INTRODUCTION

Vitamin K-dependent protein S and C4b-binding protein (C4BP)¹ form a tight 1:1 molar complex in human plasma (1, 2). Protein S (molecular mass = 75 kDa; see Fig. 1A) participates in the anticoagulant pathway as a cofactor to activated protein C. C4BP (see Fig. 1B) is a regulator of the classical complement pathway and is a highly glycosylated multimeric protein of high molecular mass (570 kDa). The primary structures of the two proteins reveal that they consist of multiple modules. Protein S contains one Gla-module, a thrombin-sensitive disulfide loop, four EGF-like modules and a C-terminal domain that is homologous to the sex hormone binding globulin (SHBG). The SHBG domain contains three glycosylation sites, two of which are conserved throughout different species (3-8). C4BP consists of 7-8 polypeptide chains that are linked together by disulfide bonds between cysteines located in the C-terminal part of each chain. The major isoform has seven -chains, each containing eight short consensus repeat (SCR) modules, and a single -chain containing three SCR modules. One of the minor isoforms of C4BP has no -chain and does not interact with protein S (9-11).

The structures of protein S and C4BP are known at the electron microscopy level (12, 13), showing that C4BP looks like a spider with the SCR modules arranged as beads on a string; whereas protein S is more compact. High resolution structures are available for several homologous modules in other proteins, for example EGF-like modules from factor IX and X (14-16), a pair of EGF-like modules from fibrillin-1 (17), a Gla-EGF module pair from factor X (18), and three single and one pair of SCR modules from factor H (19-21). Recent additions to the field are crystal structures of coagulation factor IX (22) and of factor VII in complex with tissue factor (23).

The interaction between C4BP and protein S is so strong (K_D = 0.1-0.6 nM in 150 mM NaCl, 2 mM Ca²⁺, pH 7.5) (8, 24-27) that the entire pool of -chain-containing C4BP in human plasma is complexed to protein S. In normal individuals, the total protein S concentration exceeds that of the -chain-containing isoforms of C4BP by 30%, ensuring an uncomplexed protein S fraction with anticoagulant activity. The cause for thrombosis in some patients is a consequence of total protein S concentrations not exceeding the -chain-containing C4BP isoform levels and protein S complexed to C4BP has no anticoagulant activity. C4BP is an acute-phase protein, and its concentration can increase as much as 4-fold during the acute-phase inflammatory response. It has recently been demonstrated, however, that it is primarily the concentration of the C4BP isoform lacking the -chain that increases during the acute-phase inflammatory response, resulting in a maintained anticoagulation system in a normal individual and indicating a differential regulation of the - and -chains (28). It has been suggested that the biological role for the interaction between protein S and C4BP is to anchor the complex of C4BP and C4b to phospholipid membranes at the site of injury (where the coagulant and anticoagulant protein complexes are bound via their Gla-modules) to prevent inflammation. Thus, it is important to identify the domains that are mediating this protein-protein interaction.

Electron microscopy studies suggest that protein S binds to the -chain of C4BP (13). Binding of protein S to the -chain has also been shown using recombinant truncated -chain proteins containing SCR modules 1-3 (29) or SCR 1-2 (30) as well as a construct with the first SCR of the -chain fused to SCR 2-8 of the -chain (30). A study using synthetic peptides has lead to the proposal that residues 31-45 in the first SCR-module of the C4BP -chain are essential for the interaction (31). A structural prediction has been presented for the -chain of C4BP (31, 32), revealing as a potential binding site for protein S a solvent-exposed hydrophobic patch in the first SCR-module lined with a positively charged area on an otherwise negatively charged surface.

The portion of protein S that binds to C4BP has not been conclusively specified. Many attempts have been made using either synthetic peptide fragments or site-directed mutagenesis of the protein S molecule or chimeric proteins. The smallest part that has been demonstrated to bind C4BP with full native affinity is the SHBG-like domain, as inferred from binding studies using protein hybrids of protein S and coagulation factor IX (27). Two synthetic peptides have been reported to compete with protein S for binding to C4BP when used in very large excess over protein S, one comprising residues 413-433 (33) and one comprising residues 605-614 (34, 35). Truncated protein S variants lacking residues 607-635 or 577-635 have drastically reduced affinities for C4BP (36). A 10³-fold reduction in affinity has been reported for a similar variant lacking residues 583-635 (37), meaning either that some residues in the 583-635 sequence are directly involved in the binding to C4BP or that the C-terminal 52 residues are necessary for the correct folding or presentation of another part that contains the binding site. Other mutants that have been found to have reduced affinity for C4BP have an extra alanine inserted at position 611 or Cys-598 deleted (36). It is also known that bovine and human protein S (82% identical; see Refs. 3 and 4) bind to human C4BP with about the same affinity, whereas the homologous proteins SHBG (26% identical to the SHBG-like domain of human protein S; see Ref. 38) and growth arrest-specific protein 6 (Gas 6) (44% identical to human protein S; see Ref. 39) do not show any detectable binding to C4BP.²

We have chosen to address the problem of identifying the C4BP binding site in protein S using the phage display method. Sequences with high affinity for the C4BP -chain were selected from two random peptide libraries. One library contains hexapeptides and the other pentadecapeptides, and the peptides are displayed on the surface of the filamentous phage (M13) (40, 41). We have used two alternative approaches to select for -chain binders, one using a recombinant fragment comprising the two N-terminal SCR modules of the -chain as a target and the other using intact C4BP as a target plus a recombinant protein composed of only -chains to remove -chain binding sequences.

MATERIALS AND METHODS

C4BP was purified from human plasma as described (1). Two recombinant proteins called SCR-_1,2 and Rec. , respectively, were used in the screening procedure. SCR-_1,2 was expressed in a prokaryotic expression system and contained the two N-terminal SCR modules from the C4BP -chain, whereas Rec. , which was expressed in a eukaryotic expression system, was composed of C4BP molecules lacking the -chain. Both recombinant proteins were expressed and purified as described (11, 30). Avidin, streptavidin, and BSA were from Pierce (Rockford, Illinois).

Prior to biotinylation, buffer amines were removed from the C4BP and SCR-_1,2 by concentration and multiple washes with phosphate-buffered saline buffer in Centricon concentrators (Grace, Beverly, Massachusetts) with molecular mass cut-offs of 30 kDa, for C4BP, and 3 kDa, for SCR-_1,2. NHS-LC-biotin was first dissolved in H₂O and then added to the proteins. The C4BP concentration was 0.5 mg/ml, and a 10-fold molar excess of NHS-LC-biotin was added. The SCR-_1,2 concentration was 0.3 mg/ml, and a 5-fold molar excess of NHS-LC-biotin was added. The mixture was allowed to react for 30 min at room temperature, and the unreacted NHS-LC-biotin was removed by separation on a desalting column or by multiple washes in the concentrators. The amount of biotinylation was 0.5 biotin/protein for SCR-_1,2 and 2-4 biotin/protein for C4BP, as judged by using 2-(4-hydroxyazobenzene)benzoic acid reagent included in the NHS-LC-biotinylation kit from Pierce.

The 6-mer phage display library was amplified from the 2 × 10⁸ clone library previously reported by Scott and Smith (40). The 15-mer library was amplified from the 2 × 10⁸ clone library previously reported by Nishi et al. (41). The randomized peptide sequences were fused to the N terminus (between residues 5 and 6) of the phage coat protein pIII.

The linear peptides (BD4, BD6, and SL1-SL7) were synthesized on a MilliGen 9050 Plus synthesizer ("continuous flow peptide synthesis") using Fmoc chemistry with active esters (pentafluorophenyl esters). The first amino acid in the synthesis (the C-terminal amino acid) was coupled to the resin PEG-PS Support^TM from Millipore (polyethylene glycol-polystyrene). After the synthesis, the resin was rinsed and dried. The peptide was released from the resin by cleavage for 2 h under N₂ gas in the darkness using 92-95% TFA containing different scavengers depending on the amino acid composition of the peptide. The resin was removed by filtering and washed with concentrated TFA. After concentration, the peptide was precipitated and washed 4 times in cold diethyl ether. The ether was evaporated, and the peptide was dissolved in 0.1% TFA/H₂O (or in 50-75% acetic acid for the SL1, SL2, SL4, SL6, and SL7 peptides that were difficult to dissolve in 0.1% TFA). The peptide was purified on an HPLC (Waters 600E System Controller, Waters 486 Tunable Absorbance Detector) C8 column (Kromasil 5, 100A C8, 250 × 21.2 mm) using a linear gradient of 0.1% TFA/H₂O and 0.1% TFA, 80% acetonitrile/H₂O. The peptide was concentrated by speedvac and lyophilization.

The peptides BD4 and BD6 were reduced (in 0.1 M Tris, pH 8.3, with 0.1 M dithiothreitol and 6 M guanidine-HCl for 2 h at room temperature at a peptide concentration of 10 mg/ml) prior to HPLC purification (as described above). After purification, they were folded to form a disulfide bond between the two cysteines in each peptide (in 0.1 M Tris, pH 8.3, with 1 mM EDTA, 3 mM cysteine-HCl, and 0.3 mM cystin under N₂ gas for 16 h at room temperature at a peptide concentration of 0.1 mg/ml). The peptides were subject to a second HPLC purification (as above) after folding.

All chemicals were of the highest grade commercially available. Buffers and all other solutions were autoclaved or sterile-filtered prior to use. Sterilized labware was used throughout. The following abbreviations for autoclaved buffers are used in the text: TBS, 50 mM Tris, 0.15 M NaCl, 2 mM CaCl₂, pH set to 7.5 with HCl; TBS/Tween, TBS with the addition of 0.5% Tween; TBS/NaN₃, TBS with the addition of 0.02% NaN₃; HC, 10 mM Hepes, 0.15 M NaCl, 3.4 M EDTA, 0.005% Tween 20, pH 7.4; and phosphate-buffered saline, 0.1 M sodium phosphate buffer, pH 7.0 with 0.15 M NaCl. The sensorchips CM5 and amine coupling kit containing N-hydroxysuccinimide (NHS), N-ethyl-N-(3-diethylaminopropyl)carboxydiimide, and ethanolamine hydrochloride were from Pharmacia Biosensor AB (Uppsala, Sweden). The surfactant Tween 20 was from Riedel de Haen. NHS-LC-biotinylation kit was from Pierce. Chemicals for peptide synthesis were from Millipore.

The biopanning experiments were done in 35-mm polystyrene dishes (Falcon) at 4 °C. 10 µl of the original library was used in the first round. In each subsequent round, the input was 100 µl of an amplified eluate from the previous round. The amount of biotinylated C4BP was 10 µg in the first round, 1 µg in the second, 0.1 µg in the third, 0.01 µg in the fourth, and 0.001 µg in the fifth round. Each round of biopanning started with avidin or streptavidin coating as follows. 1 µg of avidin or streptavidin was added to each dish containing 1 ml of 0.1 M NaHCO₃ (pH unadjusted), and the dishes were rocked for 12 h. After discarding the avidin/streptavidin solution, the remaining protein adsorption sites on the plastic were blocked by BSA; 400 µl of a solution containing 1 mg/ml BSA and 10 µg/ml avidin or streptavidin was added and allowed to react for 1 h. Dishes were then washed 6 times with TBS/Tween. In the first three rounds ("P+LS" method), the biotinylated ligate was allowed to react for 2 h with the immobilized avidin/streptavidin (in 400 µl TBS/Tween) prior to 6-fold washing (TBS/Tween), addition of 0.1 µg biotin (in 400 µl TBS/Tween), and addition of the phages (~10¹¹ physical particles in 100 µl of TBS/NaN₃); whereas in rounds four and five ("PL+S" method), the biotinylated ligate was allowed to equilibrate for 24 h with the phages (~10¹¹ physical particles in 100 µl TBS/NaN₃) prior to dilution with 400 µl of TBS/Tween and binding to streptavidin/avidin on the dish. After 10 washes with TBS/Tween, the binding phages were eluted with 400 µl of 0.2 M glycine buffer at pH 2.2. The eluate was neutralized by mixing with 75 µl of 1 M Tris/HCl, pH 9.1. The phage eluate from each round of biopanning was amplified in E. coli and purified from the culture supernatant using two PEG precipitation steps to provide the input phages for the next round of biopanning. The number of transducing units were counted for input and output phages of each round to provide an estimate of the yield in each round of biopanning.

Selective Removal of Peptides Binding to the C4BP -Chain

C4BP interacts with many different proteins. In addition to the high affinity protein S-binding site on the -chain, there are lower affinity binding sites on the -chain for C4b, serum amyloid P component, heparin, and factor VIII. A fraction of the captured peptides might thus bind to sites on the -subunit. Therefore, in an extra (sixth) round of biopanning, a large excess of recombinant C4BP containing only -chains (Rec. ) was added to the C4BP phage mixture to capture -chain binding phages. The input phages were in 10-fold excess over intact C4BP, and the Rec.  was in 100-fold molar excess over the -chains in intact C4BP. The same amount of Rec.  was added in one of the washes. In contrast to intact C4BP, the recombinant protein was not biotinylated whereby, in principle, predominantly -chain binding phages should remain attached to the dish after extensive washing.

Affinity purification of the phage bound 6- and 15-mer libraries against C4BP was also performed using magnetic beads. Biotinylated C4BP was coupled to streptavidin-coated magnetic beads (Dynabeads M-280, 112.06, Dynal AS, Oslo, Norway). The beads were separated from buffers in the different steps using a magnetic particle concentrator (Dynal MPC-M, Dynal AS, Oslo, Norway). The beads were first washed three times with TBS/Tween. 90 µg of biotinylated C4BP was mixed with 2 mg beads in 400 µl of TBS/NaN₃ with 1 mg/ml BSA for 30 min at room temperature. They were washed once, 1 µl of 10 mM biotin was added, and the beads were then washed four times with TBS/Tween. 10 µl of an original phage display library or 100 µl of an amplified library was mixed with C4BP-coupled beads in 400 µl of TBS/Tween and allowed to equilibrate overnight on a rocker. Four rounds were carried out, the first two rounds with 500 µg beads and the last two with 30 µg beads. Rec.  was added in rounds three and four. The beads were washed 10-fold with TBS/Tween prior to elution with 400 µl of 0.2 M glycine buffer at pH 2.2. Neutralization with 75 µl of 1 M Tris/HCl, pH 9.1, amplification in E. coli, and purification using two PEG precipitation steps provided input phages for the next round.

The amino acids sequences of affinity purified phage-bound peptides were derived by sequencing the DNA corresponding to the N-terminal part of gene III coat protein. The nucleotide sequence of the primer was 5HO-CCCTCATAGTTAGCGTAACG-OH 3. The primer is 86 base pairs away from the insert. It is complementary to a sequence coding for a region in the phage coat protein pIII, which is 29 residues C-terminal to the insert. Sequencing was performed with the PRISM Sequence Terminator double-stranded DNA sequencing kit (Applied Biosystems) using an Applied Biosystems Model 373 DNA sequencer.

Microtiter plates were coated with C4BP, 50 µl/well, 10 µg/ml in 0.075 M sodium carbonate buffer, pH 9.6. The plates were incubated overnight at 4 °C and then washed with TBS, pH 7.5, containing 0.1% Tween 20. After quenching (TBS, pH 8.0, containing 0.05% Tween 20, 3% fish gelatin, and 0.02% NaN₃, 100 µl/well, for 30 min) and washing, increasing concentrations of the peptides (0.1-3000 µM) or plasma-purified human protein S in TBS containing 10 mM EDTA were added together with a trace amount of ¹²⁵I-labeled protein S in a final volume of 50 µl and left at 4 °C overnight. The wells were then washed, and the amount of bound protein S detected using a -counter.

The surface plasmon resonance studies were performed using a BIAcore^TM apparatus from Pharmacia Biosensor AB. Immobilization of C4BP to the dextran-coated gold surface of a sensorchip was performed at a flow rate of 5 µl/min, using HC as flow buffer. Equal volumes of 0.1 M NHS and 0.1 M N-ethyl-N-(3-dietylaminopropyl)carboxydiimide were first mixed, after which 30 µl of the mixture was flown over the sensorchip surface to activate the carboxymethylated dextran. C4BP was then injected over the sensorchip (40 µl of a 60 µg/ml solution in 10 mM NaHAc, at pH 4.75), after which unreacted NHS-ester groups were blocked by 15 µl of 1 M ethanolamine, pH 8.5. The system was regenerated by addition of 15 µl of 0.1 M HCl, which removes all non-covalently bound molecules. The immobilized amount of C4BP was 8000 response units (RU). Protein S association was monitored with 50 nM human protein S in HC buffer in a continuous flow of 1 µl/min during 45 min. The ability of each peptide to inhibit the protein S binding was studied by following the association to C4BP for mixtures of protein S and peptide. In BIAcore^TM, the total amount of bound material is measured, and since the peptides are 25-fold smaller than protein S, inhibition of protein S binding by peptide binding will drastically lower the observed response during the association phase. The percent protein S bound, X, was calculated as

(Eq. 1)

where S is the measured response and S_max is the response obtained in the absence of peptide.

Fluorescence spectra were recorded on a SPEX Fluorolog spectrometer using a 1 × 1-cm cuvette. The excitation bandwidth was 2 nm, and the emission bandwidth was 3 nm. Emission spectra between 300 and 425 nm (step 1 nm) were recorded with excitation at 270, 275, 280, 285, and 290 nm. For polarization measurements, two polarizors were placed in the excitation and emission light paths. The components with the polarizors set parallel and perpendicular to one another were recorded separately, with excitation at 270 nm and emission at 325-345 nm (step 2.5 nm). Each component was taken as the average of 10 scans.

Near UV circular dichroism (CD) spectra were recorded on a JASCO-720 spectropolarimeter using a thermostated cuvette with a 1-cm path length. Spectra were recorded between 300 and 250 nm, using a wavelength step of 1 nm, response time of 4 s, and scan rate of 10 nm/min. Four scans were recorded and averaged for each spectrum.

Homology search was performed with the homemade program HOMOFILE. This program tests all possible alignments of a peptide with a protein sequence. The output is a file that lists, for each residue in the protein (as starting position of the alignment), the identity, near identity, and high similarity scores. The identity is taken as strict identity, and the criteria for near identity and high similarity are listed in Table I. The output files were either imported one by one to Kaleidagraph^TM for plotting or averaged by another homemade program, AVEHOM, the output of which is a file that lists, for the starting position of each alignment, the average identity, near identity, and high similarity scores.

Table I. Criteria for homology scoring Residue Near identity High similarity A G C P C A P F Y H I L Y W G A H F I L V Y I L F H L M L I F H I M M L I P A C V V H P Y F F H W W F Y D N E E N Q E Q D D N Q K R R Q N D Q D E Q S T Q E N D E N R K R K K Q S T N T T S N S

RESULTS

Human C4BP was used to affinity purify phage-bound peptides from two different phage display libraries, one with displayed hexapeptides and one with pentadecapeptides. Both libraries contain inserts at the N terminus of the coat protein pIII of M13. These inserts are randomized at the genetic level, and each phage contains up to five copies of one particular peptide in 6-mer or 15-mer library. In parallel, phage-bound peptides were also affinity purified against the SCR-_1,2-fragment, which contains the protein S binding site of C4BP (Fig. 1). Binding sequences were selected from these peptide libraries using biopanning in polystyrene dishes (both targets) or magnetic beads (C4BP). DNA sequencing was performed after rounds four, five, and six of the biopanning procedure and after round four of the affinity purification using the beads. All the derived amino acid sequences are summarized in the supplementary material.

For each affinity selected peptide C4BP that had been sequenced, a homology search was performed using the program HOMOFILE as described under "Materials and Methods." The identity, near identity, and similarity scores were plotted as a function of sequence. Alignments for individual peptides were in turn compared with each other to identify regions in the human protein S sequence where high scores were obtained for several peptides. Four such regions were identified, and the corresponding linear peptides were synthesized using Fmoc chemistry (peptides SL1, SL2, SL3, and SL4). However, only one peak appeared (around residue 450) when the program AVEHOM was used to average the similarity scores over all phage-bound peptides identified from affinity purification (see Fig. 2).³ Two additional peptides, BD4 (residues 405-437) and BD6 (595-628), corresponding to the two disulfide loops in the SHBG domain were synthesized on the basis of previous reports of binding (33-35). A second series of synthetic peptides provided three peptides (SL5, SL6, SL7) that extend and overlap with one peptide (SL2) that was identified as an inhibitor in initial screening experiments (see below). The amino acid sequences of the synthesized peptides are listed in Table II.

Results of homology searches against human protein S of peptides obtained by affinity purification of bacteriophage display libraries against C4BP and SCR-_1,2.

Table II. Synthetic peptides Notation Sequence hPS numbering BD4 LDGCIRSWNLMKQGASGIKEIIQEKQNKHCLVT 405 -437 BD6 YNGCMEVNINGVQLDLDEAISKHNDIRAHSCPSV 595 -628 SL1 KPENGLLETKVYFAGFPRK 374 -392 SL2 EKGSYYPGSGIAQFHIDYNNVS 439 -460 SL3 SDQQSHLEFRVNNLEKSTPLK 527 -550 SL4 DKAMKAKVATYLGGLPDVPFSAT 567 -589 SL5 LVTVEKGSYYPGSGIAQ 435 -451 SL6 SGIAQFHIDYNNVSSAEGWHVN 447 -468 SL7 LVTVEKGSYYPGSGIAQFHIDYNNVSSAEGWHVN 435 -468

The synthetic peptides were tested for their ability to displace binding of a ¹²⁵I-labeled protein S tracer to immobilized C4BP (Fig. 3A). SL2, SL6, and SL7 were found to completely inhibit binding of the protein S tracer to C4BP, whereas none of the other peptides had any effect on the protein S-C4BP interaction. Half-maximum inhibition was seen at 100-200 µM for the three inhibiting peptides.

Peptide inhibition of the protein S-C4BP interaction showing the amount of protein S bound (relative to the amount bound in the absence of peptide) versus concentration of the nine peptides listed in Table II, SL1 (), SL2 (), SL3 (), SL4 (+), SL5 (), SL6 (), SL7 (), BD4 (), and BD6 ().

The ability of the synthetic peptides to inhibit the binding of human protein S to C4BP was also studied using surface plasmon resonance on a BIAcore^TM system. For six of the peptides (SL1, SL3, SL4, SL5, BD4, and BD6), we observed the same response as with protein S alone, even when the peptides were in 6000-fold molar excess over protein S (300 µM peptide, 50 nM protein S). However, three peptides, SL2, SL6, and SL7, prevent the binding of protein S to C4BP with half-maximum inhibition at 30-120 µM peptide concentration (Fig. 3B).

Residues 447-460 are in common in all three peptides with inhibitory action.

The interaction of the SL6 and SL7 peptides with the C4BP fragment SCR-_1,2 was studied by fluorescence spectroscopy. SCR-_1,2 contains two tryptophan residues, and one tryptophan is present in each of the SL6 and SL7 sequences. The emission spectra of SCR-_1,2 alone, SL6 alone and a 1:1 mixture are shown in Fig. 4A. Two titrations of SCR-_1,2 with SL6 and SL7, respectively, were monitored using polarizers and recording the parallel (I) and perpendicular (I) emission components separately. The polarization, p, was calculated as

(Eq. 2)

The polarization of a 15 µM SCR-_1,2 solution as a function of SL7-concentration is shown in Fig. 4B. There is an initial increase in the polarization up to ~1 molar equivalent of added peptide, and then p gradually decreases due to the appearance of free peptide. The data was fitted by assuming a 1:1 stoichiometry and that the observed polarization is a weighted average of the polarizations of free peptide, free SCR-_1,2, and the SCR-_1,2·peptide complex, where the weighting takes into account the different numbers of chromophores in the three species. The variable parameters in the fit were P_S, P_L, P_C, C_S, and K_D, where P_S, P_L, and P_C are the polarizations of free SCR-_1,2, free peptide, and the SCR-_1,2·peptide complex, respectively; C_S is the total SCR-_1,2 concentration; and K_D is the dissociation constant of the SCR-_1,2·peptide complex. C_L is the total peptide concentration. The fitted equation was derived as,

(Eq. 3)

where

(Eq. 4)

and yielded good fits to the experimental points as shown in Fig. 4B. Computer analysis was performed separately for each of the two titrations (SCR-_1,2 with SL6 and SL7, respectively). The best fits in both cases were obtained with values of K_D between 0.01 and 0.1 µM. The error square sum of the fits were, however, only 2-3-fold higher for a K_D of 1 µM. The dissociation constant of the SCR-_1,2·peptide complex obtained from the polarization titrations can therefore not be specified more precisely than K_D  1 µM. The titration with SL6 was performed using an independent batch of SCR-_1,2 and initially had a lower value of the polarization. However, the increase on adding peptide was of similar magnitude, and the overall shape of the curve was very similar to that shown in Fig. 4B. We believe the small difference in starting polarization is due to variations in the amount of correctly folded SCR-_1,2 in different batches.

The interaction between SCR-_1,2 and different peptides was studied using CD spectroscopy. As shown in Fig. 5A, the near UV spectrum of an equimolar mixture of SCR-_1,2 and SL6 is distinctly different from the sum of spectra recorded for SCR-_1,2 and SL6 individually. The difference spectrum (mixture minus spectral sum) has a positive peak around 270-280 nm. Similar results were obtained also with the peptides SL2 and SL7 but not with the peptides BD4 and BD6. The peak intensity, _d, of the difference spectrum as a function of added peptide is shown in Fig. 5B for SL6, SL7, BD4, and BD6. Curves calculated using the following equation,

(Eq. 5)

where

(Eq. 6)

agreed well with data for SL6, SL7, and SL2 (data not shown) when K_D, the dissociation constant of the SCR-_1,2·peptide complex, was around 1 µM (Fig. 5B). In Equation 5, _dmax is the ellipticity difference at saturation, C_L is the total peptide concentration, and C_S is the total SCR-_1,2 concentration. V is the total volume. V_o is the total volume before the start of the titration.

DISCUSSION

The phage display technique is fruitful for mapping the epitopes recognized by monoclonal antibodies (40), and in the study of protein-protein, protein-DNA (42, 43), and intra-protein (44) interactions. The technique has also proven useful in assessing protease specificity (45), criteria for protein folding (46), as well as designing D-peptide drugs (47, 48). The amino acid sequence identified in a phage display experiment toward a specific target is not always unambiguously assigned to a given stretch of the protein that binds to the target in vivo. One reason for this is that the native binding sequence often does not have the highest possible affinity but rather the most suitable affinity for its biological context. Another reason is that the library, although very diverse, is not 100% complete, and the natural binding sequence might not be present. A third reason is that the library is not completely random on the amino acid level due to codon usage considerations and constraints. A fourth reason is that the native binding epitope may not be a contiguous primary sequence, rather, the binding epitope may be influenced by the three-dimensional structure. With these obstacles in mind, we have made an attempt to identify the regions in protein S that have the highest homologies with the phage displayed sequences that were selected by affinity purification in the present work.

Visual inspection of individual selected display peptides both to each other or to protein S failed to provide a convincing alignment, in contrast to what is commonly observed in other phage display experiments. A more objective alignment of each individual peptide sequence to protein S, using homology searches (HOMOFILE) followed by averaging the scores obtained for all peptides (AVEHOM), revealed a peak around position 450 (Fig. 2), which is significantly higher than the noise.

Surprisingly, few of the phage display peptides selected by affinity purification against C4BP or SCR-_1,2 showed any striking similarity to the first disulfide loop in the SHBG-like domain of protein S or to the second disulfide loop, both of which have been previously suggested as likely locations of the C4BP binding site (33-35). Therefore, to specifically address this discrepancy, two peptides corresponding to these disulfide loops were synthesized (BD4, residues 405-437, and BD6, residues 595-628) and tested for inhibition of the protein S-C4BP interaction. Consistent with the phage display results, we found that these peptides have no inhibitory effect even when employed in 6000-fold excess over protein S. This is further corroborated by the failure of BD4 and BD6 to show any binding to SCR-_1,2 in binding experiments monitored by CD spectroscopy.

Certain experimental details may account for the conflicting BD4 and BD6 peptide binding results obtained by Walker (34), Fernández and Griffin (33), and by those reported herein. The level of confirmation of peptide purity was lower in the previous studies; whereas all peptides used in our studies were greater than 99% pure as measured by HPLC, and their identities were confirmed by mass spectrometry. A potentially significant difference is that, in our studies, the peptides were both prepared and kept under conditions to promote intramolecular disulfide formation (as present in protein S); whereas in the previous reports, the peptides were likely in the reduced form containing two chemically reactive thiol groups. This difference can have a dramatic effect on binding measurements because we have noticed that peptides with reactive thiols have a tendency to cause C4BP unfolding and aggregation. This will cause a reduction in binding of protein S that is not due to competition for the binding site, but rather a decrease in the amount of functional C4BP. Consistent with this explanation is the inconsistency in the paper by Fernández and Griffin (33) where the most potent inhibitor of the PS-C4BP interaction (peptide 408-434 containing two free thiols) binds to C4BP with the same affinity as protein S and yet, in competition experiments, had to be employed in 200-500-fold excess of protein S to cause half-maximal inhibition.

Inhibition was, however, observed for peptides synthesized on the basis of the phage display results. Residues 447-460 are present in all three peptides that are capable of inhibiting the protein S-C4BP interaction (SL2 = 439-460, SL6 = 447-468, and SL7 = 435-468). The CD and fluorescence polarization titrations suggest that these peptides interact directly with C4BP with a dissociation constant K_D  1 µM. The dissociation constant of the peptide-C4BP complex is thus much higher than the dissociation constant of the complex of Ca²⁺-free protein S and C4BP (K_D = 6.5 nM; see Ref. 27), a reasonable result in view of the inhibition experiments in microtiter wells (performed in the absence of calcium, in EDTA-containing buffer) requiring a significant excess of peptide over protein S to produce half-maximum inhibition of the binding of the protein S to C4BP. The difference in K_D observed for the peptide and protein might in part be due to post-translational modifications of the protein.

Indeed, one of the three glycosylation sites in protein S (Asn-458) lies within the C4BP-binding region identified here, and another one (Asn-468) is close by (Fig. 6). The glycosylation sites at residues 458 and 468 are found in several species (human, rhesus monkey, bovine, mouse, porcine, rabbit, and rat), whereas the third site at Asn-489 is present only in human and monkey protein S. Gas 6 and the SHBG are two proteins that are homologous to protein S but do not bind to C4BP.² An alignment of the amino acid sequences of these proteins with protein S show distinct differences in the 447-460 region (Fig. 6). Gas 6 contains a 4-amino acid insertion and SHBG has a one residue deletion and a different pattern of hydrophobic and hydrophilic residues. In addition, both Gas 6 and SHBG lack the consensus sequence (N-X-S/T) for carbohydrate attachment at the positions corresponding to residues 458, 468, and 489 in protein S. Our future studies will investigate the role that glycosylation plays in the C4BP-protein S interaction. The N-linked carbohydrate at position 458 is not essential for binding to C4BP. A variant of human protein S, the so called protein S-Heerlen, which lacks this carbohydrate moiety due to a SerPro mutation at position 460 (49), still interacts with C4BP (50).