The polymerization pocket "a" within the carboxyl-terminal region of the gamma chain of human fibrinogen is adjacent to but independent from the calcium-binding site.

The carboxyl-terminal region of the gamma chain of fibrinogen is involved in calcium binding, fibrin polymerization, factor XIIIa-mediated cross-linking, and binding to the platelet fibrin(ogen) receptor. Protein fragments encoding amino acids Val143 to Val411 (rFbggammaC30) or Val143 to Leu427 (gamma'C30) from the carboxyl end of the gamma or gamma' chains, respectively, of human fibrinogen were expressed in yeast (Pichia pastoris) and characterized as to their cross-linking by factor XIIIa, polymerization pocket, and calcium-binding site. rFbggammaC30 and gamma'C30 were both readily cross-linked by factor XIIIa, but only rFbggammaC30 was capable of inhibiting thrombin-induced platelet aggregation. Two mutants, gammaC30-Q329R and gammaC30-D364A, which were based on the three-dimensional structure of the polymerization pocket within rFbggammaC30 and on information derived from naturally occurring mutant fibrinogens, were also expressed and characterized. rFbggammaC30 inhibited (desAA)fibrin polymerization in a dose-dependent manner, while the two mutant forms did not. Similarly, rFbggammaC30 and gamma'C30 were protected from plasmin degradation by the presence of Ca2+ or the peptide Gly-Pro-Arg-Pro, indicating that a functional Ca2+-binding site and polymerization pocket are contained within each of these fragments. The mutant fragments, however, were protected from plasmin only by metal ions, while no protective effect was conferred by GPRP or by any other peptide tested. These results indicate that the polymerization pocket "a", which binds the peptide GPRP, functions independently from the nearby calcium-binding site and that amino acids Gln329 and Asp364 play a crucial role in fibrin polymerization.

The carboxyl-terminal region of the ␥ chain of fibrinogen is involved in calcium binding, fibrin polymerization, factor XIIIa-mediated cross-linking, and binding to the platelet fibrin(ogen) receptor. Protein fragments encoding amino acids Val 143 to Val 411 (rFbg␥C30) or Val 143 to Leu 427 (␥C30) from the carboxyl end of the ␥ or ␥ chains, respectively, of human fibrinogen were expressed in yeast (Pichia pastoris) and characterized as to their cross-linking by factor XIIIa, polymerization pocket, and calcium-binding site. rFbg␥C30 and ␥C30 were both readily cross-linked by factor XIIIa, but only rFbg␥C30 was capable of inhibiting thrombin-induced platelet aggregation. Two mutants, ␥C30-Q329R and ␥C30-D364A, which were based on the three-dimensional structure of the polymerization pocket within rFbg␥C30 and on information derived from naturally occurring mutant fibrinogens, were also expressed and characterized. rFbg␥C30 inhibited (desAA)fibrin polymerization in a dose-dependent manner, while the two mutant forms did not. Similarly, rFbg␥C30 and ␥C30 were protected from plasmin degradation by the presence of Ca 2؉ or the peptide Gly-Pro-Arg-Pro, indicating that a functional Ca 2؉ -binding site and polymerization pocket are contained within each of these fragments. The mutant fragments, however, were protected from plasmin only by metal ions, while no protective effect was conferred by GPRP or by any other peptide tested. These results indicate that the polymerization pocket "a", which binds the peptide GPRP, functions independently from the nearby calcium-binding site and that amino acids Gln 329 and Asp 364 play a crucial role in fibrin polymerization.
Fibrinogen is a large glycoprotein composed of six polypeptide chains (␣, ␤, ␥) 2 held together by disulfide bonds. By electron microscopy, the molecule appears as a trinodular structure (1) in which the central nodule contains the amino termini of all six chains (2). The two distal nodules contain the carboxyl-terminal regions of the ␥ and ␤ chains. Upon cleavage by thrombin, fibrinopeptides A and B are released from the amino termini of the ␣ and ␤ chains, resulting in fibrin monomers that polymerize spontaneously. The newly exposed amino termini constitute polymerization sites A and B that are complementary to polymerization pockets "a" and "b", respectively (3)(4)(5)(6). The "a" pocket is located within the carboxyl region of the ␥ chain (7,8), while the location of the "b" pocket remains controversial. It has been proposed to arise upon the alignment of the D domains of two fibrin molecules (6), to be contained in the carboxyl-terminal region of the ␤ chain (9,10), and to involve the carboxyl-terminal region of the ␣ chain (11).
Calcium promotes the polymerization of fibrin monomers (12)(13)(14)(15) and the cross-linking of fibrin by factor XIIIa (16). It also protects fibrinogen fragment D against plasmic degradation (17). Fibrinogen binds three calcium ions per molecule (18,19), including one in the carboxyl-terminal region of each ␥ chain. The third calcium is present in the central nodule, provided that the ␣ chains are intact (20). The precise location of the calcium-binding site within the carboxyl end of the ␥ chain was described recently (21). The calcium ion is situated within the P domain of the molecule and is liganded by two water molecules, the side chain carboxyl groups of Asp 318 and Asp 320 , and the main chain carbonyl oxygens of Phe 322 and Gly 324 . The location of the third calcium within the fibrinogen molecule remains unclear. Binding studies using peptide analogs that mimic the new amino termini of the ␣ and ␤ chains of fibrin (GPRP and GHRP, respectively) have shown that calcium also modulates the binding of GHRP to fibrinogen but does not significantly affect that of GPRP (9,22). Since polymerization deficiencies have been observed in fibrinogens with defective calcium binding properties and vice versa (23), one could speculate that the two phenomena might be dependent on each other.
Apart from its role in calcium-binding and fibrin polymerization, the ␥ chain of fibrinogen is involved in factor XIIIamediated cross-linking of the fibrin polymers to form ␥-␥ dimers (24) and in binding to platelets through the cell surface receptor GPIIb/IIIa (25). The ␥ chain also mediates the binding of fibrin(ogen) to endothelial cells through the cell surface receptor ICAM-1 (26) and its binding to leukocytes through an interaction with the integrin MAC-1 (27).
Approximately 10% of the circulating fibrinogen in normal human blood contains a ␥Ј chain instead of the typical ␥ chain (28). This ␥Ј variant chain arises by alternative processing and polyadenylation of the mRNA for the ␥ chain (29,30). This creates a new carboxyl terminus for the ␥Ј chain in which the last 4 amino acids of the ␥ chain are replaced by a new sequence of 20 amino acids in the ␥Ј chain. Unlike the wild-type fibrinogen, recombinant homodimeric ␥Ј-fibrinogen does not support platelet aggregation (31), presumably because of an impaired interaction with the platelet fibrinogen receptor GPIIb/IIIa.
In the present study, the biological properties of the recombinant wild-type carboxyl-terminal region of the ␥ chain of human fibrinogen (Val 143 -Val 411 ) were compared with those of a similar fragment from the ␥Ј (Val 143 -Leu 427 ) chain and with those of two site-specific mutants (Q329R and D364A). It was found that all four recombinant ␥ species bound calcium but that only the wild-type ␥ and ␥Ј possessed a functional polymerization pocket. This indicates that the "GPRP-binding" polymerization pocket and the "calcium-binding" site are distinct and independent from one another.

Materials
The Pichia pastoris expression system was from Invitrogen (San Diego, CA). Human fibrinogen (plasminogen-free) and D-Phe-Pro-Arg chloromethylketone, HCl were from Calbiochem. Batroxobin (reagent grade) was purchased from American Diagnostica Inc. (Greenwich, CT). Pepstatin A was from Boehringer Mannheim. Apyrase (potato, grade V), heparin (porcine mucosal, grade I), prostaglandin E 1 , N-ethylmaleimide (NEM), 1 phenylmethylsulfonyl fluoride, biotin, and human plasmin were from Sigma. Yeast extract, peptone, casamino acids, and yeast nitrogen base (without amino acids) were all from Difco. Aprotinin was a gift from Novo Nordisk (Copenhagen, Denmark). Bovine serum albumin (Pentex, fraction V) was purchased from Miles (Kankakee, IL). Human thrombin was a gift from Dr. K. Fujikawa (University of Washington, Seattle, WA), and recombinant human factor XIII was graciously provided by Dr. P. Bishop (ZymoGenetics, Seattle, WA). Geneticin (G418 sulfate) was from Life Technologies, Inc. The chromogenic substrate S-2238 was purchased from Helena Laboratories (Beaumont, TX). The synthetic peptides GPRPamide, GPRP, GHRP, and GGYR were purchased from Sigma, while GPRVVER, GPRVVERH, and GPRVVERHQ were synthesized by the Peptide Synthesis Facility (Department of Pharmacology, University of Washington).

Protein Expression
Vector Construction-The human fibrinogen ␥ chain cDNA was isolated from pHI␥2 (32). The 5Ј-and 3Ј-ends were modified using polymerase chain reaction to create restriction sites suitable for subcloning into the pPIC9k vector (kindly provided by Dr. Michael Romanos, Wellcome Laboratories, UK). For ␥C30-Q329R and ␥C30-D364A, single amino acid substitutions were incorporated by polymerase chain reaction mutagenesis (33). For the ␥ЈC30 construct, the ␥Ј 3Ј-cDNA sequence was incorporated into the 3Ј-oligonucleotide used for polymerase chain reaction. All expression vector inserts were verified by DNA sequence analysis. The construct was transformed into the methylotrophic yeast P. pastoris (strain GS115 or SMD1168) by electroporation, according to the manufacturer's protocol (Invitrogen, San Diego, CA). Selected His ϩ colonies were screened further for high copy number by G418 selection as described by Scorer et al. (34). The yeast clone that grew fastest on G418 was chosen for protein expression.
Fragment D Preparation-Human fibrinogen was treated with 1 mM NEM and then dialyzed against 50 mM Tris-HCl, pH 7.4, 150 mM NaCl. The protein was then diluted to 4.5 mg/ml in 10 ml of the same buffer containing 5 mM CaCl 2 and subjected to plasmin digest (0.04 units/ml, final concentration) for 2 h at room temperature. The reaction was stopped by the addition of aprotinin (10 mg/ml final concentration). Fragment D was separated from fragment E by anion exchange chromatography on a Pharmacia Mono Q column in 20 mM Tris-HCl, pH 8.0, eluted with a NaCl gradient in the same buffer. The fractions containing fragment D were pooled, and the protein was precipitated with 80% ammonium sulfate. The protein pellet was resuspended in 1 ml of 50 mM Tris-HCl, pH 7.4, 150 mM NaCl and loaded onto a gel filtration column (Superdex 200, Pharmacia) equilibrated and developed in the same buffer. An extinction coefficient (E 0.1% ) of 2.0 and a molecular mass of 95 kDa were used for the determination of fragment D concentration.

Protein Characterization
All proteins were subjected to amino-terminal sequence analysis using an Applied Biosystems 447A protein sequencer. An extinction coefficient (E 0.1% ) of 2.2 was determined for rFbg␥C30 by amino acid analysis on a sample of pure protein of known A 280 . The same extinction coefficient was used for the other ␥ species. Mass spectroscopy was performed either by Dr. J. Hoffman (Zymogenetics, Seattle, WA) or at the UW Mass Spectroscopy Facility (Dept. of Biochemistry, University of Washington).

Fibrinogen Clotting Inhibition Assay
Human fibrinogen was treated with 1 mM NEM to inactivate any trace of FXIIIa, dialyzed against 50 mM Tris-HCl, pH 7.4, 150 mM NaCl. The clotting reaction, taking place in a cuvette, in 1 ml of the same buffer containing 1.0 M fibrinogen (0.34 mg/ml) and 5 mM CaCl 2 , was initiated by the addition of the snake venom protein batroxobin (0.001 units/ml, final concentration). The course of the reaction was followed by measuring the turbidity at 340 nm every 2 s for 20 min, using a 8452A diode array spectrophotometer (Hewlett Packard). Various amounts of rFbg␥C30, fragment D, bovine serum albumin, and mutant ␥C30 proteins were added to the reaction in the same buffer.

Fibrinopeptide A Assay
Fibrinopeptide A concentration was measured in the fibrinogen clot supernatants by competitive enzyme-linked immunoassay, using the Asserachrom FPA kit from Diagnostica Stago (Asnières-sur-Seine, France), according to the manufacturer's protocol.

Plasmic Protection Assay
The proteins were diluted to 1 mg/ml in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl. For the plasmin protection experiments using metal ions, the samples were treated with 0.5 mM EGTA to eliminate any trace of calcium from the protein preparation, after which 5 mM of CaCl 2 , MgCl 2 , ZnCl 2 , MnCl 2 , or TbCl 3 was added. For the plasmin protection experiments with peptides, the samples all contained 5 mM EDTA with and without 5 mM peptide (GPRPamide, GPRP, GHRP, GGYR (an unrelated peptide), GPRVVER, GPRVVERH, and GPRVVERHQ). The assays, performed at 37°C, were initiated by the addition of human plasmin (0.02 units/ml, final concentration). At various times, an aliquot was withdrawn, and the reaction was stopped by mixing with EDTA (10 mM) and SDS-PAGE loading buffer. The samples were then analyzed by SDS-PAGE (15% gel) stained with Coomassie Blue. The protein band corresponding to the undigested fragment was quantitated by scanning densitometry using an AlphaImager 2000 (Alpha Innotech Corp., San Leandro, CA). No digestion was observed when plasmin was omitted, whether in the presence of Ca 2ϩ or EDTA.

Determination of the Plasmin Cleavage Sites
rFbg␥C30 was diluted to 1 mg/ml in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA. Human plasmin was added at 0.02 units/ml, and aliquots were withdrawn at various times for SDS-PAGE and aminoterminal sequence analysis. For fragment D, the protein was diluted to 3 mg/ml in the same buffer, and plasmin was added at 0.2 units/ml. At 0, 1, and 16 h, aliquots were withdrawn for SDS-PAGE in duplicate. The fragments on one gel were stained with Coomassie Blue, and those on the other gel were transferred onto polyvinylidene difluoride membranes (Millipore) for amino-terminal sequencing.

Factor XIIIa-mediated Cross-linking
Factor XIII (1.5 mg/ml in 20 mM sodium borate buffer, pH 7.5, 3 mM CaCl 2 ) was activated with human thrombin (approximately 74 g/ml) for 5 min at room temperature. The reaction was stopped by the addition of D-Phe-Pro-Arg chloromethylketone, HCl at 5 mg/ml. No residual thrombin was detected by the hydrolysis of the chromogenic substrate S-2238. rFbg␥C30 was diluted at 7 mg/ml in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM CaCl 2 , and factor XIIIa was added at a final concentration of 0.1 mg/ml. The reaction proceeded at room temperature, with gentle rocking, for 20 h. An aliquot was withdrawn at various times and mixed with EDTA (10 mM) and SDS-PAGE loading buffer. The samples were analyzed on a 10% gel stained with Coomassie Blue.

Platelet Aggregation Inhibition Assay
Blood (50 ml) was freshly drawn from a healthy male volunteer and was anticoagulated with citrate. The platelets were prepared as described before (35). The platelet preparation was resuspended in a total volume of 1.5 ml. Platelet aggregation was measured by following the light transmittance using a PACKS-4 Platelet Aggregation Chromogenic Kinetic System (Helena Laboratories, Beaumont, TX). Each aggregation assay was performed at 37°C, in a final volume of 250 l, with stirring and contained 150 l of platelets, 5 mM CaCl 2 , and various amounts of rFbg␥C30 or ␥ЈC30 in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl. The reactions were initiated by the addition of 20 nM human thrombin.

RESULTS
The carboxyl-terminal region of the ␥ chain of human fibrinogen (rFbg␥C30), encompassing residues Val 143 -Val 411 , was expressed and secreted from the yeast P. pastoris at levels of approximately 100 mg/liter. The nonglycosylated recombinant fragment was purified by cation exchange chromatography and migrated as a single band when subjected to SDS-PAGE (Fig.  1A). Peak I shown in Fig. 1A was employed in all subsequent experiments with rFbg␥C30. The 30-kDa protein showed the expected amino-terminal amino acid sequence of VQIHDITG.
Mass spectroscopic analysis of rFbg␥C30 purified in the absence of protease inhibitors, however, revealed that it was partially degraded at the carboxyl terminus. This degradation occurred predominantly past His 400 . Subsequently, the addition of several protease inhibitors (phenylmethylsulfonyl fluoride, NEM, aprotinin, and pepstatin A) during the purification procedure allowed the isolation of full-length nonproteolyzed protein. The use of the carboxypeptidase-deficient yeast strain SMD1168 also appeared to improve the overall quality of the material, although expression levels were decreased 2-3-fold.
The carboxyl-terminal region of the ␥Ј chain, from Val 143 to Leu 427 (␥ЈC30), was also expressed and purified (Fig. 1B). In this case, the material in the second peak was used for characterization, since the material in the first peak appeared partially proteolyzed. Two ␥ chain fragments with specific mutations (␥C30-Q329R and ␥C30-D364A) were also expressed in the same system, at levels of ϳ50 -75 mg/ml. The elution profiles for the two mutant fragments were essentially identical to that of rFbg␥C30 (data not shown). The selection of the mutants was guided by naturally occurring mutations reported in the literature (23) and by the examination of the three-dimensional structure of the rFbg␥C30-GPRP complex (48). The first mutant was analogous to Fibrinogen Nagoya (36), in which glutamine 329 was replaced by arginine (Q329R). Aspartate 364 was changed to alanine (D364A) in the second mutant. This mutant was selected on the basis of the crystal structure of the complex of rFbg␥C30 with the peptide GPRP, since Asp 364 interacts strongly with the charged amino terminus of the GPRP peptide. The expression of a third mutant, ␥C30-D320S, was attempted using the same system. Based on the crystal structure (21), this mutation should have disrupted the calcium-binding site within rFbg␥C30, and it was of interest to determine how this would affect its binding to the GPRP peptide. Very low expression of the ␥C30-D320S mutant was detected; however, purification of the mutant proved impossible, probably due to heterogeneous folding of the fragment (data not shown).
The biological functions and characteristics of the various ␥C30s were determined by several different assays. Fragment D obtained by plasmin degradation of fibrinogen has been shown to inhibit fibrin polymerization (37). rFbg␥C30 is a structural unit within fragment D; therefore, the inhibition of clotting by the addition of rFbg␥C30 was examined to assess the folding of the molecule and its biological integrity. The inhibition of (desAA)fibrin polymerization by the various fragments (Fig. 2, A, B, C, and D) was assessed using two methods. The time required to reach half of the maximal turbidity was measured, since it is representative of the overall clotting reaction (Fig. 2E). The maximum slopes of the turbidity curves, which reflect polymerization plus lateral aggregation of the fibers (38), were also plotted as a function of the inhibitor concentration (Fig. 2F). The data show that both rFbg␥C30 and fragment D inhibited clotting in a dose-dependent manner. However, fragment D inhibited (desAA)fibrin polymerization approximately 3-8 times more effectively than did rFbg␥C30, depending on which parameters are compared. These results indicated that rFbg␥C30, like fragment D, contains a functional polymerization site. rFbg␥C30 hinders the linear elongation of protofibrils made of (desAA)fibrin monomers, most likely by binding to the amino termini of the ␣ chains of (desAA)fibrin. This in turn affects the lateral aggregation of the fibers. Neither ␥C30-Q329R nor ␥C30-D364A affected the reaction significantly, indicating the lack of a functional polymerization pocket within the mutants. The bovine serum albumin control did not affect clotting at any of the concentrations tested (0 -120 M). The amount of fibrinopeptide A (FpA) released in all of the clotting assays was similar and showed no significant variation (data not shown). These observations indicate that rFbg␥C30 does not affect FpA release directly and that the inhibition of clotting can be attributed to a specific blocking of the polymerization reaction.
Fragment D is protected against plasmin degradation by the presence of calcium ions (17) or a peptide resembling the amino terminus of the fibrin ␣ chain (GPRP) (39). Several metal ions and synthetic peptides were tested for their ability to protect rFbg␥C30, ␥ЈC30, and the mutants against plasmin digestion. Calcium ions bound to rFbg␥C30 and readily protected it from proteolytic degradation (Fig. 3A). The metal ions Zn 2ϩ , Tb 3ϩ , and Mn 2ϩ but not Mg 2ϩ also offered protection against plasmin degradation (data not shown). In the presence of EDTA, rFbg␥C30 was completely degraded in 2 h (Fig. 3B).
Several synthetic peptides were also assayed for their ability to protect rFbg␥C30 from plasmin digestion. Most of these synthetic peptides were modeled after the newly exposed amino terminus of the ␣ chain following the removal of FpA by thrombin. This region is also referred to as polymerization site A. The peptide GHRP mimics the amino terminus of the ␤ chain of fibrin, following removal of fibrinopeptide B by thrombin or polymerization site B. An unrelated peptide, GGYR, was employed as a control. The results indicated clearly that rFbg␥C30 and ␥ЈC30 bind the GPRP peptide in both its amide and carboxylate form and that the peptide-protein complexes were protected from plasmin digestion, even in the presence of EDTA (Fig. 4, A and B). In contrast, the ␤ chain-derived GHRP and the unrelated peptide GGYR offered no protection against plasmin. The longer peptides, which were identical to the amino-terminal sequence of the fibrin ␣ chain, delayed the degradation of rFbg␥C30 by plasmin, albeit considerably less efficiently than the GPRP peptides. Both of the ␥C30 mutants, Q329R and D364A, were protected by metal ions in the same manner as rFbg␥C30 (data shown for Ca 2ϩ only), but they were not protected against plasmin digestion by any of the peptides under these conditions (Fig. 4, C and D). These results were in agreement with the clotting inhibition assays and confirmed that rFbg␥C30 possesses a functional polymerization pocket that interacted with polymerization site A. Furthermore, this interaction was disrupted in the mutants ␥C30-Q329R and ␥C30-D364A. Interestingly, neither mutation appeared to affect significantly the binding of metal ions such as Ca 2ϩ . Earlier experiments showed that rFbg␥C30 binds Tb 3ϩ in the same site that binds Ca 2ϩ (21). The data presented here do not indicate whether the other metal ions interact with the molecule at the same site. It was established that the peptide GHRP does not protect rFbg␥C30 against plasmin degradation. The possibility, however, that the peptide interacted with the protein without affecting its sensitivity to plasmin cannot be eliminated.
The location of the plasmin cleavage sites within rFbg␥C30 was determined by amino-terminal sequence analysis and compared with the sites cleaved within fragment D under the same conditions. Fragment D was cleaved following residues Lys 302 , Lys 356 , and Lys 373 . In contrast, rFbg␥C30 was cleaved following residues Lys 212 , Lys 266 , Lys 356 , and Arg 275 , with no detectable hydrolysis at Lys 302 or Lys 373 . The order in which these cleavages occurred could not be determined.
The biological activities of rFbg␥C30 and ␥ЈC30 were also assessed by factor XIIIa-mediated cross-linking and by their ability to inhibit platelet aggregation. Both functions require an intact carboxyl terminus region of the ␥ chain. The time course of the cross-linking of rFbg␥C30 is shown in Fig. 5A. As the reaction proceeded, monomeric rFbg␥C30 disappeared as dimers and higher polymers of rFbg␥C30 were generated. With increasing time, the monomeric band became a tight doublet, with the intensity of the top band decreasing as that material was cross-linked while the bottom band persisted. This was probably due to the presence of some partially proteolyzed rFbg␥C30. Fig. 5B shows the same reaction with ␥ЈC30. The ␥Ј variant was also readily cross-linked by factor XIIIa, at a rate similar to that of rFbg␥C30. No cross-linking of either molecule occurred, however, in the absence of Ca 2ϩ (data not shown). The cross-linking was specific and was restricted to the carboxyl-terminal residues of the fragments, since no cross-linking was observed when material proteolyzed at the carboxyl terminus was used.
Finally, the ability of the platelet receptor to recognize rFbg␥C30 was determined by using a platelet aggregation assay. The aggregation of thrombin-activated platelets, monitored by the increase in light transmittance over time, was delayed in a dose-dependent manner by the addition of rFbg␥C30 (Fig. 6A). In contrast, the ␥ЈC30 mutant did not affect aggregation at any of the concentrations tested (Fig. 6B). These results indicated that the inhibition observed was due to competition of rFbg␥C30 with the endogenous fibrin(ogen) for binding to the platelet receptor and that this phenomenon was specific to the sequence at the carboxyl terminus of the ␥ chain.

DISCUSSION
The carboxyl-terminal region of the human fibrinogen ␥ chain has been expressed in yeast in a secreted form. This region of the molecule contains many functionally important features of fibrin(ogen). The structure of rFbg␥C30, as determined by x-ray diffraction, identified the location of the single calcium-binding site within this fragment (21). Subsequently, the three-dimensional structure of rFbg␥C30 complexed with the peptide GPRP showed the location of the polymerization pocket "a," as illustrated in Fig. 7. In that complex, the calcium- binding site, which is identical to that in the uncomplexed structure, and the GPRP-binding pocket are close to each other. However, the metal ion and the peptide do not share any amino acid ligands.
In the experiments reported here, rFbg␥C30 inhibited (desAA)fibrin polymerization without affecting FpA release. This confirmed that the 30-kDa protein was biologically functional in that it possessed a polymerization pocket. The inhibition of polymerization by rFbg␥C30, however, was less effective than that observed when equimolar amounts of human fibrinogen fragment D were used. Accordingly, it is not clear whether the polymerization pocket within rFbg␥C30 binds less tightly to (desAA)fibrin or whether other molecular interactions provided by the ␤ and ␣ chains within fragment D are responsible for this difference. Thus far, attempts to express the carboxylterminal region of the ␤ chain of human fibrinogen using this system have been unsuccessful. 2 Therefore, we have been unable to reconstitute a recombinant equivalent of fragment D for comparative studies.
Plasmin protection assays showed that both calcium and the GPRP peptides can protect rFbg␥C30 against degradation, further supporting the presence of a functional polymerization pocket within the recombinant fragment. The results obtained here indicate that the two sites are independent of each other, since binding of GPRP occurs even in the absence of calcium. This is consistent with the fact that the binding of either Ca 2ϩ or GPRP to rFbg␥C30 is sufficient to confer resistance to plasmin digestion.
The differences in plasmin cleavage patterns between rFbg␥C30 and fragment D suggest that the ␥ chain within fragment D is in a somewhat different environment than is rFbg␥C30. Clearly, the Lys 302 -Phe 303 peptide bond is accessible to plasmin hydrolysis in fragment D but not in rFbg␥C30, while Lys 266 -Val 267 and Arg 275 -Tyr 276 are exposed only in rFbg␥C30. The three-dimensional structure of rFbg␥C30 in the presence of Ca 2ϩ (21) shows that Lys 212 , Lys 266 , and Arg 275 are all exposed on the surface of the molecule and should be accessible to plasmin. Lys 212 and Arg 275 are involved in salt links to other amino acids. Presumably, these sites are protected within fragment D, either by the proximity of the ␣ or ␤ chain, by D:D contact, or else they are partially buried and therefore not accessible. Alternatively, they may become accessible to plasmin as a result of cleavages at other sites. The side chain of Lys 302 in rFbg␥C30 is projecting into the solvent, but the backbone of the peptide bond is partially hidden under the side chain of Phe 303 , which is also exposed to solvent. This arrangement may explain the lack of cleavage observed here. The aromatic side chains of Phe 303 and Phe 304 are on the surface of the rFbg␥C30 molecule. The same residues in fragment D may be involved in a molecular interaction not seen in the recombinant fragment, that would expose the Lys 302 -Phe 303 peptide bond to hydrolysis. Lys 302 and Phe 303 are part of a ␤ turn that is adjacent to polymerization pocket "a". Similarly, the side chain of rFbg␥C30 Lys 373 is involved in a salt link to Asp 252 , which may render this site unrecognizable as a substrate for plasmin. Comparison of the three-dimensional structure of the human fibrinogen fragment D (86 kDa) (40,41) with that of rFbg␥C30 should soon provide more information on these differences.
The polymerization pocket of rFbg␥C30 was also assessed for its ability to bind several peptides modeled after the amino termini of the ␣ and ␤ chains of fibrin. The short peptides GPRPamide and GPRP were as efficient as calcium ions in protecting rFbg␥C30 against plasmin degradation. The GHRP peptide mimicking the amino terminus of the fibrin ␤ chain offered no protection at all in the absence of Ca 2ϩ . This is in agreement with results obtained by others with fibrinogen (39,42) or (desAA)fibrin monomers (15). The longer peptides (encoding the true sequence of the ␣ chain, as opposed to the peptide analog GPRP, which binds to fibrinogen more tightly than does the wild-type sequence) were markedly less protective than was GPRP. Earlier studies had shown that GPRP bound human fibrinogen more tightly than GPR, GPRV, or GHRP (43).
These differences are difficult to explain based on the structure of the rFbg␥C30-GPRP complex structure alone. It appears that the GHRP and GPRVVER peptides could be modeled into the polymerization pocket, although the hydrogen bond network at this site would be altered. We cannot establish, from these data, whether or not GHRP actually bound to rFbg␥C30. However, if GHRP did interact directly with rFbg␥C30, it did not elicit the conformational change that results in increased resistance to proteolysis by plasmin. Clearly, fibrin polymerization involves more than the initial binding between the GPR amino terminus of the fibrin ␣ chains and the carboxyl-terminal regions of ␥ chains in adjacent fibrin molecules.
The two mutants, ␥C30-Q329R and ␥C30-D364A, were protected by Ca 2ϩ against plasmin degradation, but neither mutant was protected by any of the peptides tested. In the uncomplexed rFbg␥C30 structure, the Gln 329 side chain is hydrogenbonded to two water molecules in the polymerization pocket. When GPRP binds, the side chain of Gln 329 shifts to accommodate the peptide arginine (Fig. 7). The substitution of glutamine by arginine at this site would add a bulky arginine side chain to the polymerization pocket and should preclude binding of the GPRP peptide. This would explain the observed impairment of in vitro fibrin polymerization caused by this substitution in Fibrinogen Nagoya, although the heterozygous individual has no history of hemorrhage nor thrombosis (36). Similarly, the side chain of Asp 364 forms hydrogen bonds and a salt link to the side chain of Arg 375 in rFbg␥C30. Upon complex formation, the side chain of Asp 364 forms a strong salt link with the charged amino group of the peptide glycine residue. The replacement of Asp 364 by alanine abolished the ability of ␥C30-D634A to form a complex with GPRP, as anticipated. Because we were unable to study the ␥C30-D320S mutant, which should have been defective in calcium binding, we cannot determine how this mutation would affect the binding of GPRP to the polymerization pocket. However, disruption of the calciumbinding site by mutation may well result in the improper folding of the molecule, which in turn could result in impaired fibrin polymerization. During the course of this study, the molecular defects of Fibrinogen Melun (44) and Fibrinogen Matsumoto (45) were reported as the replacement of aspartate 364 by valine and histidine, respectively. The D364V mutation was reported to cause deep and superficial venous thrombosis within the affected family, by an as yet undetermined mechanism, while the D364H mutation caused delayed fibrin polymerization without any bleeding or thrombotic tendencies. Efforts to determine the structure of ␥C30-Q329R and ␥C30-D364A by x-ray crystallography are currently under way.
rFbg␥C30 was shown to act as a substrate for the transglutaminase factor XIIIa. It should also serve as a suitable surrogate substrate for the study of factor XIIIa cross-linking and of the factors influencing this reaction. The ability of rFbg␥C30 to interfere with thrombin-triggered platelet aggregation indicated that the protein is recognized by the platelet fibrinogen receptor GPIIb/IIIa and that it effectively blocks further interaction of the receptor with fibrin(ogen). The variant ␥ЈC30, however, did not bind the platelet receptor, which is in agreement with the results obtained by Farrell et al. (31).
The three-dimensional structure of the rFbg␥C30-GPRP complex (48), showed that the Ca 2ϩ ion lies approximately 9 Å away from the polymerization pocket and does not share ligands with the peptide. Results presented here demonstrated that the polymerization pocket "a" within the ␥ chain of fibrin-(ogen) is functionally independent from the Ca 2ϩ -binding site and that occupancy of one site or the other is sufficient to induce a conformational change that results in resistance to plasmin degradation. These results are in agreement with a recent study that showed normal calcium binding properties for five mutant fibrinogens with abnormal fibrin polymerization (49).
Addendum-Shortly after submission of this manuscript, the study of a similar fibrinogen fragment was reported by Medved et al. (47). The results obtained with the fragment expressed in Escherichia coli and refolded in vitro are in agreement with those described here.