A Recombinant Mutant Vascular Endothelial Growth Factor-C that Has Lost Vascular Endothelial Growth Factor Receptor-2 Binding, Activation, and Vascular Permeability Activities*

The vascular endothelial growth factor (VEGF) and the VEGF-C promote growth of blood vessels and lymphatic vessels, respectively. VEGF activates the endothelial VEGF receptors (VEGFR) 1 and 2, and VEGF-C activates VEGFR-3 and VEGFR-2. Both VEGF and VEGF-C are also potent vascular permeability factors. Here we have analyzed the re-ceptorbindingandactivatingpropertiesofseveralcysteinemutantsofVEGF-Cincludingthose(Cys 156 and Cys 165 ), which in other platelet-derived growth factor/VEGF family members mediate interchain disulfide bonding. Surpris-ingly, we found that the recombinant mature VEGF-C in which Cys 156 was replaced by a Ser residue is a selective agonist of VEGFR-3. This mutant, designated D N D C156S, binds and activates VEGFR-3 but neither binds VEGFR-2 nor activates its autophosphorylation or downstream signaling to the ERK/MAPK pathway. Unlike VEGF-C, D N D C156S neither induces vascular permeability in vivo nor stimulates migration of bovine capillary endothelial cells in culture. These data point out the critical role of VEGFR-2-mediated signal transduction for the vascular permeability activity of VEGF-C and strongly suggest that the redundant biological effects of

The vascular endothelial growth factor (VEGF) and the VEGF-C promote growth of blood vessels and lymphatic vessels, respectively. VEGF activates the endothelial VEGF receptors (VEGFR) 1 and 2, and VEGF-C activates VEGFR-3 and VEGFR-2. Both VEGF and VEGF-C are also potent vascular permeability factors. Here we have analyzed the receptor binding and activating properties of several cysteine mutants of VEGF-C including those (Cys 156 and Cys 165 ), which in other platelet-derived growth factor/VEGF family members mediate interchain disulfide bonding. Surprisingly, we found that the recombinant mature VEGF-C in which Cys 156 was replaced by a Ser residue is a selective agonist of VEGFR-3. This mutant, designated ⌬N⌬C156S, binds and activates VEGFR-3 but neither binds VEGFR-2 nor activates its autophosphorylation or downstream signaling to the ERK/MAPK pathway. Unlike VEGF-C, ⌬N⌬C156S neither induces vascular permeability in vivo nor stimulates migration of bovine capillary endothelial cells in culture. These data point out the critical role of VEGFR-2-mediated signal transduction for the vascular permeability activity of VEGF-C and strongly suggest that the redundant biological effects of VEGF and VEGF-C depend on binding and activation of VEGFR-2. The ⌬N⌬C156S mutant may provide a valuable tool for the analysis of VEGF-C effects mediated selectively via VEGFR-3. The ability of ⌬N⌬C156S to form homodimers also emphasizes differences in the structural requirements for VEGF and VEGF-C dimerization.
The PDGF/VEGF 1 family of growth factors currently in-cludes seven members: PDGF-A, PDGF-B (1, 2), VEGF (3,4), placenta growth factor (PlGF) (5), VEGF-B/VEGF-related factor (6,7), VEGF-C/VEGF-related protein (8,9), and c-fos-induced growth factor/VEGF-D (10). All members of the family share a common structure in that they contain eight characteristically spaced cysteine residues in the core domain. PDGF-A and PDGF-B promote the growth of several cell types, whereas VEGF, PlGF, and VEGF-C regulate almost exclusively endothelial cells, which express the corresponding receptors. VEGF binds VEGFR-1 (Flt-1) and VEGFR-2 (Flk-1/KDR), whereas PlGF and VEGF-B bind only VEGFR-1 (3,12,13). 2 VEGF-C and VEGF-D are ligands and activators of VEGFR-3 (8,9,11). Mature VEGF-C and VEGF-D, which are generated by proteolytic processing of precursor polypeptides, also activate VEGFR-2. VEGF and VEGF-C are distinct in their specificity toward endothelial cells. VEGF specifically stimulates proliferation of the endothelial cells of blood vessels (14), whereas VEGF-C preferentially promotes growth of lymphatic endothelia (15,16). On the other hand, there are certain similarities in the biological activities of VEGF and the mature form of VEGF-C in that both factors are potent inducers of vascular permeability (8,17,18). In addition, at higher concentrations VEGF-C, similarly to VEGF, also stimulates proliferation and migration of vascular endothelial cells in culture (8,18). These data addressed a question of whether certain redundancy in VEGF and VEGF-C activities might be mediated via VEGFR-2, which is used by both of these two growth factors.
The previously known members of the PDGF/VEGF family form homodimers via disulfide bonds between the second and fourth of the eight conserved cysteine residues. These bonds are crucial for the dimerization and biological activity of VEGF but not for the activity of PDGF-BB (19 -21). Unlike these factors, the recombinant mature VEGF-C forms mostly noncovalent homodimers. It also contains an unpaired extra cysteine residue (Cys 137 ) located between the first and second conserved cysteine residues (18). Another unpaired Cys residue (Cys 87 ) is located in the N-terminal VEGF-C propeptide, which is cleaved off during the proteolytic processing of VEGF-C precursor. To evaluate the significance of disulfide bonds for VEGF-C homodimerization and activity, we converted selected cysteine residues to serine residues either in the VEGF-C precursor or in the recombinant "processed" VEGF-C (⌬N⌬C, described in Joukov et al. (18)) and analyzed these mutants for their abilities to bind and to activate VEGFR-3 and VEGFR-2. This led to the identification of a ligand for VEGFR-3 that is devoid of VEGFR-2 stimulating properties and does not possess VEGFlike activities in vitro or in vivo.
Generation of VEGF-C Mutants-The replacement of Cys with Ser residues was carried out using the Altered Sites II in vitro Mutagenesis System (Promega) essentially as described (17). The template was a cDNA encoding either full-length wt or recombinant mature VEGF-C fused in-frame with the signal peptide and containing His 6 tag at the C terminus (⌬N⌬C). The mutant constructs in the pALTER vector were digested with HindIII and NotI, subcloned into HindIII/NotI-digested pREP7, and used to transfect 293 EBNA cells.
48 h after transfections the culture media were changed to DMEM/ 0.1% bovine serum albumin, and incubation was continued for an additional 48 h. The conditioned media were concentrated 30-fold using Centriprep-10 filtration devices (Amicon). The ⌬N⌬C and ⌬N⌬C156S proteins were purified from conditioned media using Ni-NTA Superflow resin (Qiagen), according to the protocol of the manufacturer. Conditioned medium of the mock transfected cells was similarly treated. ⌬N⌬C and ⌬N⌬C156S were produced and purified from the Pichia yeast expression system as described (18).
Binding, Receptor Autophosphorylation, and MAPK Activation Studies-Concentrations of the mutant proteins were estimated from serially diluted conditioned media by Western blotting and comparison with known amounts of recombinant yeast protein. The ability of VEGF-C mutants to compete with [ 125 I]⌬N⌬C for VEGFR-2 and VEGFR-3 and to stimulate their tyrosine autophosphorylation were analyzed as described (8,18). To study ERK/MAPK activation, PAE/ VEGFR-3 and PAE/VEGFR-2 cells were serum starved for 12 h prior to treatment with equal amounts (ϳ100 ng/ml) of purified factor or 10% fetal calf serum for 5 min, after which ERK1 and ERK2 phosphorylation was analyzed by Western blotting using phospho-specific p44/42 MAPK antibodies. Blots were then stripped and reprobed with p44/42 MAPK antibodies to detect total protein. p44/42 and phospho-specific p44/42 antibodies were purchased from New England BioLabs and used according to the manufacturer's instructions.
Analysis of VEGF-C Biological Activity-To eliminate the effect of trace amounts of co-purified VEGF, the conditioned media, purified ⌬N⌬C, ⌬N⌬C156S, and the material similarly purified from the mock transfected cells were pretreated for 1 h at room temperature with anti-human VEGF neutralizing antibody (R & D systems, 2.5 ng/1 ng of VEGF-C). Analysis of the migration and permeability assays were carried out as described (8,18).
Analysis of Dimerization of VEGF-C Mutants-The ability of VEGF-C mutants to form disulfide linked dimers was analyzed in SDS-PAGE in nonreducing conditions with subsequent Western blotting and detection using the antiserum 882. Chemical cross-linking of the 35 S metabolically labeled VEGF-C mutants with disuccinimidyl suberate (Pierce) was carried out as described (18).

Binding of VEGF-C Mutants to VEGFR-3 and VEGFR-2-
Taking into account the importance of the second and the fourth characteristic cysteine residues for VEGF activity, we replaced these residues either separately or in combination with Ser residues in the VEGF-C precursor and in the recombinant mature VEGF-C (⌬N⌬C). The mutants were expressed in 293 EBNA cells, and the conditioned media were studied for their ability to compete with [ 125 I]⌬N⌬C for VEGFR-3 and VEGFR-2 binding. Replacement of Cys 165 alone or together with Cys 156 in ⌬N⌬C or in wt VEGF-C abolished the ability of the proteins to bind VEGFR-3 or VEGFR-2 ( Fig. 1 and data not shown). On the other hand, the C83S,C137S mutant displaced [ 125 I]⌬N⌬C from both receptors, indicating that these unpaired cysteine residues are not critical for the receptor binding properties of VEGF-C. ⌬N⌬C165S and  The receptors were immunoprecipitated and analyzed by Western blotting using anti-phosphotyrosine antibodies. Phosphorylated receptors are indicated by arrows. Analysis of the conditioned media used for the stimulations by Western blotting and detection using the antiserum 882 is shown in the lower panel. B, activation of ERK1 and ERK2 MAPK through VEGFR-3 and VEGFR-2 by ⌬N⌬C and ⌬N⌬C156S. Serum starved PAE/VEGFR-3 and PAE/VEGFR-2 cells were treated with equal amounts (ϳ100 ng/ml) of the indicated purified factor or 10% FCS added to the DMEM. Cells were lysed after 5 min, and ERK1 and ERK2 phosphorylation was analyzed by Western blotting using phospho-specific p44/42 MAPK antibodies. Blots were then stripped and probed with p44/42 MAPK antibodies to confirm equivalent amounts of total protein (lower panels). The phosphorylated and total fractions are indicated as pp and p, respectively. Note that the p44/42 MAPK antibodies recognize the p44 form only very weakly.
⌬N⌬C156S Is a Selective Activator of VEGFR-3-The recombinant ⌬N⌬C and ⌬N⌬C156S proteins were purified from the conditioned media, and their abilities to stimulate the tyrosine phosphorylation of VEGFR-3 and VEGFR-2 were compared. In agreement with the receptor binding studies, ⌬N⌬C156S was as active as ⌬N⌬C in the stimulation of VEGFR-3 tyrosine phosphorylation ( Fig. 2A, upper panel), but unlike VEGF or ⌬N⌬C, ⌬N⌬C156S did not stimulate tyrosine phosphorylation of VEGFR-2 ( Fig. 2A, middle panel). Western blotting analysis confirmed that similar amounts of the ⌬N⌬C and ⌬N⌬C156S proteins were used in the analysis ( Fig. 2A, lower panel). These data indicate that ⌬N⌬C156S is a selective ligand and an activator of VEGFR-3 but that it lacks activity toward VEGFR-2.
⌬N⌬C156S Is a Selective Activator of VEGFR-3 ERK/MAPK Signaling-Mitogenic signals from growth factor receptors are frequently relayed via the ERK/MAPK pathway into the nucleus. Purified recombinant ⌬N⌬C and ⌬N⌬C156S produced in a Pichia expression system were used to determine MAPK pathway activation of cells expressing either VEGFR-2 or VEGFR-3 (Fig. 2B). The growth factor-treated cells were lysed, and activated MAPK was detected using Western blotting with antibodies against the phosphorylated forms of ERK1 and ERK2. At a concentration of 100 ng/ml, VEGF-C showed rapid activation of the ERK1 and ERK2 MAPK in VEGFR-2 and VEGFR-3 expressing cells; however, ⌬N⌬C156S activated these kinases only via VEGFR-3. At the concentrations used, both forms of VEGF-C appeared to be equally potent in activating the MAPK through VEGFR-3. The amounts of total MAPK protein were confirmed to be similar in the treated and untreated cells, as shown by staining the filter with p44/p42 MAPK antibodies made against a synthetic peptide of rat p42, which in our hands give a weaker signal from p44.
⌬N⌬C156S Lacks VEGF-like Effects-Similar to VEGF, the mature VEGF-C increases vascular permeability in vivo, and at higher (compared with VEGF) concentrations also stimulates the migration and proliferation of BCE cells in vitro (8,18). We therefore compared ⌬N⌬C and ⌬N⌬C156S for their abilities to induce these biological responses.
In agreement with our previous data (8,18), the pure recombinant mature VEGF-C increased vascular permeability in the Miles assay, and this effect could not be blocked by the antihuman VEGF neutralizing antibodies (Fig. 3A). On the other hand, these antibodies considerably decreased the activity obtained from the conditioned medium of mock transfected cells, indicating that the 293 EBNA cells produce endogenous VEGF (data not shown). To eliminate possible effects of trace amounts of co-purified VEGF, the material purified from the 293 EBNA cells was treated with the VEGF neutralizing antibodies prior to analysis of its ability to promote vascular permeability. When subsequently analyzed, ⌬N⌬C dose-dependently increased vascular permeability, whereas ⌬N⌬C156S was completely inactive (Fig. 3A, compare with M).
As can be seen from Fig. 3B, ⌬N⌬C dose-dependently stimulated the migration of BCE cells in collagen gel, whereas ⌬N⌬C156S had no significant activity in this assay. Taken together, these data indicate that the inability of ⌬N⌬C156S to activate VEGFR-2 correlates with the lack of vascular permeability and endothelial cell migration inducing activities.
⌬N⌬C156S Forms Partially Disulfide-bonded Homodimers-Replacement of the second and/or the fourth cysteine residues of VEGF abolishes its dimer formation and biological activity (21). We investigated the dimeric nature of the VEGF-C mutants. No homodimers were obtained when ⌬N⌬C156S,C165S or ⌬N⌬C165S were chemically cross-linked (Fig. 4A, lanes 1-4). On the other hand, about half of both cross-linked ⌬N⌬C (18) and ⌬N⌬C156S (lane 6) migrated as dimers. This indicates that ⌬N⌬C156S forms homodimers. Moreover, unlike ⌬N⌬C, which forms preferentially noncovalently bound dimers, a fraction of ⌬N⌬C156S was disulfide bonded, as detected by SDS-PAGE in nonreducing conditions (Fig.  4B). These data suggest that homodimerization is required for VEGFR-3 activation by VEGF-C and indicate that the inability of ⌬N⌬C156S to activate VEGFR-2 and to induce VEGF-like effects is not due to an inability of this mutant to form homodimers.

DISCUSSION
Here we describe a VEGF-C point mutant that is active toward VEGFR-3 but, unlike wt VEGF-C, is unable to bind to and to activate signaling through VEGFR-2. This mutant (⌬N⌬C156S) was generated by replacement of the second conserved Cys residue of the recombinant processed VEGF-C by a Ser residue. ⌬N⌬C156S was inactive in the vascular permeability assay and did not increase migration of the capillary endothelial cells, indicating that these VEGF-like effects of VEGF-C require VEGFR-2 binding. Interaction with VEGFR-2 has been shown to be a critical requirement for the full spectrum of biological responses induced by VEGF (22,27,28). Taking into account that VEGFR-2 is the only known receptor shared by VEGF and VEGF-C, one can speculate that the ability of VEGF-C to increase vascular permeability and the ability to stimulate the migration of capillary endothelial cells are mediated via VEGFR-2 and that the activation of VEGFR-2 is sufficient to induce these biological effects. Moreover, downstream signaling from VEGR-2 requires activation of the MAPK pathway through at least ERK1 and ERK2 (29). However, the possibility remains that there are additional, as yet unknown receptors for VEGF and VEGF-C, which could mediate the vascular permeability effect of VEGF-C instead of or in addition to VEGFR-2.
Interestingly, the structural requirements for binding and activation of VEGFR-2 by VEGF and VEGF-C are different. Despite the prominent similarity between the mature VEGF-C and VEGF 121 , none of the basic amino acid residues shown to be critical for VEGF binding to VEGFR-2 (28) are conserved in VEGF-C. Unlike fully processed VEGF-C, which forms noncovalent homodimers (18), VEGF needs to be a covalent dimer for efficient receptor binding and activation (21). Both the second and fourth cysteine residues are critical for the dimerization and biological activity of VEGF. In the case of VEGF-C, only the fourth conserved cysteine residue (Cys 165 ) is critical, whereas elimination of the second one (Cys 156 ) did not affect dimer formation. Instead, a small fraction of the ⌬N⌬C156S molecules acquired the ability to form disulfide-linked homodimers, which apparently were also inactive toward VEGFR-2. Taken together, these data indicate that VEGF-C homodimerization is necessary but not sufficient for VEGFR-2 activation and for induction of the VEGF-like effects of VEGF-C. The inability of the ⌬N⌬C156S,C165S and ⌬N⌬C165S proteins to form homodimers and to bind and to activate VEGFR-3 suggests that VEGFR-3 activation also requires the dimerization of VEGF-C.
VEGF-C, unlike VEGF, stimulates the growth of lymphatic vessels (15). It would be interesting to know whether the selective activation of VEGFR-3 is sufficient for this effect or whether activation of both receptors is required. The ⌬N⌬C156S mutant provides a suitable tool to answer this question and to delineate the signaling pathways involved in the various activities of VEGF-C. The ERK/MAPK pathway may be involved in the permeability, proliferative, and migration-inducing activities of VEGF-C toward capillary endothelial cells. If both VEGFR-3 and VEGFR-2 are needed for the full range of biological activities of VEGF-C toward the lymphatics, one needs to consider the possibility that ⌬N⌬C156S would inhibit the effects of wt VEGF-C. On the other hand, if VEGFR-2 is able to mediate the VEGF-like responses of VEGF-C independently of VEGFR-3, as seems likely because BCE cells do not express VEGFR-3, 3 then ⌬N⌬C156S would be a valuable and highly selective VEGFR-3 agonist. These issues will be further studied.