Vascular endothelial growth factor receptor-1 and neuropilin-2 form complexes.

The products of the neuropilin-1 (Np-1) and neuropilin-2 (Np-2) genes are receptors for factors belonging to the class 3 semaphorin family and participate in the guidance of growing axons to their targets. In the presence of heparin-like molecules, both receptors also function as receptors for the heparin-binding 165-amino acid isoform of vascular endothelial growth factor (VEGF(165)). Both receptors are unable to bind to the 121-amino acid isoform of vascular endothelial growth factor (VEGF(121)), which lacks a heparin-binding domain. Interestingly, complexes corresponding in size to (125)I-VEGF(121).neuropilin complexes are formed when (125)I-VEGF(121) is bound and cross-linked to porcine aortic endothelial cells co-expressing VEGFR-1 and either Np-1 or Np-2. These complexes do not seem to represent complexes of (125)I-VEGF(121) with a truncated form of VEGFR-1, presumably formed as a result of the presence of Np-1 or Np-2 in the cells, because such truncated forms could not be detected with anti-VEGFR-1 antibodies. Antibodies directed against VEGFR-1 co-immunoprecipitated the (125)I-VEGF(121).Np-2 sized cross-linked complex along with (125)I-VEGF(121).VEGFR-1 complexes from cells expressing both VEGFR-1 and Np-2 but not from control cells, indicating that VEGFR-1 and Np-2 associate with each other. To perform the reciprocal experiment we have expressed in porcine aortic endothelial cells a Np-2 receptor containing an in-frame myc epitope at the C terminus. Surprisingly, the myc-tagged Np-2 receptor lost most of its VEGF(165) binding capacity but not its semaphorin-3F binding ability. Nevertheless, when Np-2myc was co-expressed in cells with VEGFR-1, it partially regained its VEGF(165) binding ability. Antibodies directed against the myc epitope co-immunoprecipitated (125)I-VEGF(165).Np-2myc and (125)I- VEGF(165).VEGFR-1 complexes from cells co-expressing VEGFR-1 and Np-2myc, indicating again that VEGFR-1 associates with Np-2. Our experiments therefore indicate that Np-2, and possibly also Np-1, associate with VEGFR-1 and that such complexes may be part of a cell membrane-associated signaling complex.

The various forms of the growth factors belonging to the VEGF 1 family (VEGF, PlGF, VEGF-B, VEGF-C, and VEGF-D) act as inducers and modulators of angiogenesis in vivo (1)(2)(3). The active forms of the VEGF family members are synthesized as homodimers (4,5) or as heterodimers with other VEGF family members such as PlGF (6). Targeted disruption of the VEGF gene has shown that angiogenesis is severely disrupted even in heterozygous animals containing a single functional allele of the VEGF gene. It is therefore believed that the maintenance of correct levels of VEGF in vivo is critical for the development of the cardiovascular system (7,8). Five splice forms of human VEGF ranging from 121 to 206 amino acids (VEGF 121 -VEGF 206 ) have been characterized (4,5,9,10). These differ primarily in the presence or the absence of the heparin-binding domains encoded by exons 6 and 7, giving rise to forms that differ in their heparin and heparan-sulfate binding ability (11). Likewise, other VEGF family members such as PlGF and VEGF-B are also expressed in several forms that differ in their heparin binding ability. For example, the peptide encoded by exon 6 of PlGF is found only in PlGF-2 and confers a heparin binding ability to PLGF-2, whereas PlGF-1 does not bind to heparin (12).
All VEGF isoforms bind to the tyrosine-kinase receptors VEGFR-1 (flt-1) (13) and VEGFR-2 (KDR/flk-1) (14). The binding of VEGF to VEGFR-2 initiates intracellular signal transduction (1,(15)(16)(17)(18) and is correlated with the induction of endothelial cell proliferation, migration, and in vivo angiogenesis (19,20). By contrast, the activation of VEGFR-1 does not seem to result in the induction of cell proliferation or angiogenesis, although exceptions have been observed (21,22). However, the activation of VEGFR-1 seems to enhance cell migration (20,(23)(24)(25). Both receptors play pivotal roles in embryonic vasculogenesis and angiogenesis. Embryos lacking the VEGFR-2 gene die before birth because differentiation of endothelial cells does not take place and blood vessels do not form (26). In contrast, disruption of the gene encoding the VEGFR-1 receptor did not prevent the differentiation of endothelial cells in homozygous animals, but the development of functional blood vessels from these endothelial cells was severely impaired (27).
Endothelial cells also contain another type of VEGF receptor possessing a lower mass than either VEGFR-2 or VEGFR-1 (28,29). It was subsequently found that these smaller VEGF receptors of the endothelial cells are isoform-specific receptors that bind VEGF 165 but not VEGF 121 (30). Additional experiments revealed several types of prostate and breast cancer-derived cell lines that express unusually large amounts of these iso-form-specific receptors (31). The receptors were purified from such cells using affinity chromatography on VEGF 165 affinity matrices followed by partial protein sequencing and were found to be the products of the neuropilin-1 (Np-1) gene (32). It was subsequently observed that the heparin-binding form of placenta growth factor (PlGF-2) and VEGF-B, two additional members of the VEGF family of growth factors, are also able to bind to Np-1 (33,34). When the role of Np-1 as a VEGF receptor was discovered, it was already known that Np-1 functions in the nervous system as receptor for sema-3A. sema-3A is a Np-1 agonist that causes repulsion of growing tips of axons (35,36). It was recently observed that sema-3A is also able to inhibit migration of endothelial cells (37). These results indicate that signaling via Np-1 affects angiogenesis and possibly the development of the cardiovascular system. Targeted disruption of the Np-1 gene resulted in severe cardiovascular defects, confirming these suspicions (38 -40). Np-1 is part of a gene family that includes the closely related receptor Np-2. In the nervous system Np-2 is activated by another class 3 semaphorin, sema-3F, which also induces the repulsion of axons that express Np-2 (41). We have recently observed that Np-2 is also able to bind VEGF 165 and PlGF-2 but not VEGF 121 . However, unlike Np-1, Np-2 was also able to interact with the VEGF 145 form of VEGF. VEGF 145 lacks the peptide encoded by exon 7 of VEGF, which is included in VEGF 165 , but contains instead the heparin-binding domain encoded by exon 6 of the VEGF gene (9,42).
The neuropilins have a short intracellular domain and are unlikely to function as independent receptors. Indeed, no responses to VEGF 165 were observed when cells expressing either Np-1 or Np-2 but no other VEGF receptors were stimulated with VEGF 165 (32,42). It was recently found that plexins form complexes with neuropilins and that these complexes mediate signal transduction by sema-3A (43,44). It is possible that neuropilins associate with additional cell surface molecules to form complexes that transduce VEGF signaling. We present evidence indicating that Np-2 and possibly also Np-1 form complexes with the VEGFR-1 receptor and that the formation of these complexes changes the binding characteristics of neuropilins so that they are now able to bind VEGF 121 , a splice form that is not recognized by neuropilins in cells that do not express VEGFR-1.

EXPERIMENTAL PROCEDURES
Materials-Antibodies directed against the intracellular tyrosinekinase domain of VEGFR-1 were obtained from Santa Cruz Biotechnology. Immunoprecipitating antibodies that bind to the extracellular part of VEGFR-1 were generated as previously described (45). Antibodies directed against the myc epitope were purchased from Santa Cruz Biotechnology. Np-2 and Np-1 expressing PAE cells were generated as previously described (42). A sema-3F expression plasmid was kindly given to us by Dr. David Ginty by permission from Dr. Marc Tessier-Lavigne. This construct contains the semaphorin-3F cDNA fused inframe to alkaline phosphatase at the N terminus of sema-3F (41). LipofectAMINE was purchased from Life Technologies, Inc.
Generation of Anti-Np-2 Antibodies-A 780-base pair Np-2 Sph-1/ Pst-1 cDNA fragment was ligated into the bacterial expression vector pQE-30 (Qiagen). This plasmid was used for the production of the recombinant, His 6 -tagged 30-kDa peptide according to the manufacturer's instructions. The peptide was cleaned from bacterial cell extracts using nickel affinity chromatography according to the instructions of the vendor and purified again using SDS/PAGE. The gel was electroblotted onto nitrocellulose, and the band containing the peptide was cut out, solubilized in Me 2 SO, and used to immunize rabbits. The antiserum was purified on a protein A affinity column followed by affinity purification on a column to which the recombinant peptide was coupled using a previously described method (46). The antibody was eluted from the column using 0.1 M glycine at pH 3. The antibody generated in this manner recognized Np-2 specifically in Western blots and did not recognize other proteins even when these proteins contained a His 6 tag but did not immunoprecipitate Np-2 (data not shown).
Cell Lines and cDNAs-PAE cells (20) were kindly provided by Dr.
Carl Heldin. The PAE⅐VEGFR-1 cell lines were generated by transfecting PAE cells with the VEGFR-1 expression vector BCMGSneo-hu-flt-1 (23), and selection of VEGFR-1 expressing cell lines was done using 0.5 mg/ml G418. The cells were continuously maintained in medium containing 0.25 mg/ml G418. Cell lines expressing VEGFR-1 and Np-2 were generated by co-transfection of VEGFR-1 expressing cells with the PECE/Np-2 expression vector and with the pBabe-puro vector (47). Stable cell lines were isolated by double selection with 0.5 g/ml puromycin and 0.25 mg/ml G418. For the generation of PAE cells expressing VEGFR-1 and Np-1, the VEGFR-1 expression vector BCMGSneo-huflt-1 was transfected into the previously described PAE⅐Np-1 cells (32), and VEGFR-1 expressing cell lines were selected using 0.5 mg/ml G418. Transfections were carried out using LipofectAMINE according to the instructions of the vendor. Human umbilical vein-derived endothelial cells were cultured as previously described (30). Construction and Expression of Np-2myc in PAE Cells-The primer 5Ј-GCTCTAGAGGGCCCTCACAGATCCTCCTCTGAGATGAGTTTTT-GTTCAGCCTCGGAGCAGCACTTTTG-3Ј containing the myc epitope (underlined) and the primer 5Ј-CAACCTCAGGGTCTGGCGCC-3Ј were used to introduce a myc epitope after the last amino acid of Np-2a22 (41). The primers were used to amplify the modified C terminus using the Np-2 cDNA (42) as a template. The myc epitope-containing fragment was ligated back into the Np-2 cDNA using a unique NarI site and an Xba1 site donated by the plasmid (pcDNA3.1/hygro). Following sequencing, the complete expression vector (pcDNA3.1-hygro/Np-2myc) was transfected into PAE or PAE⅐VEGFR-1 cells using LipofectAMINE. Stable cell lines were selected using hygromycin (0.3 mg/ml). In the case of the PAE⅐VEGFR-1⅐Np-2myc cells the selective medium also contained G418 (0.5 mg/ml).
Binding and Cross-linking-VEGF 121 and VEGF 165 were produced using the baculovirus expression system and iodinated as described (30,48). Binding of 125 I-VEGF 165 or 125 I-VEGF 121 to cells was carried out essentially as previously described (42). The water-soluble cross-linker BS 3 was used to cross-link bound VEGFs to receptors. BS 3 was dissolved in PBS to a final concentration of 0.2 mM and applied to cells. The cross-linking procedure was done essentially as previously described (30). All experiments were performed at least twice with similar results.
Immunoprecipitation of Complexes Cross-linked to 125 I-VEGF 121 or 125 I-VEGF 165 Using Anti-VEGFR-1 or Anti-myc Epitope Antibodies-Anti-VEGFR-1 antibodies directed against the extracellular domain of VEGFR-1 (45) or commercial anti-myc epitope antibodies were used in immunoprecipitation experiments. 125 I-VEGF 121 or 125 I-VEGF 165 was bound and cross-linked to cells expressing various VEGF receptors. Following cross-linking, the cells were lysed using lysis buffer (20 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 0.5 mM Na 3 VO 4 , 1 mM dithiothreitol, and protease inhibitors (2 g/ml of leupeptin and aprotinin and 1 mM phenylmethylsulfonyl fluoride)) for 10 min. at 4°C. The lysate was cleared by centrifugation, and protein content was measured using the Bio-Rad protein assay according to the instructions of the vendor. Equal amounts of protein from the different lysates were taken for immunoprecipitation. The lysates were incubated overnight at 4°C with anti-myc or anti-VEGFR-1 antibodies. Protein A-Sepharose was added the next day and incubated with the antibody for 2 h at 4°C, and the beads were subsequently washed three times with cold PBS. SDS/PAGE sample buffer was then added to the beads, and the beads were boiled for 3 min. The supernatant was then separated using SDS/PAGE followed by autoradiography and phosphorimaging analysis. All experiments were performed at least twice with similar results.
Western Blot Analysis-Detection of VEGFR-1 by Western blot analysis was performed using a commercial anti-VEGFR-1 antibody (Santa Cruz Biotechnology) directed against the intracellular tyrosine-kinase domain of the receptor using the ECL system (Amersham Pharmacia Biotech) as previously described (9). Detection of Np-2 in Western blots was performed similarly using our affinity purified anti-Np-2 rabbit derived polyclonal antibodies. All experiments were performed at least twice with similar results.
Production and Binding of sema-3F-The production of sema-3F was done essentially as described (41). The Np-2 expressing PAE cells or parental PAE cells were grown in gelatin-coated 48-well dishes to confluence. The cells were washed once with PBS and incubated with 0.1 ml of conditioned medium from transiently transfected COS-7 cells expressing alkaline phosphatase (AP)-tagged sema-3F. The conditioned medium was supplemented with 1 mg/ml gelatin, 10 mM HEPES buffer, pH 7.3, 2 g/ml leupeptin, 2 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride prior to the binding experiment. The cells were incubated with the conditioned medium at room temperature for 2 h, washed three times with PBS, and then fixed with 4% paraformalde-hyde for 20 min at room temperature. After fixation, the dishes were washed three times with PBS and incubated at 65°C for 1 h to inactivate endogenous phosphatases. The cells were rinsed with AP substrate buffer containing 0.05 M Na 2 CO 3 , 1 mM MgCl 2 , pH 9.8, and detection of alkaline phosphatase activity was done using the same buffer containing 1 mg/ml p-nitrophenyl phosphate for 3 or 10 h at room temperature. Absorption of generated nitrophenol was measured in a spectrophotometer at 405 nm. All experiments were performed at least twice with similar results.

Np-2 Appears to Gain a VEGF 121 Binding Ability When
Co-expressed with VEGFR-1-We have previously found that recombinant Np-2 is unable to bind 125 I-VEGF 121 when it is expressed on its own in PAE cells (Fig. 1A, lane 1) (42). When 125 I-VEGF 121 was bound and cross-linked to PAE cells expressing either VEGFR-1 (PAE⅐VEGFR-1 cells) or VEGFR-1 and Np-2 (PAE⅐VEGFR-1⅐Np-2 cells), two diffuse complexes of ϳ190 and ϳ210 kDa were formed, corresponding to the expected masses of 125 I-VEGF 121 ⅐VEGFR-1 complexes (Fig. 1A,  lanes 2 and 3). Surprisingly, we have found that two additional labeled complexes corresponding in mass to the expected masses of 125 I-VEGF 121 ⅐Np-2 complexes (ϳ140 and ϳ160 kDa) were formed in PAE⅐VEGFR-1⅐Np-2 cells (Fig. 1A, lane 3, arrow). This observation suggested that the presence of VEGFR-1 affects the ligand binding specificity of Np-2, enabling 125 I-VEGF 121 binding, and that VEGFR-1 and Np-2 may perhaps form complexes. This effect on 125 I-VEGF 121 binding seems to be specific to VEGFR-1 because we could not detect binding of 125 I-VEGF 121 to Np-2 when 125 I-VEGF 121 was bound and crosslinked to PAE cells co-expressing VEGFR-2 and Np-2 (not shown). On closer examination we have also noticed that a very weak cross-linked complex corresponding in size to a 125 I-VEGF 121 ⅐Np complex is usually formed when 125 I-VEGF 121 is bound and cross-linked to PAE⅐VEGFR-1 cells (see faint bands at the level of the arrows in Fig. 1A (lane 2) and in Fig. 1B (lane  1)). This complex may perhaps represent binding of 125 I-VEGF 121 to residual endogenous neuropilins and was never observed in parental PAE cells or in PAE cells transfected with empty expression vectors.
The related Np-1 receptor is also unable to bind 125 I-VEGF 121 when it is expressed in PAE cells on its own (32). However, cross-linked complexes corresponding in size to 125 I-VEGF 121 ⅐Np-1 complexes were observed following the binding of 125 I-VEGF 121 to cells co-expressing VEGFR-1 and Np-1, indicating that Np-1 and VEGFR-1 may also form complexes (Fig. 1B, lane 2, arrow). Indeed, such complexes have been recently observed using a completely different experimental approach (49).
Western Blot Analysis Does Not Reveal Smaller Forms of VEGFR-1 in Cells Expressing Both VEGFR-1 and Np-2-Although the ϳ140and ϳ160-kDa complexes observed in the previous experiment correspond in size to 125 I-VEGF 121 ⅐Np-2 complexes, it was still possible that these cross-linked complexes do not represent complexes of 125 I-VEGF 121 and Np-2 but rather complexes formed as the result of the binding of 125 I-VEGF 121 to truncated forms of VEGFR-1. Such truncated VEGFR-1 forms could perhaps be generated as a result of the presence of Np-2 in the cells. To exclude this possibility, extracts from parental PAE cells, PAE⅐VEGFR-1 cells, and PAE⅐VEGFR-1⅐Np-2 cells were analyzed for the presence of shorter forms of VEGFR-1 using commercial antibodies directed against the tyrosine-kinase domain of VEGFR-1. The cells were solubilized and subjected to SDS/PAGE chromatography followed by Western blot analysis. Two VEGFR-1 high molecular weight forms of ϳ170 and ϳ190 kDa were easily detected in PAE⅐VEGFR-1 and in PAE⅐VEGFR-1⅐Np-2 cells but not in Np-2-expressing PAE cells (Fig. 2). A band of ϳ140 kDa was observed in all of the PAE-derived cells including the Extracts were clarified by centrifugation and separated using SDS/ PAGE. The proteins were electroblotted to nitrocellulose and probed with a commercial anti-VEGFR-1 antibody (Santa Cruz Biotechnology.). A secondary anti-rabbit antibody coupled to peroxidase was used to detect bound primary antibody. Bound secondary antibody was detected using the ECL method.

VEGFR-1 and Neuropilin-2 Form Complexes
parental nontransfected cells (not shown). This band probably represents nonspecific binding of the antibody to an unknown antigen and has a mass that is higher than that of Np-2 (Fig. 2). These experiments therefore indicate that it is unlikely that PAE⅐VEGFR-1⅐Np-2 cells express smaller forms of VEGFR-1 as a result of the presence of Np-2 in these cells.
Antibodies Directed against VEGFR-1 Co-immunoprecipitate a Labeled Complex Corresponding in Mass to a 125 I-VEGF 121 ⅐Np-2 Complex-The experiments described above suggested that VEGFR-1 may be able to form complexes with Np-1 and Np-2. To test this hypothesis directly, we used antibodies directed against VEGFR-1 to co-immunoprecipitate Np-2 from PAE cells co-expressing VEGFR-1 and Np-2 receptors. 125 I-VEGF 121 was bound and cross-linked to PAE⅐Np-2, PAE⅐VEGFR-1, or PAE⅐VEGFR-1⅐Np-2 cells. The cells were then lysed, and radiolabeled complexes were immunoprecipitated using anti-VEGFR-1 antibodies. This method takes advantage of the high sensitivity afforded by the use of 125 Ilabeled VEGF but cannot be used to determine whether complex formation is VEGF-dependent or not (50). This experiment revealed that the ϳ140and ϳ160-kDa 125 I-VEGF 121containing complexes seen in the experiment shown in Fig. 1 could be immunoprecipitated along with the 125 I-VEGF 121 ⅐VEGFR-1 complex from PAE⅐VEGFR-1⅐Np-2 cells but not from PAE⅐VEGFR-1 cells (Fig. 3B, lane 5 versus lane 6). The anti-VEGFR-1 antibodies did not immunoprecipitate a similar 125 I-VEGF 165 ⅐Np-2 complex from cells expressing Np-2 but no VEGFR-1, indicating that the anti-VEGFR-1 antibodies do not cross-react with Np-2 (Fig. 3B, lane 4). It should be noted that the PAE⅐Np-2 cells used in this control experiment contain large amounts of Np-2 as revealed by 125 I-VEGF 165 binding/ cross-linking (Fig. 3A, lane 1). These results therefore indicate that VEGFR-1 forms complexes with Np-2.
Np-2 Tagged with a myc Epitope at the C Terminus Loses its VEGF 165 Binding Ability-To obtain further independent experimental evidence for the formation of complexes between VEGFR-1 and Np-2, we have attempted to immunoprecipitate VEGFR-1 from cells expressing both VEGFR-1 and Np-2 using our anti-Np-2 affinity purified antibodies. However, our antibodies turned out to be poor precipitating antibodies. We have therefore tagged Np-2 by expressing a myc epitope in-frame at the end of the intracellular C-terminal domain of Np-2. When we expressed the Np-2myc construct in PAE cells the cDNA directed the expression of tagged Np-2 as revealed in Western blots employing anti-Np-2 antibodies (Fig. 4A, lane 2). The amount of Np-2myc in the transfected cells was about 5-fold lower than the amount of Np-2 in our Np-2 expressing PAE cells (Fig. 4A, lane 1). However, in contrast to our expectations, we could not detect specific 125 I-VEGF 165 binding to the Np-2myc expressing PAE cells (Fig. 4B). Only at very high concentrations was there some residual specific binding. The maximal specific binding observed was at least 40-fold less than the specific binding of 125 I-VEGF 165 to PAE⅐Np-2 cells (Fig. 4B). Surprisingly, the PAE⅐Np-2myc cells were still able to bind sema-3F. The amount of sema-3F bound per cell was about 5-fold lower than the binding of sema-3F to PAE⅐Np-2 cells (Fig. 4C) and therefore in good agreement with the experiment shown in Fig. 4A. These experiments therefore indicate that the attachment of the myc epitope selectively inhibits the binding of 125 I-VEGF 165 to Np-2 but does not affect significantly the binding of sema-3F to Np-2myc. This experiment indicates indirectly that sema-3F and VEGF 165 may bind to different domains on the extracellular part of Np-2. This conclusion was further supported by an experiment showing that the binding of sema-3F to PAE⅐Np-2 cells could not be inhibited by the inclusion of 2 g/ml of VEGF 165 in the binding reaction (Fig. 5).

Np-2myc Partially Regains Its 125 I-VEGF 165 Binding Properties in the Presence of VEGFR-1, and Antibodies Directed against the myc Epitope Tag of Np-2 Co-precipitate VEGFR-1-
The previous experiments have indicated that the addition of the myc tag inhibits the VEGF binding ability of Np-2. Nevertheless, when 125 I-VEGF 165 was bound and cross-linked to PAE cells co-expressing Np-2myc and VEGFR-1, we noted that 125 I-VEGF 165 formed complexes corresponding in size to VEGF 165 ⅐Np-2myc complexes (Fig. 6A, lane 2). It therefore seems that the presence of VEGFR-1 enables Np-2myc to regain at least part of the 125 I-VEGF 165 binding ability of untagged Np-2. We have therefore attempted to co-immunoprecipitate VEGFR-1 from cells co-expressing Np-2myc and VEGFR-1 using an antibody directed against the myc epitope. 125 I-VEGF 165 was bound and cross-linked to the cells, and the anti-myc antibody was then used in the immunoprecipitation. As expected from the previous experiments, the anti-myc antibody was able to co-precipitate 125 I-VEGF 165 ⅐Np-2myc and 125 I-VEGF 165 ⅐VEGFR-1 complexes from the cells (Fig. 6B, lane 3). Because 125 I-VEGF 165 did not bind to Np-2myc in cells lacking VEGFR-1, no complex could be precipitated from such cells, nor did the anti-myc antibody precipitate any cross-linked complexes from cells co-expressing VEGFR-1 and native Np-2 following the cross-linking of 125 I-VEGF 165 to such cells (Fig. 6B,  lanes 2 and 4). These observations therefore support the results obtained using the anti-VEGFR-1 antibodies (Fig. 3B, lane 6) indicating that VEGFR-1 is able to form complexes with Np-2. in the presence of 1 g/ml heparin. Binding was carried out 2 h at 4°C. Bound 125 I-VEGF was cross-linked (CL) to the cells, which were subsequently lysed as described. Equal amounts of protein representing about 10% of the lysates were chromatographed on a 6% SDS/PAGE gel, which was subsequently dried and autoradiographed. B, the remaining 90% of the lysate from each of the binding/cross-linking reactions described under A was subjected to immunoprecipitation (IP) using an anti-VEGFR-1 antibody as described under "Experimental Procedures." Immunoprecipitates were solubilized in SDS/PAGE sample buffer, chromatographed on a 6% SDS/PAGE gel, and autoradiographed. Lane 4, immunoprecipitate from a PAE⅐Np-2 cell lysate to which 125 I-VEGF 165 was bound and cross-linked. Lane 5, immunoprecipitate from a PAE⅐VEGFR-1 cell lysate to which 125 I-VEGF 121 was bound and crosslinked. Lane 6, immunoprecipitate from a PAE⅐VEGFR-1⅐Np-2 cell lysate to which 125 I-VEGF 121 was bound and cross-linked.

DISCUSSION
Np-1 and Np-2 were originally found to function as receptors for several class 3 semaphorins that repel growing tips of axons during the development of the nervous system (35,36,41). These discoveries were followed by experiments that have demonstrated that Np-1 and Np-2 can function in addition as receptors for VEGF 165 , one of the heparin-binding forms of the angiogenic factor VEGF (32,42). These experiments indicated that the neuropilins may play a role in cardiovascular biology, in addition to their role in the nervous system. In the case of Np-1 these expectations were verified when it was shown that targeted disruption of the Np-1 gene results in severe cardiovascular defects (38). In agreement with this observation it was found that the Np-1 agonist sema-3A inhibits migration of endothelial cells (37), but the consequences of the binding of VEGF to Np-1 are not completely clear as yet. In contrast, mice lacking functional Np-2 receptors are viable, and no vascular defects were reported so far (51,52). Nevertheless, the absence of vascular defects in these gene targeted mice does not necessarily preclude a physiological role for these receptors in vascular biology because the absence of a phenotype may be explained by redundancy with other signaling pathways.
We have not been able to demonstrate any biological responses to VEGF 165 in PAE cells expressing either recombinant Np-1 or recombinant Np-2 receptors and no other types of VEGF receptors (32,42). These observations suggested that for the transduction of VEGF signals the neuropilins may perhaps have to associate with other membrane-bound proteins. Neuropilins possess short intracellular domains, and it was demonstrated that binding of sema-3A to Np-1 is not sufficient for induction of sema-3A mediated growth cone collapse (53). It was indeed found that neuropilins form complexes with plexin receptors to be able to transduce semaphorin signals (43,44). Our binding/cross-linking experiments and co-immunoprecipitation experiments indicate that in addition, Np-2 can form complexes with VEGFR-1. Our experiments also suggest that Np-1 too can associate with VEGFR-1. This was recently verified in a manuscript that was published during the preparation of this manuscript in which complexes between Np-1 and VEGFR-1 were observed using completely different methods (49).
The mechanism by which VEGFR-1 enables the binding of VEGF 121 to Np-1 and Np-2 is unclear. The binding of VEGFR-1 to the neuropilins may induce a neuropilin conformation that binds VEGF 121 with increased affinity. It is possible that VEGF 121 binds initially to VEGFR-1, placing the bound VEGF 121 in close proximity to neuropilins in cells co-expressing both receptor types and effectively increasing the affinity of the neuropilins toward VEGF 121 . The effect of VEGFR-1 on VEGF 121 binding may therefore be similar to the potentiating effect that heparin-like molecules have on the binding of VEGF 165 to neuropilins (32,33,42).
VEGFR-1 and Np-2 may be able to form complexes prior to the addition of VEGF. Alternatively, it is possible that VEGFR-1 binds to Np-2 only subsequent to the binding of were grown to confluence in 5-cm dishes (ϳ3 ϫ 10 6 cells/dish) and lysed. Lysates were chromatographed on a 6% SDS/PAGE gel. The amount of lysate loaded in each lane was equivalent to 2 ϫ 10 6 cells. The proteins were transferred to nitrocellulose by electroblotting and probed with an affinity purified anti-Np-2 antibody. A secondary antibody coupled to peroxidase was used to detect bound primary antibody. Bound secondary antibody was detected using the ECL detection method. B, PAE cells expressing Np-1, Np-2, or Np-2myc were grown in gelatin-coated 24-well dishes to confluence. Increasing concentrations of 125 I-VEGF 165 were bound in duplicates to the cells in the presence of 1 g/ml heparin at 4°C for 2 h as described to determine total binding. Nonspecific binding was measured in the presence of 1 g/ml VEGF 165 . At the end of the binding, the cells were washed three times with PBS containing 1 mg/ml bovine serum albumin and lysed by the addition of 0.5 ml of 0.5 N NaOH. Aliquots of 0.4 ml were then counted in a ␥ counter. Nonspecific binding was subtracted from total binding to calculate specific binding values. The nonspecific binding did not exceed 15% of the total binding values. The experiment was repeated twice, and the error bars represent the standard deviations from the mean. C, control PAE cells transfected with the pBabePuro plasmid (42) as well as PAE cells expressing recombinant Np-2, Np-2myc, VEGFR-1 and Np-2, or VEGFR-1 and Np-2myc were grown to confluence in gelatin-coated 48-well dishes. The cells were incubated with conditioned medium containing AP-tagged sema-3F as described. Following washing, heat-resistant alkaline phosphatase activity of cell bound sema-3F was measured as described. Binding to pBabe-puro vector transfected PAE cells was 0.08 optical density units (O.D.) and was subtracted from the total binding values to calculate the specific binding values shown. The experiment was repeated twice in triplicate with similar results, and the error bars represent the standard deviations from the mean.

VEGFR-1 and Neuropilin-2 Form Complexes
VEGF to VEGFR-1. We have not been able to differentiate between these two possibilities. We have attempted to detect co-immunoprecipitated Np-2 or VEGFR-1 using Western blot analysis, but we have failed regardless of whether the cells were exposed or not to VEGF prior to the immunoprecipitation. It is possible that these experiments failed because the sensitivity of the assays was insufficient or because the VEGFR-1⅐Np-2 complexes are sensitive to the detergents used during the solubilization of the cells, making the detection of VEGFR-1⅐Np-2 complexes by less sensitive techniques than the technique we have used difficult. To circumvent these problems we have therefore used antibodies to immunoprecipitate recombinant receptors that have been covalently cross-linked to 125 I-VEGF prior to the immunoprecipitation using a previously described method (50). The method we used utilized the high sensitivity afforded by the use of 125 I but did not allow us to determine whether complex formation between Np-2 and VEGFR-1 was VEGF-dependent or not.
We have no data regarding the biological significance of VEGFR-1⅐Np-2 complexes at this stage. The formation of complexes between Np-2 and VEGFR-1 may contribute to VEGFinduced signal transduction by VEGFR-1. If that assumption is correct, then it may provide a clue to a puzzling observation. Mice deficient in VEGFR-1 expression die before birth because of severe cardiovascular defects (27). In contrast, mice that retain the extracellular and trans-membrane domains of VEGFR-1 but lack the signaling tyrosine-kinase domain develop normally (54). It is unclear how the extracellular domain of VEGFR-1 is able to restore the normal embryonic development of mice. It is possible that the extracellular domain is required for VEGF sequestration, so as to limit the activity of VEGF. Alternatively, the extracellular domain may associate with another membrane protein to form a signaling holo-receptor. It is possible that Np-2 and Np-1 may perhaps participate in the formation of such a putative VEGFR-1 containing holo-receptor.
In the course of our experiments we have found that when a myc epitope is inserted in-frame after the conserved SEA terminal tripeptide of Np-2a, the modified Np-2myc receptor loses most of its VEGF 165 binding ability. It was shown that Np-1 FIG. 5. VEGF 165 does not inhibit the binding of sema-3F to Np-2. PAE cells expressing Np-2, VEGFR-1, or VEGFR-1 and Np-2 as well as control PAE cells transfected with the pBabePuro plasmid (42) were grown in gelatin-coated 48-well dishes to confluence. The cells were incubated with conditioned medium containing AP-tagged sema-3F at room temperature for 3 h. The binding was performed in the absence (black columns) or the presence of 2 g/ml VEGF 165 (open columns). The cells were then washed and assayed for bound heat stable AP activity as described. The experiment was repeated twice in triplicate with similar results. The error bars represent the standard deviations from the mean. O.D., optical density units. were grown to confluence in 10-cm dishes. 125 I-VEGF 165 (10 ng/ml) was bound to the cells for 2 h at 4°C in the presence of 1 g/ml heparin. At the end of the experiment, bound 125 I-VEGF was cross-linked to the cells using BS 3 as described. Following cross-linking, the cells were lysed, and immunoprecipitation was carried out using anti-myc antibodies as described under "Experimental Procedures." Precipitated 125 I-labeled complexes were solubilized using SDS/PAGE sample buffer and chromatographed on a 6% SDS/PAGE gel, and the gel was then dried and autoradiographed. and Np-2 can form homodimers and heterodimers (55). Formation of such dimers may perhaps be required for high affinity binding of VEGF to neuropilins. The insertion of the myc epitope may perhaps interfere with dimer formation and consequently with VEGF binding. Interestingly, the VEGF 165 binding ability of Np-2myc was restored to some extent in cells co-expressing VEGFR-1, perhaps because following complex formation a high affinity VEGF-binding conformation of Np-2myc is favored. Interestingly, the sema-3F binding properties of Np-2 were not affected by the introduction of the myc epitope, perhaps because the sema-3F-binding domain of Np-2 seems to be distinct from the VEGF-binding domain as suggested by the results of the competition experiments.
To conclude, our experiments indicate that VEGFR-1 forms complexes with Np-2 and possibly also with Np-1. The presence of VEGFR-1 changes the specificity of VEGF binding, allowing VEGF 121 to bind to Np-2. However, the biological function of these VEGFR-1⅐Np-2 complexes is still unclear. In addition our experiments indicate that changes in the intracellular domain of Np-2 can affect VEGF binding to Np-2 and provide evidence indicating that VEGF and semaphorins bind to nonoverlapping sites in the extracellular part of Np-2.