A mutant form of vascular endothelial growth factor (VEGF) that lacks VEGF receptor-2 activation retains the ability to induce vascular permeability.

Vascular endothelial growth factor (VEGF) is a major mediator of vasculogenesis and angiogenesis both during development and in pathological conditions. VEGF has a variety of effects on vascular endothelium, including the ability to stimulate endothelial cell mitogenesis, and the potent induction of vascular permeability. These activities are at least in part mediated by binding to two high affinity receptors, VEGFR-1 and VEGFR-2. In this study we have made mutations of mouse VEGF in order to define the regions that are required for VEGFR-2-mediated functions. Development of a bioassay, which responds only to signals generated by cross-linking of VEGFR-2, has allowed evaluation of these mutants for their ability to activate VEGFR-2. One mutant (VEGF0), which had amino acids 83-89 of VEGF substituted with the analogous region of the related placenta growth factor, demonstrated significantly reduced VEGFR-2 binding compared with wild type VEGF, indicating that this region was required for VEGF-VEGFR-2 interaction. Intriguingly, when this mutant was evaluated in a Miles assay for its ability to induce vascular permeability, no difference was found when compared with wild type VEGF. In addition we have shown that the VEGF homology domain of the structurally related growth factor VEGF-D is capable of binding to and activating VEGFR-2 but has no vascular permeability activity, indicating that VEGFR-2 binding does not correlate with permeability activity for all VEGF family members. These data suggest different mechanisms for VEGF-mediated mitogenesis and vascular permeability and raise the possibility of an alternative receptor mediating vascular permeability.


From the Ludwig Institute for Cancer Research, Post Office Box 2008, Royal Melbourne Hospital, Victoria 3050 Australia
Vascular endothelial growth factor (VEGF) is a major mediator of vasculogenesis and angiogenesis both during development and in pathological conditions. VEGF has a variety of effects on vascular endothelium, including the ability to stimulate endothelial cell mitogenesis, and the potent induction of vascular permeability. These activities are at least in part mediated by binding to two high affinity receptors, VEGFR-1 and VEGFR-2. In this study we have made mutations of mouse VEGF in order to define the regions that are required for VEGFR-2-mediated functions. Development of a bioassay, which responds only to signals generated by cross-linking of VEGFR-2, has allowed evaluation of these mutants for their ability to activate VEGFR-2. One mutant (VEGF0), which had amino acids 83-89 of VEGF substituted with the analogous region of the related placenta growth factor, demonstrated significantly reduced VEGFR-2 binding compared with wild type VEGF, indicating that this region was required for VEGF-VEGFR-2 interaction. Intriguingly, when this mutant was evaluated in a Miles assay for its ability to induce vascular permeability, no difference was found when compared with wild type VEGF. In addition we have shown that the VEGF homology domain of the structurally related growth factor VEGF-D is capable of binding to and activating VEGFR-2 but has no vascular permeability activity, indicating that VEGFR-2 binding does not correlate with permeability activity for all VEGF family members. These data suggest different mechanisms for VEGF-mediated mitogenesis and vascular permeability and raise the possibility of an alternative receptor mediating vascular permeability.
Vascular endothelial growth factor (VEGF) 1 is an endothelial cell mitogen that plays an important role in angiogenesis both in the developing embryo and in the growth and spread of tumors (1)(2)(3)(4). VEGF (also called VPF for vascular permeability factor) was originally described as a factor responsible for the accumulation of plasma protein-rich fluid in the ascites of tumor patients due to an ability to induce vascular hyperpermeability (5). Studies of VEGF have demonstrated that it has a permeability enhancing capability, on a molar basis, 50,000 times greater than that of histamine (6).
Structurally, VEGF exists as a dimeric glycoprotein of M r 34 -42,000 and is related to the platelet-derived growth factor (PDGF) family of molecules, having a conserved cystine knot motif in each monomer (7). Although VEGF is the product of a single gene, differential RNA splicing produces at least five isoforms in the human (8 -11). Other related but distinct factors have since been described and designated VEGF-B (12), VEGF-C (13), VEGF-D (14), and placenta growth factor (PlGF) (15), which also share the common central region (VEGF homology domain; VHD) containing the cystine knot motif.
VEGF mediates its effects via at least two receptors, Flt-1(VEGFR-1) and Flk-1(VEGFR-2), both of which have been shown to have high affinity binding sites for VEGF (16 -19). The responses of these receptors to VEGF are quite different. VEGFR-2 displays ligand-dependent phosphorylation in intact cells and mediates mitogenesis and chemotaxis when transfected into porcine endothelial cells (20). In contrast, VEGFR-1 shows minimal tyrosine phosphorylation in response to VEGF, and binding of VEGF does not lead to a mitogenic signal (20,21). The phenotypes of VEGFR-1-and VEGFR-2-deficient mice are consistent with the critical importance of these receptors to the development of the vascular system (22,23). Mice carrying germ-line mutations in either VEGFR-1 or VEGFR-2 die before embryonic day 10 (E10) due to a failure to organize the vasculature in the case of VEGFR-1 or a complete failure to develop endothelial cells in the case of VEGFR-2. Recent studies have identified neuropilin-1 (NP-1), a receptor of the collapsin/semaphorin family, as an isoform-specific VEGF receptor (24). NP-1 is structurally unrelated to VEGFR-1 and VEGFR-2 and plays a role in neural cell guidance, suggesting a broader role for VEGF in non-endothelial cells.
The signaling mechanism by which VEGF mediates vascular permeability remains essentially uncharacterized, although recent work has shown that nitric oxide may play a role (25,26).
Nevertheless, some studies have suggested that VEGFR-2 is the receptor responsible for the permeability activity (25,27). The permeability induced by VEGF is transient and reversible, is not associated with mast cell degranulation or infiltration of inflammatory cells, and is not inhibited by antihistamines (5,28). Recently, it has been shown that VEGF increases microvascular permeability within tumors by induction of clusters of small vesicles and vacuoles within the cytoplasm. These structures have been termed vesicular-vacuolar organelles (VVO) (29,30). They span the cytoplasm and can form channels to connect the endothelial lumen to the tissue space, thereby allowing a point of exit for plasma and plasma proteins. The receptor signaling pathway that mediates formation of VVOs is at present not clear, nor are the roles of VEGFR-1, VEGFR-2, or NP-1 in this process. Other recent studies have suggested that VEGF can alter the tight junctions of endothelial cells, which may provide an additional mechanism for VEGF permeability (31,32).
In this study we have used site-directed mutagenesis to determine the residues of mouse VEGF important for mediating functions via VEGFR-2 and to determine their effect on endothelial mitogenesis and vascular permeability. We found that mutations that mapped to the third variable (V3) domain of VEGF had the most profound effect on VEGFR-2 binding. A mutant that had this region substituted with the V3 domain from human placenta growth factor (PlGF) showed reduced binding to VEGFR-2 and an inability to induce mitogenesis or activation of a VEGFR-2 bioassay. Nevertheless, this mutant was found to be equivalent to wild type VEGF when evaluated for its ability to induce vascular permeability in a Miles assay. Furthermore, another member of the VEGF family, VEGF-D, was shown to be incapable of inducing vascular permeability, even though it has the ability to bind and activate VEGFR-2. These data imply that the two biological effects of VEGF, mitogenesis and vascular permeability, are mediated by different mechanisms and that permeability may be signaled through a receptor other than VEGFR-2.
Antibodies-mAbs to the extracellular domain of the tie2 receptor tyrosine kinase are described elsewhere (33). mAb 4H3 to the extracellular domain of VEGFR-2 was used in the bioassay and for flow cytometry analysis. A rabbit antiserum made to mouse VEGF is described below.
Transient Expression of Proteins in COS and 293EBNA Cells-Wild type VEGF, PlGF, and mutant VEGF constructs were transiently expressed in COS cells using the DEAE-dextran transfection protocol as described (34) or 293 cells using a lipid-mediated transfection (Fugene, Roche Molecular Biochemicals). Conditioned media were collected from COS cells on day 6 -7 post-transfection (24 -48 h for 293 cells), and cell debris was removed by centrifugation prior to testing or affinity chromatography.
FACS Analysis-VEGFR-2/EpoR bioassay cells were washed and resuspended in PBS/BSA to a concentration of 10 6 cells per ml. Cells were stained with monoclonal antibodies to VEGFR-2 and tie2 (control) according to methods described elsewhere (35).
SDS-PAGE and Western Blotting Analysis-Immunoprecipitates, purified VEGF, or mutant VEGFs were combined with 2ϫ SDS-PAGE sample buffer (with or without ␤-mercaptoethanol), boiled, and resolved by SDS-PAGE (36). The proteins were then transferred to membrane (Immobilon-P, Millipore, Bedford, MA), and the nonspecific binding sites were blocked by incubation in 3% BSA, 100 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.02% Tween 20. Blots were then incubated with a 1/2000 dilution of the anti-mouse VEGF serum for 2 h at room temperature. After washing in buffer (3% BSA, 100 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.02% Tween 20) the blots were incubated with 0.5 Ci of 125 Iprotein A for 1 h at room temperature. After washing, the signal was detected by autoradiography using XAR film (Eastman Kodak Co.). Alternatively, blots were probed with anti-rabbit Ig horseradish peroxidase conjugate instead of the protein A and developed using chemiluminescence (ECL, Amersham Pharmacia Biotech).
VEGF Clone-A cDNA clone encoding mouse VEGF 164 was isolated by polymerase chain reaction using mouse colon mRNA as the template. After subcloning, the full-length sequence was confirmed by dideoxy sequencing to be identical to that previously reported (37). The fragment, containing the entire open reading frame of VEGF, was subcloned into the expression vectors pCDM8 and pcDNA1/Amp and transiently expressed in COS cells (34). Conditioned medium from day 6 -7 transfected COS cells specifically induced vascular permeability in the Miles assay (38), bound to purified VEGFR-2 extracellular domain immobilized on a biosensor (BIAcore, Amersham Pharmacia Biotech), and specifically activated the VEGFR-2 bioassay described below (38). This VEGF cDNA was then subcloned into the expression vector pEE6 and was stably transfected into CHO K1 cells. Expressing clones were selected in the presence of 25 M MSX. The mutant VEGF0 was also subcloned into the pEE6 vector and expressed in CHO cells.
Purification of Mouse VEGF-Mouse VEGF and VEGF0 were purified from media conditioned with a CHO cell line transfected with VEGF 164 (CHO-VEGF 164 ) or VEGF0 (CHO-VEGF0) cDNA. Conditioned media (250 -500 ml) were subjected to anion exchange chromatography on a 5-ml Q-Sepharose column (Amersham Pharmacia Biotech), and the flow-through was then loaded onto a 5-ml heparin-Sepharose column (heparin-Sepharose CL-6B, Amersham Pharmacia Biotech) in 150 mM NaCl, 20 mM Tris-HCl, pH 7.5. After washing, the bound material was eluted in 1 M NaCl/20 mM Tris-HCl, pH 7.5. Further purification was achieved by reversed phase liquid chromatography on a Brownlee RP-300 column (30 ϫ 21 mm internal diameter, Applied Biosystems, CA) equilibrated with 0.15% trifluoroacetic acid and eluted using a linear 60-min gradient to 60% CH 3 CN, 0.125% trifluoroacetic acid at a flow rate of 1 ml/min. Purified material was analyzed by SDS-PAGE under reducing and non-reducing conditions and corresponded to the reported molecular weight for VEGF 164 and VEGF0. N-terminal amino acid sequence analysis confirmed its identity. Amino acid analysis was also used to confirm the identity and concentration of the purified material.
Production of Antisera to Mouse VEGF-Purified mouse VEGF (75 g) was combined with Freund's complete adjuvant (CSL, Melbourne, Australia) and used as a primary immunogen for raising antibodies in rabbits (in collaboration with Dr. Peter Rogers, Monash University). Further immunizations were carried out using Freund's incomplete adjuvant. Rabbits were test bled after the third immunization, and the titer and specificity of the anti-VEGF response were evaluated by inhibition of VEGF-mediated proliferation of the VEGFR-2/EpoR bioassay cells (effective at a dilution of 1/20,000), and Western blotting, respectively.
Generation of a Soluble Mouse VEGFR-2 Extracellular Domain-To obtain constructs encoding the VEGFR-2 extracellular domain, we employed site-directed mutagenesis to generate specific in-frame restriction enzyme sites within the clone. The full-length clone of mouse VEGFR-2 (Nyk/Flk-1 receptor (39)) was subcloned into the mammalian expression vector pcDNA1/Amp (Invitrogen, San Diego, CA) using the BstXI restriction enzyme site. Single-stranded UTP ϩ DNA was generated using the M13 origin of replication and used as a template to generate VEGFR-2 cDNA containing specific restriction enzyme sites. A BglII site was introduced just 3Ј of sequences encoding the final ␤ strand of the seventh Ig-like domain to produce a VEGFR-2 construct containing the entire extracellular domain (VEGFR-2-EX). An oligonucleotide linker sequence encoding the FLAG marker peptide and a termination codon (IBI/Kodak; 5Ј-GATCTGACTACAAGGACGACGAC-GATGACAAGTGAATCGATA-3Ј encoding (N)Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys-STOP(C)) was inserted at the BglII site of the mutant VEGFR-2 cDNA at the junction of the regions encoding the extracellular domain and the transmembrane domain. The VEGFR-2-EX-FLAG insert was then subcloned into the XbaI site of the pEE6 CHO cell expression vector for large scale protein production as described above. This construct was transfected into CHO-K1 cells and positive clones selected in medium containing 25 M MSX. Clones expressing VEGFR-2-EX-FLAG were identified by immunoprecipitation of [ 35 S]methioninelabeled cells with anti-FLAG monoclonal antibody (M2, IBI/Kodak) and analysis by SDS-PAGE.
Mutants of VEGF-Mutants of mouse VEGF were generated by oligonucleotide directed mutagenesis according to the methods of Kunkel (40) using the expression vector pcDNA1/Amp (Invitrogen, San Diego, CA). Mutants are designated by the single letter amino acid code denoting the original amino acid first, then the relevant position in the mature VEGF polypeptide (without leader sequence), followed by the amino acid resulting from the mutation (e.g. I90Y, denotes isoleucine at position 90 changed to a tyrosine). The following mutants containing single amino acid substitutions were produced Q36D, D40S, E43S, Y44H, K47S, E66H, E72T, K83A, H85G, Q86D, I90Y, and G91V. An additional mutant (VEGF0) was produced, which had sequence in the V3 domain of VEGF substituted with the corresponding V3 sequence of human PlGF (RIKPHQSQHIGE to RIRSGDRPSIGE; see Fig. 1). Mutant VEGF cDNAs were sequenced to ensure no additional mutations had been introduced. Proteins were produced by transient expression in COS cells and 293EBNA cells or by stable expression in CHO cells. Conditioned medium was either tested directly or concentrated by affinity chromatography as described below.
Affinity Chromatography-VEGFR-2-EX-FLAG was purified from conditioned medium by affinity chromatography on M2 (anti-FLAG) gel. Conditioned medium (200 ml) was passed over the M2 column before washing with 10 column volumes of wash buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.02% Tween 20) and then 50 mM triethylamine, pH 10.0, 150 mM NaCl, 0.02% Tween 20, and a further 10 column volumes of wash buffer. Bound material was then eluted with 25 g/ml FLAG peptide (N-Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys-C) in wash buffer. Elution with FLAG peptide gave about 90 -95% pure VEGFR-2-EX-FLAG by SDS-PAGE and silver staining. For immunizations and coupling to the sensor chip for analysis with a biosensor, these proteins were further purified by ion exchange chromatography (Mo-noS, Amersham Pharmacia Biotech) to a single homogeneous species corresponding to VEGFR-2-EX-FLAG.
Heparin-Sepharose Affinity Chromatography-Conditioned media containing VEGF or VEGF mutants were concentrated on heparin-Sepharose-CL4B (5 ml of conditioned medium/100 l of beads; Amersham Pharmacia Biotech) by incubation at 4°C for 4 -16 h with rotation. The beads were then washed with 100 mM sodium phosphate, pH 7.2, and the bound protein was eluted with buffer containing 800 mM NaCl. Eluates were then buffer exchanged into PBS using Centricon 10 concentrators or diluted such that the salt concentration was reduced to 150 mM in an appropriate buffer for use in the bioassay.
Quantitation of VEGF and VEGF Mutants-VEGF and VEGF mutants that had been concentrated by affinity chromatography were quantitated using a competitive radioimmunoassay with 125 I-VEGF. Purified mouse VEGF was iodinated using either the Chloramine T or IODO-GEN protocols as described previously (41). 96-well microtiter plates were coated with a 10 g/ml solution of protein A (Amersham Pharmacia Biotech), washed with PBS, and nonspecific binding sites blocked by incubation with 3% BSA in PBS. The plates were then incubated with a 1/400 dilution of anti-mouse VEGF rabbit antiserum for 1 h at 4°C, followed by extensive washing with buffer (PBS 3%, BSA 0.02%, Tween 20). The plates were then incubated with a dilution of the sample containing VEGF or VEGF mutants and 2 ϫ 10 5 cpm of 125 I-VEGF. In addition, dilutions of a known concentration of purified mouse VEGF derived from CHO cells (previously quantitated by OD 280 and amino acid analysis) were also combined with 125 I-VEGF and used to generate a standard curve to enable accurate determination of the amount of VEGF present in the samples. VEGF mutants were also analyzed by SDS-PAGE and Western blotting to confirm the results generated by the competitive radioimmunoassay.
Establishment of a VEGFR-2/EpoR Bioassay-A bioassay was established in which Ba/F3 cells were stably transfected with a chimeric molecule containing the extracellular domain of the mouse VEGFR-2 and the transmembrane and cytoplasmic domain of the mouse erythropoietin receptor (EpoR; kindly supplied by Dr. D. Hilton, Walter and Eliza Hall Institute for Medical Research, Melbourne, Australia). The chimera was made by introducing a BglII restriction enzyme site at the junction of the regions encoding the extracellular and transmembrane domain of the mouse EpoR cDNA (subcloned into pcDNA1/Amp) using site-directed mutagenesis. Prior to this, a silent mutation was introduced into the EpoR cDNA in a region encoding the cytoplasmic domain of the EpoR to eliminate a naturally occurring BglII site. The fragment of the EpoR encoding the transmembrane and cytoplasmic domains was then ligated to VEGFR-2-BglII to give a cDNA encoding a fusion protein consisting of the VEGFR-2 extracellular domain and the transmem-brane and cytoplasmic domains of EpoR. This construct was subcloned into the expression vector pBOS and co-transfected into the Ba/F3 cell line with pgk-Neo at a ratio of 20:1. Transfected cells were selected in G418, and VEGFR-2 expressing cell lines selected by FACS analysis using mAb 4H3, which was directed to the VEGFR-2 extracellular domain. Cell lines expressing higher levels of VEGFR-2-EpoR were selected by growing the cells in either 5 g/ml mAb 4H3 (an activating mAb) or 25 ng/ml recombinant VEGF.
Binding Assays with Soluble VEGFR Extracellular Domains-Constructs encoding the extracellular domain of VEGFR-1 and VEGFR-2 fused to the Fc portion of human IgG1 were used for binding assays. VEGFR-1-and VEGFR-2-Ig cDNA (kindly supplied by K. Alitalo, E. Korpelainen & Y. Gunji, Helsinki) were expressed in 293EBNA cells grown in Dulbecco's modified Eagle's medium containing either low Ig serum or 0.2% BSA. The fusion proteins were then immunoprecipitated from the conditioned medium using protein A-Sepharose beads. These beads were then combined with 900 l of medium from CHO-VEGF or CHO-VEGF0 cells biosynthetically labeled with 35 S-Cys/Met and 100 l of 5% BSA, 0.2% Tween 20, 10 g/ml heparin. The Sepharose beads were then washed twice with binding buffer (0.5% BSA, 0.02% Tween 20, 1 g/ml heparin) at 4°C, once with PBS, and boiled in SDS-PAGE sample buffer before resolving the proteins by SDS-PAGE.
Biosensor Analysis-The purified extracellular domain of VEGFR-2 (VEGFR-2-EX-FLAG) was coupled to the carboxymethylated dextran layer of the sensor ship using standard NHS-EDC chemistry for analysis of binding using an optical biosensor (BIAcore2000) (42). The residual activated ester groups were blocked by treatment with 1 M ethanolamine hydrochloride, pH 8.5, followed by washing with 10 mM diethylamine to remove non-covalently bound material. Samples for analysis were diluted in HBS (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% Tween 20). The viability of the bound VEGFR-2-EX-FLAG was assessed by binding of the anti-VEGFR-2 extracellular domain 4H3 mAb and purified VEGF.
Purification of VEGF-D⌬N⌬C-FLAG and VEGF-D-FULL-N-FLAG-The VHD of VEGF-D, VEGF-D⌬N⌬C, was purified using M2 affinity chromatography as described previously (14). Material used for the Miles assay was evaluated in the VEGFR-2 bioassay and shown to be active for binding VEGFR-2. VEGF-D-FULL-N-FLAG was produced in 293EBNA cells and purified by M2 chromatography as described elsewhere (43).
Miles Vascular Permeability Assay-The Miles vascular permeability assay (38) was performed using anesthetized guinea pigs as described previously (44). The amount of dye extracted from the skin with formamide was quantitated by spectrophotometry at 620 nm.
Human Microvascular Endothelial Cell (HMVEC) Proliferation Assay-HMVECs were grown in EBM-2, 5% FBS, and growth supplements. For the assay, cells were removed with trypsin, washed, and resuspended in complete medium and aliquoted at 10 4 cells/well in a 24-well plate. Cells were allowed to adhere for 16 h at 37°C after which the appropriate dilution of VEGF, VEGF mutant, or control sample was added in EBM-2, 22% FBS plus supplements but without growth factors. After 96 h of growth at 37°C the cellular proliferation was quantitated by cell counting.

Alignment of VEGF and PlGF
Sequences-VEGF and PlGF are two related members of the cystine knot family of growth factors. Both bind a common receptor, VEGFR-1, whereas only VEGF binds VEGFR-2. We hypothesized that examination of the sequence differences between VEGF and PlGF could provide details about the regions of VEGF specific for its interaction with VEGFR-2. In addition, knowledge of the related factor PDGF, which has been shown to interact with the PDGF ␤ receptor via sequences within the third variable (V3) domain of the ligand (45,46), suggests that the equivalent domain in VEGF may be an important site for receptor interaction. Alignment of mouse and human VEGF and PlGF sequences shows conservation of location of the six cysteine residues of the cystine knot motif, as well as conservation of many residues in the predicted ␤ sheets and variable loops of the two factors ( Fig.  1). However, there are a number of regions, especially in the variable domains V1 and V3, that differ between VEGF and PlGF, notably in amino acid composition, charge, and hydropathy. These areas were targeted for further analysis; the amino acid substitutions generated are listed under "Materials and Methods." VEGF Mutants, Production, and Quantitation-A series of VEGF mutants with single amino acid substitutions, reflecting the differences between the primary sequences of VEGF and PlGF, was produced using site-directed mutagenesis. A mutant (VEGF0) in which part of the V3 region of mouse VEGF was replaced with that of human PlGF was also made. In this mutant the sequence KPHQSQH at position 83-89 of mouse VEGF has been changed to RSGDRPS, effectively substituting the V3 domain of human PlGF for that of VEGF (Fig. 1). Mutants were transiently expressed in COS or 293 cells and theparin-Sepharose chromatography. The resulting preparations were then quantitated by a competitive radioimmunoassay and Western blotting. The VEGF0 mutant was also expressed in CHO cells and purified by affinity chromatography. As mutations did not change the number of amino acids in the polypeptides, there was little detectable difference in the relative migration of the mutants by SDS-PAGE (Fig. 2). Evaluation of the mutants by SDS-PAGE under non-reducing conditions demonstrated that these mutations had no effect on the ability of the VEGF molecules to form homodimers, with the major detectable species of M r ϳ45,000, consistent with the size of the biologically active VEGF homodimer ( Fig. 2A). Under reducing conditions the mutants migrated in the same manner as the wild type VEGF monomer with an M r of approximately 21,000 -22,000 (Fig. 2B) and were recognized by the anti-VEGF antisera.
Evaluation of Mutants in the VEGFR-2 Bioassay-As endothelial cells express several different receptors for VEGF, which could potentially complicate the evaluation of VEGFR-2 binding and activation, we established an assay that would only be responsive to signaling via VEGFR-2 but would also test for one aspect of VEGFR-2 activation, namely receptor cross-linking. A chimeric molecule containing the extracellular domain of mouse VEGFR-2 and the transmembrane and cytoplasmic domains of the mouse erythropoietin receptor (EpoR) was produced and stably expressed in the IL-3-dependent cell line Ba/F3. The chimeric molecule is used because members of the receptor-type tyrosine kinase family signal poorly in hema-topoietic cells like Ba/F3 (47). This cell line does not express endogenous VEGFR-2 nor does it respond to VEGF stimulation. The parental cell line is dependent on IL-3 for survival and does not respond to VEGF. However, when expressing the chimeric receptor the cells can be rescued with VEGF in the absence of IL-3. Cells expressing the chimeric molecule were initially selected by flow cytometry using monoclonal antibody (mAb; 4H3) specific for the extracellular domain of VEGFR-2. Highly expressing cell lines were subsequently selected by limiting dilution cloning in either mAb 4H3 or purified mouse VEGF (Fig. 3A). One cell line expressing the chimeric receptor 18a was used for the VEGFR-2 bioassay as it expresses high levels of the VEGFR-2/EpoR chimera (Fig. 3B). The bioassay is designed to test whether a given VEGF molecule can bind and cross-link VEGFR-2 and therefore would not only provide information on VEGF-VEGFR-2 interaction but also on the abil-FIG. 2. Western blotting analysis of VEGF mutants. Wild type VEGF or mutant VEGF cDNA was expressed in COS cells, and the conditioned media were concentrated by heparin-Sepharose chromatography as described under "Materials and Methods." 50 ng of each VEGF protein was combined with SDS-PAGE sample buffer, boiled, and loaded onto a 15% polyacrylamide gel. Proteins were resolved under either non-reducing (A) or reducing (B) conditions. Proteins were then transferred to nitrocellulose membranes that were probed with the polyclonal antisera raised to purified mouse VEGF. Blots were then probed with a goat anti-rabbit horseradish peroxidase-conjugated secondary antibody and the gel developed using the ECL detection system. Dimeric VEGF polypeptides are indicated by large arrowheads, and the smaller arrowheads indicate the VEGF monomer.
FIG. 1. Alignment of the predicted amino acid sequences of mouse VEGF, human VEGF, mouse PlGF, and human PlGF in the region of the cystine knot motif. The mouse VEGF sequence is numbered from the first residue of the secreted protein, whereas the PlGF sequences are numbered from their initiation methionines. The sequences are reported elsewhere (15,37,56). ␤ strands (␤1-7), ␣ helices (␣1-2), and variable domains (V1-3) are indicated. The shaded sequence represents the region of mouse VEGF that was replaced by the corresponding region of human PlGF in the VEGF0 mutant.
ity of the VEGF protein to cross-link the extracellular domains, which in the case of the native receptor results in signaling. There also exists the possibility that factors can bind and cross-link the VEGFR-2 extracellular domain but fail to recruit the appropriate downstream signaling molecules. In the absence of IL-3, VEGFR-2 bioassay line 18a responded to VEGF or mAb 4H3 in a dose-dependent manner (Fig. 3, C and D). VEGF can stimulate the VEGFR-2 bioassay in a dose-dependent manner, whereas the non-binding but structurally related factor PlGF fails to induce proliferation (Fig. 3C). Some VEGF mutants (Q36D, D40S, E43S, Y44H, K47S, E66H, and E72T) had little or no effect on the ability of the molecule to induce proliferation in the VEGFR-2 bioassay, presumably because these mutations are not critical for the interaction of VEGF with VEGFR-2 (data not shown). However, a subset of the VEGF mutants, in particular those that mapped to the V3 domain of VEGF, exhibited reduced ability to stimulate proliferation in the VEGFR-2 bioassay (Fig. 4). The VEGF0 mutant, which had part of the VEGF V3 domain substituted for the V3 domain of PlGF, had almost a 100-fold reduction in its ability to stimulate proliferation in the bioassay, which indicated that the VEGF binding determinants had been disrupted in this mutant (or it lacked the ability to cross-link the receptors). When VEGF mutants which contained single amino acid substitutions in the V3 domain were evaluated for their ability to induce proliferation, most showed some reduction in their ability to stimulate the bioassay cell line, when compared with VEGF itself. The most significant were K83A, H85G, and I90Y, whereas G91V gave only slight reduction in activity.
Binding of VEGF Mutants to VEGFR-2-To evaluate whether the VEGF0 mutant, which failed to activate the VEGFR-2 bioassay, lacked the ability to bind VEGFR-2, binding studies were performed with soluble receptor extracellular domains and with immobilized VEGFR-2-EX-FLAG using the BIAcore. VEGF0 was able to bind the extracellular domain of VEGFR-1 (human) to the same level as wild type VEGF; however, the mutant displayed a significantly reduced level of binding to the dimeric human VEGFR-2-Ig construct indicating that its binding sites for VEGFR-2 had been substantially altered (Fig. 5A). This result was confirmed by analysis of purified VEGF and VEGF0 for binding to immobilized VEGFR-2-EX-FLAG (monomeric) on the optical biosensor (Fig. 5B). VEGF bound to the VEGFR-2 domain in a dose-dependent  (10 4 ) were plated in a 96-well plate with dilutions of VEGF (starting concentration at 10% dilution: 100 ng/ml), human PlGF (starting concentration 100 ng/ml), WEHI-3D (10% conditioned medium). D, mouse VEGF, anti-VEGFR-2 mAb 4H3 or medium control were tested for their ability to induce proliferation of the VEGFR-2/EpoR cells in a [ 3 H]thymidine assay. Washed cells (10 4 ) were plated in a 96-well plate with dilutions of VEGF (starting concentration at 10% dilution: 100 ng/ml) and anti-VEGFR-2 mAb 4H3 (500 ng/ml). Cells plated in "Medium" alone were also assessed in the bioassay, to establish a base-line level of response. C and D, the graphs represent the concentration of stimulating factor versus incorporated counts (cpm). The values represent means Ϯ S.D. manner, whereas VEGF0 demonstrated no detectable binding even at 10 g/ml. These data suggest that the binding sites for VEGFR-2 have been significantly disrupted in VEGF0 leading to reduced binding. The minor binding component seen in the immunoprecipitation experiment may be due to the use of dimeric constructs.
Evaluation of Mutants in an Endothelial Cell Proliferation Assay-VEGF0 was tested in an endothelial cell proliferation assay to determine if the reduction in activation of VEGFR-2 in the bioassay was seen with full-length VEGFR-2. Studies by others have demonstrated that the mitogenic activity of endothelial cells is via VEGFR-2 and not VEGFR-1 (48). VEGF and VEGF0 were evaluated for their ability to induce proliferation of HMVEC cells after 96 h stimulation using high concentrations of growth factor. Fig. 6 shows that VEGF can effectively stimulate the growth of HMVEC, compared with medium alone, and this activity is fully titratable (not shown). In contrast, VEGF0, even at concentrations of 400 ng/ml, was unable to induce proliferation of HMVEC cells, which was consistent with the result of the VEGFR-2 bioassay. VEGF0 also failed to induce the proliferation of bovine aortic endothelial cells (not shown).
Evaluation of Mutants in the Miles Vascular Permeability Assay-To examine whether mutations of VEGF that affect activation of VEGFR-2 have an impact on its ability to induce microvascular permeability, mutant VEGF proteins were evaluated in the Miles assay using guinea pigs and Evans Blue dye as an indicator. Wild type and mutant forms of VEGF, previously evaluated in the VEGFR-2 bioassay, were tested over a range of concentrations. As the sensitivity of the Miles assay is lower than that of the bioassay, samples were tested at 100, 25, 12, and 6 ng per sample area. Interestingly, when tested over a range of concentrations the mutants, in general, showed little reduction in their ability to induce vascular permeability. Most notably the VEGF0 and K83A mutants, which had exhibited reduced ability to stimulate in the VEGFR-2 bioassay and endothelial mitogenesis assay, showed no reduction in their ability to induce vascular permeability (Fig. 7). The only mutant that appeared to show a reduction in permeability activity was the I90Y mutation, which also showed reduction in the bioassay. Vector alone and 4H3 mAb (not shown) gave no significant signal in the Miles assay. These data suggest that the permeability activity of VEGF seen in the Miles assay is not dependent on structural regions that are critical for crosslinking the extracellular domain of VEGFR-2 and inducing mitogenesis in endothelial cells.
Evaluation of VEGF-D in the Miles Permeability Assay-VEGF-D is a member of the VEGF family which is 31% identical in amino acid sequence to VEGF (14). The VHD of VEGF-D (designated VEGF-D⌬N⌬C-FLAG) binds and activates VEGFR-2 and VEGFR-3 and induces mitogenesis of endothelial cells (14). The full-length form of VEGF-D (VEGF-D-FULL-N-FLAG) also binds to VEGFR-2 but at a substantially reduced affinity (ϳ300-fold less than for VEGF-D⌬N⌬C (43)).
To determine whether these VEGFR-2 ligands also induced vascular permeability, we assayed the purified VEGF-D⌬N⌬C-FLAG and VEGF-D-FULL-N-FLAG over a range of concentrations (100 -10,000 ng/ml) in the Miles assay. Control VEGF was also tested and, as expected, induced strong permeability at concentrations below 100 ng/ml. In contrast VEGF-D⌬N⌬C-FLAG and VEGF-D-FULL-N-FLAG produced no significant induction of vascular permeability over a range of concentrations (Fig. 8). In order to confirm that the purified VEGF-D⌬N⌬C-FLAG and VEGF-D-FULL-N-FLAG bound to VEGFR-2, samples were tested in the bioassay and in binding assays using the biosensor (data not shown). This study indicates that not all VEGFR-2 binding growth factors induce vascular permeability. DISCUSSION The molecular mechanisms whereby VEGF induces vascular permeability are not well understood. In particular the cellsurface receptor(s) responsible for coordinating this rapid extravasation of plasma proteins are essentially uncharacterized. In this study we have examined the structure-function relationship of mouse VEGF using site-directed mutagenesis. Re-

FIG. 5. Analysis of binding of the VEGF0 mutant to VEGFR-2.
A, mutant VEGF0 was assessed for binding to VEGFR-2 by immunoprecipitation of metabolically labeled VEGF and VEGF0 from the conditioned medium of CHO cells expressing VEGF and VEGF0 cDNAs, using receptor-Ig constructs VEGFR-1-Ig and VEGFR-2-Ig according to the "Materials and Methods." Washed precipitates were eluted and analyzed by SDS-PAGE under reducing conditions. Levels of binding of VEGF0 to the Ig fusion proteins were quantitated using the Image-Quant program and expressed as a relative binding index. The relative binding index for VEGF to VEGFR1 or VEGFR2 is set at 100. The relative binding index of the VEGF0 mutant is therefore a ratio of the binding of the VEGF0 mutant to VEGFR1 or VEGFR2 when compared with VEGF. Bkg denotes the relative binding index for a mock-transfected and labeled cell conditioned medium immunoprecipitated with VEGFR-2-Ig. B, analysis of VEGF0 binding to immobilized mouse VEGFR-2 extracellular domain using the BIAcore. Purified VEGFR-2-EX-FLAG was immobilized onto a BIAcore chip surface using standard chemistry and analyzed for binding of purified VEGF and VEGF0. The values were expressed as response units (RU). Control mAb 4H3 gave a binding response of 3232 response units at 10 g/ml. gions in and around the V3 domain of the VHD were found to be critical for binding to the major mitogenic receptor VEGFR-2. Interestingly, mutants altered in the V3 region that did not bind VEGFR-2 were still capable of inducing vascular permeability, thus demonstrating that the two biological functions of VEGF, induction of mitogenesis and vascular permeability, can be distinguished by mutations within the V3 domain. These results imply that vascular permeability and mitogenesis induced by VEGF are mediated through the recruitment of different receptors.
In order to evaluate VEGF mutants in an assay system that would be responsive to the binding and cross-linking of VEGFR-2 only, we avoided the use of in vitro cell lines that may express other VEGF receptors. This approach excludes the possibility that other receptors could initiate signaling alone or as part of a heterodimeric complex with VEGFR-2. The approach we followed, of developing a factor-dependent cell line expressing a chimeric RTK/cytokine receptor, was initially described by others (47) as a means of detecting the cross-linking/ activation of the ligand binding domains of orphan receptors. The system has been shown to work effectively for other receptors that rely on dimerization for signal transduction (e.g. EGFR (47) and Tie2 in our own laboratory). 2 Screening of the mutant VEGF molecules in the bioassay demonstrated that the V3 domain between ␤ strands 5 and 6 is critical for binding to and activation of VEGFR-2. In particular, the point mutations H85G and I90Y caused dramatic reductions in the level of activity seen in the bioassay. When the loop containing residues 83-89 was replaced with the equivalent loop from human PlGF, to generate VEGF0, a similar reduction in activity was observed, suggesting that key substitutions had been made. The VEGF0 mutant contains substitutions at each position throughout the amino acid sequence 83-89 and appears to be one of the most severely affected of the mutants consistently giving low stimulation in the bioassay. Our results are consistent with studies from other laboratories that have examined the binding determinants of human VEGF for VEGFR-2 (48,49). In particular, they are also consistent with the large "hot spot" for interaction with VEGFR-2 on the crystal structure of human VEGF, which includes residues Ile-83, Lys-84, and Pro-85 (82-84 in the mouse) of the V3 region (50). Our results with the VEGF0 mutant are also in agreement with a previous suggestion that residues in the loop between ␤ strands 5 and 6 of human VEGF, specifically residues 85, 86 and 89, might influence the conformation of the loop and thus determine the difference in binding to VEGFR-2 between VEGF and PlGF. One could argue that flexibility in this region may therefore be important for binding and activation of VEGFR-2 for mitogenic signaling but have less significance for  7. Permeability induced by wild type VEGF, mutant VEGFs, and controls in the Miles assay. The figure represents skin patches taken from guinea pigs in which the Miles assay has been performed. Guinea pigs are given an intracardiac injection of Evans blue dye followed by intradermal injections of aliquots of expressed, purified, and quantitated protein (100 l). The amount (ng) of VEGF/VEGF mutant protein (quantitated by the competitive radioimmunoassay) injected is indicated above the sample area. After 20 min the animals are sacrificed and the skin patches taken. The Vector sample represents the protein affinity purified from the equivalent amount of conditioned medium from cells transfected with vector lacking VEGF sequence.
the induction of permeability. A mAb that cross-linked VEGFR-2 in the bioassay also failed to induce permeability (data not shown), which would support the hypothesis that not all VEGFR-2-binding proteins can induce permeability. However, one must consider the caveat that this mAb may be able to cross-link in vitro but not in vivo.
Permeability induced by VEGF, at least in tumor vessels, appears to be mediated by the formation of vesicular-vacuolar organelles (VVO), groups of vesicles/vacuoles that are arranged at intervals through the cell (29,30,51). Upon stimulation with VEGF these vesicles interconnect with each other and the plasma membrane to create a passage for fluid and plasma proteins to pass from the blood to the interstitium. VVO are more abundant in tumor vessels than in normal vessels and become more abundant when blood vessel permeability is induced by injection of VEGF. However, it is still possible that some of the increased permeability induced by VEGF may occur by formation of gaps between endothelial cells.
The receptor(s) involved in transducing the signal for vascular permeability are as yet uncharacterized. Both VEGFR-1 and VEGFR-2 bind VEGF, and recent data indicate that neuropilin-1 present on endothelial cells and other non-endothelial cell populations can also specifically bind VEGF (24). Given that the induction of vascular permeability occurs readily and extremely rapidly (in less than 5 min) in the peripheral vessels of mature animals, a receptor with constitutive expression on resting endothelium is likely to be involved. However, VEGFR-1 and VEGFR-2 appear to be down-regulated on adult quiescent endothelium compared with fetal endothelium (19,52,53) suggesting that these receptors may not be involved in the induction of vascular permeability. Recent experiments describing a mutant of VEGF-C (27) which has lost VEGFR-2 binding and permeability activity were suggestive of a role for VEGFR-2 in vascular permeability. Furthermore, we and others (44,54) have recently characterized viral VEGF molecules from the orf virus family which bind VEGFR-2 and neuropilin-1, or in some cases VEGFR-2 alone, but not VEGFR-1 and VEGFR-3, and induce vascular permeability. In combination these results are suggestive of a role for VEGFR-2 in vascular permeability, although they do not rule out the involvement of an alternative receptor that can bind these mutants and the viral VEGF molecule. VEGFR-1 which binds in addition to VEGF, PlGF and VEGF-B would appear not to be a strong candidate as both PlGF and VEGF-B lack significant permeability activity (25,43). NP-1 would also appear not to play a role in vascular permeability as it binds only VEGF 165 , whereas the non-binding VEGF 121 has been shown to induce vascular permeability (55). Our findings with the VEGF mutants and the VHD of VEGF-D suggest that VEGFR-2 by itself is not sufficient for inducing vascular permeability. This leads to two possible conclusions. First, the use of an alternative receptor that has high affinity binding for VEGF and that itself induces the signal for permeability. An alternative explanation would be the use of a heterodimeric complex involving VEGFR-2 in which only low affinity binding of VEGF to VEGFR-2 is required, with the other receptor providing the high affinity interaction. Both of these alternatives are not inconsistent with the results generated by the VEGF0 mutant.
Recent work (25,26) has shown a link between the nitric oxide/prostaglandin pathway and vascular permeability induced by VEGF. These findings are not inconsistent with the findings presented here and may suggest that the receptormediating permeability acts via nitric oxide or that they both converge on the same downstream signaling molecules.
In summary, the data presented here show that two of the biological functions of VEGF, namely induction of mitogenesis and of vascular permeability, can be distinguished by mutations that alter the binding of VEGF to VEGFR-2. A VEGF mutant that induces permeability but not mitogenesis would be of practical value to examine its effects on the morphology changes of endothelial cells during vascular permeability as distinct from changes that may occur due to cell division. The data presented here also suggest the possibility that one or more as yet unidentified VEGF receptors may exist that are responsible for vascular permeability. The recent unexpected description of NP-1 as a receptor for VEGF demonstrates that other VEGF receptors may indeed exist.