Site-directed Mutagenesis in the B-Neuropilin-2 Domain Selectively Enhances Its Affinity to VEGF165, but Not to Semaphorin 3F*

Neuropilins (NRPs) are 130-kDa receptors that bind and respond to the class 3 semaphorin family of axon guidance molecules (SEMAs) and to members of the vascular endothelial growth factor (VEGF) family of angiogenic factors. Two NRPs have been reported so far, NRP1 and NRP2. Unlike NRP1, little is known about NRP2 interactions with its ligands, VEGF165 and SEMA3F. Cell binding studies reveal that VEGF165 and SEMA3F bind NRP2 with similar affinities, 5.2 and 3.9 nm, respectively, and are competitive NRP2 ligands. Immunoprecipitation studies show that the B (b1b2) extracellular domain of NRP2 is sufficient for VEGF165 binding, whereas SEMA3F requires both the A (a1a2) and B domains. To identify residues of B-NRP2 involved in VEGF165 binding, point mutations were introduced by site-directed mutagenesis. VEGF165 is a basic protein. Reduction of the electronegative potential of B-NRP2 by exchanging acidic residues for uncharged alanine (B-NRP2 E284A,E291A) in the 280–290 b1-NRP2 loop resulted in a 2-fold reduction in VEGF165 affinity. Conversely, enhancing the electronegative potential (B-NRP2 R287E,N290D and R287E,N290S) significantly increased VEGF165 affinity for B-NRP2 by 8- and 6.6-fold, respectively. The mutagenesis did not affect SEMA3F/B-NRP2 interactions. These results demonstrate that it is possible to alter VEGF165 affinity for NRP2 without affecting SEMA3F affinity. They also identify NRP2 residues involved in VEGF165 binding and suggest that modifications of B-NRP2 could lead to potentially high affinity selective inhibitors of VEGF165/NRP2 interactions.

Neuropilins (NRP1 and NRP2) 3 are transmembrane receptors for members of the class 3 semaphorin family of neuronal guidance mediators (SEMA3A-G) and for the vascular endo-thelial growth factor (VEGF) family of angiogenic factors. NRPs play a prominent role in neuronal wiring, normal blood vessel development, and tumor angiogenesis (1)(2)(3)(4)(5). NRP1 and NRP2 are often differentially expressed. In the neuronal system, NRP1 is expressed on sympathetic and sensory neurons, whereas NRP2 is detected on sympathetic neurons. Binding of SEMA3 to NRPs and complexing with plexins results in axon repulsion and growth cone collapse (6 -8). There is a specificity in SEMA3s binding to NRPs. For example, SEMA3A acts via NRP1, whereas SEMA3F is functional via NRP2 (9 -11).
Recently, it has been demonstrated that SEMA3E acts independently of NRPs and is instead a direct ligand for plexin D1 (12).
In the vascular system, NRP1 is confined to the arterial compartment, whereas NRP2 is expressed by venous and lymphatic endothelial cells (EC) (13)(14)(15)(16). It is apparent that NRP1 promotes angiogenesis in vivo. NRP1 and NRP1/NRP2 mouse knockouts have major defects in yolk sac and embryonic vasculature (17,18). NRP2 mutant mice show abnormal development of small lymphatic vessels and capillaries, suggesting a selective role for NRP2 in lymphatic vessel development (13). In zebrafish embryos, NRP1 knockdown with morpholinos results in loss of the angiogenic intersegmental vessels (19,20).
The ability of two structurally disparate proteins, VEGF 165 and SEMA3s, to bind to one single receptor (NRP) has engendered considerable interest in NRP structure. That VEGF 165 and SEMA3A (Collapsin-1) are competitive inhibitors in EC binding, EC migration, and dorsal root ganglia collapse assays suggests possible overlapping binding sites on NRP1 (23). NRP1 and NRP2 have about 40% amino acid identity but share similar domain structure (1) composed of a large extracellular domain (ϳ840 amino acids), a very short transmembrane region (25 amino acids) and a short cytoplasmic sequence (42 amino acids) in which no signaling consensus sequences have been identified so far. The NRP extracellular domain contains: (i) an A domain consisting of two a-domain repeats (a1a2), (ii) a B domain consisting of two b-domains repeats (b1b2), and (iii) a C domain, most likely involved in NRP oligomerization (35)(36)(37)(38)(39). Our previous studies using pull-down methods indicated that VEGF 165 binds solely to the B domain of NRP1 (40). Domain deletion studies have also demonstrated that the B-NRP1 domain is essential for VEGF 165 binding (41). In addition, both A-and B-NRP1 domains are needed for SEMA3A binding (37,41).
Compared with NRP1, the structural features of NRP2 necessary for ligand binding have not been reported yet. In this report, we show that SEMA3F competes with VEGF 165 for NRP2 binding and that VEGF 165 needs only the B domain for optimal binding, whereas SEMA3F needs both the A and B domains. Moreover, point mutations were made by site-directed mutagenesis to pinpoint VEGF 165 binding sites based on information provided by the crystal structure of the b1-NRP1 domain (42) and by the over 50% identity of the NRP1 and NRP2 b1 domains. VEGF 165 is a basic protein. Mutation of 2 electronegative glutamates to neutral alanines in the 280 -290 loop of b1-NRP2 decreased VEGF 165 affinity by 2-fold. On the other hand, enhancing the electronegative potential of the same b1 loop increased VEGF affinity by 8-fold. The point mutations do not affect SEMA3F binding. These studies identify specific NRP2 residues that bind VEGF 165 . Furthermore, because B-NRP2 can inhibit VEGF 165 binding to EC, further modifications of this NRP2 region to increase its affinity for VEGF 165 could lead to more potent VEGF 165 antagonists to target angiogenesis. Cell Culture-Parental porcine aortic endothelial cells (PAEC) were kindly provided by Dr. Lena Claesson-Welsh (University of Uppsala, Uppsala, Sweden) (43). PAEC NRP1 and NRP2 were established as previously described (21,34). PAEC were grown in Ham's F-12 medium containing 10% fetal bovine serum and 1% L-glutamine/penicillin G/streptomycin sulfate. Human umbilical vein endothelial cells (HUVEC) were from Clonetics/BioWhittaker (Walkersville, MD) and were cultured in endothelial basal media-2 supplemented with endothelial growth media-2 SingleQuots. HEK293 cells were from American Type Culture Collection (Manassas, VA) and cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 1% L-glutamine/penicillin G/streptomycin sulfate.

Materials-PfuTurbo
Cloning and Mutagenesis-The A-, B-, and AB-NRP2 domains were PCR amplified using the human NRP2 construct in the pcDNA3.1(ϩ) vector as a template. The sites of HindIII and EcoRV were added to the forward and reverse primers, respectively, for consequent enzymatic digestion. PfuTurbo TM High Fidelity DNA polymerase was used for the PCR (95°C for 2 min, 30 cycles: 95°C for 30 s, 58°C for 30 s, and 72°C for 1 min; 72°C for 10 min) together with the following primer pairs: A-NRP2, forward, 5Ј-CTTACCAAGCTTTGCGGAGGTCG-TTTGAATC and reverse, 5Ј-TACAGATATCTTGACCAG-GTAGTAACGCC; B-NRP2, forward, 5Ј-CTTACCAAGCT-TTGCAATGTTCCTCTGGGC and reverse, 5Ј-TACAGAT-ATCTTACAGCCCAGCACCTCCAG; AB-NRP2, forward, 5Ј-CTTACCAAGCTTTGCGGAGGTCGTTTGAATC and reverse, 5Ј-TACAGATATCTTACAGCCCAGCACCTCCAG. The PCR were carried out in a Mastercycler Gradient (Eppendorf Scientific, Westbury, NY). The PCR products were subcloned in pCRII-TOPO and an enzymatic digestion with Hin-dIII and EcoRV followed. The fragments were ligated with the DNA Ligation Kit into the pSecTag2B mammalian expression vector fused at the N terminus with the murine Ig chain coding sequence and at the C terminus with the Myc epitope and a His 6 coding sequence. The full-length, human SEMA3F vector was a gift of Dr. M. Tessier-Lavigne (UCSF, San Francisco, CA). The SEMA3F V5-His 6 -tagged was prepared by cloning from the above pSecTag construct onto the pcDNA 3.1 V5-His 6 vector. Briefly, the cDNA was amplified from the pSecTag vector including the ATG start codon and the Ig chain coding sequence by PCR (94°C for 4 min, 15 cycles: 94°C for 30 s, 52°C for 30 s, 68°C for 5 min; 68°C for 4 min) with the following primers: forward, 5Ј-GCCACCATGGAGACAGAGACACTC and reverse, 5Ј-GTCCGGAGGGTGGTGCCGGCG. The PCR product was cloned into the pcDNA 3.1 V5-His 6 vector.
Expression and Purification of NRP2 Extracellular Domains and Semaphorins-The domain constructs were transiently transfected overnight into HEK293 cells using the transfection reagent FuGENE 6. The following day the culture medium was replaced by serum-free CD 293 medium and the conditioned media were collected after 48 h. The conditioned media were centrifuged at 3500 ϫ g for 20 min, adjusted to 30 mM Na 2 HPO 4 , pH 7.6, 150 mM NaCl, filtered on a 0.22-m filter, and applied onto a 1-ml HiTrap TM chelating column loaded with NiSO 4 and attached to a fast protein liquid chromatography system. After washing with 30 mM Na 2 HPO 4 , pH 7.6, 150 mM NaCl, 5 mM imidazole, the proteins were eluted with 30 mM Na 2 HPO 4 , pH 7.6, 150 mM NaCl, 250 mM imidazole. The fractions were analyzed by SDS-PAGE followed by SYPRO Ruby staining and the positive fractions were subsequently desalted on a HiTrap desalting column. The protein concentrations were determined using the Quick Start TM Bradford dye reagent. The same experimental procedures were used for the purification of SEMA3F.
Iodination of VEGF 165 and SEMA3F and Binding to PAEC NRP2-VEGF 165 and SEMA3F were iodinated using IODO-BEADS as previously described by Soker et al. (21). Specific activities of ϳ60,000 and ϳ10,000 cpm/ng were obtained for the two proteins, respectively.
To determine the dissociation constants (K d values) of 125 I-VEGF 165 and 125 I-SEMA3F for PAEC NRP2, the cells were incubated in 96-or 48-well plates with increasing concentrations of the radioligand for 1 h at 4°C in binding buffer (Ham's F-12, 20 mM HEPES, pH 7.5, 0.5% bovine serum albumin, 1 g/ml heparin sodium salt). The supernatant was aspirated, the cells were thoroughly washed with cold PBS, lysed with 0.2 N NaOH, and the cell-associated radioactivity was measured in a ␥-counter. The values at each concentration point were corrected for the nonspecific binding obtained on the parental PAEC. The dissociation constants (K d values) were calculated from average values of at least two experiments using the Software GraphPad nonlinear regression fitting parameters for one-site binding.
For the competition-binding experiments, PAEC NRP1 and NRP2 were incubated in 48-well plates with 125 I-VEGF 165 (5 ng/ml) in binding buffer at increasing concentrations of cold SEMA3F (0 to 250 nM) for 1 h at 4°C or PAEC NRP2 were incubated with 125 I-SEMA3F (25 ng/ml) in binding buffer at increasing concentrations of cold VEGF 165 (0 to 200 nM). After-ward, the cells were washed with cold PBS, lysed with 0.2 N NaOH, and cell-associated radioactivity was measured in a ␥-counter. The experiment was repeated in duplicates two to three times.
Co-immunoprecipitation of 125 I-VEGF 165 and SEMA3F with the NRP2 Domains-125 I-VEGF 165 (5 ng/ml) was preincubated in binding buffer (PBS, pH 7.4, 0.5% bovine serum albumin, 0.1% Tween 20, 1 g/ml heparin sodium salt) with equal amounts of purified Myc-His 6 -tagged domains at room temperature for 2 h. The complexes were immunoprecipitated overnight at 4°C with an anti-Myc antibody. Protein G slurry (20 l) was added to the immunocomplexes and an incubation of 1 h at 4°C followed. The beads were washed with PBS, boiled in SDS sample buffer for 10 min at 95°C, and analyzed by SDS-PAGE on a 15% gel followed by autoradiography. For quantification, an aliquot of the samples was measured in a ␥-counter. In the case of SEMA3F, 25 g/ml of V5-tagged SEMA3F were incubated with equal amounts of A-, B-, and AB-NRP2 in binding buffer for 2 h at room temperature. The complexes were immunoprecipitated with an anti-Myc antibody overnight at 4°C. Protein G slurry (20 l) was added to the immunocomplexes and an incubation of 1 h at 4°C followed. The beads were washed with PBS, boiled in SDS sample buffer for 10 min at 95°C, and analyzed by SDS-PAGE on a 7.5% gel followed by Western blot analysis with an anti-V5 antibody. The intensity of the bands was quantified with ImageJ. The dissociation constants (K d values) of the different domains and B-NRP2 mutants for VEGF 165 and SEMA3F were calculated. Five ng/ml of 125 I-VEGF 165 or 25 ng/ml of 125 I-SEMA3F V5-His 6 -tagged were preincubated in binding buffer at room temperature with increasing concentrations (0 to 50 nM) of the different proteins. Immunoprecipitation of the complexes was carried out and the Protein G bead-associated radioactivity was measured in a ␥-counter. The values obtained were plotted and the curves were fitted using the Origin5.0 software (Microcal Inc., Northampton, MA) with nonlinear regression fitting parameters for one-site binding.
Inhibition of 125 I-VEGF 165 and 125 I-SEMA3F Binding to PAEC NRP2 and HUVEC by the NRP2 Domains-The radioligands at a concentration of 5 ng/ml ( 125 I-VEGF 165 ) or 25 ng/ml ( 125 I-SEMA3F) were preincubated in binding buffer (Ham's F-12, 20 mM HEPES, pH 7.5, 0.5% bovine serum albumin, 1 g/ml heparin sodium salt) with increasing concentrations of the discrete NRP2 domains (0, 10, 50, 300, and 1000 nM) at room temperature for 2 h. Afterward the complexes were added to the cells in 48-well plates and incubated at 4°C for 1 h. The supernatants were aspirated, the cells were washed with cold PBS, lysed in 0.2 N NaOH, and the cell-associated radioactivity was measured in a ␥-counter. The results are expressed as percentages of the cell-bound radioligand alone. IC 50 values were calculated from the graphs (using Origin5.0) as the concentrations of the respective domain necessary to reduce the amount of cell-bound radioligand to 50% of the initial value.

RESULTS
VEGF 165 and SEMA3F Are Competitive NRP2 Ligands-Porcine aortic endothelial cells (PAEC) do not express VEGF 165

B-NRP2 Mutants with Selective Enhanced Affinity for VEGF 165
receptors or semaphorin receptors and can be engineered to overexpress one receptor at a time, for example, NRP2 or NRP1 (21,34) . The K d of VEGF 165 and SEMA3F for NRP2 were calculated from the binding curves of 125 I-VEGF 165 and 125 I-SEMA3F for PAEC NRP2. The K d for the 2 ligands were fairly similar, 5.2 Ϯ 3.7 nM for VEGF 165 (Fig. 1A) and 3.9 Ϯ 1.5 nM for SEMA3F (Fig. 1B), respectively. To determine whether VEGF 165 and SEMA3F are competitive inhibitors, 125 I-VEGF 165 (5 ng/ml, 0.12 nM) was incubated with PAEC NRP2 or PAEC NRP1 at increasing unlabeled SEMA3F concentrations (0 to 250 nM) (Fig. 1C). 125 I-VEGF 165 binding to PAEC NRP2 was inhibited by up to 80% at 100 nM cold SEMA3F (830-fold molar ratio of SEMA3F/ 125 I-VEGF 165 ). On the other hand, inhibition of 125 I-VEGF 165 binding to PAEC NRP1 was only 15% at the same concentration of cold SEMA3F. The diminished ability of SEMA3F to inhibit 125 I-VEGF 165 binding to NRP1 compared with NRP2 is consistent with SEMA3F having a 10-fold lower affinity for NRP1 compared with NRP2 (10).
In a complementary experiment, PAEC NRP2 were incubated with 125 I-SEMA3F (25 ng/ml, 0.13 nM) and increasing concentrations of cold VEGF 165 (0 to 200 nM). VEGF 165 was able to compete with 125 I-SEMA3F for NRP2 binding sites on PAEC NRP2. At the highest concentration of unlabeled VEGF 165 (200 nM, about 1540-fold molar ratio VEGF 165 / 125 I-SEMA3F) 53% of 125 I-SEMA3F was displaced from the cell surface by VEGF 165 (Fig. 1D). The binding of 125 I-SEMA3F to PAEC NRP1 was so low, almost null as expected, that competition with cold VEGF 165 could not be measured and was moot.  (40). To identify the VEGF 165 binding sites on NRP2, discrete Myc-His 6 -tagged extracellular domains of NRP2, A, B, and AB, were cloned, expressed, and purified from HEK293 conditioned media ( Fig. 2A). Whereas the B-NRP2 domain is the main site of interaction between NRP2 and VEGF 165 , SEMA3F interacts only weakly with B-NRP2 and requires both the A and B domains of NRP2 for optimal interaction (Fig. 2C). Equal amounts of V5-His 6 -tagged SEMA3F (5 g) were incubated with equal amounts of each of the NRP2 domains and these proteins were co-immunoprecipitated with anti-Myc antibody, followed by Western blot with an anti-V5 antibody. Relatively little SEMA3F interaction was observed with A-NRP2 (lane 2) or with B-NRP2 (lane 3). However, a substantial amount of SEMA3F was pulled down with AB-NRP2, 15-and 10-fold more than with A-and B-NRP2, respectively (Fig. 2C, lane 4  versus 2 and 3, respectively).
The co-immunoprecipitation results (Fig. 2, B and C) were confirmed by analyzing the binding curves of the different NRP2 domains for the two ligands. For VEGF 165 , a dose-dependent binding to B-NRP2 and AB-NRP2, but not to A-NRP2, was observed (Fig. 2D). The K d values were 5.9 Ϯ 0.6 nM for B-NRP2 and 8.3 Ϯ 1.3 nM for AB-NRP2 but not detectable for A-NRP2, suggesting that the A domain did not make a significant contribution to VEGF 165 binding. Interestingly, the B domain alone (5.9 nM) bound VEGF 165 as well as the whole NRP2 protein did (5.2 nM). Binding curves showed dose-dependent binding of SEMA3F to AB-NRP2, lower to B-NRP2 but none to A-NRP2 (Fig. 2E). The values of the K d values were 1.4 Ϯ 0.5 and 2.6 Ϯ 0.4 nM, respectively, for B and AB-NRP2, and not detectable for A-NRP2.
Even though the affinity of SEMA3F for B-NRP2 was 4-fold higher than VEGF 165 values, the capacity of B-NRP2 for VEGF 165 was significantly higher than for SEMA3F. Indeed, at saturation about 30% of the total VEGF 165 was bound to B-NRP2, whereas only about 4% of the total SEMA3F was bound to B-NRP2. Taken together these data suggest a preferential interaction of B-NRP2 with VEGF 165 rather than with SEMA3F.
Effects of NRP2 Domains on VEGF 165 and SEMA3F Binding to EC-Domain-mediated inhibition of VEGF 165 and SEMA3F binding to NRP2 was also tested out at a cellular level using PAEC NRP2 (Fig. 3). A-NRP2 was not effective in displacing 125 I-VEGF 165 from PAEC NRP2. On the other hand, about 80% of the 125 I-VEGF 165 was displaced from the cell surface by both B-and AB-NRP2 at the final concentration of 1000 nM (Fig. 3A). 125 I-SEMA3F displacement from the PAEC NRP2 was most effective using AB-NRP2 and much less effective using A-NRP2 or B-NRP2 (Fig. 3B). Similar results for both ligands were obtained on HUVEC (not shown). IC 50 values for these inhibitions are reported in Table 1. Taken together, the competitionbinding experiments on EC confirm the immunoprecipitation results.
Engineering B-NRP2 Mutants with Enhanced VEGF 165 Binding Affinity-As shown above, the B-NRP2 domain by itself or together with the A-NRP2 domain inhibits 125 I-VEGF 165 binding to PAEC NRP2 (Fig. 3A). One goal of this study was to identify residues of B-NRP2 involved in VEGF 165 binding. A NRP2 structure has not been solved so far. However, the crystal structure of the b1 domain of NRP1 (b1-NRP1) has been reported (42) and recently, the complete NRP1 B domain (b1b2) has been solved as well (44). The b1 domains of NRP1 and NRP2 share a high sequence identity (52%) that allowed us to take advantage of the structural information of b1-NRP1 to perform mutagenesis studies in the b1-NRP2 domain. The crystal structure of b1-NRP1 reveals a marked polarity in the charge distribution with an electropositive surface (42) possibly responsible for the interaction with heparin (40) and an electronegative area, possibly a site for interactions with Semaphorin3A and VEGF 165 , both of which are basic proteins. A region in the b1-NRP2 280 -290 acidic loop with homology to a corresponding sequence in b1-NRP1 was chosen first as a site for making mutations (Fig. 4A).
To identify residues of B-NRP2 possibly involved in VEGF 165 binding, two glutamate residues of the loop, Glu-284 and Glu-291, which correspond to NRP1 Glu-282 and Asp-289, respectively, were exchanged with neutral alanine by site-directed mutagenesis to reduce the B-NRP2 electronegative potential to obtain B-NRP2 E284A,E291A (designated as POS) (Fig. 4A). The construct was transiently transfected into HEK293 cells and the collected conditioned media were purified as described under "Experimental Procedures." The relative K d values of the mutant for VEGF 165 was determined by immunoprecipitation assays ( Table 2). The double alanine mutation increased the K d for VEGF 165 by 2-fold, from 5.9 nM of the wild type (wt) B-NRP2 to 12.0 nM ( Fig. 4B and Table 2). By contrast, the K d of this mutant for SEMA3F was virtually unchanged (1.05 versus 1.44 nM of the WT B-NRP2) (Fig. 4C and Table 2).
On the other hand, when the electronegative potential of the B-domain was increased significant improvements in K d values for VEGF 165 occurred. In particular, arginine 287 (Arg-287) was changed to glutamate and asparagine 290 (Asn-290) was changed to serine or aspartate to obtain double mutants antibody. The bottom panel shows a densitometric quantification of the bands (ImageJ); D and E, binding of 125 I-VEGF 165 and 125 I-SEMA3F to NRP2 domains. 125 I-VEGF 165 (5 ng/ml; ϳ70,000 cpm total) (D) or 125 I-SEMA3F (25 ng/ml, ϳ55,000 cpm total) (E) were preincubated at room temperature for 2 h with increasing concentrations of the Myc-tagged A-, B-, and AB-NRP2 domains, immunoprecipitated by Myc antibody, and bound radioactivity was determined. The capacity of binding was also measured. At saturation, the percentages of total 125 I-VEGF 165 bound to B-and AB-NRP2 (about 27 and 32%, respectively) are both significantly higher than the corresponding percentages of bound 125 I-SEMA3F (3.5 and 15%, respectively). Thus the capacity of B-NRP2 for VEGF 165 is about 8-fold higher than for SEMA3F. B-NRP2 R287E,N290S and R287E,N290D (designated as NEG1 and NEG2, respectively) (Fig. 4A). The K d values of the two mutants for VEGF 165 were 0.89 and 0.74 nM, respectively, showing an enhancement in affinity of 6.6-and 8-fold, respectively, compared with WT B-NRP2 (Fig. 4B and Table 2). Thus, these increases in electronegativity increase VEGF 165 affinity by almost a log.
The K d values of these B-NRP2 domain mutants for SEMA3F were determined and were not significantly changed. The K d values of NEG1 and NEG2 for SEMA3F were 1.78 and 0.98 nM, respectively, compared with 1.44 nM for WT B-NRP2, showing that SEMA3F binding was not affected (Fig. 4C and Table 2).
The ability of the mutants to displace 125 I-VEGF 165 and 125 I-SEMA3F from the surface of PAEC NRP2 was deter-mined as well. Consistent with the dissociation constants, the IC 50 values of mutants NEG1 and NEG2 (166 and 184 nM, respectively) for 125 I-VEGF 165 showed an improvement of  Table 1.  NEG2). B and C, binding assays: increasing concentrations (0 -10 nM) of B-NRP2 wild type (wt) and mutants POS, NEG1, and NEG2 were preincubated, respectively, with 5 ng/ml 125 I-VEGF 165 (70,000 cpm total) (B) or 25 ng/ml of 125 I-SEMA3F V5-tagged (55,000 cpm total) (C) for 2 h at room temperature. The complexes were immunoprecipitated with an anti-Myc antibody and the associated radioactivity was determined in a ␥-counter. The mutants NEG1 and NEG2 have lower K d values for VEGF 165 than the WT B-NRP2 as shown in Table 2. The K d values for SEMA3F on the other hand are unchanged.  Fig. 3 (A and B) using the Software Origin5.0 (Microcal Inc.) as the concentrations of the respective domain necessary to reduce the amount of cell-bound radioligand to 50% of the initial value.

NRP2 domain IC 50 (nM)/ 125 I-VEGF 165 IC 50 (nM)/ 125 I-SEMA3F
A  5A and Table 2). No changes were observed between the IC 50 values of B-NRP2 WT and the mutants in the case of SEMA3F (Fig. 5B and Table 2).

DISCUSSION
NRP1 and NRP2 play an important role in the normal development of the neuronal and vascular systems and in pathological tumor progression, tumor angiogenesis, and metastasis.
NRPs are atypical in that they bind two disparate families of ligands, the VEGFs that are pro-angiogenic and the class-3 semaphorins that, besides repelling axons, are anti-angiogenic. This ability to bind two unrelated ligands suggests that the NRP structure must have a defining role in mediating VEGF 165 and semaphorin binding and activity. The pair of NRP1 ligands, VEGF 165 and Collapsin-1 (chicken homolog for the human SEMA3A), competed for NRP1 binding and showed opposite effects on a dorsal root ganglia collapse assay and on an in vitro angiogenesis assay (23). The interactions of VEGF 165 and semaphorins with NRP2 are not as well studied; thus, we decided to look more closely at these interactions, with an emphasis on NRP2 structural requirements for binding.
In this report we have 1) determined the K d values for VEGF 165 and SEMA3F binding to NRP2-expressing cells and showed competition between these 2 ligands for NRP2 binding; 2) determined that whereas the B-NRP2 domain is sufficient for VEGF 165 binding, both the A-and B-NRP2 domains are necessary for SEMA3F binding; 3) used site-directed mutagenesis to identify amino acid residues in the B-NRP2 domain involved in VEGF 165 , but not SEMA3F, binding; and 4) showed that these (up to 8-fold) higher affinity B-NRP2 mutants have enhanced inhibition of VEGF 165 binding to endothelial cells but no effect on SEMA3F binding compared with WT B-NRP2.
PAEC NRP2 were used to determine the affinities of VEGF 165 and SEMA3F for NRP2, because they have the advantage of expressing NRP2 but not any of the VEGF 165 receptor tyrosine kinases that could compromise interpretation of VEGF 165 binding results. The calculated dissociation constants for VEGF 165 and SEMA3F were similar, 5.2 and 3.9 nM, respectively. These K d values were higher than the previously reported K d of 0.13 nM for VEGF 165 /NRP2 (45) and 0.09 nM for SEMA3F/ NRP2 (10). For SEMA3F, one possibility is that Chen et al. (10) expressed the a5 isoform of NRP2, whereas we have expressed the a22 isoform. There are also differences in cell lines used for NRP2 overexpression, namely PAEC versus COS-7. Different cell lines with different cell surface proteoglycan expression patterns might result in different dissociation constants. It should be noted that there have also been discrepancies reported for VEGF 165 binding to NRP1, with reported K d values of 0.18 (45) and 14.4 nM (46), respectively. SEMA3A and VEGF 165 compete for binding to NRP1 (23). Accordingly, the possibility of competition between SEMA3F or VEGF 165 with NRP2 was examined. Cold SEMA3F could efficiently compete with 125 I-VEGF 165 for binding to PAEC  Table 2. The B-NRP2 mutants NEG1 and NEG2 are more effective than WT B-NRP2 in inhibiting 125 I-VEGF 165 binding to PAEC NRP2. No significant differences were observed between the mutants and the WT B-NRP2 in the case of 125 I-SEMA3F.

TABLE 2 K d values of B-NRP2 wild type (wt) and mutants for the binding to 125 I-VEGF 165 and 125 I-SEMA3F and IC 50 values for the inhibition of 125 I-VEGF 165 and 125 I-SEMA3F binding to PAEC NRP2 by B-NRP2 wt and mutants, respectively
The K d values were obtained by fitting the binding curves in Fig. 4 (B and C) using the Origin5.0 (Microcal Inc.) with nonlinear regression fitting parameters for one-site binding. The IC 50 values were calculated from the curves in Fig. 5 (A and B) using the Origin5.0 as the concentrations of the respective domain necessary to reduce the amount of cell-bound radioligand to 50% of the initial value.

B-NRP2 Mutants with Selective Enhanced Affinity for VEGF 165
NRP2, with 80% displacement at 100 nM, whereas binding to PAEC NRP1 was displaced by only 15% at the same concentration. This result is consistent with previous observations that SEMA3F is a much more specific ligand, by 10-fold, for NRP2 than for NRP1 (10). In complementary experiments, cold VEGF 165 was not as effective in inhibiting 125 I-SEMA3F binding to PAEC NRP2. The maximal inhibition was 53%. Previously it has been reported that VEGF 165 did not inhibit SEMA3F binding to PAEC NRP2 at all (47). It may be that because SEMA3F, but not VEGF 165 , binds A-NRP2, VEGF 165 may not be an optimal competitor of SEMA3F-NRP2 interaction because it cannot compete at the A domain level. The competition of VEGF 165 and SEMA3F for binding to NRP2 suggested that there are overlapping binding sites on this receptor. The extracellular domain of NRP2 has 3 subdomains, A, B, and C, the last involved in receptor dimerization (35)(36)(37)(38)(39)(40)(41).
To determine NRP2 ligand binding sites, discrete NRP2 extracellular domains (A-, B-, and AB-NRP2) were cloned and used in pull-down experiments. The conclusion was that VEGF 165 requires only the B-NRP2 domain for binding, whereas SEMA3F requires both the A and B domain. Previously we had shown similar results for NRP1 (40). VEGF 165 binding to NRP1 showed that the B-NRP1 domain was sufficient for binding VEGF 165 . However, differently than NRP2, the additional presence of A-NRP1 (AB-NRP1) resulted in markedly improved VEGF 165 binding to NRP1. Our VEGF 165 NRP2 binding data are consistent with results recently published by Kärpänen et al. (48).
Semaphorin 3A has a SEMA domain that binds to A-NRP1, an immunoglobulin (Ig) domain, and a C-terminal basic domain that interacts with both A-and B-NRP1 (37). Mutations in the N-terminal a-domain repeat of NRP1 (a1-NRP1) completely disrupts interaction with SEMA3A while maintaining VEGF 165 binding capacity (11). In our analysis of SEMA3F-NRP2 interaction, both A and B (AB-NRP2) were necessary for optimal interaction with SEMA3F consistent with previous analysis showing that SEMA3A requires both A-NRP1 and B-NRP1 for optimal binding (41). Comparisons of affinity constants showed that VEGF 165 affinity for the B domain alone was about the same as full-length NRP2 expressed on cells. Similarly, the K d values for SEMA3F binding to the isolated AB domain and full-length NRP2 were very similar. These results show that data obtained by binding of discrete domains can be quite accurate. Further support for identifying the role of the A and B domains in VEGF 165 and SEMA3F binding came from IC 50 values of cell-based competition binding studies. Both ABand B-NRP2 domains inhibited binding of VEGF 165 to PAEC NRP2, whereas only AB-NRP2 inhibited SEMA3F binding. Similar results were obtained with HUVEC, a non-engineered EC type expressing NRP2 (not shown).
Affinity is not the only measure of efficacy of ligand/NRP2 interactions. Our binding experiments show that the percentages of VEGF 165 bound to B and AB-NRP2 at saturation (about 27 and 32%, respectively) were both significantly higher than the corresponding percentages of bound SEMA3F (3.5 and 15%, respectively). Taken together these data suggest a preferential interaction of B-NRP2 with VEGF 165 rather than with SEMA3F.
By all accounts, VEGF 165 binds to the B-NRP2 domain. To identify amino acids in B-NRP2 involved in VEGF 165 binding, advantage was taken of the published crystal structure of the b1 domain of NRP1 (42) and of the high sequence identity (52%) between b1-NRP1 and b1-NRP2. The b1 domain has the characteristic folding of discoidin family protein members. It consists of eight ␤ strands, a five-stranded anti-parallel ␤-sheet packed against a three-stranded anti-parallel ␤-sheet, giving rise to a ␤-sandwich fold (42). Six loops (L1-L6) extend from the top of the sandwich and define a groove (42), which has been recently demonstrated to be the binding site for Tuftsin (a VEGF exon 8 peptide analogue) (44).
The b1 structure reveals a marked polarity in charge distribution, with electropositive and electronegative regions. Because VEGF 165 and SEMA3F are basic, the electronegative region is a possible binding site for these two ligands. Site-directed mutagenesis of B-NRP2 in the 280 -290 loop was used to substitute amino acids with the aim of altering B-NRP2 charge and looking for effects on VEGF 165 binding. Mutagenesis sites in b1-NRP2 were chosen so that the corresponding positions in b1-NRP1 were occupied by identical/homologous residues, and thus less likely to induce structural alterations. Exchange of electronegative residues to neutral ones showed a 2-fold reduced affinity for VEGF 165 compared with B-NRP2 WT. The affinity for SEMA3F was unchanged. On the other hand, increasing the electronegative charge by two different amino acid substitutions decreased the K d s to 0.89 and 0.74 nM, respectively, compared with 5.9 nM for B-NRP2 WT, an increase in VEGF 165 affinity of 6.6-and 8-fold, respectively. In contrast, the K d s of the mutants for SEMA3F (1.78 and 0.98 nM, respectively) were only slightly altered compared with B-NRP2 WT (1.44 nM). These results are significant in identifying NRP2 amino acid residues involved in VEGF 165 binding, 2 glutamate residues (Glu-284 and Glu-291) in b1-NRP2. Interestingly, they are involved in VEGF 165 binding but not SEMA3F binding. Thus, whereas VEGF 165 and SEMA3F are competitive inhibitors for NRP2 binding, there appear to be some non-overlapping residues.
The co-crystallization of B-NRP1 with Tuftsin has been recently reported (44). Tuftsin (TKPR) is partially homologous to the VEGF exon 8 sequence (CDKPRR). The co-crystallization data suggest that Tuftsin establishes contacts with residues Tyr-297, Asp-320, Ser-346, Thr-349, and Tyr-353. No contacts were observed with residues of the 280 -290 loop. However, VEGF 165 interaction with NRP1 involves not only VEGF exon 8 but the 44-amino acid VEGF exon 7 as well (49,50). In particular, residues 22-44 of VEGF exon 7 with the addition of the first cysteine of VEGF exon 8 have been shown to be critical for VEGF 165 -NRP interactions (50). Thus, it is likely that the B-NRP2 mutants we have generated interact with VEGF exon 7. A VEGF exon 7/NRP co-crystal has not yet been reported. Future site-directed mutagenesis studies will identify other amino acid residues involved in VEGF 165 binding.
VEGF inhibitors are currently being developed for anti-angiogenesis cancer therapy. Examples showing clinical efficacy include VEGFR-2 kinase inhibitors (PTK787, Novartis), antibodies against VEGF (Bevacizumab, Genentech), and VEGF-Trap, a decoy soluble receptor consisting of the second Ig domain of VEGFR-1 fused to the third Ig domain of VEGFR-2 (Aventis/Regeneron) (51)(52)(53). These antiangiogenesis therapies target VEGF/VEGFR interactions. However, it is evident that NRPs promote tumor progression in animal models (24,25). In addition, high expression levels of NRP1 and NRP2 correlate with poor outcomes clinically (27)(28)(29)(30). Therefore, targeting NRPs might be a viable strategy. A peptide (ATWLPPR, A7R) that inhibits VEGF 165 interaction with NRP1 inhibited VEGF 165 -induced in vitro angiogenesis and tumor growth of MDA-MB231 (54). Furthermore, a bicyclic peptide corresponding to the sequence of VEGF 165 exons 7 and 8 was able to inhibit VEGF 165 interaction with NRP1 and VEGF 165 signaling, e.g. VEGFR-2, phospholipase C␥, and extracellular signal-regulated kinase (ERK) phosphorylation (55). Recently, it has been shown that targeting tumors with a combination of NRP1blocking antibodies together with anti-VEGF antibodies resulted in a significantly increased tumor growth inhibition compared with anti-VEGF alone (26). Previously, we showed that soluble NRP1 (essentially the A and B domains) inhibited VEGF 165 binding to ECs, and VEGF 165 -induced VEGFR-2 phosphorylation (56). Overexpression of sNRP1 inhibited tumor growth and induced apoptosis, possibly by sequestering VEGF (56).
We have developed a strategy of increasing the electronegative potential of B-NRP2 to generate mutant proteins with increased affinity for VEGF 165 and unaffected SEMA3F binding. This is a crucial property because these mutants could be selective inhibitors of the pro-angiogenic and pro-tumor progression properties of VEGF 165 without affecting the reported anti-angiogenic and anti-metastasis features of SEMA3F. The mutants are better inhibitors of VEGF 165 binding to EC than WT B-NRP2. Mutant K d values for VEGF 165 so far are as low as 0.74 nM, lower than soluble Flk (16 nM) (57) but 1-2 logs higher than soluble Flt (20 pM) (58). Future goals are to improve the NRP2 affinity to VEGF 165 and to characterize these mutants for their potential of inhibiting VEGF 165 NRP2-dependent bioactivity in view of developing new anti-angiogenesis therapies.