The High Affinity Calcium-binding Sites in the Epidermal Growth Factor Module Region of Vitamin K-dependent Protein S*

Vitamin K-dependent protein S, a cofactor of the anticoagulant enzyme-activated protein C, has four epidermal growth factor (EGF)-like modules, all of which have one partially hydroxylated Asp (EGF 1; β-hydroxyaspartic acid) or Asn (EGF 2, 3, and 4; β-hydroxyasparagine) residue. The three C-terminal modules have a typical Ca2+ binding sequence motif that is usually present in EGF modules with hydroxylated Asp/Asn residues. Using the chromophoric Ca2+ chelators Quin 2 and 5,5′-Br2BAPTA, we have now determined the Ca2+affinity of recombinant fragments containing EGF modules 1–3, 1–4, 2–3, and 2–4. EGF modules 1–4 and 2–4 each contains two very high affinity Ca2+-binding sites, i.e. with dissociation constants ranging from 10−10 to 10−8 m in the absence of salt and from 10−8 to 10−6 m in the presence of 0.15 m NaCl. In contrast, in EGF 1–3 and EGF 2–3, the Ca2+ affinity is 2–4 orders of magnitude lower. EGF 4 thus appears to have the highest Ca2+ affinity, and furthermore it seems to influence the Ca2+ affinity of its immediate N-terminal neighbor EGF 3 by a factor of approximately 230. In addition, EGF 4 seems to influence the Ca2+ affinity of EGF 2 by a factor of approximately 25. The Ca2+ affinity of the binding sites in EGF modules 3 and 4 in fragments EGF 1–4 and EGF 2–4 is 103–105-fold higher than in the corresponding isolated modules, implying important contributions to the Ca2+ affinity of each module from interactions with neighboring modules. This difference is much higher than the approximately 10-fold difference previously found in similar comparisons of EGF modules from fibrillin. However, the modules studied in protein S and fibrillin appear to have the similar Ca2+ligands. The structural basis for the difference in Ca2+affinity is not yet understood.

Protein S is a vitamin K-dependent plasma protein that functions as a cofactor to activated protein C, a regulator of blood coagulation, by enhancing the activated protein C-mediated rates of degradation of factors Va and VIIIa (1)(2)(3). Protein S consists of an N-terminal ␥-carboxyglutamic acid (Gla) 1 -con-taining module that is followed by a module with a thrombinsensitive peptide bond and four epidermal growth factor (EGF)like modules, whereas the C-terminal half of the molecule is occupied by a module that is homologous to steroid hormonebinding proteins (4 -7). Calcium binding to vitamin K-dependent clotting factors, including protein S, is complex, Ͼ10 calcium ions being bound to 2 or 3 types of sites that profoundly influence the function of these proteins (8). The Gla module binds Ca 2ϩ and endows protein S with membrane affinity, whereas the EGF module-containing part of protein S appears to be involved in the interaction with activated protein C (9,10).
The EGF modules in protein S are members of a large family of similar modules that are found in extracellular and membrane mosaic proteins in species ranging from Caenorhabditis elegans to Drosophila melanogaster and humans (7,11,12). Several of these proteins are involved in cell differentiation, blood coagulation and fibrinolysis, and in the complement and fibrinolytic systems. The EGF modules are approximately 45 amino acids long and contain 6 cysteine residues that are paired in a characteristic manner: 1 to 3, 2 to 4, and 5 to 6, with a double-stranded ␤-sheet as the main structural feature. EGF modules provide a structural scaffold that supports proteinprotein interactions, serves as spacer units, and orients adjacent modules relative to each other in a way that sustains biological activity.
EGF modules 2-4 of protein S belong to a large subgroup of such modules that has a Ca 2ϩ binding sequence motif DI/VDE (or variants thereof) before the first Cys residue (13,14), and the motif XD*/N*XXXXY/FX between the third and fourth Cys residue that is required for hydroxylation of Asp/Asn (denoted by an asterisk in the sequence) to erythro-␤-hydroxyaspartic acid or erythro-␤-hydroxyasparagine (12,15,16). The Asp*/ Asn* and Tyr/Phe residues are adjacent in the major doublestranded ␤-sheet. The two sequence motifs appear to be coupled and phylogenetically conserved in Ca 2ϩ binding EGF modules. There are however exceptions, such as the first EGF module of protein S, which contains the hydroxylation motif but lacks the Ca 2ϩ consensus sequence before the first Cys residue.
Calcium binding to an EGF module was first observed in protein C and subsequently in factors IX and X (17)(18)(19)(20). The Ca 2ϩ affinity of isolated EGF-like modules is low and roughly equal (K d ϭ 0.5-8 mM at 0.15 M NaCl) (19,20,22), 2 although the reported Ca 2ϩ affinities for EGF sites in intact proteins vary from moderate (K d ϭ 0.1 mM) to very strong (K d Յ 10 nM) (23)(24)(25)(26)(27). This is in accord with a large body of experimental data for different types of Ca 2ϩ -binding proteins. These data show that the observed Ca 2ϩ affinity of a specific site is not governed solely by local interactions between the calcium ion and its coordinating oxygens provided by the backbone and side chains of the protein as well as water molecules. Major contributions are provided by other parts of the protein, and long range interactions can have profound effects on affinity (28,29). In coagulation factors IX and X, the adjacent Gla-containing module increases the affinity of the EGF site approximately 10-fold to a K d Ϸ 1 ϫ 10 Ϫ4 M, making the site essentially saturated at physiological Ca 2ϩ concentrations (26). We have found that the isolated third and fourth EGF modules from protein S bind Ca 2ϩ with a K d of 5.2 mM and 0.6 mM, respectively. 2 However, in intact protein S there appears to be four very high affinity Ca 2ϩ -binding sites (K d Ϸ 10 Ϫ9 -10 Ϫ7 M), at least one of which appears to be located in the EGF module region (24).
Fibrillin, a structural component of connective tissue microfibrils, contains 47 EGF modules, 43 of which have the Ca 2ϩ binding motif (22,23,31). In pairs of EGF modules from fibrillin, others have found the Ca 2ϩ affinity of the C-terminal module to be K d Ϸ 0.35 mM (22,27), which is approximately 25-fold higher than the Ca 2ϩ affinity found in the isolated modules. Yet the N-terminal module appeared not to contribute additional ligands to the calcium ion in the C-terminal module (27). Downing et al. (33) proposed the increased affinity to be attributable to shielding of the calcium ion from solvent and the provision of a more defined binding site by the interdomain interface.
To shed light on the effect of an adjacent EGF module on the Ca 2ϩ affinity of its neighbor, we have expressed fragments containing two, three, or four Ca 2ϩ binding EGF modules from human protein S in Spodoptera cells using baculovirus. The expression and characterization of the fragments are described in an earlier report. 3 We have now determined the Ca 2ϩ affinity of the recombinant modules using a Ca 2ϩ complexation method and 1 H NMR spectroscopy. EGF modules 3 and 4 each contains one very high affinity Ca 2ϩ -binding site (ՅK d of 10 Ϫ10 -10 Ϫ8 M), whereas there is no site of comparable affinity in EGF 1 or 2. Moreover, the Ca 2ϩ -binding site in EGF 3 is much stronger in the presence of EGF 4. Module-module interactions thus have a major effect on the Ca 2ϩ affinity of these module oligomers and, presumably, even more so in the intact proteins. However, the nature of these interactions has yet to be eluciated in molecular detail.

EXPERIMENTAL PROCEDURES
Materials-The fluorescent Ca 2ϩ chelators Quin 2 and 5,5Ј-Br 2 BAPTA were from Fluka, Buchs, Switzerland and Molecular Probes Europe BV, Leiden, Holland, respectively. Chelex 100 resin 200 -400 mesh (sodium form) was from Bio-Rad. Chemicals for amino acid analysis and protein sequence analysis were obtained from Beckman and Applied Biosystems, respectively. Other reagents were of reagent grade or better.
Proteins-The human protein S fragments EGF 1-3, EGF 1-4, EGF 2-3, and EGF 2-4 were expressed by baculovirus in SF 9 cells as described previously. 3 Immunochemical and qualitative Ca 2ϩ binding experiments indicated that the recombinant proteins had attained their native conformation.
Amino Acid Hydrolysis-Amino acid compositions were determined after 24 h of hydrolysis in 6 M HCl in vacuo using orthophthalaldehyde as described previously (35). Amino acid analysis after hydrolysis was also used to determine the protein concentration in the samples used in Ca 2ϩ titrations.
Calcium-binding Measurements Using Quin 2 and 5,5Ј-Br 2 BAPTA-Calcium-free human protein S EGF fragments were titrated in the presence of two different chromophoric chelators, Quin 2 and 5,5Ј-Br 2 BAPTA, to cover the range of dissociation constants between 7 ϫ 10 Ϫ10 and 10 Ϫ5 M (36, 37). The tetrapotassium salt of 5,5Ј-Br 2 BAPTA was used to determine dissociation constants in the range of 10 Ϫ8 -10 Ϫ5 M (38), and Quin 2 was used for dissociation constants between 10 Ϫ10 and 10 Ϫ7 M (24). All titrations were carried out at room temperature and at pH 7.5 using a 1-ml quartz cell with a 1-cm path length (acid-washed before use). The chelators were dissolved in 2 mM Tris-HCl buffer with or without 0.15 M NaCl at pH 7.5 at concentrations of 22-27 M as determined in the Ca 2ϩ -saturated form by measuring the absorbance at 239 nm using ⑀ 239 ϭ 1.6 ϫ 10 4 liters⅐mol Ϫ1 ⅐cm Ϫ1 (5,5Ј-Br 2 BAPTA) (38) or ⑀ 239 ϭ 4.2 ϫ 10 4 liters⅐mol Ϫ1 ⅐cm Ϫ1 (Quin 2) (36). The 2 mM Tris-HCl buffer, pH 7.5, was made with 18 ⍀ water (electrical resistance) and stored in a plastic bottle in the presence of a dialysis tube containing Chelex 100. The Ca 2ϩ content in these buffers did not exceed 0.5 M. The EGF fragments EGF 1-3 (46 -56 M), EGF 1-4 (20 -29 M), and EGF 2-4 (14 -25 M) were dissolved in 2 mM Tris-HCl buffer, pH 7.5, either with or without 0.15 M NaCl, and the pH was checked and corrected when necessary with 0.1 M HCl or NaOH. Before using the pH meter, it was washed with 0.1 M EDTA, chelator, and finally 18 ⍀ water to avoid Ca 2ϩ contamination. Before titration, a small aliquot of each EGF fragment solution was removed for determination of the protein concentration by amino acid analysis. To obtain the initial Ca 2ϩ content in EGF 1-4 and EGF 2-4 samples, aliquots were removed for Ca 2ϩ determination by atomic absorption spectroscopy. Titrations were performed by the sequential addition of 2 l aliquots from a 3.1 mM CaCl 2 solution (determined by atomic absorption spectroscopy) followed by absorbance measurements at 263 nm on a Cary 4E spectrophotometer.
Calculations of the Calcium Dissociation Constants from the Quin 2 and 5,5Ј-Br 2 BAPTA Measurements-The macroscopic binding constants were determined by an iterative least squares fit directly to the measured data (i.e. to the absorbance at 263 nm as a function of total Ca 2ϩ concentration), as described by Linse et al. (39). The reported macroscopic dissociation constants are the inverse of the binding constants (K d ϭ 1/K). Fits using two (K 1 and K 2 ) or three (K 1 , K 2 , and K 3 ) macroscopic binding constants were attempted for fragments EGF 1-3 and EGF 2-4 and three (K 1 , K 2 , and K 3 ) or four (K 1 , K 2 , K 3 , and K 4 ) for fragment EGF 1-4. Fixed parameters in the fitting procedure were the Ca 2ϩ dissociation constant of the chelator (KDQ), the chelator concentration at each titration point i (CQ i ), the total Ca 2ϩ concentration at point i including initial and added Ca 2ϩ (CATOT i ) and protein concentration at point i (CP i ). The following dissociation constants were used:  (34). CQ i , CP i and CATOT i were corrected for the dilution due to the Ca 2ϩ addition. Variable parameters in the fits are the macroscopic binding constants (K 1 , K 2 . . . K N ) and the absorbances in the Ca 2ϩ -free, AMAX, and Ca 2ϩ -saturated solution, AMIN, respectively. For each set of values of the variables, the Newton-Raphson method was used to solve the free Ca 2ϩ concentration Y at each titration point i using the following equation.
This equation states that the free Ca 2ϩ equals the total Ca 2ϩ minus the chelator-bound Ca 2ϩ and the protein-bound Ca 2ϩ . The absorbance could then be calculated as, where CQ o is the initial chelator concentration. Thus, the changes in absorbance were due to Ca 2ϩ binding to the chelator. Minimization of the Error Square Sum-The error square sum, ESS, was obtained by summing over all titration points (24), The variable parameters were iterated (in a separate procedure) to obtain an optimal fit of calculated to experimental data as deemed by the minimum in ESS.
Calcium-binding Measurement by 1 H NMR@-1 H nuclear magnetic resonance spectroscopy was used to determine the Ca 2ϩ dissociation 3 Y. Stenberg, B. Dahlbä ck, and J. Stenflo, submitted for publication. constant of fragment EGF 2-3. The Ca 2ϩ titration was carried out on a Varian Unity Plus 600 spectrometer at 599.89 MHz. The standard one-pulse experiment was performed with 1.5 s presaturation pulse to reduce the water resonance. Each spectrum is made up of 4096 complex data points taken from 512 accumulated scans using a spectral width of 6410 Hz. The Ca 2ϩ dissociation constant was determined by measuring changes in peak amplitudes induced by the addition of Ca 2ϩ . Only resonances in the methyl region (spectra not shown) were observed.
The titration was carried out at ϩ27°C and pH 7.5. Lyophilized EGF 2-3 fragment was dissolved in 99.9% D 2 O containing 10% trifluoroethanol to a protein concentration of 93.5 M, and the pH was adjusted to 7.5 Ϯ 0.05 with NaOD or DCl. To determine the fragment concentration, an aliquot was removed before titration for determination of the protein concentration by amino acid analysis. The initial Ca 2ϩ concentration was determined from the changes in the absorption at 263 nM in a Ca 2ϩ titration in the presence of 5,5Ј-Br 2 BAPTA. Stock solutions of CaCl 2 were made in 99.9% D 2 O, and the pH was adjusted to 7.5 with NaOD or DCl. The titration was performed by 14 sequential additions of 4 -5 l aliquots of CaCl 2 from 1 to 200 mM stock solutions (as determined by atomic absorption spectroscopy). After each addition, the pH was adjusted to its original value by adding 0.5-5.0 l of NaOD or DCl. The Ca 2ϩ sites in EGF 2-3 were saturated to approximately 90% when 5 fragment equivalents of Ca 2ϩ had been added.
Calculation of the Calcium Dissociation Constant from the 1 H NMR Measurement-A one-site binding equation was used when calculating the Ca 2ϩ binding to EGF 2-3.
where Ia and Ib are the signal amplitudes in the Ca 2ϩ -bound and Ca 2ϩ -free form, respectively. V o is the initial volume, V i is the volume at titration point i, and p i is the fraction of the sample in the Ca 2ϩ bound form. The Ca 2ϩ dissociation constant was calculated by iterative fitting of the calculated intensities to the experimental ones as a function of total Ca 2ϩ concentration. Equal weight was given to all titration points, and V o /V i corrects for the dilution due to the addition of Ca 2ϩ and pH. Dissociation constant and intensities at zero and saturating concentrations of Ca 2ϩ were allowed to vary during the iterative fit (see Fig. 4). The error square sum, ESS, was obtained by summing over all points in the titration (24), ESS ϭ ͑I calc Ϫ I obs ͒ 2 (Eq. 6)

RESULTS
Calcium-binding Measurements-To determine the affinity of the Ca 2ϩ -binding sites in the EGF modules of protein S, methods such as equilibrium dialysis were deemed unsuitable due to difficulties encountered in preventing the high affinity sites from picking up Ca 2ϩ from the environment (24). Instead, Ca 2ϩ was allowed to partition between the protein and a Ca 2ϩ binding chromophore in spectrophotometrically monitored titrations. The method, which requires minimal manipulation of the sample and ensures accurate mesurements, was performed as outlined by Linse et al. (36,37). By using the two Ca 2ϩ chelating chromophores, Quin 2 and 5,5Ј-Br 2 BAPTA, a range of dissociation constants from Ϸ10 Ϫ10 to 10 Ϫ5 M could be covered, Quin 2 covering those of 10 Ϫ10 -10 Ϫ7 M under low ionic strength (32) and 5,5Ј-Br 2 BAPTA covering those of 10 Ϫ8 -10 Ϫ5 M (38). The dissociation constants were obtained directly from the primary data, i.e. the absorbance measurements, as a function of the total Ca 2ϩ concentrations by performing iterative least squares fit to the data. The titrations of EGF 1-3 and EGF 1-4 (Quin 2 at low salt concentration) were performed twice. However, due to lack of material, the titrations of the remaining constructs were only performed once.
In a previous report, qualitative 45 Ca 2ϩ blotting experiments suggested EGF 4 to be crucial for high affinity Ca 2ϩ binding. 3 5,5Ј-Br 2 BAPTA was used in titrations of EGF 1-4 and 2-4 at low ionic strength and Quin 2 at both low ionic strength and 0.15 M NaCl. The results of the 5,5Ј-Br 2 BAPTA titrations performed at low ionic strength are shown in Fig. 1 together with the optimal curve obtained from least squares fit to the data points. Two macroscopic dissociation constants were obtained for EGF 2-4, and three constants were obtained for EGF 1-4. The affinity of one or two (EGF 1-4, EGF 2-4) of these sites appeared to be very high and outside the range of 5,5Ј-Br 2 BAPTA measurements. The fragments may have additional Ca 2ϩ site(s) with dissociation constant(s) above 10 Ϫ5 M.
In titrations of EGF 1-4 and EGF 2-4 with Quin 2, it was observed that both fragments picked up traces of Ca 2ϩ . Therefore, a small aliquot was removed from the cuvette after the addition of chelator and the pH adjustment but before the Ca 2ϩ titration. The Ca 2ϩ concentration in this sample was measured by atomic absorption spectroscopy. In the presence of Quin 2, the initial Ca 2ϩ content was 0.87 mol of Ca 2ϩ /mol of EGF 1-4 and 0.63 mol of Ca 2ϩ /mol of EGF 2-4 despite precautions taken to avoid Ca 2ϩ contamination. Similar observations have been made with intact protein S (24). The initial Ca 2ϩ content was lower when 5,5Ј-Br 2 BAPTA was present instead of Quin 2: 0.17 mol of Ca 2ϩ /mol of EGF 1-4 and 0.27 mol of Ca 2ϩ /mol of EGF 2-4. The initial Ca 2ϩ content was taken into account in the calculations of the binding constants. The results of the titrations of EGF 1-4 and of EGF 2-4 in the presence of Quin 2 and 0.15 M NaCl are shown in Fig. 2 together with the curves fitted to the data. Two Ca 2ϩ dissociation constants were obtained for EGF 2-4 and three for EGF 1-4 at low salt, whereas in the presence of 0.15 M NaCl, three sites were detected in EGF 1-4 but only one in EGF 2-4. The Ca 2ϩ contamination in EGF 1-4 and EGF 2-4 was not noted in the 5,5Ј-Br 2 BAPTA titrations due to the high Ca 2ϩ affinity of the fragments as compared with 5,5Ј-Br 2 BAPTA (K d ϭ 1 ϫ 10 Ϫ7 M). Accordingly, most of the Ca 2ϩ in EGF 1-4 and EGF 2-4 was still bound to the fragments even after the addition of chelator, and therefore no initial decrease in the absorbance maximum was observed. Quin 2 has a higher Ca 2ϩ affinity than 5,5Ј-Br 2 BAPTA and will therefore bind more of the initially present Ca 2ϩ , resulting in a larger decrease in the absorbance (not shown). The affinity for the second calcium ion in EGF 2-4 in the presence of 0.15 M NaCl is too low to be measured accurately in Quin 2 titrations (Table I).
In Ca 2ϩ blotting experiments, fragment EGF 1-3 yielded only a weak band, whereas EGF 2-3 was not seen at all, and EGF 1-4 and EGF 2-4 gave intensely stained bands. 3 The site(s) in EGF 1-3 and EGF 2-3 also proved to have too low affinity for Ca 2ϩ to allow titration with Quin 2. In titrations with 5,5Ј-Br 2 BAPTA, a fragment concentration almost double that of the chelator was required to obtain accurate measurements. Titrations were made both with and without 0.15 M NaCl in the buffer. The initial Ca 2ϩ concentration in EGF 1-3 was approximately 0.14 mol of Ca 2ϩ /mol of fragment. One of the titrations performed at low ionic strength is shown in Fig.  3 together with the optimal curve obtained by least squares fit to the data points. The optimal fits to two titrations with low salt were equally good, as judged by the error square sums, and yielded two dissociation constants (Table I). Due to a larger decrease in Ca 2ϩ affinity of the fragment as compared with the chelator, no dissociation constants from titrations performed in the presence of 0.15 M NaCl were obtained, as the Ca 2ϩ affinity was below the detection limit of the method.
Titrations of EGF 2-3 with 5,5Ј-Br 2 BAPTA demonstrated the Ca 2ϩ affinity of the fragment to be below the detection limit of the method. The initial Ca 2ϩ concentration in the EGF 2-3 fragment was approximately 0.15 mol of Ca 2ϩ /mol of fragment. Calcium titrations were instead monitored by 1 H NMR spectroscopy. Calcium binding induced amplitude changes in the methyl resonance region. The corresponding peaks were integrated and related to the Ca 2ϩ concentration. The data was calculated using a one-site binding equation (Fig. 4). The dissociation constant was calculated and is shown in Table I. DISCUSSION Protein S is the only vitamin K-dependent coagulation factor that is not a serine protease. Moreover, the N-terminal part of protein S has a modular structure that differs from that of factors VII, IX, and X and protein C. Whereas these proteins all have two EGF modules immediately C-terminal of the Gla module, protein S has a module with a thrombin-sensitive peptide bond between the Gla module and the four EGF modules. In factors VII, IX, and X and protein C, only the Nterminal EGF module has the Ca 2ϩ binding and Asp/Asn-␤hydroxylation sequence motifs. In protein S, all four EGF modules have the hydroxylation motif with partial hydroxylation of Asp in the first module and of Asn in the three following   FIG. 3. EGF 1-3 titration curve. Absorbance at 263 nm versus total Ca 2ϩ concentration (CATOT) from titration in the presence of 5,5Ј-Br 2 BAPTA and EGF 1-3; ‚, chelator alone and E, chelator ϩ EGF 1-3. The titration was performed in Tris buffer at pH 7.5 with no NaCl. The optimal curve obtained by least squares fit to the data points is shown.
ones. The first EGF module of protein S is atypical, however, in that it has the ␤-hydroxylation motif but lacks the characteristic N-terminal motif required for Ca 2ϩ binding (Fig. 5).
The present studies performed on recombinant human EGF modules from protein S demonstrated the presence of very high affinity Ca 2ϩ binding sites in the EGF module region of the protein. The studies also demonstrated the high Ca 2ϩ affinity to be independent of the Gla module. This is in good agreement with previous Ca 2ϩ blotting experiments performed on bovine and human protein S (24). 3 The affinity of one of the Ca 2ϩ sites in EGF 1-4 and 2-4 was too high to allow an accurate estimate of the binding constant with the Quin 2 method, i.e. K d Ͻ 10 Ϫ10 M (in the absence of NaCl). This site, as well as a second high affinity site (K d Ϸ 2-3 ϫ 10 Ϫ8 M in the absence of NaCl), appeared to be located in the third or fourth EGF module, as judged by the Ca 2ϩ titrations. In the presence of 0.15 M NaCl, the Ca 2ϩ affinity of the sites in EGF 1-4 and 2-4 decreased 10 -1000-fold, and the affinity of the site in EGF 1-3 became too low to be measured accurately with 5,5Ј-Br 2 BAPTA. This is consistent with all the binding sites in highly charged environments. Similar salt effects have been found in studies of Ca 2ϩbinding sites in calmodulin and calbindin D 9K (34,39). In contrast, the effect of salt on the Ca 2ϩ affinity of the isolated EGF 3 (K d ϭ 5.2 mM in the absence of salt and 6.1 mM in the presence of 0.15 M NaCl) was negligible and for EGF 4 (K d ϭ 0.6 mM in the absence of salt and 8.6 mM in the presence of 0.15 M NaCl) reduced the affinity by about 10-fold, as has previously been observed in the isolated EGF modules from factors IX and X (14,19). 2 The Ca 2ϩ affinity of EGF 3 and 4 in EGF 1-4 and EGF 2-4 in the presence of 0.15 M NaCl is 10 3 -and 10 5 -fold higher than in the corresponding synthetic isolated modules. 2 The present data show that a major part of the increase in Ca 2ϩ affinity for EGF 3 and EGF 4, when going from isolated module to intact protein, stems from interactions with a neighboring EGF module in the protein.
A noteworthy result in this study was that EGF 3 has a Ca 2ϩ -binding site manifesting higher Ca 2ϩ affinity in fragment EGF 2-4 than it does in fragment EGF 1-3, (K d Ϸ 2 ϫ 10 Ϫ8 M versus K d Ϸ 7 ϫ 10 Ϫ6 M); i.e. EGF 3 appears to have an at least 350-fold lower Ca 2ϩ affinity in EGF 1-3 than in EGF 2-4. It has been shown in several cases that the Ca 2ϩ affinity of an EGF module is reduced if the protein segment N-terminal to it is removed. This may be an effect of stabilization of the site, since one or two of the Ca 2ϩ ligands may be located N-terminal to the module. The data obtained for the site in EGF 3 demonstrates an influence on Ca 2ϩ affinity from the module on the C-terminal site. However, we suspect that EGF 4 has higher affinity than EGF 3 because of extra negative charge in its N-terminal region (see below). If EGF 3 has the highest affinity site, then its affinity would have been increased by the addition of EGF 4. Also EGF 2 has ϳ25-fold higher Ca 2ϩ affinity in EGF 1-4 than in EGF 1-3; this is also influenced by EGF 4. EGF 1 seems to influence EGF 2 but to have no influence on EGF 3 or 4. To clarify the interactions between modules in detail, access to recombinant EGF 3-4 would have been helpful. Unfortunately, at this stage we could not investigate the Ca 2ϩ binding constants further, as attempts to express a construct encompassing modules 3-4 were fraught with difficulty because the recombinant protein yielded only a smear and high molecular weight oligomers when analyzed by SDS-polyacrylamide gel electrophoresis (attempts to express EGF 1-2 were beset with similar problems).
In this context, it should be borne in mind that EGF 4 in human protein S has the sequence EDIDE (positions 201-205; Fig. 5), whereas bovine protein S has the sequence DDVDE in the corresponding region, i.e. in both species, EGF 4 has an extra negative charge that is not found in most other Ca 2ϩ binding EGF modules. This may contribute to the very high affinity of the Ca 2ϩ site but does not in itself suffice to explain the difference in Ca 2ϩ affinity compared with for instance, fibrillin (K d Ϸ 0.35 mM in the C-terminal module in a module pair) (27). The Ca 2ϩ binding properties of other EGF modules with the extra negative charge have yet to be reported. EGF modules 2, 3, and 4 from protein S are similar to most other Ca 2ϩ binding EGF modules, e.g. from thrombomodulin and fibrillin, in that the sequence motif DI/VDE before the first Cys residue in the module is conserved. The other coagulation factors form a unique group in that they all have the sequence DGDQ before the first Cys residue in the N-terminal EGF module (7,12). Moreover, they have erythro-␤-hydroxyaspartic acid/Asp where EGF 2, 3, and 4 from protein S have erythro-␤-hydroxyasparagine/Asn. Yet, the Ca 2ϩ affinity of the isolated EGF modules from factors IX and X is in the same range as those from protein S and fibrillin, i.e. K d Ϸ 1 mM (14,19,30). The function of the conserved Gly residue is not known. At this stage it can only be speculated that the adjacent C-terminal ␣-helix in the Gla module can not accomodate the bulky side chain of an Ile/Val residue in this position for steric reasons but requires a small residue such as glycine.
The present and previous studies demonstrate that EGF modules can provide extremely versatile Ca 2ϩ -binding sites in extracellular and membrane proteins. The wide span of Ca 2ϩ affinities, from K d values in the nanomolar to the millimolar range, is striking. How the structure accounts for the differences in Ca 2ϩ affinity, in addition to the apparent variations in net negative charge in the vicinity of the site, is not clear in the absence of a high resolution structure for protein S. The same Ca 2ϩ ligands were identified in the determination of the structure of the Ca 2ϩ form of the isolated EGF modules from factor X (NMR spectroscopy) and factor IX (x-ray crystallography) (21,33). The EGF module pair from fibrillin binds Ca 2ϩ with approximately 10-fold higher affinity than does the isolated modules from factors IX and X (27,30). The x-ray structure of the module pair did not lead to the identification of additional Ca 2ϩ ligands. Instead, Downing et al. proposed the increased Ca 2ϩ affinity to be attributable to shielding of the calcium ion from the solvent and the provision of a better-defined binding site by the intermodular surface (33). The present findings for protein S suggest that EGF 4 increases the Ca 2ϩ affinities of the sites in EGF 3 and EGF 2 in a manner that remains to be elucidated. In this context, it is also of interest that EGF 1 is crucial for the inhibitory activity of EGF 1-4 on the interaction between activated protein C and protein S. 3 However, EGF 1-4 is 10-fold more active in this respect than EGF 1-3, suggesting extensive module-module interactions among all four EGF modules). 3 Determination of the solution structure of a pair of Ca 2ϩ binding EGF modules from fibrillin has shown them to be oriented in a near linear arrangement that is stabilized by Ca 2ϩ ligation. The orientation of the EGF modules in protein S is not known. Nor do we know the biological effects of the high variation in Ca 2ϩ affinity between the EGF modules in protein S. Elucidation of the importance of Ca 2ϩ binding in modulemodule interactions appears to be a prerequisite for an understanding of the function of these modules in coagulation factors as well as in the more complex receptors.