Structural Requirements of Anticoagulant Protein S for Its Binding to the Complement Regulator C4b-binding Protein*

The vitamin K-dependent anticoagulant protein S binds with high affinity to C4b-binding protein (C4BP), a regulator of complement. Despite the physiological importance of the complex, we have only a patchy view of the C4BP-binding site in protein S. Based on phage display experiments, protein S residues 447–460 were suggested to form part of the binding site. Several experimental approaches were now used to further elucidate the structural requirements for protein S binding to C4BP. Peptides comprising residues 447–460, 451–460, or 453–460 of protein S were found to inhibit the protein S-C4BP interaction, whereas deletion of residues 459–460 from the peptide caused complete loss of inhibition. In recombinant protein S, each of residues 447–460 was mutated to Ala, and the protein S variants were tested for binding to C4BP. The Y456A mutation reduced binding to C4BP ∼10-fold, and a peptide corresponding to residues 447–460 of this mutant was less inhibitory than the parent peptide. A further decrease in binding was observed using a recombinant variant in which a site for N-linked glycosylation was moved from position 458 to 456 (Y456N/N458T). A monoclonal antibody (HPSf) selective for free protein S reacted poorly with the Y456A variant but reacted efficiently with the other variants. A second antibody, HPS 34, which partially inhibited the protein S-C4BP interaction, reacted poorly with several of the Ala mutants, suggesting that its epitope was located in the 451–460 region. Phage display analysis of the HPS 34 antibody further identified this region as its epitope. Taken together, our results suggest that residues 453–460 of protein S form part of a more complex binding site for C4BP. A recently developed three-dimensional model of the sex hormone-binding globulin-like region of protein S was used to analyze available experimental data.

The major isoform of the human complement regulator C4bbinding protein (C4BP) 1 circulates in a 1:1 high affinity noncovalent complex with the vitamin K-dependent anticoagulant protein S (KD, 0.1-0.6 nM), thus bringing the complement and coagulation systems into close interplay (1)(2)(3)(4)(5). C4BP attenuates the classical complement pathway by serving as a decayaccelerating factor for the C4b-C2a complex and as a cofactor to factor I in the proteolytic degradation of C4b (6). Protein S is an anticoagulant, acting as a cofactor to activated protein C (APC) in the proteolytic degradation of the activated forms of coagulation factors V (7,8) and VIII (9 -11). Protein S and APC form a complex on the surface of negatively charged phospholipid membranes, and protein S is involved in localizing and orienting the active site of APC toward its substrates (7,12). Protein S has also been reported to exert a direct anticoagulant activity independent of APC (13)(14)(15)(16). The physiological importance of the anticoagulant function of protein S is supported by the association between heterozygous protein S deficiency and an increased risk of thrombosis (17). Upon binding to C4BP, protein S loses its APC cofactor activity, whereas the functions of C4BP are not perturbed (18 -20). It has been suggested that protein S helps anchor C4BP to negatively charged phospholipid exposed on cell surfaces at sites of injury, thereby assisting regulation of inflammation (1).
C4BP and protein S are multidomain proteins. Protein S contains a ␥-carboxyglutamic acid (Gla) domain, a thrombinsensitive loop (thrombin-sensitive region), four epidermal growth factor-like domains, and a COOH-terminal region that is homologous to sex hormone-binding globulin (SHBG). The SHBG-like region comprises two laminin globular (LG) domains (LG1 and LG2), a fold present in the COOH-terminal part of the laminin ␣ chain, and many other extracellular matrix proteins (21)(22)(23).
LG2 contains three glycosylation sites, two of which are conserved in several species (3, 24 -28). C4BP is an approximately 570-kDa glycoprotein composed of 6 -8 polypeptide chains connected at their COOH-terminal ends by disulfide bridges, giving the oligomer a spider-like shape, as revealed by high resolution electron microscopy (29). The major isoform of C4BP comprises seven ␣ chains and a ␤ chain, whereas the minor isoform has no ␤ chain and does not interact with protein S (30). Although C4BP is an acute phase protein, it is primarily the form lacking the ␤ chain that increases during the acute phase inflammatory response so that levels of free protein S remain stable (31,32). The ␣ and ␤ chains are composed of repeating domains of about 60 amino acids denoted complement control protein (CCP) domains. The binding site for protein S is contained in CCP1-CCP2 of the ␤ chain (33)(34)(35). Using a molecular model of the ␤ chain in combination with recombinant ␤ chain expression and site-directed mutagenesis, it has been shown that a solvent-exposed hydrophobic patch in CCP1 lined by a positively charged area on an otherwise negatively charged surface forms the key binding site for protein S (35).
Several studies have demonstrated that the C4BP-binding site in protein S is fully contained in the two LG domains (2,36). Using recombinant chimeric proteins created between protein S and the structurally related protein Gas6, it was recently shown that both LG domains contribute independently to the interaction (37). Synthetic peptides corresponding to protein S residues 413-434 (38), 447-460 (39), and 605-614 (40,41) have been reported to compete with protein S for binding to C4BP. Recombinant truncated protein S variants lacking the COOH-terminal 28 -58 residues demonstrated very low affinity for C4BP (4,42). In addition, specific substitutions of amino acids Lys 423 , Lys 427 , and Lys 429 with polar amino acids resulted in a 5-10-fold reduction in the affinities (43).
In the present study, we have continued characterizing the binding site in protein S for C4BP, focusing on the region encompassing residues 447-460, which, based on phage display experimentation, was suggested to be involved in C4BP binding (39). Using Ala scanning mutagenesis, peptide inhibition assays, surface plasmon resonance, and monoclonal epitope mapping, the involvement of this region in C4BP binding was elucidated. To gain better insight into the characteristics of the C4BP-binding site, the now presented experimental data and those on record were evaluated on a recently created three-dimensional model for the SHBG-like region of protein S (44).

MATERIALS AND METHODS
Reagents-Rabbit polyclonal antibodies against human protein S (PK-anti-hPS) and mouse monoclonal antibodies against human protein S (HPS 54, HPS 34, HPS 67, HPS 21, and HPS 42) have been described previously (45). HPS 54 was conjugated with HRP as described previously (46). Protein S and C4BP were prepared from human plasma as reported previously (47,49). A monoclonal antibody specific for Gla residues (M3B) (50) was a kind gift of Drs. Mark Brown and Johan Stenflo. HRP was obtained from Roche Molecular Biochemicals. 1,2-Phenylene diamine tablets and HRP-conjugated goat anti-mouse IgG were obtained from DAKO. N-Glycosidase F was from Roche Molecular Biochemicals.
Synthetic Peptides-Five peptides (Table I) with acetylated NH 2 termini and amidated COOH termini were synthesized on a MilliGen 9050 Plus synthesizer and purified by high pressure liquid chromatography, as described previously (39).
Site-directed Mutagenesis-The cDNA encoding human protein S (in vector pcDNA3; Invitrogen) was mutated using the QuikChange kit (Stratagene) and a series of oligonucleotides containing the desired mutation, as described previously (51). A total of 13 cDNAs were produced, encoding variants designated S447A, G448A, I449A, Q451A, F452A, H453A, I454A, D455A, Y456A, N457A, N458A, V459A, and S460A. In a fourteenth variant, a new glycosylation consensus sequence was created around position 456 through the substitutions Y456N and N458T. The double mutant (Y456N/N458T) lacked the wild-type carbohydrate side chain at position 458. The mutations were confirmed by DNA sequence analysis using an ABIprism Taq polymerase-based sequencing kit with fluorescent dye terminators (PerkinElmer Life Sciences).
Transient Eukaryotic Cell Expression-Vectors encoding wild-type protein S, the various Ala variants, and the Y456N/N458T double variant were used to transfect monkey kidney COS-1 cells by the DEAE-dextran method (51). Expression levels were determined with an ELISA essentially as described previously (46), except that the wells were coated with PK-anti-hPS, and samples were incubated overnight. The expression levels from confluent 10-cm Petri dishes with 10 ml of added Optimem were found to vary between 35 and 150 ng/ml/24 h.
Most mutant expression levels were comparable to those seen for wildtype protein S. Exceptions were F452A (50% of wild type; p ϭ 0.02), I454A (76% of wild type; p ϭ 0.07), Y456A (35% of wild type; p ϭ 0.009), and S460A (75% of wild type; p ϭ 0.04), which demonstrated lower expression. The reported values are the mean of three independent experiments and were compared with the expression of wild-type protein S using a paired Student's t test. Conditioned media containing recombinant proteins were concentrated in Centricon concentrators (Amicon) and stored at Ϫ20°C until further analysis. SDS-PAGE and Western blotting were performed following standard procedures. To deglycosylate the recombinant proteins, 5-10 l of the concentrated culture medium containing ϳ1 g/ml protein S were treated with Nglycosidase F (0.5 unit/sample) under reducing and denaturing conditions and then analyzed by Western blotting using a polyclonal protein S antiserum Stable Eukaryotic Cell Expression-The cDNAs encoding wt protein S and the D455A and Y456A protein S variants were used to transfect human embryonic kidney 293 cells using the Lipofectin method, and stable cell lines resistant to G418 were established, as described previously (51). The recombinant proteins grown in the presence of vitamin K were collected in Optimem and purified using an immobilized calciumdependent monoclonal antibody (HPS 21) directed against the Gla domain essentially as described previously (51). The expression level was determined with the ELISA; wt protein S and the D455A mutant were present in ϳ2-3 mg/liter, whereas the expression of the Y456A variant was 10 -20-fold less. The purified proteins were analyzed by SDS-PAGE and Western blotting using polyclonal protein S antibodies or the M3B monoclonal antibody recognizing Gla residues (50). The concentration of the proteins was determined by amino acid analysis after acid hydrolysis in 6 M HCl, and the Gla content was measured after base hydrolysis using methods outlined previously (26).
Surface Plasmon Resonance Studies-Surface plasmon resonance experiments were carried out using a BIAcore 1000 system. Immobilization was performed using 10 mM Hepes, 0.15 M NaCl, 3.4 mM EDTA, and 0.005% Tween 20, pH 7.4, as flow buffer and a flow rate of 5 l/min. Equal volumes of 0.1 M N-hydroxysulfosuccinimide and 0.4 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide were mixed, and 40 l of this solution were injected to activate the carboxymethylated dextran. Then 40 l of either 25 g/ml monoclonal antibody against HPS 34 in 10 mM sodium acetate, pH 4.75, or 60 g/ml C4BP in 10 mM sodium acetate, pH 4.5, were injected. Unreacted N-hydroxysulfosuccinimide-ester groups were deactivated by injecting a 20-l pulse of 1 M ethanolamine hydrochloride, pH 8.5, and uncoupled protein was removed with 20 l of 0.1 M HCl. Flow rates of 5-20 l/min yielded identical results for the binding reactions. Surface plasmon resonance data were fitted as described previously (2,39).
Binding of protein S to HPS 34 was performed using 10 mM Hepes, 0.15 M NaCl, 2 mM CaCl 2 , and 0.005% Tween 20, pH 7.4, as flow buffer. Wild-type recombinant protein S was injected at 12.5, 25, 50, 100, and 200 nM concentrations. In addition, 50 nM of a 1:1 complex between protein S and C4BP was injected. The association phase was monitored for 12 min, and the dissociation into pure buffer was followed for 4 h. The remaining bound protein S was then removed by washing with 25 l of 0.1 M HCl. Flow rates of 5-30 l/min were tested and gave identical results. Binding experiments were also performed using the F452A, I454A, and Y456A mutants. The data were fitted as described previously (2).
Binding of protein S to C4BP was performed in a flow buffer comprising 10 mM Hepes, 0.15 M NaCl, 2 mM CaCl 2, and 0.005% Tween 20, pH 7.4. Wild-type protein S and variants were injected at concentrations of 1-20 nM. The association phase was monitored for 15 min, and the dissociation into pure buffer was followed for 300 min. Bound protein S was then removed by washing with 25 l of 0.1 M HCl. Flow rates of 5-30 l/min were tested and gave identical results.
Peptide inhibition of protein S binding to C4BP was performed using 10 mM Hepes, 0.15 M NaCl, 3.4 mM EDTA, and 0.005% Tween 20, pH 7.4, as flow buffer. Each peptide was injected at concentrations ranging from 1 to 400 M together with 60 nM protein S. The association phase was monitored for 15 min, and bound protein S or peptide was then removed by washing with 25 l of 0.1 M HCl. Flow rates of 5-30 l/min were tested and gave identical results.
Enzyme-linked Ligandsorbent Assay Plate Assay-Conditioned medium containing each recombinant protein S variant was tested for direct binding to immobilized C4BP using the enzyme-linked ligandsorbent assay method, as described previously (46), except that an overnight incubation step was used. The concentration of protein S in each sample was standardized before performing the assays. In brief, wells were coated with purified C4BP (10 g/ml), blocked with bovine serum albumin, and washed with 50 mM Tris-HCl, 150 mM NaCl, 2 mM CaCl 2 , and 0.1% Tween 20, pH 7.5. Conditioned medium containing wild-type protein S or the variants (serially diluted with 50 mM Tris-HCl, 150 mM NaCl, 2 mM CaCl 2 , and 0.1% bovine serum albumin, pH 7.5) was added, and the plates were incubated overnight at 4°C. The wells were washed, and HRP-labeled antibody HPS 54 directed against the epidermal growth factor 1-like module of protein S (45) was used for detection of bound protein S.
The apparent dissociation constant, K D app , for the interaction was calculated by fitting the absorbance data to the formula A n ϭ A/(1 ϩ (K D app /C)), where A n is the observed absorbance, A is the maximum absorbance obtained with wild-type protein S, and C is the concentration of protein S. It was assumed that the amount of protein S bound was negligible compared with the total concentration of protein S, such that C reflects free protein S.
Protein S Binding to Monoclonal Antibody HPS 34 -Conditioned medium containing wild-type protein S or the variant proteins was tested for binding to immobilized HPS 34 by ELISA. Bound protein S was detected with the antibody HPS 54 as described previously (46). Apparent dissociation constants were calculated as described above. The ability of HPS 34 to inhibit the binding of human protein S to immobilized C4BP was tested using the enzyme-linked ligandsorbent assay method (46). In brief, aliquots of 5 nM plasma-purified protein S were preincubated with various concentrations of HPS 34 (up to a 500-fold molar excess) for 30 min at room temperature, and then the samples were added to C4BP-coated wells and incubated for 1 h at room temperature. After washing, HRP-labeled HPS 54 was used as the detecting antibody.
Epitope Mapping by Phage Display-Phage display experiments using immobilized HPS 34 as a target and random linear 15-mer peptides displayed on the bacteriophage surface were performed as described previously (39). The selected peptides were aligned against the protein S sequence using the HOMOFILE and AVEHOM programs (39).

Inhibition of the Protein S-C4BP Interaction by Synthetic
Peptides-Five peptides (Table I) were tested for their ability to compete with protein S for binding to C4BP using surface plasmon resonance analysis. Peptides corresponding to residues 447-460, 451-460, and 453-460 of protein S were found to inhibit protein S binding to C4BP (Fig. 1) at concentrations comparable to those observed previously for peptides 439 -460 and 447-468 (39). The 447-460/Y456A peptide required an ϳ3-fold higher concentration to give half-maximum inhibition, and the 447-458 peptide showed no inhibition at the concentrations tested.
Expression and Characterization of Protein S Variants-Thirteen recombinant protein S variants were generated by replacing each amino acid in the 447-460 region of protein S ( Fig. 2A) with Ala. The transiently expressed mutants were analyzed by SDS-PAGE and detected by Western blotting using a polyclonal antibody (PK-anti-hPS) (Fig. 2B). All migrated as single bands. The N458A and S460A variants exhibited an increased mobility, consistent with loss of the carbohydrate moiety present at residue Asn 458 in wild-type protein S (52). After deglycosylation, the different recombinant proteins demonstrated migration rates similar to that of deglycosylated wild-type protein S, suggesting that all recombinant proteins were glycosylated to the expected degree. Furthermore, the Western blotting patterns of the deglycosylated recombinant proteins were identical to that of deglycosylated plasma-derived protein S (data not shown). Recombinant protein S was tested in Western blotting and ELISA techniques with a panel of carefully characterized monoclonal antibodies that reacted with conformation-dependent epitopes in different domains to investigate the structural integrity of the proteins and their correct folding. The antibodies tested were HPS 21 (reacting with a calcium-dependent epitope in the Gla domain), HPS 67 (reacting with a calcium-dependent epitope in thrombin-sensitive region), and HPS 54 (reacting with a calcium-dependent epitope in epidermal growth factor 1-like module). These three antibodies demonstrated calcium-dependent recognition of the recombinant protein S, suggesting the first three domains of protein S to be correctly folded and to bind calcium similarly to plasma-derived protein S (results not shown). Three of the protein S variants, wt protein S, D455A, and Y456A, were also expressed in stable cell lines, purified, and analyzed by SDS-PAGE and Western blotting using polyclonal anti-protein S antibodies as well as an antibody recognizing Gla residues. The proteins migrated to the expected positions and were recognized by both antibodies (Fig. 2C). The concentration of Gla residues was determined after base hydrolysis, and all three recombinant proteins yielded the same number of Gla/mol as plasma-derived protein S (ϳ6 residues/mol), which was tested in parallel.
Binding of Protein S Variants to C4BP-The Ala-substituted variants were tested for their ability to bind to immobilized C4BP in microtiter plates. A statistically significant difference relative to wild-type protein S was obtained only for the Y456A variant (Fig. 2D). Binding to immobilized C4BP was also studied under flow conditions using surface plasmon resonance (Fig. 2E). Most of the variants exhibited binding characteristics comparable to those of wild-type protein S, but the dissociation rates were increased 10-fold for the Y456A variant and 3-fold for the I454A variant. To further characterize the interaction between the recombinant Y456A variant and C4BP, the purified proteins derived from the stable cell lines were used. In this experiment, the Y456A variant was compared with wt protein S and the D455A variant. The wt protein S and D455A yielded similar results, with identical k off values of 2.8 ϫ 10 Ϫ5 s Ϫ1 , whereas the k on values were 2.6 and 1.5 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively, resulting in K D values of 0.11 and 0.19 nM, respectively. The Y456A variant demonstrated a decreased association rate (k on Ϸ 5 ϫ 10 4 M Ϫ1 s Ϫ1 ) and an increased dissociation rate (k off Ϸ 1.4 ϫ 10 Ϫ4 s Ϫ1 ), resulting in a K D of 1.4 nM. Thus, the Y456A variant bound to C4BP with ϳ10-fold lower affinity than wt protein S.
Effect of a Shifted Glycosylation Site-Because the Y456A substitution in protein S decreased binding to C4BP, we sought to further elucidate the importance of this region for binding by shifting the natural glycosylation site from Asn 458 to residue 456. This was achieved by the double substitution Y456N/ N458T (Fig. 3A). The mobility of the variant in SDS-PAGE gels was identical to that of the wild-type protein (Fig. 3B), indicating that the number of carbohydrate side chains remained unaffected (Figs. 2B and 3B). In contrast, a faster mobility was observed for the N458A variant, consistent with the loss of one carbohydrate side chain. After deglycosylation using N-glycosidase F, the different protein S variants migrated like deglycosylated wild-type protein S. The expression level of the double mutant Y456N/N458T (20 ng/ml/24 h) was 5-fold lower than that of the wild-type protein, but sufficient material was ob-tained for binding analysis. Y456N/N458T bound efficiently to the PK-anti-hPS antibody (Fig. 3A), but its affinity for C4BP was significantly more reduced than that of the Y456A variant (Fig. 3C), suggesting that the carbohydrate chain attached to position 456 covers part of the recognition site.

Recognition of Protein S Variants by Two Monoclonal Antibodies with Epitopes in the SHBG-like Region of Protein S-A
monoclonal antibody from the Asserachrom free protein S kit that specifically recognizes the free form of protein S (HPSf) was tested for its ability to recognize the protein S variants. Only one mutant, Y456A, was poorly recognized by the monoclonal antibody used in the ELISA (Fig. 4), i.e. similar to the result on C4BP binding obtained with this series of protein S variants (Fig. 2). A second monoclonal antibody, HPS 34, which reacts with an epitope in the SHBG-like region of protein S (45), was also tested for its ability to interact with the mutants. In contrast to the very selective effect on recognition of C4BP by HPSf seen only with Y456A, several mutations affected binding of HPS 34. N458A and S460A were not recognized by HPS 34. Drastic reductions in affinity for HPS 34 (by a factor of 100 or more) were seen for the three variants Q451A, I454A and Y456A, whereas H453A, D455A, and V459A yielded less severe effects (Fig. 5A). The results demonstrate that the HPS 34 epitope is located in the 451-460 region. Surface plasmon resonance was used to measure the affinity and kinetics of the protein S-HPS 34 interaction (Fig. 5, B and C). A dissociation rate constant (k off ) for the protein S-HPS 34 complex of 2 ϫ 10 Ϫ4 s Ϫ1 and an association rate constant (k on ) of 2 ϫ 10 4 M Ϫ1 s Ϫ1 were calculated, yielding an equilibrium dissociation constant (K D ) of 10 nM. Hence, HPS 34 bound protein S approximately 100-fold more weakly than did C4BP. No binding to HPS 34 was detected when protein S in complex with C4BP was injected (Fig. 5C), suggesting that C4BP binding blocks the HPS 34 epitope on protein S. HPS 34 Epitope Mapping Using Phage Display-Linear 15mer peptides were affinity-purified from a random phage display library with immobilized HPS 34 as a target and aligned against the protein S sequence. Many of the selected peptides contained a high proportion of tryptophan and proline residues. Alignment against the sequence of human protein S produced high similarity scores for all peptides when the first residue of the 15-mer peptide was located between residues 442 and 460. When the sequence alignments were overlaid, a significant overlap of the 15-mer sequences was obtained for residues 451-460 (data not shown). This suggested that the HPS 34 epitope encompasses some of the residues in the 451-460 region.
Partial Inhibition of C4BP-Protein S Interaction by HPS 34 -The HPS 34 epitope mapping results suggested the possibility that binding sites for C4BP and HPS 34 on protein S could partially overlap. A microtiter plate assay with immobilized C4BP was used to test whether HPS 34 and C4BP competed for protein S. In this assay, increasing concentrations of HPS 34 were added over a fixed concentration of fluid phase protein S, and the amount of protein S bound to C4BP was then measured. An approximately 60ϫ molar excess of HPS 34 was required to obtain close to 50% inhibition of C4BP-protein S complex formation (Fig. 6). No further inhibition was observed even at a 500-fold excess of the antibody. A control experiment using HPS 42 instead of HPS 34 suggested that no nonspecific inhibition of the protein S binding occurred due to the presence of a high molar excess of antibody (Fig. 6). The HPS 42 epitope The detecting antibody of the Asserachrom free protein S kit (HPSf) was incubated with the recombinant protein S variants that had been bound to immobilized polyclonal antibodies to protein S. The various recombinant protein S variants were added to the microtiter plates at a concentration of 32 ng/ml. The antibody selectively reacted poorly with the Y456A mutant, whereas all other protein S variants were recognized by the antibody.

FIG. 3. Effect of a shift in glycosylation site.
A, using mutagenesis, the glycosylation site at Asn 458 of protein S was shifted to residue 456 by a double substitution (Y456N/N458T). B, Western blot analysis of the glycosylation-shifted Y456N/N458T double variant and control protein S. Top panel, the proteins were immunoprecipitated from conditioned medium with PK-anti-hPS and Western blotted using antibody HPS 54 and alkaline phosphatase-conjugated anti-IgG for detection. Bottom panel, conditioned medium containing 5 ng of protein S was treated with N-glycosidase F and analyzed as described in the Fig. 2 legend. C, binding of wild-type protein S and the variants Y456A and Y456N/N458T to immobilized C4BP (top) or PK-anti-hPS (bottom) was measured in microtiter plate assays. Bound protein S was detected with HRP-labeled HPS 54.
is located in the thrombin-sensitive region, and it does not affect the binding of HPS 34 or C4BP (45). The inhibitory effect of HPS 34 on protein S binding to C4BP was only seen when relatively short incubation times were used (up to ϳ90 min). This suggests that at equilibrium, HPS 34 was displaced from protein S by C4BP due to a much higher affinity of the C4BPprotein S complex as compared with the HPS 34-protein S complex.

DISCUSSION
The very high affinity of the interaction between protein S and C4BP, its hydrophobic nature, and the very slow rate of dissociation of the complex suggest that the two molecules circulate in blood as a stable complex (2,35,37). In protein S, both LG modules are involved in the binding of C4BP, suggesting that important residues for the interaction are distributed on both modules (37). Three specific regions of protein S have been suggested to be involved, residues 413-434 in the LG1 module and residues 447-460 and 605-614 in the LG2 module (39,40,43). Involvement of the 447-460 region in C4BP binding was initially established using phage display technology (39). Phages interacting with the ␤ chain of C4BP displayed sequences similar to the 447-460 region. Using peptides comprising protein S residues 447-460 and shorter peptides, the sequence capable of inhibiting C4BP-protein S interaction was narrowed down to residues 453-460. Then, an Ala scanning strategy was devised to further evaluate the importance of this segment and the specific involvement of the different amino acid residues. A transient expression system was used for screening purposes, and the recombinant proteins were characterized with monoclonal antibodies recognizing calcium-dependent conformational epitopes located in the first three domains. In addition, the presence of carbohydrate side chains in the LG domains was investigated with enzymatic deglycosylation. The results of these experiments indicated that the recombinant proteins were correctly folded and posttranslationally modified. The conclusion that the recombinant proteins were correctly folded was also supported by the demonstration of intact C4BP binding ability for almost all of the recombinant protein S variants. The only substitution that resulted in a decreased affinity of protein S for C4BP binding was Y456A, although an effect on the rate of dissociation was seen also for the I454A substitution. The impaired binding of this mutant to C4BP was confirmed with purified recombinant protein that was expressed in a stable cell line, an approach that allowed the demonstration of the presence of posttranslationally carboxylated Gla residues. The reduced affinity for C4BP of the Y456A mutant appeared to be due to a combination of destabilization of the bound conformation (observed as an increased off-rate) and direct interaction between the tyrosine ring and C4BP. The effect of the Y456A mutation was paralleled by a reduced inhibitory action of the 447-460 (Y456A) peptide as compared with the peptide with the wild-type sequence. Tyrosine is a common amino acid in protein-protein interaction sites because of its mixed hydrophobic/aromatic nature and ability to form hydrogen bonds. For instance, a tyrosine residue in bovine insulin-like growth factor-binding protein 2 was recently identified as a determinant of the interaction with the insulin-like growth factor because a Tyr3 Ala mutation reduced the affinity by a factor of 3.5-4 (53). As in the situation studied here, an increase in both on-rate and off-rate was observed; hence, the effects on kinetics were larger than the effect on the overall affinity constant.
Protein S is glycosylated at three positions, N458, N468, and N489, all in the SHBG region (52). S460A is a naturally occurring mutation in phenotypic protein S deficiency (54). The Heerlen polymorphism (S460P) also lacks glycosylation at Asn 458 . In the present work, the two glycosylation-deficient variants, N458A and S460A, were observed to bind to C4BP with an affinity similar to that of wild-type protein S, confirming the results on record (52,55). Hence, the carbohydrate at position 458 does not appear to influence the affinity for C4BP. We shifted the consensus sequence for N-glycosylation to localize it to residue 456. This resulted in a variant that was glycosylated at position 456, and binding to C4BP was almost abolished (Fig. 3). Hence, moving the bulky carbohydrate moiety by as little as 2 residues, from position 458 to 456, was sufficient to block a surface of importance for the interaction with C4BP. The fact that a low level of specific binding to C4BP was still detected in the Y456N/N458T mutant supports the idea that multiple binding sites located on both LG modules of protein S produce a high affinity recognition surface for C4BP. The presence of a carbohydrate site chain at position 456 in the Y456N/N458T mutant indicated that residue Y456 is fully or partially solvent-exposed (Fig. 7). The strategy of using Nglycosylation for the inhibition of protein-protein interactions to test putative binding sites seems to be informative and could be extended to other systems, as has been recently shown (48).
The Y456A substitution in protein S resulted in decreased recognition by a commercially available antibody (HPSf) that is specific for the free form of protein S. The rest of the Ala mutations had no effect on the binding of protein S to HPSf or C4BP, suggesting that the epitopes for C4BP and HPSf on the protein S surface are similar. In contrast, monoclonal antibody HPS 34, recognizing the SHBG-like region of protein S (45), reacted poorly against several of the Ala mutants in the 453-460 region of protein S (Q451A, H453A, I454A, D455A, Y456A, N458A, and S460A), indicating that its epitope covers most of this region. This conclusion was further supported by phage display experiments using HPS 34 as target. The sequences displayed on the isolated phages showed highest similarity to the 451-460 region. The carbohydrate side chain at Asn 458 could also be part of the epitope because mutants N458A and S460A did not bind HPS 34. The affinity of protein S for HPS 34 is 2 orders of magnitude lower than the affinity for C4BP and has a half-life of 1 h. A modest inhibitory effect of HPS 34 on the C4BP binding to protein S was found only when HPS 34 and protein S were preincubated using a large molar excess of the antibody, whereas no binding of HPS 34 to the C4BPprotein S complex was detected with surface plasmon resonance. However small, the effect of HPS 34 on C4BP binding to protein S and the fact that its epitope is located in the 453-460 area support the view that the interaction sites of HPS 34 and C4BP on protein S partially overlap. Interestingly, the results with HPS 34 are similar to those obtained with another monoclonal antibody (LJ-56) raised against a peptide comprising residues 420 -437. LJ-56 failed to recognize variants K423E and K429E but had only a small effect on C4BP binding. The LJ-56 antibody only bound to free protein S and did not bind to protein S in complex with C4BP. Mutations may abolish antibody binding, but effects on C4BP binding may be only modest.
The lack of three-dimensional structural data of protein S has limited the interpretation of the experimental results obtained thus far. We recently created a three-dimensional model for the SHBG-like region of human protein S (44) based on x-ray structures of the related protein SHBG and of the LG domains of laminin (22,23). Therefore, it is now possible to analyze the available experimental data in the light of this predicted structure (Fig. 7). In protein S, the epidermal growth factor-like module 4 and LG1 modules are connected by a FIG. 7. Proposed three-dimensional structure for the SHBG-like region of protein S. The two LG modules are shown with their ␤ strands drawn as an arrow, whereas the remaining loops and short helices are presented as tubes. The glycosylated Asn residues are yellow, and a common sugar motif was grafted onto the structure to provide approximate information about the overall surface that could be covered by such side chains. Two peptide segments suggested to represent a binding region for C4BP are shown as green ribbon. The segment investigated in the present study is presented as a blue ribbon, with the C␣ atom of all the residues that were mutated to Ala depicted by a blue sphere.