Localization of a hydrophobic binding site for anticoagulant protein S on the beta -chain of complement regulator C4b-binding protein.

C4b-binding protein (C4BP) is a plasma glycoprotein involved in regulation of the complement system. C4BP consists of seven alpha-chains and one unique beta-chain, all constructed of repeating complement control protein (CCP) modules. The beta-chain, made up of three CCPs, binds tightly to vitamin K-dependent protein S, a cofactor to anticoagulant activated protein C. When bound to C4BP, protein S loses its activated protein C cofactor function. In this study, we have mutated potentially important amino acids located at the surface of CCP1 of the beta-chain to probe the protein S-C4BP interaction. The substitutions were designed after analysis of a homology-based three-dimensional structure of the beta-chain and were L27T/F45Q, I16S/V18S, V31T/I33N, I16S/V18S/V31T/I33N, L38S/V39S, and K41E/K42E. The mutants were expressed in a prokaryotic system, purified using an N-terminal His-tag, refolded using an oxido-shuffling system, and tested in several assays for their ability to bind protein S. Our data define Ile(16), Val(18), Val(31), and Ile(33) as crucial for protein S binding, with secondary effects from Leu(38) and Val(39). In addition, Lys(41) and Lys(42) contribute slightly to the interaction. Our results further confirm that surface hydrophobicity analysis may be used to identify ligand recognition sites.

and contains two disulfide bridges and a central antiparallel ␤-sheet (5). CCP domains are present in numerous proteins both within and outside the complement system (6). Some CCP-containing molecules have been investigated by NMR or x-ray crystallography, e.g. vaccinia virus complement control protein (7), ␤ 2 -glycoprotein I (8), and CD46 (9). The knowledge of the three-dimensional structure of these domains enabled us to construct a homology-based model of C4BP.
The ␣-chains of C4BP, consisting of eight CCPs, bind complement protein C4b (1). A key recognition site for C4b on C4BP has been recently ascribed to a cluster of positively charged amino acids on the interface of CCP1-CCP2 of the ␣-chain (10). The unique C4BP ␤-chain, essentially made up of three CCPs, binds protein S, an anticoagulant molecule that acts mainly as cofactor to activated protein C in the degradation of coagulation factors Va and VIIIa. C4BP and protein S form a high affinity, noncovalent complex with a 1:1 molecular ratio, which is greatly enhanced by calcium (11). The C-terminal sex hormone globulin binding-like region of protein S is involved in the interaction with C4BP (12,13). This domain in protein S is expected to have calcium-binding site(s), whereas it has never been shown or proposed that C4BP interacts with any metal ion (12,13).
In plasma, ϳ70% of protein S is in complex with C4BP. Only free protein S functions as an activated protein C cofactor (11). In contrast, C4BP in complex with protein S can still exert its regulatory functions on the complement system (14). C4BP regulates the plasma availability of free protein S since the concentration of free protein S represents the molar excess of protein S over C4BP (15). The biological importance of the protein S-C4BP interaction is emphasized by the fact that only the concentration of free protein S can clearly be linked to thrombotic risk in patients suffering from protein S deficiencies (16). The physiological purpose of the interaction between C4BP and protein S is not yet fully understood. However, protein S, being a vitamin K-dependent protein, has a very high affinity for negatively charged phospholipids; and therefore, it could localize C4BP to surfaces where such phospholipids are exposed (17,18).
Our group has shown that the ␤-chain of C4BP (19) contains the protein S-binding site (20); more precisely, CCP1 is required for binding to occur (21). It was recently suggested by van de Poel et al. (22,23) that CCP2 also contributes to the binding to a small extent. van de Poel et al. used a different approach to study the binding. In their investigation, chimeras were constructed composed of individual CCP modules (or the different CCPs in combination) fused to the N-terminus of a modified tissue plasminogen activator. They found that CCP2 increased the affinity for protein S ϳ5-fold.
Ferná ndez and Griffin (24) used synthetic peptides to probe the protein S-C4BP interaction and suggested residues 31-45 on C4BP to be important for protein S binding. This hypothesis was supported in a subsequent report (25), where it was found that preincubation of C4BP with monoclonal antibody 6F6 (directed against a region located nearby residues 31-45) inhibited the C4BP and protein S interaction. It was then concluded that the antibody interfered with the interaction most likely because of steric hindrance (25).
We have previously shown that binding of protein S to C4BP varied only to a small extent with the concentration of salt, in a manner implying a significant contribution from hydrophobic interactions, with minor roles played by electrostatic forces (26). These experiments were the first to confirm the hypothesis that a hydrophobic cluster at the surface of CCP1 may be the main binding site for protein S (27).
In this study, we have mutated potentially important amino acids located in this hydrophobic cluster on C4BP CCP1. The mutations were chosen based on the homology-based, computer-generated three-dimensional structure of the C4BP ␤-chain and previous experimental data. In addition, two lysines that could be responsible for the slight electrostatic component seen in the interaction were also mutated and studied. Wild-type CCP1 and CCP2 of the ␤-chain and the mutants were expressed in a prokaryotic system, purified on a nickel-Sepharose column, and refolded. All recombinant proteins were then tested for their ability to bind protein S. We found that substitution of four hydrophobic amino acids by polar residues in the first CCP of the ␤-chain decreased the apparent affinity for protein S 100-fold. Our results not only provide insights into the nature of the protein S-C4BP interaction, but also have implications for the prediction of binding sites at the surface of other CCP modules and may be valuable for the understanding of protein-protein recognition.

Cloning Procedure
The prokaryotic expression vector used for expression of the recombinant C4BP ␤-chain was pET-26b(ϩ) (Novagen). It carries an Nterminal pelB signal sequence and a C-terminal 6-His tag. CCP1 and CCP2 of the C4BP ␤-chain were cloned by polymerase chain reaction from full-length cDNA of the ␤-chain into pET-26b(ϩ) using the following primers: 5Ј-ATC CAT GGG ATC AGA TGC AGA GCA C-3Ј and 5Ј-CGC TCG AGA CTT TTG CAG ATG GGA AA-3Ј. This construct was then used as a template, and the mutations were introduced using the QuickChange site-directed mutagenesis kit (Stratagene). The templates and sense primers used for mutagenesis as well as the resulting mutants are shown in Table I. The various ␤-chain constructs were then transformed into Escherichia coli DH5␣ bacteria, and mutations were confirmed using an automated DNA sequencer (PerkinElmer Life Sciences).

Expression and Purification of Recombinant Proteins
cDNAs coding for the recombinant proteins were transformed into and expressed in E. coli strain BL21(DE3). About 1 ml of overnight culture of the transformed bacteria grown in Luria broth containing 30 g/ml kanamycin was used to inoculate 500 ml of the same medium. The bacteria were grown by shaking at 37°C until the absorbance at 600 nm was ϳ0.7. Expression of protein was induced by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and incubation was continued for 3 h. The culture was then centrifuged at 7000 ϫ g for 25 min at 4°C, and the bacterial pellet was resuspended in 100 ml of cold phosphate-buffered saline. After incubation for 15 min at room temperature, with lysozyme added to a final concentration of 100 g/ml, the bacteria were sonicated at 10 micron peak to peak and centrifuged in the same way. The pellet obtained was suspended in 6 M guanidine HCl, 20 mM Tris-HCl (pH 8.0), and 10 mM reduced glutathione and sonicated and centrifuged as described above. The supernatant was then applied to a nickel-nitrilotriacetic acid Superflow column (2.6 ϫ 12 cm, QIA-GEN) equilibrated with the same buffer. The column was washed with 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 20 mM imidazole, and the protein was eluted with 10 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 100 mM EDTA. Fractions containing protein were chosen by measurement of the absorbance at 280 nm and pooled. Tris-HCl (pH 8.3) and dithiothreitol were then added, both to a final concentration of 100 mM. After incubation for 2 h at 4°C, the sample was diluted in 50 mM Tris-HCl (pH 8.3), 3 mM cysteine, and 0.3 mM cystine so that the absorbance at 280 nm was equal to 0.1, and folding of the protein was accomplished by overnight dialysis at 4°C against the same buffer. Iodoacetamide was then added to the dialyzed sample to a final concentration of 5 mM, and dialysis was continued overnight at 4°C against 50 mM Tris-HCl (pH 8.0) and 10% glycerol. The dialyzed sample was then applied to a MonoQ column (2 ml, Amersham Pharmacia Biotech) equilibrated with the same buffer; protein was in the flowthrough. The protein was concentrated using an Amicon concentrator. Finally, the protein was dialyzed against 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 10% glycerol and stored at Ϫ70°C until further use. Exact concentrations of recombinant protein were determined by analysis of amino acid composition after hydrolysis in 6 M HCl for 24 h.

Structural Analysis of Recombinant Proteins
Binding of Monoclonal Antibodies-All recombinant proteins were tested for their ability to bind to seven monoclonal antibodies raised against the recombinant wild-type ␤-chain using a standard procedure. Microtiter plates were coated with 50 l of purified antibody at 10 g/ml in 75 mM sodium carbonate (pH 9.6) at 4°C overnight. The plate was then washed three times with wash buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% (w/v) Tween 20, and 2 mM CaCl 2 ), quenched in wash buffer with 3% fish gelatin for 2 h, and washed as described above. Increasing amounts of recombinant ␤-chain (wild-type and mutant) plus trace amounts of 125 I-labeled recombinant wild-type ␤-chain were added for 5 h at room temperature. The plates were then washed five times, and bound radioactivity was measured in a ␥-counter.
Gel Filtration-All recombinant proteins were analyzed by gel filtration (Superose 12 HR 10/30, Amersham Pharmacia Biotech). Fifty g of each protein was applied to the column, previously equilibrated with Tris-buffered saline (pH 8.0). The flow rate used was 0.5 ml/min. During each run, the absorbance of the eluate at 280 and 214 nm was constantly monitored.
Mass Spectrometry-Mass spectrometry, carried out at the Protein Analysis Center of the Karolinska Institute (Stockholm, Sweden), was performed on all recombinant proteins using quadrupole time-of-flight (Q-TOF) mass spectrometry (Micromass) (28). The proteins were dialyzed against 2% HAc. Samples were subsequently analyzed by nanoelectrospray mass spectrometry in 1% acetic acid and 60% acetonitrile.
Circular Dichroism-Recombinant proteins were dialyzed against 10 mM sodium phosphate (pH 7.4) before analysis. Approximately 50 g of each protein was analyzed in the far-UV region (185-250 nm). The resolution was 1 nm; the speed was 10 nm/min; and the response was measured every 8 s. Sensitivity was 20 millidegrees.
Electrophoretic and Blotting Techniques-Recombinant proteins were run on 15% SDS-polyacrylamide gel under reducing (ϳ1 g) and nonreducing (ϳ0.5 g) conditions for silver staining and under nonreducing conditions for radioligand blotting (ϳ2 g). For radioligand blotting, the proteins were transferred from the gel to a polyvinylidene difluoride membrane. The membrane was then incubated for 1 h at room temperature in a quenching solution composed of buffer A (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 0.5% (w/v) Tween 20) supplemented with 3% fish gelatin. The buffer was changed to buffer A with 2 mM CaCl 2 and trace amounts of 125 I-labeled protein S, and the membrane was incubated overnight at 4°C. The membrane was then washed with buffer A, dried, and exposed in a cassette. Finally, the membrane was scanned using a PhosphorImager (Molecular Dynamics). In addition, the intensities of the bands detected were estimated by densitometry using ImageQuant software (Molecular Dynamics).

Binding Assays
Direct Binding-Microtiter plates were coated with 50 l of protein (plasma-purified C4BP or recombinant wild-type or mutant ␤-chain) at 10 g/ml in 75 mM sodium carbonate (pH 9.6) at 4°C overnight. The plate was then washed three times with wash buffer and quenched in wash buffer with 3% fish gelatin for 2 h. The plate was washed as described above, and protein S was added at increasing concentrations (0 -240 nM final concentration) with trace amounts of 125 I-labeled protein S at 4°C overnight. Finally, the plate was washed five times with wash buffer, and bound radioactivity was determined in a ␥-counter.
Competition Assay-Microtiter plates were coated with 50 l of plasma-purified C4BP at 10 g/ml in 75 mM sodium carbonate (pH 9.6) at 4°C overnight. The plate was then washed three times with wash buffer, quenched in wash buffer with 3% fish gelatin for 2 h, and washed as described above. Increasing amounts of plasma-purified C4BP or recombinant ␤-chain (wild-type and mutant) plus trace amounts of 125 I-labeled protein S were added overnight at 4°C. The next day, the plates were washed five times with the same buffer, and bound radioactivity was measured in a ␥-counter.

Expression and Purification of Recombinant Proteins-
To study a possible binding site for protein S on the C4BP ␤-chain, the following mutations were introduced in the first CCP of the ␤-chain: L27T/F45Q, I16S/V18S, V31T/I33N, I16S/V18S/V31T/ I33N, K41E/K42E, and L38S/V39S. The mutations were chosen after analysis of a homology-based three-dimensional model of the ␤-chain (Fig. 1) and previous experimental data on the protein S-C4BP interaction (21, 24 -27). The residues selected for mutagenesis were all solvent-exposed, and their replacement should be structurally well tolerated, as removal of hydrophobic residues from the surface generally tends to stabilize a protein. This results from the fact that surrounding water molecules should be able to form hydrogen bonds with the newly introduced polar amino acids. Furthermore, substitution of amino acids expected to be positively charged at physiological pH (Lys 41 and Lys 42 ) with two negatively charged residues at the surface of CCP1 should not alter the structure of the domain. Indeed, the two Lys side chains point in opposite directions in the three-dimensional model, in part due to charge-charge repulsion. This phenomenon is also expected for the two resulting Glu residues.
All proteins (recombinant wild-type and mutant) were expressed in a prokaryotic system and purified utilizing a Cterminal His-tag. Since the expressed proteins were localized to inclusion bodies in the BL21(DE3) bacteria, the almost pure but misfolded protein obtained was then refolded using an oxido-shuffling system (30,31) and subsequently purified on a MonoQ column. All recombinant proteins were of similar apparent molecular mass (14 kDa) as judged by nonreducing SDS-PAGE (Fig. 2B). (A slight increase in apparent size was seen under reducing conditions for all recombinant proteins ( Fig. 2A).) Approximately 3 mg of purified protein was obtained from each liter of bacterial culture.
Characterization of Recombinant Proteins-Introduction of mutations did not affect the expression levels or electrophoretic mobilities of corresponding proteins compared with the wildtype protein. Furthermore, all constructs bound with similar apparent affinity to seven monoclonal antibodies raised against the recombinant wild-type protein. Results for two different antibodies are shown as an example of the binding curves obtained (Fig. 3).
The exact masses of all recombinant proteins (except L38S/ V39S) were analyzed by mass spectrometry. All masses were the precise expected match given the change in mass due to the changes of amino acids (Table II).
Upon gel filtration, all proteins eluted with a major single peak at a volume of ϳ15 ml, corresponding to a protein size of 13 kDa, as judged by a standard curve obtained from proteins with a known molecular mass, indicating that aggregates were not present (data not shown). The amount of protein eluted in the major peak, compared with the total amount of protein eluted, varied between 63 and 99%, the lowest being for the L27T/F45Q mutant. The major eluted peak for the mutant with four hydrophobic amino acids mutated, I16S/V18S/V31T/I33N, contained 84% of the total eluted protein.
Circular dichroism analysis of all recombinant proteins gave very similar spectra, once again confirming that introductions of mutations did not cause folding changes. The signal was not possible to judge below 205 nm due to background noise. Results are presented as strength of the signal relative to the The predicted CCP1 module is shown using ribbon representation. A synthetic peptide (residues 31-45) expected to form a binding site for protein S is colored green (24). Part of the proposed epitope (residues 51-65) for monoclonal antibody 6F6, blocking protein S binding due to steric hindrance, is colored orange (25). The residues mutated in this study are shown in red. Two potential N-linked glycans (yellow) in the first CCP of C4BP are approximated as an oligosaccharide (five sugar units) core structure. The side chains of Asn 47 and Asn 54 are displayed and colored blue.
Radioligand Blotting-To assess the binding of protein S to the various mutants, we used radioligand blotting, in which unreduced wild-type and mutant proteins immobilized on a polyvinylidene difluoride membrane were allowed to bind 125 Ilabeled protein S (Fig. 5, one representative experiment is shown). Mutant L27T/F45Q bound with similar or increased strength compared with the recombinant wild-type ␤-chain. I16S/V18S, V31T/I33N, I16S/V18S/V31T/I33N, and L38S/V39S all lost the binding ability for protein S as judged by the absence of bands on the blot after analysis by the PhosphorImager. The K41E/K42E mutant displayed weaker binding than the recombinant wild-type ␤-chain. For better quantification of differences between mutants, the intensities of all bands were estimated by measurement of density using ImageQuant software. The results are shown in Table III; each value represents the mean Ϯ S.D. of three different experiments. No binding was detectable when proteins were reduced (data not shown), implying that the two characteristic disulfide bonds present in the CCP modules were appropriately formed and, as expected, are crucial for the domain folding and thus for the interaction with protein S.
Direct Binding Assay-To further confirm the results obtained by radioligand blotting, a ligand binding assay was performed (Fig. 6A). Increasing amounts of unlabeled and 125 Ilabeled protein S were added to immobilized plasma-purified C4BP or the recombinant ␤-chain (wild-type or mutant). After washing, bound radioactive protein S was measured using a ␥-counter. The recombinant wild-type ␤-chain construct bound protein S with similar affinity as plasma-purified C4BP. This result also supports the structural integrity of the recombinant molecule and further confirms the fact that this expression system combined with the proper refolding technique leads to the production of a protein sharing the same characteristics as plasma-purified C4BP, but with the absence of glycosylation. It is known, however, that the carbohydrate side chains are not  important for protein S binding, as a truncated recombinant wild-type ␤-chain composed of three CCP modules was able to bind to protein S in a similar fashion as plasma-purified C4BP (20), which is also confirmed in the present study. As in the radioligand blotting, the L27T/F45Q mutation seemed to increase the binding of C4BP to protein S. For the remaining mutants, I16S/V18S, V31T/I33N, I16S/V18S/V31T/I33N, and L38S/V39S, the binding was essentially lost. K41E/K42E again displayed weaker binding, never reaching more than ϳ70% of binding of 125 I-labeled protein S, compared with the wild-type ␤-chain. Competition Assay-In the competition assay, increasing amounts of plasma-purified C4BP or recombinant ␤-chain (wild-type or mutant) were allowed to compete with immobilized C4BP for binding of fluid-phase 125 I-labeled protein S. After washing, bound radioactive protein S was measured using a ␥-counter. The wild-type ␤-chain bound protein S equally well compared with plasma-purified C4BP (Fig. 6). In this assay, mutant K41E/K42E bound in a similar fashion compared with the wild-type ␤-chain (Fig. 6C). L38S/V39S displayed ϳ10-fold less apparent affinity (Fig. 6C), whereas I16S/ V18S, V31T/I33N, and I16S/V18S/V31T/I33N had an ϳ100-fold lower apparent affinity (Fig. 6B).

DISCUSSION
In this study, we show that a key binding surface for protein S is centered on Ile 16 , Val 18 , Val 31 , and Ile 33 on CCP1 of the ␤-chain of C4BP. This cluster of solvent-exposed hydrophobic residues on the first CCP of the ␤-chain, together with two lysines, became apparent during analysis of a predicted threedimensional model structure for C4BP (26,27) based upon the NMR structure reported by Norman et al. (5). Since relatively large patches of solvent-exposed hydrophobic residues tend to destabilize the native state of a protein, possibly by shifting the folding equilibrium toward denaturation (32), it could have been argued that this region of C4BP was not appropriately modeled and was a computational artifact. However, we have also predicted the structure of this first CCP using other experimental templates (e.g. ␤ 2 -glycoprotein I, CD46, and vaccinia virus complement control protein) and found that this exposed hydrophobic cluster is essentially present in all models (data not shown). Because of this observation and the data showing that the protein S-C4BP interaction is not significantly altered by the presence of increasing concentrations of NaCl, it was most likely that the solvent-exposed cluster indeed plays a significant role in protein S binding. The mutations introduced were chosen based on the analysis of the homology-based three-dimensional structure of the ␤-chain of C4BP (Fig. 1). Substitution of solvent-exposed residues by more polar ones should be favorable to the protein stability and/or folding, and we do not expect that structural problems could be induced because of the mutations. For example, the Ile 16 , Val 18 , Val 31 , Ile 33 , and Leu 38 cluster has a surface area of ϳ300 Å 2 . With a hydrophobic solvation free energy of ϳ20 cal/mol/Å 2 , the energetic cost of exposing a patch of 300 Å 2 is high and ϳ6 kcal/mol.
The substitutions used were I16S/V18S, V31T/I33N, I16S/ V18S/V31T/I33N, L38S/V39S, L27T/F45Q, and K41E/K42E. Our results strongly suggest that the protein S-binding site is indeed composed of a hydrophobic patch containing Ile 16   The residues involved in protein S binding are shown in Fig. 7. These amino acids form a cluster located in the direct vicinity of the second CCP ( Figs. 1 and 6). van de Poel et al. (22,23) showed that CCP2 seems to have a weak contribution to the binding between protein S and C4BP. It is possible that the large sex hormone globulin binding-like domain of protein S not only binds to CCP1, but also interacts with CCP2 of the ␤-chain. Another explanation for the influence of CCP2 on the binding of protein S could be that the second CCP sterically facilitates the binding of protein S to CCP1. However, our data are in agreement with our previous results stating that CCP1 contains the key binding site for protein S (21), as the substitution of four residues dramatically alters the interaction, fully consistent with the structural observations. The ␤-chain is glycosylated, and two consensus sequences for N-linked glycosylation are present on the first CCP module (33). The difference between the expected molecular mass of the ␤-chain based on its amino acid composition (26.4 kDa) and the apparent molecular mass as judged by SDS-PAGE (45 kDa) implies that most or all of the glycosylation sites are occupied (33). This is further emphasized by the fact that, after digestion with endoglycosidase F, an enzyme that removes N-linked car- bohydrates, the apparent molecular mass on SDS-PAGE of the ␤-chain is 29 kDa (33). It is known that oligosaccharide moieties can contribute to protein-protein interactions (34). For instance, the N-linked glycan attached to the second CCP of CD46 is essential for virus binding (35). However, it has been shown that sugars are not important for the binding of protein S to C4BP (20). Therefore, assuming one or both Asn residues (at positions 47 and 54) to be glycosylated, the glycan should be located outside the key binding surface for protein S. To study the spatial arrangement of these sugars relative to the defined protein S-binding site, glycan core structures were modeled as shown in Figs. 1 and 7. The glycan molecule was taken from the x-ray structure of CD46 (9) and grafted onto C4BP Asn 47 and Asn 54 . Because we show that the key binding residues for protein S are Ile 16 , Val 18 , Val 31 , and Ile 33 and since the glycans are away from this region, the predicted structure and experimental data are again in full agreement.
Ferná ndez and Griffin (24) used synthetic peptides to probe the protein S-C4BP interaction and suggested that residues 31-45 are important for protein S binding. Our results in part confirm their observations since we show that Leu 38 and Val 39 have great influence on the protein S-C4BP interaction. However, we show that also Ile 16 and Val 18 are crucial for the binding of protein S to C4BP. Furthermore, changing the hydrophobic residues Leu 27 and Phe 45 to polar residues did not lessen the binding to protein S. Rather, the substitution seemed to slightly enhance the binding, suggesting that these residues directly or indirectly repulse protein S.
The role of exposed hydrophobic residues in protein-protein interaction is not unique for the CCPs of C4BP. Our data are consistent with the analysis of other macromolecular interactions, as in many systems, a solvent-exposed hydrophobic cluster has been observed (36 -38). For instance, it has been proposed upon analysis of the CD46 x-ray structure that the CD46-measles virus hemagglutinin interaction is dependent on a critical set of hydrophobic residues at the protein interfaces and that this reaction resembles the CD4-HIV gp120 interaction (9). Recent investigations have highlighted the importance of solvent-exposed hydrophobic patches (ranging from 200 to 1200 Å 2 ) (39,40). Many different types of binding surfaces are expected to be found in nature to allow proper folding of the molecules and to render high, medium, or low affinity as well as specificity. Thus, it is obvious that some protein interfaces are very rich in charged residues, whereas others display solventexposed hydrophobic clusters. The importance of hydrophobic contacts for protein interactions is appealing, as such interactions appear to glue two molecules together. However, these interactions may not be very specific and so should be complemented by hydrophilic interactions. This situation is expected in the protein S-C4BP interaction, as it seems that the key hydrophobic binding surface is supplemented by hydrophilic interactions. Long-range electrostatic interaction may not be the only force that can affect association. Important contributions to the binding free energy involve also desolvation (i.e. the removal of solvent from nonpolar and polar atoms). Indeed, when a large apolar surface is exposed to solvent, other longrange attractive forces are expected to contribute significantly to protein assemblies (41). There seem to be many reasons to maintain large or small hydrophobic clusters within a binding site area of transient or very stable molecular complexes, despite the overall energetic cost of such a structural feature.
Investigation of the protein S-C4BP interaction is of importance for numerous reasons. First, protein S deficiency is a risk factor for thrombosis, and a better understanding of the protein S-C4BP interaction could be valuable for better diagnosis and treatment of coagulation disorders. Second, CCP modules are present in numerous proteins involved in various important biological processes. Thus, analysis of a specific CCP domain could provide information that may be a general consensus for CCP modules. Third, a better understanding of protein-protein interaction guides the design of approaches aimed at the prediction of hot spots at the surface of a molecule. This is of importance because many research projects involve characterization of binding sites, which then helps the understanding of molecular mechanisms and thus the generation of new therapeutic compounds or diagnostic tools.
In conclusion, we have shown a large hydrophobic patch on the ␤-chain of C4BP to be crucial for binding of protein S to C4BP. In addition, Lys 41 and Lys 42 could contribute to the modest electrostatic component playing a role in this interaction. These data are the first reported that clearly pinpoint the specific amino acids on CCP1 responsible for the interaction between protein S and C4BP. Our data are also in agreement with what has been observed for other CCP-containing molecules as well as in several macromolecular assemblies. Furthermore, our investigation emphasizes the rational of using computer-based molecular modeling to predict the three-dimensional structure of a protein and its potential in the design of experiments aimed at a better understanding of the relationships between structure and function.