ2 (20), PlGF 152 (21), and both splice isoforms of VEGF-B (22). In the case of VEGF 165 , it was hypothesized that Npn-1 acts as a potentiator of ligand binding to VEGFR-2, thereby explaining the differential activity of VEGF 165 versus VEGF 121 in various endothelial cell signaling assays (18). However, this conclusion was based solely on affinity labeling data generated in cells The blotting antibodies are R2.2 and Npn-1 number 30 for VEGFR-2 and Npn-1, respectively. The binding experiment has been repeated at least three times with similar results (Table I). The Western blot analysis has been repeated at least six times with similar results. engineered to co-express Npn-1 with VEGFR-2. Affinity labeling experiments are not accurate indicators of receptor-ligand binding affinity because the results can be biased by changes in cross-linking efficiency of the ligand to the receptor that result from conformational alterations upon ligand binding to the receptor component within the complex versus that observed when the receptor component is expressed alone. Additionally, the amount of radioligand observed to cross-link a receptor component can be influenced by alterations in receptor expression between cell lines that express the receptor alone versus those that co-express multiple receptor components. Indeed, we observe a similar increase in binding to VEGFR-2 in COS-1 cells transiently expressing VEGFR-2 in concert with Npn-1 versus that observed in COS-1 cells that express VEGFR-2 alone, but this increased binding to VEGFR-2 can be explained by increased expression of VEGFR-2 in the presence of Npn-1 that often occurs in this system (Fig. 3).
Further support for the contention that formation of the VEGFR-2⅐Npn-1 receptor complex does not result in an alteration in VEGF 165 binding affinity comes from a detailed analysis of the saturation isotherms in cells expressing Npn-1 alone or in concert with VEGFR-2 ( Fig. 6 and Table I). The binding affinity of [ 125 I]VEGF 165 is similar in cells expressing Npn-1 alone or VEGFR-2 alone, and there is no increase in affinity observed upon co-expression of VEGFR-2 on an Npn-1 background ( Fig. 6 and Table I). These data demonstrate that formation of the VEGFR-2⅐Npn-1 co-receptor complex in the presence of ligand (Figs. 3, 4, and 7) does not result in formation of a subpopulation of high affinity binding sites. In this respect, the VEGFR-2⅐Npn-1 receptor complex is similar to that observed for the glial cell line-derived neurotrophic factor receptor ␣ subunit and the Ret receptor tyrosine kinase (RTK) co-receptor complex, where ligand binding is also observed to the non-receptor tyrosine kinase ␣ subunit in the absence of the RTK, and co-expression of the ␣ subunit with the RTK does not result in an increase in ligand binding affinity (46). Hence, Npn-1 does not appear to function as an affinity converter for VEGF 165 in concert with VEGFR-2, as has been observed for other multi-component receptor systems (47). This is in contrast to the role played by Npn-1 in the regulation of ligand binding affinity of Sema3A to the Npn-1⅐Plexin1 co-receptor complex (26).
Despite a lack of effect on ligand binding affinity, Npn-1 appears to be responsible for the discrimination of signaling efficiency elicited by VEGFR-2 in response to VEGF 165 versus VEGF 121 . Previous attempts at blocking Npn-1 involvement in VEGF 165 signaling utilized a recombinant version of the Npn-1-binding domain of VEGF (16,18). This glutathione S-transferase-Exon 7-8 fusion protein competes for [ 125 I]VEGF 165 binding to both VEGFR-2 and Npn-1, (18) and reduces signaling in response to both VEGF 165 and VEGF 121 (16). Because VEGF 121 does not bind to Npn-1 (18,23), it is possible that the antagonistic effect of this protein is due to direct antagonism of signaling through VEGFR-2. Indeed, the results of the deletion FIG. 7. The PlGF 152 Exon 6 peptide competes for binding at the VEGFR-2⅐Npn-1 complex as well as at Npn-1 alone. A, experimental design is the same as for Fig. 4, with the following exceptions.
The tracer (420 pM [ 125 I]VEGF 165 (still greater than the K d ; see Table I)) is added simultaneously in the presence or the absence of the indicated concentration of unlabeled VEGF 165 or the PlGF 152 Exon 6 peptide, and the NP1ECD4 antibody was used to immunoprecipitate (IP) Npn-1. Both VEGF 165 and the PlGF 152 peptide compete for binding at the VEGFR-2⅐Npn-1 complex. The PlGF 152 Exon 6 peptide also competes for binding at the Npn-1 bands (lane 6). This experiment has been repeated once with similar results. B, experimental design is the same as for Fig. 2  analysis on the activity of this fusion protein (16) are inconsistent with the known structure of the heparin-binding region of VEGF 165 (48,49), suggesting that this reagent may not represent the Npn-1-binding domain of VEGF 165 and, by inference, that the protein is not an Npn-1 antagonist. Similar to what was observed for the glutathione S-transferase-Exon 7-8 protein, we also find that a peptide derived from the Exon 6 region of PlGF 152 (21) can compete for the binding of [ 125 I]VEGF 165 to Npn-1, with a proportional competition for [ 125 I]VEGF 165 binding to VEGFR-2 in HUVEC cells (Fig. 7A). The proportional decrease in binding to VEFGR-2 in HUVEC probably represents a decrease in binding to VEGFR-2 in the VEGFR-2⅐Npn-1 complex because this concentration of the PlGF 152 Exon 6 peptide does not substantially compete for binding at VEGFR-2 when it is expressed in the absence of Npn-1 (Fig. 7B). Notably, this peptide only antagonizes the signaling of VEGF 165 and does not affect VEGF 121 -mediated phosphorylation of VEGFR-2 (Fig. 8). Together, these data suggest that a blockade of binding to the VEGFR-2⅐Npn-1 complex can antagonize VEGF signaling in an isoform-specific manner and that isoform-specific binding to the VEGFR-2⅐Npn-1 complex can explain the increased potency of VEGF 165 versus VEGF 121 in HUVEC. These data stand in contrast to that observed for PlGF 139 and PlGF 152 , where access to the complex of VEGFR-1 with Npn-1 (41) does not appear to confer a signaling advantage for the Npn-1-binding PlGF isoform (PlGF 152 ) (21).
If formation of the VEGFR-2⅐Npn-1 complex does not result in a higher affinity state of the receptor, how is the increased potency of VEGF 165 relative to VEGF 121 explained? In Fig. 9 we propose a mechanism wherein formation of the VEGFR-2⅐Npn-1 complex can serve to potentiate signaling through VEGFR-2. VEGFR-2 is thought to bind to VEGF 121 across the dimer interface, at the opposite poles of the ligand dimer (50 -52) (Fig. 9A). VEGF 165 binds to VEGFR-2 and Npn-1 through distinct epitopes, with the VEGFR-2 binding epitope similar to that observed with VEGF 121 , and the Npn-1 binding epitope occurring symmetrically in the Exon-7-8 regions of the VEGF 165 dimer (Fig. 9B). Unlike VEGF 121 , VEGF 165 can cluster VEGFR-2 through its ability to bind VEGFR-2 in a manner similar to VEGF 121 as well as through the preformed Npn-1⅐VEGFR-2 complex that binds via Exon 7.
The model incorporates the experimental data we have de-scribed but also incorporates data from a variety of other receptor systems. It has been demonstrated that the concentration of receptors at the cell surface can affect the position of the dose-response curve for G-protein-coupled receptors (53)(54)(55), as well as for RTK (56,57) ligands, with detection of the ligandmediated response occurring at progressively lower ligand concentrations as receptor expression increases. By analogy, the ability of VEGF 165 to bind to both VEGFR-2 and Npn-1 may serve to increase the local concentration of VEGFR-2 upon binding VEGF 165 relative to that achieved with VEGF 121 (Fig.  9) because Npn-1 has the potential to form a ligand-independent complex with VEGFR-2 (Fig. 5). It is also possible that Npn-1 could function to increase the local ligand concentration in the vicinity of VEGFR-2 (58) and that the multivalency of VEGF 165 relative to VEGF 121 could serve to increase the avidity of the ligand to VEGFR-2, resulting in increased activity at lower ligand concentrations for VEGF 165 relative to VEGF 121 . The latter provides an explanation for the limited ability of VEGF 121 to compete for the binding of [ 125 I]VEGF 165 in cells co-expressing VEGFR-2 with Npn-1 (Fig. 2). There is precedence for the preclustering of different members of multi-component receptors (59 -61), wherein ligand binding to different receptor clusters can have different signaling consequences (61,62). Such precedence is also available for the receptor tyrosine kinases. In the case of the Ephrin receptors, the oligomeric state of the ligand has been demonstrated to confer signaling specificity through the EphB1 and EphB2 receptors in endogenous endothelial cell systems, with cellular attachment and recruitment of the low molecular mass proteintyrosine phosphatase to the receptor cytoplasmic domain only occurring in higher order ligand-receptor oligomers (63). Both platelet-derived growth factor receptor ␤ and VEGFR-2 have been demonstrated to exist in a ligand-independent complex with the ␣ v ␤ 3 integrin through an interaction of the RTK with the ␤ 3 extracellular domain, although association of VEGFR-2 with ␣ v ␤ 3 requires the presence of the ␣ v integrin receptor subunit (64). This finding indicates that oligomerization of receptor subunits need not occur through enzyme-substrate interactions of the cytoplasmic domains. The attachment of endothelial cells to vitronectin potentiates signaling through VEGFR-2, and an antibody to the integrin ␤ 3 subunit that does not antagonize cellular adhesion is capable of inhibiting the FIG. 8. The PlGF 152 peptide antagonizes VEGF 165 signaling without affecting VEGF 121 signaling in HUVEC cells. The experimental design is the same as for Fig. 1, with the exception that the PlGF 152 peptide was added to the indicated samples at a final concentration of 100 M. If the PlGF 152 peptide were acting solely as an antagonist of VEGFR-2, it would shift both the VEGF 165 and VEGF 121 dose-response curves to the right. In contrast, only the VEGF 165 response curve is affected. This implies that the PlGF 152 Exon 6 peptide is specifically affecting signaling through the VEGFR2⅐Npn-1 complex. The nonlinear regression analysis (Prism) of the phosphotyrosine (pY)/R2 signal ratio reveals the following: VEGF 165 EC 50 ϭ 283 pM; VEGF 165 ϩ PlGF 152 Exon 6 peptide EC 50 ϭ 1.06 nM; VEGF 121 EC 50 ϭ 1.84 nM; VEGF 121 ϩ PlGF 152 Exon 6 peptide EC 50 ϭ 2.38 nM. This experiment has been repeated once with similar results.
ligand-stimulated phosphorylation of VEGFR-2 as well as signaling downstream of VEGFR-2 (65). Because VEGFR-2 has the potential to be preassociated with ␣ v ␤ 3 on the cell surface (64) and because the anti-integrin ␤ 3 antibody does not interfere with the ability of [ 125 I]VEGF 165 to bind to VEGFR-2 (65), the mechanism for the inhibitory effect of this antibody on VEGFR-2-mediated signaling may also be due to an interference with formation of higher order VEGFR-2 oligomers, similar to what we have hypothesized for the consequences of inhibition of VEGF 165 binding to Npn-1 within the VEGFR-2⅐Npn-1 receptor complex.
In summary, we have demonstrated that Npn-1 forms a co-receptor complex with VEGFR-2 in the presence of VEGF 165 ligand, both in heterologous systems as well as in an endogenous endothelial cell system. Surprisingly, formation of this complex does not result in formation of a higher affinity state of the receptor for VEGF 165 ; therefore the difference in the potency of VEGF 165 versus VEGF 121 to stimulate VEGFR-2 cannot be explained by alterations in receptor affinity that are specific to VEGF 165 . Using a peptide derived from the Npn-1binding region of PlGF 152 (21), we demonstrate that antagonism of VEGF 165 binding to Npn-1 interferes with signaling specifically through the VEGFR-2⅐Npn-1 complex, without an effect on signaling of VEGF 121 through VEGFR-2. Because a concurrent up-regulation of VEGF 165 with VEGFR-2 and Npn-1 correlates with increased vascular density in certain pathologies (66,67), it is tempting to speculate that antagonism of signaling through the VEGFR-2⅐Npn-1 receptor complex may attenuate pathological angiogenesis without affecting the function of VEGFR-2 in the quiescent vasculature (68,69).