Localization of Phosphatidylserine Binding Sites to Structural Domains of Factor X a *

Binding of short chain phosphatidylserine (C6PS) enhances the proteolytic activity of factor X a by 60-fold J., B. R. (1996) Biochemistry 35, 7482–7491). In the present study, we locate three C6PS binding sites to different domains of factor X a using a combination of activity, circular dichroism, fluorescence, and equilibrium dialysis measurements on proteolytic and biosynthetic fragments of factor X a . Our results demonstrate that the structural responses of human and bovine factor X a to C6PS binding are somewhat different. Despite this difference, data obtained with fragments from both human and bovine factor X a are consistent with a common hypothesis for the location of C6PS binding sites to different structural domains. First, the (cid:1) -carboxyglutamic acid (Gla) domain binds C6PS only in the absence of Ca 2 (cid:2) ( k d (cid:1) 1 m M

The substantial effects of soluble phosphatidylserine (C6PS 1 ) on the kinetics of prothrombin activation by factor X a (1) and on the structure of factor X a , as documented here, indicate that phosphatidylserine (PS) may act as an allosteric regulator of prothrombin activation. PS located on the cytoplasmic face of resting platelet plasma membranes is exposed on the surface of activated platelet vesicles (2,3). The implication of this PS exposure and of the effect of PS on factor X a and on its ability to catalyze activation of prothrombin is that PS may act as a second messenger in regulating thrombin formation. Because of the crucial role of thrombin in hemostasis, the exposure of PS may be a crucial regulatory step in blood coagulation. To better define this regulatory process, it is important to know the locations of the PS binding sites on factor X a .
The organization of factor X into structural domains is illustrated below in Fig. 1. Factor X consists of two peptides. The light chain consists of an N terminus ␥-carboxyglutamic acidrich region (Gla module) and two Cys-rich cassette modules. The heavy chain consists of the serine protease catalytic domain. The two cassette modules of the light chain show strong sequence and structural homology to epidermal growth factor (EGF) (4) and are thus referred to as EGF N and EGF C , where N and C indicate the domain nearer to the N and C termini, respectively. Crystal structures of Gla domain-less factor X a (GDFX a ) have been published (5)(6)(7). In the most recent of these (7), the EGF cassette modules extend from the catalytic domain to make an extended molecule. In the structure of the analogous serine protease, factor IX a , the EGF N module is bent at the inner-EGF C hinge region to right angles with the EGF C , which is tucked along the catalytic module (8). It may be that the EGF modules form a hinge region that modulates the global structure of factor X a . The factor IX a structure also differs from that of GDFX a in containing the Gla domain. Although the Gla domain is critical for membrane binding and may modulate the structure of the EGF modules, little is known about the structure of Gla in whole factor X a . We have only a model structure of factor X a Gla domain based on the prothrombin Gla domain (9).
Binding of Ca 2ϩ to factor X is reportedly required for activation by factor VII a /tissue factor or by factor IX a /VIII a (10,11) and for the activity of factor X a (12). Ca 2ϩ binding is also required for PS regulation of factor X a proteolytic activity (1). Ca 2ϩ binds mainly to the Gla module (13), but there also appears to be a high affinity Ca 2ϩ binding site (k d ϳ 160 M) in the catalytic domain (9, 14 -16) and a lower affinity Ca 2ϩ binding site (k d ϳ 0.7-1.2 m M) on the isolated first EGF-like module (4,12,16,17). A Ca 2ϩ -dependent interaction between the EGF-like and Gla modules appears to enhance the affinity of the site on the EGF-like module to the point that it is tighter (17,18) (k d ϳ 120 M) than the catalytic domain site. Consistent with this, nuclear magnetic resonance shows that Ca 2ϩ binding tightens the fold of the isolated EGF N domain and bends Gla and EGF N domains toward each other around a hinge located in the Gla domain, referred to as a helical or hydrophobic stack (19).
Despite considerable information about Ca 2ϩ binding to factor X a , we have virtually no information about the location of PS binding sites on this key enzyme. The aims of this work have been to locate the lipid regulatory site(s) in factor X a and to identify the structural domains of factor X a necessary to see the C6PS regulatory effect on factor X a activity.

Methods
Preparation of Factor X a -Bovine factor X was isolated from a barium citrate precipitate obtained from freshly collected bovine plasma (20,21). Human factor X for stoichiometry and CD measurements was purified from recovered human plasma obtained from the American Red Cross, according to the method of Dahlback et al. (22). Factor X obtained as above was analyzed by SDS-PAGE, concentrated (Centricon-10 concentrator supplier), and then stored at Ϫ70°C at a concentration of about 1 mg/ml in 5 mM Tris, 20 mM sodium citrate, 0.6 M NaCl, pH 7.4. A final purification of factor X was performed 1 day before an experiment by high-performance liquid chromatography on a PerkinElmer Life Sciences Isopure LC system using a Mono Q HR 5/5 ion exchange column (Amersham Biosciences, Inc., Norwalk, CN). The purified factor X was dialyzed into buffer (50 mM Tris, 175 mM NaCl, pH 7.4) for activation. Factor X (10 M) with 5 mM Ca 2ϩ was activated at 25°C with RVV-X that had been covalently linked to agarose beads (10,23). Factor X a was purified by high-performance liquid chromatography on a Mono Q column, and the isolated protein was analyzed by SDS-PAGE electrophoresis. Factor X a concentration was measured by determining the rate of S-2765 hydrolysis in a plate reader-based assay (1), using active site-titrated factor X a to construct the standard curve (24).
Purification of Domain Fragments of Bovine Factor X a -Isolation and purification of fragments of bovine factor X a and several of its structural domains (Gla, EGF N , Gla-EGF N , and Gla-EGF NC ) by controlled trypsin digestion were described previously (18,25,26).
Preparation of GDFX a , E 2 FX a , and Y99T-The RSV-PL4 expression vector was used to express factor X, GDFX, and a construct missing both the Gla and first EGF domain (E 2 FX) in human 293 cells (16). GDFX was also prepared for stoichiometry measurements as described by Morita and Jackson (27) from isolated factor X. Purified factor X was reacted with ␣-chymotrypsin (1:400 factor X:␣-chymotrypsin) at 22°C for 45 min, a time sufficient to convert 95% of factor X to GDFX, as judged by SDS-PAGE on a 6% gel. The reaction was stopped by addition of 1 mM diisopropyl fluorophosphate, and Gla-domainless factor X was chromatographed on a Mono Q column. GDFX Y99T mutant was also prepared as described (28 -30). GDFX and its mutant and E 2 FX were activated with RVV-X as described earlier (29).
Measurement of Amidolytic Activity of Factor X a and Its Constructs-The amidolytic activities of expression products of a human factor X a cDNA and of expression products of two of its deletion mutants (GDFX a , E 2 FX a ) were measured in the presence of 3 mM Ca 2ϩ using the synthetic substrate S-2765 and the microplate reader-based assay described above. Samples, containing 20 nM protein, various concentrations of C6PS and 3 mM Ca 2ϩ in a buffer (50 mM Tris, 175 mM NaCl, pH 7.6) containing 0.6% PEG, were incubated at 37°C for 15 min before measuring activities. The amidolytic activities were estimated from measured initial rates of S-2765 hydrolysis, using a standard curve obtained with active site-titrated factor X a (1). Amidolytic activities of GDFX a and Y99T were measured using the chromogenetic substrate Spectrozyme PCa (SpPCa) also as described earlier (29). Samples containing 20 nM protein, various concentrations of C6PS (0, 400, 900 M), and 0.6% PEG (to prevent adsorption of protein to the plate) were incubated in buffer (50 mM Tris, 175 mM NaCl, pH 7.6) in polypropylene Eppendorf tubes at 37°C for 5 min before being added to a flat-bottomed polypropylene 96-well plate (Greiner America, Inc.) preincubated at 37°C. The initial rates of SpPCa amidolysis were determined on a Versamax Tunable Microplate Reader (Molecular Devices, Sunnyvale, CA) at five substrate concentrations (50, 100, 200, 400, and 600 M) and analyzed in terms of the Michaelis-Menten model using non-linear regression methods available in Sigma Plot 6.0.
Circular Dichroism Measurements-Circular dichroism (CD) spectra were generally recorded from 250 to 200 nm on an Aviv Model 620S spectrometer (Aviv Associates, Inc., Lake Wood, NJ) in a 1-cm pathlength cell at 24°C with a bandwidth of 1.0 nm. Data points were collected at every 0.5 nm with an average time of 5 s on each point. Some data were obtained down to 195 nm on an Applied Photophysics P i * spectrometer in a 1-mm path-length cell with a bandwidth of 1 nm and data collection at every 0.5 nm. Baseline CD spectra of buffer containing various concentrations of soluble C6PS were collected in the absence and in the presence of 3 mM Ca 2ϩ and were subtracted from sample spectra. The baseline-corrected digital data were processed, smoothed, and converted to molar ellipticity, ⌬⌰. We have previously determined the critical micelle concentration (CMC) of C6PS at different Ca 2ϩ and protein concentrations (1), but controls to detect micelle formation were in all cases still performed by watching for sudden drops in ellipticity in the range of 240 -250 nm. For human and bovine factor X a , 2 the CMC seen in this way was similar to the CMC reported earlier by quasi-elastic light scattering methods (1). The ellipticity ratio ⌰ 222 / ⌰ 208 (32) is used here as a convenient parameter to follow changes in the secondary structure of factor X a and its fragments upon addition of C6PS. In addition, we have estimated ␣-helix content using published software packages CDSSTR and CONTIN (33) to give context to the ⌰ 222 /⌰ 208 ratio. The ability of CD spectra taken to 200 nm to define ␣-helix content but not ␤-sheet or turn content is well documented (34). We could not collect spectra to 185 nm to perform a complete secondary structure analysis in a buffer containing NaCl, because Na ϩ increases the buffer absorbance in the deep UV (34). Na ϩ was necessary in our studies, because Ca 2ϩ binding is required for regulation of factor X a by C6PS (1), and Ca 2ϩ binding is linked to Na ϩ binding (35,36).
Fluorescence Titration of DEGR-E 2 FX a by Soluble C6PS-Fluorescence intensity measurements were carried out on an SLM 48000 spectrofluorometer (SLM Aminco, Urbana, IL). Slits were closed between measurements to avoid photodegradation of the sample. All buffer solutions were filtered using 0.2-m filters (Nalge Co., Rochester, NY). DEGR-E 2 FX a was prepared by sequential addition of 5 l of DEGR-CK (1 mg/ml in 0.02 M Tris, 0.1 M NaCl, pH 7.5) to 1 ml of about 1 M purified factor E 2 FX a . The extent of labeling at the active site was followed by the loss of enzymatic activity, as monitored by the S-2765 assay. Labeling was stopped when no activity remained. DEGR-E 2 FXa was then dialyzed against 50 mM Tris, 0.1 M NaCl, pH 7.5, to remove free reagent (37). DEGR-E 2 FX a (100 nM) in 1.0 ml of buffer (50 mM Tris, pH 7.5) was incubated in a stirred micro-cuvette (Hellma Cells, Jamaica, NY) with 0.15 mM NaCl or 3 mM Ca 2ϩ or both at 25°C for 20 min. Following additions of C6PS (1-2 l each addition for a maximum of 4% dilution) and an equilibration of at least 4 min, fluorescence intensity was recorded using an excitation wavelength of 340 nm (bandpass 8 nm) and an emission wavelength of 550 nm (bandpass 4 nm). For each addition, several intensity measurements were performed and averaged and corrected for dilution. Control experiments were per-formed in which buffer was titrated with soluble lipid in the absence of protein. The lipid solution showed very minor background fluorescence or light scattering signal (which was subtracted from sample signal) until the critical micelle concentration was reached. The critical micelle concentration for C6PS in the presence and absence of 1 M factor X a were determined previously to be 0.95 and 2.5 mM, respectively (1). The critical micelle concentrations for C6PC under similar conditions were even higher (1). Data were not analyzed above the critical micelle concentration.
Intrinsic Fluorescence of Gla-EGF NC -Gla-EGF NC (100 nM) in 50 mM Tris, 150 mM NaCl, pH 7.4, in the presence and in absence of 3 mM Ca 2ϩ was titrated with soluble C6PS, and the intrinsic fluorescence was monitored at 345 nm (bandpass 4 nm) followed by excitation at 285 nm (bandpass 8 nm). Control experiments were as mentioned for DEGR-E 2 FX a fluorescence.
Phospholipid Sample Preparation-C6PS and C6PC solutions were prepared from measured quantities of 10 mg/ml stock solutions in chloroform. The chloroform was evaporated under a stream of nitrogen. The lipid was re-dissolved in cyclohexane, and this solution was frozen on the wall of a capped test tube and then lyophilized overnight. The resulting dry powder was dispersed in the appropriate volume of buffer and vortexed thoroughly to reach a concentration of ϳ100 mM. The final concentration of this phospholipid stock solution was determined by an inorganic phosphate assay (38).
Determination of the Stoichiometry of C6PS Binding to Various Fragments of Factor X a -The stoichiometries of soluble C6PS binding to factor X a , GDFX a , Gla-EGF NC , and Gla in the absence and in the presence of 3 mM Ca 2ϩ were determined by equilibrium dialysis measurements. This procedure not only establishes stoichiometry but also confirms indirect binding results by a direct measurement. Experiments were performed using 2.0-ml Teflon dialysis cells (Spectrum Medical, Los Angeles, CA) with the two cells separated by a 2000 molecular weight cut-off membrane. Both chambers contained equal amounts of C6PS, enough to saturate Ͼ85% of the protein present in one-half of the dialysis cell at varying concentrations (30 -100 M). The lipid concentrations used depended on the crude binding constants estimated in CD titrations. Depending on the particular combination of lipid and protein concentration in a given experiment and on the stoichiometry of binding for a particular peptide, between 79 and 94% of lipid remained unbound at equilibrium. The two chambers were allowed to equilibrate at room temperature for 24 h while being rotated horizontally at 20 rpm. The protein concentration gradient between the two halves of the cell causes a difference in the total phosphate concentration between the two halves of the cell. The concentration of proteinbound C6PS was measured as the difference in total phosphate concentration (⌬P) (38) between the two chambers of the dialysis cell. Assuming a simple model of binding of lipid to n equivalent and independent sites, it is easy to show that ⌬P should vary with protein concentration as follows, where [L] is free lipid concentration and k d /n is the observed stoichiometric dissociation constant for lipid binding, assuming a single site model. For [L] Ͼ Ͼ k d /n, this is roughly a straight line with a slope proportional to n. In our experiments, we maintained [L] Ͼ k d /n, but the total lipid concentration had to remain less than the CMC of the lipid. Thus, to obtain n, we had to fit a plot of ⌬P versus protein concentration to the non-linear equation given in Equation 1, using standard nonlinear regression procedures and the program SigmaPlot (version 6 for Windows 2000; Jandel Scientific).

Data Analysis
In our experiments, soluble lipid was added to the protein solution, and the observed response was taken as representing the fraction of protein bound (f) to lipid at concentration [L] given by, where K d is the apparent stoichiometric binding constant for soluble lipid binding to the protein. Any observable value that changes from an initial value of R 0 to a final value at saturation, R sat , as a result of binding can be written as follows,

Effect of Soluble C6PS on the Amidolytic Activities of Human Factor X a and Its Deletion Constructs-
We have shown previously that C6PS enhances proteolytic activity of human factor X a by roughly 60-fold but inhibits the amidolytic activity toward S-2765 by 60% (1). The variation of amidolytic activity of expressed human factor X a , GDFX a , and E 2 FX a ( Fig. 1) in the presence of 3 mM Ca 2ϩ with soluble C6PS concentration is shown in Fig. 2. The highest C6PS concentration used (0.8 mM) was still below the CMC for C6PS in the presence of 3 mM Ca 2ϩ (ϳ2.5 mM) (1). The amidolytic activities of factor X a and GDFX a were decreased by 79 and 9%, respectively, at saturation with C6PS (Table I), but C6PS had negligible effect on the amidolytic activity of E 2 FX a . The functional responses of factor X a and its constructs to C6PS binding were reasonably well described by a single binding site model (see "Methods"), and the binding parameters are given in Table I. The apparent K d for C6PS binding to the expressed human factor X a in this experiment (39 Ϯ 6 M) was comparable to but somewhat smaller than we have reported previously for factor X a isolated from outdated human plasma (65 Ϯ 5 M) (1), and the percent inhibition was also greater (80 versus 60%). This probably reflects the slight difference between X a from human plasma and factor X a from a single cDNA clone, as used here.
Effect of Soluble C6PS on the CD Spectra of Expressed Human Factor X a and Its Constructs-The effects of soluble C6PS on the CD spectra of expressed human factor X a and its constructs, GDFX a and E 2 FX a , were studied in the absence and presence of 3 mM Ca 2ϩ . CD spectra of human factor X a are shown at various concentrations of C6PS in the presence (Fig.  3A) and in the absence of 3 mM Ca 2ϩ (Fig. 3B). Although it is reported to bind Ca 2ϩ (17), human factor X a did not undergo a detectable change in secondary structure upon addition of 3 mM Ca 2ϩ , as seen from the solid and dotted curves in Fig. 3A and from the ⌰ 222 /⌰ 208 ratio and ␣-helical content (Table I). Secondary structure analysis yielded an estimate of 11% helical content in the presence or absence of Ca 2ϩ , in good agreement with the reported helicity for the analogous factor IX a crystal structure (10.6%) (8). However, there was a substantial change in CD upon addition of C6PS to human factor X a in the presence of 3 mM Ca 2ϩ (Fig. 3A). By contrast, there was only a small change in CD upon addition of C6PS in the absence of Ca 2ϩ . The variation of ellipticity ratio (⌰ 222 /⌰ 208 ) for human factor X a with C6PS concentration is shown as an inset in Fig. 3 (A and  B) in the presence and absence of Ca 2ϩ , respectively.
The variations of ellipticity ratio of GDFX a and E 2 FX a with C6PS concentration are shown in Fig. 4 (B and C, respectively). The smooth lines through the data result from fitting the data to a single-binding-site model as described under "Methods." The stoichiometric binding constants (K d ) and percent changes in ⌰ 222 /⌰ 208 at saturation (⌬R sat /R 0 ϫ 100%), resulting in the least square fits, are reported in Table II. We stress that K d values obtained in this way cannot be interpreted as site binding constants (k d ), because the number of data points taken and their intrinsic accuracy were not sufficient to define a binding mechanism in terms of the number, affinities, and responses of different sites present on each peptide fragment. However, the data in Fig. 4 do establish, in most cases, which fragments do bind C6PS, the magnitude of the total response (shifts in secondary structure), and the fraction of sites occupied at any lipid concentration (from K d values). The ellipticity ratio of factor X a decreased by 6% upon saturation with C6PS in the absence of Ca 2ϩ and by 16% when saturated with C6PS in the presence of 3 mM Ca 2ϩ . Similarly, the ␣-helical content decreased by 1 and 3%, respectively (Table II). There was an 18% decrease in ⌰ 222 /⌰ 208 ratio of GDFX a but only a slight (0.8%) decrease in helical content in response to soluble C6PS in the presence of 3 mM Ca 2ϩ . This discrepancy between changes in ⌰ 222 /⌰ 208 and helical content suggests that GDFX a binding to C6PS involves more than a change in helical content. Neither the ⌰ 222 /⌰ 208 ratio nor the ␣-helix content changed upon addition of C6PS in the absence of Ca 2ϩ (Fig. 4B, Table I). Human GDFX a seems to bind soluble C6PS in the presence of 3 mM Ca 2ϩ , with its structural response being comparable to that of whole human factor X a , but seems not to bind C6PS in the absence of Ca 2ϩ , at least not with a measurable structural response. The ⌰ 222 / ⌰ 208 ratio of E 2 FX a decreased by 10% upon addition of saturating concentrations of C6PS in the presence of 3 mM Ca 2ϩ but remained unchanged in the absence of Ca 2ϩ . The 10% decrease in ellipticity ratio was significantly smaller than that seen either for native human factor X a (16%) or for GDFX a (16%).
Effect of Soluble C6PS on the CD Spectra of Bovine Factor X a and Its Domains-The CD spectra of bovine factor X a and its Gla, EGF N , Gla-EGF N , and Gla-EGF NC domains were also collected and analyzed in terms of ⌰ 222 /⌰ 208 and ␣-helical content. The variations of ellipticity ratio (⌰ 222 /⌰ 208 ) with C6PS concentration in the presence (closed circles) and absence (open circles) of 3 mM Ca 2ϩ for factor X a , Gla, Gla-EGF N , and Gla-EGF NC domains, are shown in Fig. 4 (A, D, E, and F,  respectively). The smooth curves through these data result from fitting the data to a single-site binding model (see "Methods"), with binding constants (K d ) and fractional changes in ⌰ 222 /⌰ 208 at saturation (⌬R sat ) reported in Table II. The ⌰ 222 / ⌰ 208 ratio of bovine factor X a decreased by 16% upon addition of saturating concentration of C6PS in the absence of Ca 2ϩ but increased by 26% in the presence of 3 mM Ca 2ϩ (Table II). These same spectra were analyzed to reveal changes in ␣-helical content of Ϫ1 and ϩ6%, respectively (Table II).
Comparison of changes in ␣-helical content with changes in ⌰ 222 /⌰ 208 ratios for human and bovine factor X a and their   1. Domain structure of factor X a and summary of results. Schematic diagram of the domain structure of human factor X a . The N terminus contains a region rich in ␥-carboxyglutamic acid responsible for membrane binding (Gla domain). The epidermal growth factor domains (EGF N and EGF C ) and the catalytic domain are also shown. The construct missing the Gla domain is referred to as GDFX a , and the domain missing both the Gla and EGF N domains is called E 2 FX a . Results in the absence of Ca 2ϩ are summarized below the factor X a diagram: bovine Gla, Gla-EGF N , and Gla-EGF NC CD spectra all responded to C6PS. Stoichiometry measurements showed that Gla domain binds to one molecule of C6PS, whereas the Gla-EGF NC domain binds to two molecules. Neither stoichiometry nor CD measurements showed an interaction of C6PS with human E 2 FX a or GDFX a . Factor X a binds one molecule of C6PS, an interaction confirmed by CD. Results in the presence of Ca 2ϩ are summarized above the factor X a diagram: bovine Gla and Gla-EGF N did not bind to C6PS, but Gla-EGF NC binds one molecule of C6PS and also showed a change in CD spectrum and intrinsic fluorescence with C6PS. Human E 2 FX a , GDFX a , and factor X a all showed altered CD spectra with C6PS. GDFX a and factor X a each bind two molecules of C6PS, and their amidolytic activities were sensitive to the presence of C6PS. Although E 2 FX a binding to C6PS was detected with CD and fluorescence, its amidolytic activity was unchanged with the addition of C6PS. fragments (Table II) makes it clear that there is a general correlation between these two parameters, but that the ⌰ 222 / ⌰ 208 ratio reflects more than just helical content. It is also clear from comparison of the responses of bovine and human factor X a to C6PS that the secondary structural changes associated with C6PS binding are different for these two proteins (Table  II). However, CD spectra for bovine X a taken under the same conditions used for the spectra in Fig. 3 were qualitatively similar to the human X a spectra shown in Fig. 3. To make a more quantitative comparison of these two proteins, we collected spectra down to 185 nm in a buffer identical to that used for Fig. 3, except it lacked NaCl and was therefore much more transparent in the deep UV spectrum (34). Dialysis of bovine factor X a into this buffer and then return to a normal 150 mM NaCl buffer over a period of 48 h had no measurable effect on the ability to bind C6PS or on the amidolytic activity of factor X a toward S-2765 substrate. Secondary structure analysis of the bovine and human protein spectra in a NaCl-free buffer revealed no difference between the two of greater than 0.9% of helical, beta, turn, or unstructured content. As expected, because Na ϩ is needed for Ca 2ϩ binding (35,36) and Ca 2ϩ is needed for C6PS binding (1), there was no significant change in secondary structure content upon addition of either Ca 2ϩ or C6PS to either protein in the buffer lacking NaCl. Thus, spectra of this quality could not be used to quantitate the different effects of C6PS on bovine and human factor X a secondary structure, but were useful to establish the expected structural similarity between these two analogous proteins.
The ellipticity ratio of the Gla domain increased by 43%, while ␣-helicity increased by 1.8% upon addition of 3 mM Ca 2ϩ (Fig. 4D at 0 mM C6PS). In the absence of Ca 2ϩ , the ellipticity ratio increased by 97% at saturation with C6PS while ␣-helicity increased by 2.1% (Table II). However, no measurable change was seen in the presence of 3 mM Ca 2ϩ . This suggests the presence of at least one Ca 2ϩ -masked soluble lipid binding site in the Gla domain of bovine factor X a but no Ca 2ϩ -dependent site. This site was not specific for PS, because C6PC also bound to the Gla domain in the absence of Ca 2ϩ (data not shown).
Ca 2ϩ bound to the EGF N domain of factor X a and induced a large decrease in the ellipticity ratio (81%) but no change in the ␣-helical content (Table II). This is not surprising, because the solution structure of bovine factor X a EGF N is reported to consist almost entirely of anti-parallel ␤-sheets and turns (39) and Ca 2ϩ binding to Gla-EGF N seems not to alter its fold (19). This is another example of how the ⌰ 222 /⌰ 208 ratio reflects more than the ␣-helical content of a protein, and, in this case, must reflect Ca 2ϩ -induced structural rearrangements that alter the electronic symmetry of EGF N . By contrast to Ca 2ϩ , C6PS had no effect on the EGF N spectrum (data not shown)  Table II.  Table II. a Calculated as the percent change in observable at saturating lipid, i.e. ⌬ sat ϭ (R sat Ϫ R 0 )/R 0 ϫ 100%. R ϭ ⌰ 222 /⌰ 208 .
b Changes in the percent helicity estimated by fitting the CD spectra by two different algorithms (33). The disparity between estimates by the two methods was always Ͻ10% of the estimates.
c Parameter uncertainty based on non-linear regression. d NA, not applicable. either in the presence or in the absence of Ca 2ϩ (Table II). The Gla-EGF N domain pair also bound to Ca 2ϩ and induced a 140% increase in ellipticity ratio (Fig. 4E). In terms of secondary structure, this translated into a large increase in ⌰ 222 /⌰ 208 and a nearly insignificant decrease in ␣-helical content (Tables I  and II). It is known from NMR studies that Gla-EGF N undergoes a significant structural reorganization in which the EGF and Gla domains fold onto each other upon binding Ca 2ϩ (19). C6PS induced a structural change in Gla-EGF N in the absence of Ca 2ϩ (⌰ 222 /⌰ 208 increased by 333% and helicity increased by 2.3%; Fig. 4E and Table II), but no C6PS-dependent change was detected in the presence of 3 mM Ca 2ϩ . The change in ⌰ 222 /⌰ 208 seen in the absence of Ca 2ϩ was much greater than seen for the Gla domain (Table II). Because EGF N showed no change in response to C6PS, this implies that C6PS binding to Gla-EGF N in the absence of Ca 2ϩ , like the binding of Ca 2ϩ in the absence of C6PS (19), involves linkage between the Gla and EGF N domains. Unlike Gla-EGF N , Gla-EGF NC interacted with soluble C6PS in the presence of Ca 2ϩ (⌰ 222 /⌰ 208 increased by a barely perceptible 10% while helicity also increased by only 1.1%) as well as in its absence (⌰ 222 /⌰ 208 and helicity increased by 121 and 2.8%, respectively).
Stoichiometry of C6PS Binding to Factor X a and Its Fragments-To test and extend our CD observations, C6PS binding to bovine and human factor X a and their fragments was also monitored by equilibrium dialysis, a direct binding measurement. Because of the large quantities of protein needed for these measurements, it was not possible to obtain complete binding isotherms by this method. However, using dissociation constants estimated from our CD data, it was possible to estimate binding stoichiometries (see Equation 1 under "Methods"). The measured stoichiometries of C6PS binding to various fragments in the presence and in the absence of 3 mM Ca 2ϩ are shown in Table III. Human factor X a binds two molecules of C6PS in the presence and one in the absence of 3 mM Ca 2ϩ , respectively, just like bovine factor X a . 2 Human Gla domainless factor X a (GDFX a ) also bound two molecules of C6PS in the presence of Ca 2ϩ but did not bind to C6PS in the absence of Ca 2ϩ . The bovine Gla domain bound one molecule of C6PS in the absence of Ca 2ϩ and did not bind C6PS in the presence of 3 mM Ca 2ϩ , consistent with the lack of any change in ⌰ 222 /⌰ 208 under these conditions (Fig. 4D). When the Gla domain was linked to the EGF NC domain, it bound two molecules of C6PS in the absence and one molecule of C6PS in the presence of 3 mM Ca 2ϩ . Stoichiometry measurements were thus all consistent with the results of our CD experiments. Together, these results provide a clear and self-consistent picture of the distribution of C6PS binding sites on factor X a (Fig. 1).
Fluorescence of Gla-EGF NC Titrated with C6PS-Because the C6PS-induced change in secondary structure in the Gla-EGF NC domain triplet was so small (Fig. 4F), we tested further for a C6PS-induced conformational change by titrating the intrinsic fluorescence of Gla-EGF NC with soluble C6PS in the absence (open circles) and presence (closed circles) of 3 mM Ca 2ϩ , with the results shown in Fig. 5. The curves passing through the data were obtained by fitting the data to the single binding site model (see "Methods"). The apparent stoichiometric binding constants for C6PS binding to Gla-EGF NC in the presence and absence of Ca 2ϩ were 203 Ϯ 63 and 470 Ϯ 74 M, respectively. These binding constant are comparable to those obtained by CD measurements (Table II). We conclude that Ca 2ϩ -dependent binding of C6PS produced conformational changes in the Gla-EGF NC domain triplet.
Fluorescence of DEGR-E 2 FX a Titrated with C6PS-Although CD data suggest that a Ca 2ϩ -requiring site might be located in the E 2 FX a fragment, we could not confirm this by direct stoichiometry measurement due to a lack of sufficient quantities of this expressed protein. To confirm the Ca 2ϩ -requiring binding site in to the E 2 FX a fragment, we labeled E 2 FX a with DEGR-CK and monitored the change of fluorescence intensity of DEGR-E 2 FX a as a function of C6PS. The binding analysis was performed with two preparations of DEGR-E 2 FX a , and the results are presented in Fig. 6, with the results from the two preparations distinguished by closed circles and closed squares. The results clearly show a saturable drop in DEGR-E 2 FX a fluorescence in the presence of Na ϩ and Ca 2ϩ , although no change was detected when either Ca 2ϩ (open triangles) or Na ϩ (open circles) were missing. This requirement for Na 2ϩ and Ca 2ϩ for C6PS binding to E 2 FX a raises the possibility that the site in the E 2 FX a fragment might be the amine binding site that is also reported to require Ca 2ϩ and Na ϩ (40). A global fit of a single-site binding model to the two data sets obtained in the presence of both Na 2ϩ and Ca 2ϩ is shown by the solid hyperbolic curve.
Effect of Soluble C6PS on the Amidolytic Activities of Expressed Human GDFXa and Its Mutant Y99T-To better establish the identity of the C6PS site located in the E 2 FX a fragment, we monitored hydrolysis of SpPCa by GDFX a and its mutant Y99T as a function of soluble C6PS, in buffer lacking Ca 2ϩ . Monnaie et al. (40) showed that an amine binding site  exists in the catalytic domain and shares at least some residues with the substrate binding site. Like binding of C6PS, this site was both Ca 2ϩ -and Na ϩ -dependent. The variation of amidolytic activities as a function of C6PS concentrations is shown in Fig. 7. As expected, because no C6PS binding site was detected in GDFXa in the absence of Ca 2ϩ , soluble C6PS did not have any effect on the rate of hydrolysis of SpPCa by GDFX a (closed triangles). However, the rate of hydrolysis of SpPCa by the Y99T GDFX a mutant decreased with the addition of C6PS, and the binding curve (open triangles) was fitted to a single-binding-site model with an apparent stoichiometric dissociation constant of 410 M. It appears that the Ca 2ϩ -requiring C6PS binding site in GDFX a becomes independent of Ca 2ϩ as a result of the Y99T mutation. This result clearly proves that residue Tyr-99, which is known to be part of a reported amine binding site and the substrate binding site (40), plays a role in C6PS binding to the Ca 2ϩ -requiring site located in the E 2 FX a fragment. The question still remains: Are the C6PS site and the substrate binding site are one and the same? To answer this question, we monitored the effect of C6PS on the k cat and K M for SpPCa hydrolysis by Y99T GDFX a in the absence of Ca 2ϩ . As shown in the inset in Fig. 7, k cat remained constant with C6PS concentration but K M increased in a hyperbolic fashion with an increase in lipid concentration. This means that C6PS competes with substrate for the substrate binding site, just as tertiary amines have been shown to do (40). We conclude that the Ca 2ϩ -requiring C6PS binding site shares some ligand-recognition residues with the substrate binding site. We report elsewhere that the Ca 2ϩ -requiring site minimally recognizes glycerolphosphorylserine and does not bind phosphatidylcholine. 3 Because the amine binding site recognizes choline (40), it is unlikely that the C6PS site and the amine site are identical, but they likely share some ligand recognition regions with each other and with the substrate binding site. DISCUSSION We have shown previously (1) that soluble C6PS enhanced factor X a 's proteolytic activity by about 60-to 70-fold. The purpose of this work was to locate the C6PS effector sites to one or more domains of factor X a . Our results have located three sites on this serine protease but suggest that only one is involved in functional regulation. These results support a reasonable hypothesis for how one molecule of regulatory C6PS and two molecules of non-regulatory C6PS are bound to factor X a . This hypothesis is summarized in Fig. 1. The arguments in favor of this hypothesis are summarized below.
One Ca 2ϩ -masked Phospholipid Site Is Located in the Gla Domain-As expected (42), the secondary structure of the Gla domain was sensitive to the presence of 3 mM Ca 2ϩ (compare open and closed circles of Fig. 4ID at 0 mM C6PS) with the change elicited by Ca 2ϩ being a 43% increase in ⌰ 222 /⌰ 208 and a 1.8% increase in ␣-helical content. However, the C6PS site in the Gla domain was masked by Ca 2ϩ , because no structural response was seen (Fig. 4D). This is surprising, because binding of Gla-containing proteins to PS-containing membranes has long been seen as mediated by a Ca 2ϩ -induced conformational change of the Gla domain (43)(44)(45)(46). It may be that binding of Gla-containing proteins to a PS-containing membrane involves adsorption of the Ca 2ϩ -conformation of the Gla domain to a membrane surface (31,(47)(48) rather than recognition of individual PS molecules by specific binding sites. By contrast, the C6PS-induced conformational change that we see in the absence of Ca 2ϩ does involve a single C6PS molecule (Table III). Because neither human (1) nor bovine factor X a 2 respond functionally to C6PS in the absence of Ca 2ϩ , it is unlikely that this Ca 2ϩ -masked site could by itself be a regulatory site.
Two C6PS Sites Exist in the Factor X a Fragment that Lacks the Gla Domain (GDFX a )-Our stoichiometry measurements showed that GDFX a binds two molecules of C6PS in the presence of 3 mM Ca 2ϩ (Table III). Both of these sites are Ca 2ϩ -dependent, because there was no detectable change in ⌰ 222 /⌰ 208 (Fig. 4B, open circles), and the stoichiometry was nearly zero (Table III) in the absence of Ca 2ϩ , meaning binding of C6PS in the absence of Ca 2ϩ is at best quite weak.
The EGF and Gla Domains Are Structurally Linked by C6PS Binding-The EGF N domain alone did not respond structurally to C6PS, in either the presence or absence of Ca 2ϩ (Table II). In the absence of Ca 2ϩ , however, when linked covalently to the Gla domain, the EGF N domain either experienced a large conformational change (⌬⌰ 222 /⌰ 208 ϭ 333%) or modified in a major way the response of the Gla domain (⌬⌰ 222 /⌰ 208 ϭ 97%) to C6PS. Based on a comparison of the structural changes for Gla and Gla-EGF N and on the low helical content of these two domains in homologous factor IX a (8), it would appear most likely that the response being monitored in Gla-EGF N is not FIG. 7. The effect of soluble C6PS on the amidolytic activities of expressed human GDFX a and its mutant Y99T. The initial rates of SpPCa amidolysis by GDFX a (closed triangles) and Y99T (open triangles) in the absence of Ca 2ϩ are plotted as a function of C6PS concentrations. Rates were measured at 37°C in a buffer containing 50 mM Tris, 175 mM NaCl, 0.6% PEG, at pH 7.6. The hyperbolic line passing through the Y99T data was obtained by fitting the data to the singlebinding-site model, which yielded a dissociation constant of 410 Ϯ 88 M. The initial rates of Y99T amidolysis were determined at 0, 400, and 900 M C6PS concentrations and at 50, 100, 200, 400, and 600 M substrate concentrations. From these data, we determined the k cat and K M values for SpPCa amidolysis by Y99T at these three lipid concentrations. The variation of k cat /k cat0 (circles) and K M (squares) with the C6PS concentration is shown in the inset. that of the Gla domain and is not a change in ␣-helix content, but is a change in the EGF N module that requires conformational linkage to the Gla domain. Like Gla, Gla-EGF N did not bind C6PS in the presence of Ca 2ϩ , so the Gla-EGF N site seen in the absence of Ca 2ϩ must involve the Ca 2ϩ -masked site in the Gla domain. Ca 2ϩ -mediated conformational linkage between the Gla and EGF N domains of factor X a is well established (14,19), and our results imply that this linkage involves a C6PS binding site as well.
Unlike the Gla-EGF N fragment, the Gla-EGF NC fragment experienced a structural change induced by C6PS both in the presence and absence of Ca 2ϩ , although the changes seen in the presence of Ca 2ϩ were qualitatively different from those seen in its absence (Figs. 4F and 5). Stoichiometry measurements showed that Gla-EGF NC binds one C6PS molecule in the presence of Ca 2ϩ and two in the absence of Ca 2ϩ (Table III). Based on these observations, Gla-EGF NC seems to have two types of C6PS sites. One is the Ca 2ϩ -masked site in the Gla domain, and one is a site that does not require Ca 2ϩ to recognize C6PS but that binds C6PS much more tightly in the presence of Ca 2ϩ than in its absence (Table II and Fig. 5). The Ca 2ϩ -dependent site requires linkage of all three N-terminal modules (Gla, EGF N , and EGF C ) for a full response. Ca 2ϩ is known to link the EGF N and Gla domains (14,19), and Ca 2ϩ masks the C6PS site in the Gla domain. For these reasons, we suggest that Ca 2ϩ links the C6PS sites in the Gla-EGF N and EGF NC domains to create one site having higher affinity for C6PS than either of the individual sites. If so, we expect that the sandwich formed by the Ca 2ϩ -linked Gla and EGF N domains (19) will form an important element of this binding site.
A Ca 2ϩ -requiring Site in the EGF C -catalytic Domain Is Probably Part of the Substrate Recognition Site-Direct measurement of stoichiometry by equilibrium dialysis measurements showed two C6PS binding sites in GDFX a in the presence of Ca 2ϩ and none in the absence of Ca 2ϩ (Table III). We know that one of these Ca 2ϩ -dependent sites must be in the EGF NC pair. A Ca 2ϩ -requiring site is in the E 2 FX a fragment (Figs. 4C and 6). Our experiment with the Y99T mutant of GDFX a showed that C6PS binding to the Ca 2ϩ -requiring site in the E 2 FX a fragment competes with binding of substrate (inset to Fig. 7). In addition, C6PS binding to GDFX a was altered by the Y99T mutation and Y99 is known to be part of the substrate recognition site (40). From these observations, we conclude that the Ca 2ϩ -requiring C6PS site in E 2 FX a at least overlaps the substrate recognition site. This is consistent with the fact that substrate recognition is also linked to Na ϩ and Ca 2ϩ binding (35,36). Because this site is located roughly 60 Å from the membrane surface (37), it is unlikely to be involved in regulating activity in vivo.
The Functional Response to C6PS Seems to Require Minimally the EGF NC Pair and the Catalytic Domain-It is known that the proteolytic activity of factor X a is enhanced roughly 60to 70-fold by binding of C6PS (1). The change in amidolytic activity of the GDFX a fragment upon titration with C6PS clearly followed a single-site-binding model (Fig. 2), suggesting either that only one site regulates activity or that the two sites are equivalent in their abilities to modulate activity. This fragment consists of a pair of EGF-like domains and a catalytic domain. Our data show that one C6PS site is located at or near the substrate recognition site (Fig. 7) and that one site is most likely in the EGF NC pair ( Fig. 1 and Tables I-III). It appears for three reasons that regulation of activity by C6PS requires the site in the EGF NC pair. First, the substantial amidolytic activity toward S-2765 of the E 2 FX a fragment did not respond to C6PS (Fig. 2), although E 2 FX a did undergo a C6PS-induced conformational change that required Ca 2ϩ (Fig. 6). Second, the K d for the Ca 2ϩ -dependent structural response of bovine Gla-EGF NC to C6PS (155 M, Table II) was similar to that for the activity response of bovine factor X a (167 M). 2 Finally, the site in the EGF NC pair can be occupied in the presence or absence of Ca 2ϩ (see Figs. 4F and 5 and Table II), but with very different K d values and structural responses under these two circumstances (Table II). In the absence of Ca 2ϩ , the response to C6PS seems to depend only on the linkage of Gla to EGF N , while in the presence of Ca 2ϩ , the complete EGF NC pair is needed. Because factor X a is active only in the presence of Ca 2ϩ , we conclude that the site that regulates activity requires both the EGF domains. The fact that the response of GDFX a activity to C6PS was only about a tenth that of whole factor X a implies that the regulatory site is in the EGF NC domain but that linkage to the Gla domain is essential for optimal changes in the catalytic domain's active site.
Bovine and Human Factor X a Have Analogous C6PS Binding Sites-Based on our results, we have noted that the bovine and human forms of factor X a show different structural responses to C6PS despite having very similar secondary structures in solution. If their responses to C6PS are different, it could also be that the location of C6PS binding sites might be different and our use of data from both bovine N-terminal fragments (Gla, Gla-EGF N , and Gla-EGF NC ) and human C-terminal biosynthetic fragments from cDNA constructs (X a , GDFX a , E 2 FX a ) would be flawed. Based on the analysis provided below that is based on careful inspection of the diagram in Fig. 1 summarizing all our observations, we argue that this is not the case.
Data obtained with the human X fragments clearly show that a single Ca 2ϩ -requiring C6PS site exists in E 2 FX a . The existence of this site in the catalytic domain was confirmed by our titration of DEGR-E 2 FX a fluorescence in the absence of Na ϩ and by titration of the activity of the Y99 mutant of GDFX a in the absence of Ca 2ϩ . Titrations of the amidolytic activity of these biosynthetic human factor X a fragments show that another site exists in the EGF NC pair and this site regulates factor X a activity. This second site was either absent or too weak to be detected in the absence of Ca 2ϩ . Because a single C6PS bound to whole factor X a and none bound to GDFX a in the absence of Ca 2ϩ , there must be a Ca 2ϩ -masked C6PS binding site in the Gla domain. The very different responses of whole factor X a and GDFX a in the presence of Ca 2ϩ (Fig. 2) to C6PS shows that linkage of the Gla and EGF NC modules is needed for a full functional response.
If we consider the experiments done with N-terminal fragments of bovine factor X a , we see clearly that a Ca 2ϩ -masked C6PS site exists in the Gla domain. We see as well that a single Ca 2ϩ -dependent (but not requiring) site exists in Gla-EGF NC . This site requires the linkage of the Gla, EGF N , and EGF C modules. Because two C6PS bind to whole bovine factor X a in the presence of Ca 2ϩ , there must be a second site in the catalytic domain.
We argue from this analysis that data obtained with bovine and human fragments lead independently to nearly the same conclusions. This means both that the bovine and human proteins interact similarly with C6PS (as expected for such highly homologous proteins) and that the two sets of data with proteins from different species support and confirm each other.