Membrane binding properties of the factor IX gamma-carboxyglutamic acid-rich domain prepared by chemical synthesis.

The fully gamma-carboxylated peptides based upon the complete and truncated Gla/aromatic amino acid stack domains of human Factor IX were prepared by solid phase peptide synthesis using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. A 47-residue peptide Factor IX-(1-47) and a 42-residue peptide Factor IX-(1-42), both containing 12 residues of L-gamma-carboxyglutamic acid, were purified by high performance liquid chromatography and oxidized to form the disulfide bond. Quantitative gamma-carboxyglutamic acid analysis of Factor IX-(1-47) and Factor IX-(1-42) indicated the presence of 12.1 and 11.2 gamma-carboxyglutamic acid residues/mol of peptide, respectively; no glutamic acid was detected. As monitored by fluorescence quenching, calcium ions induced the prototypical conformational transition in Factor IX-(1-47), but not in Factor IX-(1-42), that is observed with Factor IX. Half-maximal quenching of the intrinsic fluorescence of Factor IX-(1-47) was observed at Ca(II) concentrations of about 50 microM. Factor IX-(1-47) bound to the conformation-specific antibodies, anti-Factor IX:Mg(II) and anti-Factor IX:Ca(II)-specific in the presence of metal ions. Factor IX-(1-47) bound to phospholipid membranes, as monitored by energy transfer from intrinsic fluorophores to dansyl (5-dimethylaminonaphthalene-1-sulfonyl)-phosphatidylethanolamine incorporated into a lipid bilayer composed of phosphatidylserine:phosphatidylcholine. In contrast, Factor IX-(1-42) bound poorly to these same membranes. Factor IX-(1-47) did not inhibit Factor XIa activation of Factor IX but did inhibit the activation of Factor X by Factor IXa bound to Factor VIII in the presence of calcium ions and phospholipid. These results show that phospholipid membrane binding is a property of the Gla/aromatic amino acid stack domain and that the Factor IX-(1-47) peptide, prepared by chemical synthesis, preserves the membrane binding properties and the metal-induced conformational transitions observed in native Factor IX. These results indicate that Factor IX-(1-47) but not Factor IX-(1-42) is a suitable model for structural studies of Factor IX-membrane interaction.

The fully y-carboxylated peptides based upon the complete and truncated Gldaromatic amino acid stack domains of human Factor M were prepared by solid phase peptide synthesis using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. A 47-residue peptide Factor M41-47) and a 42-residue peptide Factor IX-(1-42), both containing 12 residues of L-y-carboxyglutamic acid, were purified by high performance liquid chromatography and oxidized to form the disulfide bond. Quantitative y-carboxyglutamic acid analysis of Factor IX4-47) and Factor M is a vitamin K-dependent plasma zymogen that plays a central role in blood coagulation (1). The mature protein is composed of a series of domains, including the Gla domain, the aromatic amino acid stack domain, two epidermal growth factor domains, and a serine protease domain (2,3). The fully *This work was supported by Grant HL42443 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
8 To whom correspondence should be addressed.
carboxylated form of human Factor IX binds calcium ions (4). Factor IX contains two classes of metal-binding sites which are defined by y-carboxyglutamic acid (5,6) and a third class which is a component of the EGF' domain (7, 8). More recently, an additional site within the protease domain has been suggested (9). The metal binding site within the EGF domain is defined by aspartic acid 64, a residue that is partially (about 30%) p-hydroxylated (10)(11)(12). Similar to the other vitamin K-dependent proteins, Factor IX and Factor IXa bind to phospholipid membranes composed of phosphatidylserine (PS)-phosphatidylcholine (PC) (13,14). Factor IX affinity for PS-PC membranes is independent of the PS concentration with PS compositions above 20-30%; a Kd of about 1-2 p~ has been measured (15). This interaction requires both calcium ions and fully carboxylated Factor IX (14). When Factor IX binds to calcium, the protein undergoes conformational transitions that can be monitored by fluorescence quenching and the expression of neoantigenic determinants. With confonnation-specific antibodies, Liebman et al. (14) demonstrated that, like prothrombin (161, Factor M undergoes two sequential conformational changes. Most divalent metal ions support the first conformational change while only calcium ions support the second conformational change. It is this final conformer that expresses phospholipid binding properties. The location of the phospholipid binding site in Factor IX is not known, but studies on the other vitamin K-dependent proteins provide insight. We have prepared mutants of prothrombin in which the first kringle domain, the second kringle domain, or both kringle domains have been deleted and shown that these mutant forms bind to phospholipid vesicles in the presence of calcium (17). These results indicate that the kringle domains, and by analogy the EGF domains in Factor E, are not implicated in phospholipid binding.
Despite considerable study of the membrane binding properties of the vitamin K-dependent proteins that spans two decades, the molecular details of the nature of protein-membrane interaction of this class of membrane-binding proteins remain elusive. Solution of the crystal structure of prothrombin fragment 1 has defined an internal core of calcium ions liganded by the carboxylate groups of y-carboxyglutamic acid (18). Most of the carboxylate oxygen atoms of y-carboxyglutamic acid residues participate in forming this core and are thus unavailable for bridging to phospholipid through calcium ions. This structure confirmed earlier suggestions that the membrane-binding sites of the vitamin K-dependent proteins did not necessarily involve y-carboxyglutamic acids as contact residues that interphosphatidylserine; PC, phosphatidylcholine; PE, phosphatidylethano-'The abbreviations used are: EGF, epidermal growth factor: PS, lamine; Fmoc, N-(9-fluorenyl)methoxycarbonyl; HPLC, high performance liquid chromatography; dansyl, 5-dimethylaminonaphthalene-lsulfonyl.
acted with the phospholipid membrane (16,19), in contrast to alternative models that had been proposed earlier (20). To determine the features of the protein-Ca(I1)-membrane complex at atomic resolution requires crystallization of the ternary complex or examination of small but functional Gla-containing protein domains in complex with membranes in the presence of Ca(I1) by two-dimensional nuclear magnetic resonance spectroscopy. In the current study we have determined whether the Factor M Gldaromatic amino acid stack domain in its fully carboxylated form is a suitable model for structural studies of Factor M-membrane interaction. We show that Factor M41-47) has membrane binding properties. This peptide can serve as a suitable analog for the characterization of the structure of the Gla domain of Factor M and its interaction with membrane surfaces.

EXPERIMENTAL PROCEDURES
Chemical Synthesis of Factor IX Residues 1 4 7 and of Factor IX Residues 142-Factor M-(1-47) and Factor IX41-42) were synthesized using FmoclN-methylpyrrolidone chemistry on an Applied Biosystems model 430A peptide synthesizer. Amino acids were coupled as 1-hydroxybenzotriazole esters onto 0.25 mmol of p-hydroxymethylphenoxymethyl polystyrene resin. Side chain protecting groups included 2,2,5,7,8-pentamethylchroman-6-sulfonyl (arginine), OtBu (aspartic acid and y-carboxyglutamic acid), trityl (cysteine), t-butoxycarbonyl (lysine), tBu (serine, threonine, and tyrosine). Following each coupling step all uncoupled a-NH, termini were acetylated. All L-y-carboxyglutamic acid residues were coupled using 1 mM of protected amino acid per cartridge. Activation was performed with 0.4 ml of methylene chloride and 1 ml of N-hydroxybenzotriazole/N-methylpyrrolidone. The coupling reaction was allowed to proceed for 36.5 min. Following coupling of each Gla residue and prior to acetylation, a ninhydrin assay was performed to determine coupling efficiency prior to proceeding with the synthesis. Residues which had coupling efficiencies of less than 90% were recoupled. Cleavage of the peptide from its solid support resin and simultaneous side chain deprotection was performed using trifluoroacetic acid. The cleavage reaction was performed in trifluoroacetic acid/l,2-ethanedithioVthioanisole/water/phenol (10:2.5:5:5:1). The cleavage reaction was allowed to proceed for 2 h at 25 "C. The resin was removed by filtration in a fritted funnel and 50 ml of methylene chloride added to the cleavage reaction supernatant and the volume reduced to about 2 ml on a Rotavapor-R (Buchi). The crude peptide was precipitated using cold anhydrous ethyl ether and washed several times with ethyl ether, then dried under vacuum. The crude deprotected peptide was purified by HPLC using a Hi-Pore 318 column (Bio-Rad, 250 x 21.5 mm). Alinear gradient of 2 0 4 0 % B (Buffer A 0.1% trifluoroacetic acid, water; Buffer B: 0.1% trifluoroacetic acid, acetonitrile) over 60 min was employed. The absorption of the peptide was monitored at 214 nm. The disulfide bond was formed by spontaneous air oxidation. A peptide solution (1 mg/lO ml) was adjusted to pH 8.5 with ammonium hydroxide. The peptide was incubated a t 23 "C for 20 h with constant stirring. The progress of the reaction was monitored using a n analytical reverse phase C18 column (Vydac, 0.46 x 25 cm; 5 pm). The above HPLC gradient was employed. Following oxidation the reaction mixture was lyophilized. The amino acid sequence was verified using a Milligen 6600 Prosequencer. The purity and molecular size of the peptides were confirmed by SDS-polyacrylamide gel electrophoresis in the presence and absence of p-mercaptoethanol, according to Schagger and von Jagow (21). The peptides were visualized with Coomassie Blue.
Materials-Phosphatidylcholine, phosphatidylserine, and dansylphosphatidylethanolamine were obtained from Avanti Polar Lipids and DEGR-CMK from Calbiochem. The chromogenic substrate specific for Factor Xa (CBS 31.39) was obtained from Diagnostica Stago. Human Factor X was purchased from Enzyme Research Laboratories. Human Analysis for y-Carboxyglutamic Acid-The y-carboxyglutamic acid content of the purified peptides was determined by a modification of the method of Kuwada and Katayama (22). Peptide (10 nmol) was hydrolyzed in 2.5 M KOH for 16 h at 110 "C. The alkaline hydrolysate, after neutralization, was derivatized with o-phthalaldehyde and ethanethiol. The derivatized amino acids were separated by high performance liquid chromatography on a Nucleosil 5SB anion exchange column (4.6 x 250 mm; Macherey-Nagel, Germany) under isocratic conditions at a flow rate of 1 ml/min. A 1:l volume to volume mixture of 0.1 M sodium citrate, pH 5.28:acetonitrile was used as the mobile phase (23). The column effluent was irradiated a t 240 nm using an Applied Biosystems model 980 fluorescence detector equipped with a 418-nm bandpass filter. Twenty-five to 200 pl of derivatized sample was injected. The y-carboxyglutamic acid content of the Factor M peptides was determined by calculating the molar ratios of aspartic acid to y-carboxyglutamic acid from the areas of the peaks corresponding to these amino acids. The molar ratios of these amino acids in the peptide sample was compared with those obtained for plasma-derived Factor M.
Preparation of Phospholipid Vesicles-Small unilamellar phospholipid vesicles (phosphatidy1serine:phosphatidylcholine:dansyl-phosphatidylethanolamine, 40:50:10) were prepared for phospholipid binding experiments by the method of Barenholz (24). Phospholipids were dried under N,, resuspended in Tris-buffered saline, and sonicated in a bath sonicator (Lab Supply) until the solution cleared. The phospholipids were subjected to centrifugation a t 40,000 x g for 30 min, then 50,000 x g for 90 min in a Beckman L8-80 M ultracentrifuge using a 70 Ti rotor. The supernatant contained the small unilamellar vesicles; phospholipid concentrations were determined by analysis for phosphorus (25). These vesicles were stored at -80 "C under nitrogen and used Immunochemical Analyses with Conformation-specific Antibodies-The binding of Factor "(1-47) and Factor IX-(1-42) to conformationspecific anti-Factor IX antibodies was studied using a solution phase radioimmunoassay. Factor IX was iodinated with NalZ5I using Enzymobeads (Bio-Rad). lZ5I-Labeled Factor IX was separated from free iodine by gel filtration on Sephadex G-25, then by immunoaffinity chromatography with anti-Factor M:Mg(II) covalently linked to Sepharose. Anti-Factor M:Ca(II)-specific antibodies and anti-Factor M:Mg(II) antibodies were purified as described previously (14). A competition radioimmunoassay was configured to study the displacement of 'T-labeled Factor IX from anti-Factor IX antibodies at increasing concentrations of Factor IX-(1-47) or Factor "(1-42). Varying concentrations of unlabeled competitor (Factor "(1-47) or Factor IX-(l42)), normal rabbit immunoglobulin (1 mg/ml), and '251-labeled Factor IX were incubated with either anti-Factor IXCa(I1)-specific antibodies or anti-Factor WMg(I1) antibodies. A final assay volume of 300 pl containing 0.15 M NaCl, 50 mM Tris, pH 7.4,l mM benzamidine, 0.1% bovine serum albumin, 0.1% Tween 20, and either 3 mM CaC1, or 5 mM MgC1, was incubated overnight at 4 "C. Following incubation, 1 ml of rabbit anti-goat immunoglobulin (25 mg in 1 ml of 0.1 M Tris, 0.15 M NaCl,O,l% NaN,, 2.5% polyethylene glycol-6000, pH 7.4) was added. The precipitate that formed was removed by centrifugation and assayed for lzeI in a Packard 5000 Series Auto-Gamma scintillation counter.

The Effect of Factor I X -( 1 4 7 ) on Factor IX Activation by Factor
XIa-The effect of Factor IX-(147) on Factor IX activation by Factor XIa was studied using SDS-gel electrophoresis and autoradiography to monitor the development of the Factor Ma light chain and the loss of Factor X. Factor XIa (0.4 nM) was added to '251-labeled Factor IX (2 nM) in a volume of 100 p1 containing 20 mM Tris-HC1, pH 7.4, 150 mM NaCl, 1 mM CaC1,. The reaction, performed in the presence or absence of Factor "(1-47) (2 p~) , was incubated at 37 "C, with aliquots removed a t various time intervals over 90 min. The reaction was stopped with the addition of Laemmli SDS gel loading buffer containing 10% p-mercaptoethanol and 10 mM EDTA. After the samples were subjected to electrophoresis in SDS gels, the gels were dried and exposed to Kodak X-Omat AR film. Autoradiographs were scanned for densitometric analysis using a n LKB UltroScan XL densitomer. The Effect of Factor IX-(147) on Factor X Activation by the Tenase Complex-The effect of Factor "(1-47) on Factor Xa generation was measured from the amidolytic activity of Factor Xa assayed using the chromogenic substrate CBS 31.39 (1.25 mM). The reaction mixtures, in a volume of 250 p1, were placed in microtiter plates in a buffer containing 20 mM Tris-HC1, pH 7.4, 150 mM NaCI, 0.1% bovine serum albumin, 5 mM CaC1,. The components of the reaction included human Factor IXa  Factor M-(1-47), at indicated concentrations between 1 and 25 p~, was added to the reaction mixture. The reaction was allowed to proceed at 25 "C for 18 min, and 25-pl aliquots were removed at 2-min intervals and placed in wells containing 10 m~ EDTA to stop the reaction. The amount of Factor Xa generated was determined using the chromogenic substrate CBS 31.39. CBS 31.39 (50 pl; 625 PM final concentration) was added to each sample in a 96-well plate, and the absorbance at 405 nm was determined over 10 min at 37 "C using a Thermomax kinetic enzyme-linked immunosorbent assay reader (Molecular Devices). From these data the rate of activation of Factor X was determined.
Phospholipid Binding Studies-The binding of the Factor "(1-47) peptide and the Factor "  peptide to phospholipid vesicles was evaluated by energy transfer measurements performed on a SLM 8000C fluorescence spectrophotometer. Fluorophores in the peptide were excited at 280 nm. When the peptide is in close proximity to a dansyl group covalently linked to phosphatidylethanolamine in the membrane, transfer of energy excites the dansyl group leading to emission at 520 nm. Because the eficiency of energy transfer has an inverse relation to the distance between the peptide and the dansyl group, the resulting dansyl emission is proportional to the amount of peptide at the membrane surface. Factor M peptides were prepared in 50 mM Tris, pH 7.4, 50 m~ NaCl, 1 mM CaCl,. Aliquots of the peptide were added to a fluorescence cuvette containing 3.0 ml of buffer with 3 p~ small unilammellar phospholipid vesicles. The sample was irradiated at 280 nm and emission monitored at 520 nm. Dilution effects were corrected using a buffer control. The reversibility of Ca(I1)-mediated peptidevesicle interaction was tested by the addition of 2 m~ EDTA.

RESULTS
Factor "(1-47) ( Fig. 1) was synthesized using FmoclNmethylpyrrolidone chemistry. All L-y-carboxyglutamic acid residues were coupled using 1 n" of protected amino acid per cartridge after activation. Following coupling of each Gla residue, a ninhydrin assay was performed to determine coupling eficiency prior to proceeding with the synthesis. Cleavage of the peptide from its solid support resin and simultaneous side chain deprotection was performed using trifluoroacetic acid. The crude deprotected peptide was purified by HPLC using a Hi-Pore 318 column monitored at 214 nm. The identity of the peak was confirmed by automated Edman degradation from residues 1 through 47. The purified peptide, containing two free sulfhydryl groups associated with cysteine 18 and cysteine 23, was oxidized by incubating the dilute peptide for 30 h at pH 8.5. The purified Factor M-(1-47) migrated as a single band in SDS gels following electrophoresis in the presence or absence of 10% P-mercaptoethanol (Fig. 2). The reduced form showed slightly reduced electrophoretic mobility. The amino acid analysis of the acid hydrolysate of Factor K(1-47) and of Factor IX41-42) yielded amino acid compositions within 10% of the theoretical expected results (Table I). Direct y-carboxyglutamic acid analyses of the alkaline hydrolysates were performed to ensure that no decarboxylation accompanied peptide synthesis, decoupling, or deprotection. Both peptides contained within 8% of the expected 12 residues of y-carboxyglutamic acid; no glutamic acid was observed. To determine the ability of the Factor M peptides to assume the metal-stabilized conformation characteristic of Factor E, the interaction of these peptides with conformation-specific an-   Background values were subtracted and the data plotted as the percent of inhibition of radiotracer binding versus the competitor concentration.
Factor IX to Factor Ma, we evaluated the effect of Factor M-(1-47) on this reaction. Factor IX was incubated with Factor XIa in the presence of 1 mM Ca(1I) in the presence or absence of Factor 1x41-47). The concentration of Factor IX-(1-47) was 2 PM, 1000-fold higher than that of Factor IX. The progress of the reaction was monitored over time by SDS-gel electrophoresis in the presence of reducing agent. The disappearance of Factor IX (Fig. 6) and the appearance of the light chain of Factor IXa (data not shown) were determined by quantitative densitometric analysis of the SDS gels. Under the conditions employed, the conversion of Factor IX to Factor IXa was nearly complete by 40 min. The presence of Factor IX41-471, even at high concentration, did not accelerate or decrease the rate of Factor IX activation (Fig. 6) or the rate of Factor IX light chain ap- aliquots removed at the indicated times and the reaction stopped with the addition of Laemmli SDS gel loading buffer containing 10% p-mercaptoethanol and 10 mM EDTA. After SDS-gel electrophoresis, the gels were dried and exposed to Kodak X-Omat AR film. Autoradiographs were quantitated by densitometry The disappearance of Factor M versus time reflects the rate of Factor IX activation.
pearance. These results suggest that Factor IX41-47) alone does not bind directly t o Factor XIa and interfere with substrate binding. It has been demonstrated previously that Factor "(1-43) inhibits the activation of Factor X by the complex of Factor Ma and Factor VI11 in the presence of Ca(I1) and phospholipid (28, 29). To determine whether Factor IX4-47) is also inhibitory in this reaction, we used the competition tenase assay system. The progress of the conversion of Factor X to Factor Xa was Factor IX-(1-47), at indicated concentrations, was added to the reaction mixture. The reaction was allowed to proceed at 25 "C for 18 min, and 25-pl aliquots were removed at 2-min intervals, and the amount of CBS 31.39. The Factor "(1-47) concentration is plotted uersus the Factor Xa generated was determined using the chromogenic substrate percent inhibition of Factor X activation.

DISCUSSION
To study the interaction of the Gla domain of Factor IX with phospholipid membranes, we have evaluated the properties and characteristics of synthetic peptides based upon the primary structure of Factor IX. Our goal was to identify the minimal region of Factor IX that retained the complete membrane binding properties that characterize Factor IX. We describe a high yield chemical synthesis of a 47 residue peptide that contains 12 residues of L-y-carboxyglutamic acid. Analysis of the purified peptide demonstrated that under the synthesis and purification conditions employed, the y-carboxyglutamic acid residues were incorporated quantitatively and did not undergo decarboxylation. This chemical synthesis, parallel to the synthesis of y-carboxyglutamic acid-containing peptides based upon the structure of Factor VI1 (30) and protein C (311, offers new opportunities for analysis of structure-function relationships in the vitamin K-dependent proteins. This approach provides a strategy to obtain large quantities of a homogeneous Gla-containing peptide of predetermined length and primary structure for purposes of structural and functional characterization. This has significant advantages over the use of proteolytic digestion products of the intact vitamin K-dependent proteins, and, particularly for Gla-containing peptides, expression of specific domains in heterologous recombinant systems. In the current work, we have established that Factor IX-(l-47) exhibits many of the membrane binding properties that characterize Factor IX. This peptide undergoes a conformational change induced by Ca(1I) ions as measured by both quenching of intrinsic fluorescence and the exposure of conformation-dependent antigenic determinants. Furthermore, this peptide bound reversibly to phospholipid vesicles composed of acidic phospholipids in the presence of Ca(I1). This peptide inhibited the enzymatic activation of Factor X by Factor IXa in a system that requires phospholipid membranes, but did not inhibit the activation of Factor IX by Factor XIa in a reaction that does not require phospholipid membranes. By these criteria, Factor IX41-47) has intrinsic calcium-dependent membrane-binding properties that parallel those of Factor IX. In contrast, Factor 1x41421, a peptide truncated within the aromatic amino acid stack domain, lacks these properties. Our results on Factor IX(1-42) confirm previous work in which bovine Factor 1x41421, prepared by proteolytic digestion of bovine Factor IX, did not bind t o phospholipid membranes in the presence of calcium ions (32). These results indicate that both the entire Gla and aromatic amino acid stack domains are required for membrane binding in the case of Factor IX. The propeptide and Gla domain of Factor M (residues -18 t o 38) are encoded by Exon 11, whereas the aromatic amino acid stack domain (residues 39-46) is encoded by Exon I11 (2,3). Thus, the phospholipid binding results of Factor IX-(1-47) indicate that the functional unit for phospholipid binding involves domains, the Gla domain and the aromatic amino acid stack domain, encoded by two separate exons. Previously published data on protein C-  have demonstrated more modestly reduced phospholipid binding solely by the Gla domain (31) encoded by Exon I1 of the protein C gene compared with the Gla-aromatic amino acid stack domain. Our results indicate that a functional lipid binding unit must include not only the Gla domain but also the aromatic amino acid stack domain to support calciummediated Factor IX-phospholipid binding. Whether the aromatic amino acid stack domain contributes residues for contact with the membrane or whether this region is critical to the proper folding and expression of other regions of the Gla domain that contact the membrane surface remains to be determined.
Previous study of the metal ion-induced conformational transitions in Factor IX that can be monitored by fluorescence quenching have shown that half-maximal quenching occurs a t about 100 PM Ca(I1) (33). Factor 1 x 4 1 4 7 ) binds to Ca(I1) and Mg(I1) ions, which induce the characteristic fluorescence quenching observed in the vitamin K-dependent proteins. Factor IX-(1-47) demonstrated half-maximal quenching at 50 JIM Ca(II), a somewhat lower concentration than observed for intact Factor IX. This likely represents the absence of conformational constraints on the Gldaromatic amino acid stack domains imposed by the EGF domain. Evidence exists for the interaction of the Gla domain and the first EGF domain or the combined effect of multiple conformational changes in distinct portions of Factor IX in response to calcium ions (34, 35); this interaction may be favored by a structure stabilized by the occupancy of metal binding sites by calcium ions and may contribute thermodynamically to the global energy of calcium binding and stabilization of the polypeptide backbone conformation, including calcium-dependent interaction between the Gla and EGF domains. The absence of these interactions in Factor 1 x 4 1 4 7 ) may be responsible for the lower Ca(1I) concentration for half-maximal fluorescence quenching. By comparison, protein C41-49) exhibited most of the calcium and membrane properties of intact protein C, but did not demonstrate calcium ion-induced fluorescence quenching (31). These authors suggest that the absence of fluorescence quenching does not preclude a metal-induced conformational transition, particularly when high levels of calcium ions are required to effect this transition in proteolytically derived peptides lacking part of the aromatic stack. It is important to emphasize that our Factor IX peptide, from residues 1 to 47, undergoes the conformational transition and the amount of calcium ion needed to effect this transition is lower than that required in the intact Factor IX. These may be intrinsic differences between Factor IX and protein C.
Anti-Factor IX:Ca(II)-specific antibodies and anti-Factor IX: Mg(I1) antibodies are polyclonal immunoaffinity-purified conformation-specific antibodies that recognize the conformational transition in Factor IX induced by metal ions (14). The location of the antigenic determinants against which these antibodies are directed has not been known, although studies on prothrombin have implicated residues 1 4 4 (36). However, the anti-Factor IX:Ca(II)-specific antibodies are known to inhibit binding of Factor IX to phospholipid membranes and thus are likely directed a t or near the phospholipid binding site on Factor IX. In the current experiments, we show that Factor IX-( 1-47) contains the antigenic determinants against which anti-Factor IX:Ca(II)-specific antibodies and anti-Factor IXMg(I1) antibodies are directed. The Ca(I1)-stabilized epitopes are expressed nearly quantitatively as compared directly with Factor IX. It would appear that Factor IX-(147) undergoes the Ca(I1)induced conformational change leading to the expression of these Ca(I1)-stabilized epitopes and that the conformational motility of this peptide favors a Factor "like native structure.
The activation of Factor IX by Factor XIa in the presence of calcium ions occurs independent of phospholipid membranes. Based upon the ability of anti-Factor IX:Ca(II)-specific antibodies to inhibit Factor IX-phospholipid binding and Factor IX activation by Factor XIa, we had previously proposed that the Gla domain of Factor IX might interact directly with Factor XIa, thus facilitating Factor IX cleavage (14). Based upon the current experiments in which Factor IX-(1-47) failed to inhibit Factor IX cleavage by Factor XIa, even at a molar excess of Factor IX-(1-47) over Factor IX of 1000-fold, this would appear not to be the case. However, we cannot rule out the possibility that the Gldaromatic amino acid stack domains and the EGF domains may be sufficient to inhibit this reaction. In contrast, Factor IX-(1-47) did inhibit the activation of Factor X by the tenase complex, a phospholipid-dependent reaction. Quantitative analysis of this inhibition indicates results parallel to those reported by Astermark et al. (28) on inhibition by bovine Factor 1x4143). The basis of this inhibition, which was observed by Astermark et al. (28) only when Factor VI11 was present but was independent of phospholipid (28), remains uncertain but may be a consequence of an interaction between the Gla domain of Factor IX with Factor VI11 (28). Similarly, the inhibition of this reaction by Factor VII-(1-49) (30) has been interpreted as resulting from a phospholipid-dependent interaction of the peptide with Factor IX. A model in which the phospholipid-dependent binding of Factor VIIa to Factor X is partially mediated by the Factor VIIa Gla domain is suggested (30). However, these potential interactions of the Gla domains of Factor IX and Factor VI1 must be taken within the context of the well established role of the Gla domains of the vitamin K-dependent proteins in facilitating the protein enzyme-substrate complexes required for blood coagulation.
Although it had been generally believed that the Gla domain and particularly individual Gla residues play identical roles in each of the vitamin K-dependent proteins, it is now becoming clear that this is not the case. For example, mutagenesis studies of the Gla residues of human prothrombin and human protein C suggest that not all Gla regions in different proteins work identically despite the marked sequence homology (36-40). Using the Factor IX numbering system, prothrombin and protein C have 3 or 4 Gla residues (Gla17, GlaZ7, Gla30, and possibly Gla26) which define an internal carboxylate core that chelates calcium that is not accessible to solvent; disruption of these residues destroys membrane binding activity and function. We suspect that these residues in Factor IX have an identical function. Gla7 is not essential for prothrombin or protein C, and based upon the naturally occurring but fully active mutant Factor IX Oxford 2 (Gla7 to Ala) (41), Gla7 is not important for Factor IX either. However, disruption of the other Gla residues in prothrombin and protein C have markedly different effects. For example, disruption of GlaI5 and GlaZ0 in prothrombin decreases functional activity but has no effect of the activity of protein C. Conversely, disruption of Gla8 or GlaZ1 in protein C eliminates activity, but the same mutation in prothrombin only partially inhibits function. Based upon these findings we have proposed (36) that some of the Gla residues may contact directly with the membrane surface, whereas others stabilize interdomain interaction between the Gla domain and the linking region adjacent to the first kringle domain. The differences in phospholipid and metal binding characteristics between Factor IX-(1-47) and protein C-(148) must be appreciated within this context. As discussed, protein C-(1-48) does not undergo calcium-induced fluorescence quenching, whereas Factor "(1-47) undergoes calcium-induced fluorescence quenching under conditions parallel to intact Factor IX. Moreover, protein C-(1-48) shows a phospholipid binding affinity that is only 3-fold greater than protein C-(1-38) (31). In contrast, Factor IX-(1-47) binds to phospholipids with high affinity, but Factor IX-(1-42) shows poor binding to phospholipid, and this binding is not reversible upon the addition of EDTA. In sum, these results serve to indicate that the structures of the Gla domains of the vitamin K-dependent proteins share many structural elements but that differences in structure reflect differences in function. Based upon the characterization of Factor IX-(l-471, this peptide but not Factor IX-(1-42) will serve as a model for the NMR study of the structure of the Factor IX Gla domain and its interaction with membrane surfaces. Recent results of twodimensional NMR experiments have led to the complete assignment of protons in Factor IX(1-47) in the absence of metal ions (42). Preliminary structural models of Factor IX(1-47) based upon the NMR constraints indicate the presence of a-helical structure and other local structures about the disulfide loop and the NH, terminus but that the structure is highly flexible and motile in the absence of Ca(I1). Further studies of this peptide in the presence of metal ions and of phospholipid membranes should provide insight into the structures that define the membrane-binding contact residues of this protein.