Recombinant Fibrinogen Vlissingen/Frankfurt IV

We synthesized a variant, recombinant fibrinogen modeled after the heterozygous dysfibrinogen Vlissingen/Frankfurt IV, a deletion of two residues, γAsn-319 and γAsp-320, located within the high affinity calcium-binding pocket. Turbidity studies showed no evidence of fibrin polymerization, although size exclusion chromatography, transmission electron microscopy, and dynamic light scattering studies showed small aggregates. These aggregates did not resemble normal protofibrils nor did they clot. Fibrinopeptide A release was normal, whereas fibrinopeptide B release was delayed approximately 3-fold. Plasmin cleavage of this fibrinogen was not changed by the presence of calcium or Gly-Pro-Arg-Pro, indicating that both the calcium-binding site and the “a” polymerization site were non-functional. We conclude that the loss of normal polymerization was due to the lack of “A-a” interactions. Moreover, functions associated with the C-terminal end of the γ chain, such as platelet aggregation and factor XIII cross-linking, were also disrupted, suggesting that this deletion of two residues affected the overall structure of the C-terminal domain of the γ chain.

Fibrinogen is a 340-kDa glycoprotein that consists of six polypeptides chains, (A␣, B␤, and ␥) 2 . The molecule is symmetric, consisting of two globular D nodules connected to a central E nodule by a coiled-coil region (1). The D nodule consists of a portion of the coiled-coil region and the C termini of both the B␤ and ␥ chains, whereas the E nodule consists of the N termini of all six polypeptides. Thrombin cleaves the N termini of both the A␣ and the B␤ chains, releasing fibrinopeptides A (FpA) 1 and B (FpB) and producing fibrin monomers that polymerize to form a fibrin clot. FpA is released first, exposing the polymerization site "A" that binds to the polymerization site "a" in the ␥ chain in the D domain of another molecule (2). This "A-a" interaction allows molecules to align into double-stranded, half-staggered protofibrils (3,4). It is unclear by what molecular mechanism FpB release affects polymerization, although a commonly accepted model proposes that FpB is released from growing protofibrils, enhancing the assembly of protofibrils into fibers (5)(6)(7). Release of FpB exposes the "B" polymerization site that binds to the "b" site (2), which may be located in the B␤ chain (8,9) or the A␣ chain (10). In vivo, the result of fibrin polymerization is an intricate fibrous network that, along with platelets, forms a hemostatic plug (11).
X-ray crystallography studies have defined several important residues of the C-terminal domain of the ␥ chain. When the peptide Gly-Pro-Arg-Pro (GPRP), which mimics the "A" site, was co-crystallized with either two covalently linked D nodules (12) or the ␥ C-terminal fragment of the D nodule (13), side chains from amino acids Gln-329, Asp-330, and Asp-364 were shown to interact with the peptide, forming the "a" site. Two of the three high affinity calcium-binding sites in fibrinogen (14,15) are in the individual C termini of the ␥ chains. These ␥ chain calcium-binding sites lie adjacent to the polymerization pocket "a" (12,13,16). The specific residues that bind calcium are Asp-318, Asp-320, Phe-322, and Gly-324 (16). The structure of the last 14 residues of the C terminus of the ␥ chain (␥398 -␥411) has been difficult to resolve. Crystallography of a 30-kDa fragment of ␥ chain provided information through residue 401 (16); the last residue resolved in the structure of the 80-kDa fragment D was 398 (12). Nonetheless, biochemical studies have shown that the sites necessary for factor XIII (FXIII) ␥ chain cross-linking and platelet aggregation reside in these C-terminal residues. Specifically, FXIII forms a covalent bond between ␥Gln-398 on one molecule and ␥Lys-406 from another molecule (17). The last four residues, ␥408 -411, are required for platelet aggregation (18).
We have synthesized the recombinant fibrinogen, ␥⌬319,320, modeled after the heterozygous dysfibrinogen Vlissingen/ Frankfurt IV (V/F IV), in which residues ␥Asn-319 and ␥Asp-320 are absent. We found that calcium binding, fibrin polymerization, and platelet aggregation were completely disrupted with the recombinant variant, in contrast to the plasma V/F IV fibrinogen, where functions were partially impaired (19). These results suggest that the overall structure of the ␥ C-terminal fragment has been altered. Medical Center, Milwaukee, WI). Plasminogen was purified from human plasma by the method previously described (20) and stored at Ϫ20°C. Streptokinase was purchased from American Diagnostica (Greenwich, CT).

Materials-All
Construction of the Vector and Expression of the Protein-The expression vector pMLP-␥ contains the entire ␥ chain cDNA and has been described (21). The 6 base pairs coding for Asn-319 and Asp-320 were deleted by oligo-directed mutagenesis using the Transformer TM sitedirected mutagenesis kit from CLONTECH Laboratories (Palo Alto, CA) based on the procedure described by Deng et al. (22). The primers used to construct the mutation have been described (23). Chinese hamster ovary cells, which expressed the human A␣ and B␤ fibrinogen chains, were co-transfected with the mutant ␥ chain expression vector and a plasmid containing the selection gene, pMSV his (see Ref. 21 for details). Individual clones, secreting high amounts of fibrinogen as determined by enzyme-linked immunosorbent assay, were grown in serum-free medium in roller bottles for large scale protein expression. The medium was harvested; phenylmethylsulfonyl fluoride was added, and the medium was stored at Ϫ20 or Ϫ70°C.
Purification of Recombinant Fibrinogen-Both ␥⌬319,320 and normal recombinant fibrinogen (24) were purified by ammonium sulfate precipitation and immunoaffinity chromatography using IF-1 monoclonal antibody (Iatron Corp., Tokyo, Japan) conjugated to Sepharose 4B (see Refs. 24 and 25 for details). Proteins were dialyzed against 20 mM HEPES, pH 7.4, 150 mM NaCl (HEPES buffer), aliquoted, and stored at Ϫ70°C. To test the purity and assembly of the recombinant fibrinogen, proteins were analyzed by SDS-PAGE under reduced and non-reduced conditions by the method of Laemmli (26).
Size Exclusion Chromatography-Reactions were initiated by adding 10 l of thrombin (4 units/ml) to 90 l of fibrinogen (0.45 mg/ml). At specified time points, 1 l of hirudin (2,000 units/ml) was added to inhibit thrombin. Samples were immediately injected onto an Amersham Pharmacia Biotech HR 10/30 column packed with Superose 6 resin according to the manufacturer's instructions; this resin has a separation range from 5 to 5,000 kDa. The column was equilibrated with the reaction buffer (HEPES buffer containing 1 mM CaCl 2 ). The products were eluted under constant flow at 1 ml/min and monitored at 280 nm. The void volume was determined with blue dextran 2,000 (Amersham Pharmacia Biotech).
Plasmin Protection Assay and Western Blot-Fibrinogen (0.1 mg/ml) was preincubated at ambient temperature for 15 min with EDTA (5 mM), CaCl 2 (5 mM), or GPRP (2 mM). Meanwhile, 100 l of plasminogen (0.11 mg/ml) was activated with 10 l of streptokinase (10,000 units/ml) for 5 min, and 2 l was added to the fibrinogen samples (100 l) and incubated overnight at 37°C. After incubation, the samples were run on a 7.5% SDS-PAGE gel, under non-reduced conditions, as described by Laemmli (26). The gel was electroblotted in 50 mM Tris, 380 mM glycine, 0.1% SDS, 20% methanol at 150 mA for 1.5 h to a nitrocellulose membrane (Micron Separations Inc, Westborough, MA). The membrane was blocked overnight at 4°C in TBS (20 mM Tris-HCl, pH 7.4, 150 mM NaCl) with 5% non-fat milk. Then it was incubated with a 1:10,000 dilution of rabbit polyclonal antiserum to fibrinogen (Dako, Carpenteria, CA) for 2 h in TBST (TBS with 0.05% Tween 20 (Aldrich)), containing 1% bovine serum albumin. The membrane was next washed three times with TBST and incubated with a 1:5000 dilution of goat antirabbit IgG conjugated to horseradish peroxidase (Calbiochem) for 1 h in TBST with 1% bovine serum albumin. The blot was washed three times and developed with ECL TM Western blotting Detection Reagent (Amersham Pharmacia Biotech).
Thrombin-catalyzed Fibrinopeptide Release-Reactions initiated by adding thrombin (either 0.01 or 0.005 units/ml) to fibrinogen (0.1 mg/ ml) in HEPES buffer containing 1 mM CaCl 2 were performed as described (27). Immediately after thrombin addition, the solutions were aliquoted, and at specified time points, the samples were boiled for 15 min to stop the reaction. The samples were centrifuged for 15 min at 13,000 ϫ g, and the supernatants were stored at Ϫ70°C. The amount of FpA and FpB in the supernatants was monitored by reverse phase high performance liquid chromatography. The FpA and FpB peak areas were calculated, and the data were fit to two successive first-order rate equations, and specificity constants, k cat /K m , were calculated as described (27). The means were compared by an unpaired t test using Statview (Abacus Concepts, Berkeley, CA).
Transmission Electron Microscopy-Reactions were initiated by adding 3 l of thrombin (4 units/ml) to 27 l of fibrinogen (0.45 mg/ml, HEPES buffer containing 1 mM CaCl 2 ). At specified time points, 5 l of the reaction was added to tubes containing 5 l of hirudin (4 units/ml). These samples were immediately diluted in 90 l of 50 mM ammonium formate, 70% glycerol, pH 7.4, and sprayed onto freshly cleaved mica. The samples were dried and rotary-shadowed with tungsten in a Denton Vacuum DV-502, as described previously (28 -30). The shadowed protein films were picked up on 400 mesh copper grids and viewed on a LEO EM 910 transmission electron microscope.
Isolation of Human Platelets and ADP-induced Aggregation-The human platelets were isolated by the method previously described (18). Briefly, human blood was drawn into citrate, and the platelet-rich plasma was obtained by centrifugation at 800 rpm for 20 min. Prostaglandin E-1 (final concentration 20 nM) was added, and the sample was centrifuged again at 1800 rpm for 20 min. The pellet containing platelets was resuspended in Tyrode's buffer, pH 7.2 (10 mM HEPES, 135 mM NaCl, 2.7 mM KCl, 12 mM NaHCO 3 , 5.5 mM glucose, 2% bovine serum albumin), and loaded onto a Sepharose CL-2B column. The platelets were eluted, and CaCl 2 was added to 1 mM and MgCl 2 added to 2 mM. For aggregation, 2 ϫ 10 8 platelets/ml were preincubated at 37°C with 250 nM fibrinogen, prior to activation with 10 M ADP. Aggregation of stirred solutions was monitored with a platelet aggregometer (Chrono-Log) as the increase in light transmission. Normal and ␥⌬319,320 fibrinogens were assayed three times with one platelet donor.
FXIII Fibrin and Fibrinogen Cross-linking and Western Blot-For fibrin cross-linking, each reaction contained 16 l of fibrinogen (0.375 mg/ml), 2 l of FXIII (0.04 mg/ml), and 2 l of thrombin (3 units/ml) in HEPES buffer containing 1 mM CaCl 2 . After incubation at ambient temperature, the reaction was stopped with the addition of reducing buffer followed by boiling. For ␥⌬319,320 fibrinogen, an additional set of experiments was performed where 16 l of fibrinogen (0.375 mg/ml) was first incubated with 2 l of thrombin (3 units/ml) for 4 h. Next, 2 l of FXIII (40 units/ml) and 1 l of thrombin (3 units/ml) were added and incubated for the specified times. For fibrinogen cross-linking, 18 l of FXIII (0.67 mg/ml) and 18 l of thrombin (20 units/ml) were mixed in the presence of HEPES buffer containing 1 mM CaCl 2 and incubated at 37°C for 35 min to activate the FXIII. Thrombin was inhibited by incubating with 12 l of hirudin (450 units/ml) at ambient temperature for 1 min. Four l of the FXIIIa mix was added to 16 l fibrinogen (0.375 mg/ml) and incubated at ambient temperature for the specified times. All cross-linked products were analyzed on 7.5% reduced gels, electrophoresed, and blotted as described for the plasmin protection assay.
Dynamic Light Scattering-The variant ␥⌬319,320 fibrinogen was further purified by gel filtration chromatography with an Amersham Pharmacia Biotech 10/50 HR column packed with Superose 6B resin. The elution buffer was HEPES buffer, and the flow rate was 1 ml/min. Peak fractions were collected, concentrated using a Millipore Ultra-free® centrifugal filter device (Millipore Corp., Bedford, MA), and used for polymerization. All fibrinogen solutions were filtered using a 0.2-m Millipore filter before use to remove dust. Light-scattering intensity versus time measurements were performed using a 35-milliwatt He-Ne laser (model 120, Spectra Physics Inc., Mountain View, CA), a redsensitive photomultiplier tube (R298P, Hamatsu, Middlesex, NJ), and a Brookhaven Instruments BI200 SM goniometer at a 90 o angle (31,32). For dynamic light scattering, a 72-channel digital correlator (BI-2030 AT, Brookhaven Instruments) was used with a "multiple sample times" option to divide each channel into four distinct real-time bases, allowing for an expanded decay time collection range. Data were acquired using the program BI-30ATN, and the resultant auto-correlation functions were analyzed employing cumulants analysis to obtain z-average translational diffusion coefficients (33)(34)(35) and the CONTIN algorithm to obtain size distributions (36). Polymerizations were performed at 0.2 mg/ml fibrinogen, 0.3 units/ml thrombin in HEPES buffer. Time-averaged intensities were taken incrementally for 4 h.

Characterization of the Purified ␥⌬319,320 Fibrinogen
The recombinant fibrinogen ␥⌬319,320 was synthesized and purified as described under "Experimental Procedures." SDS-PAGE performed under non-reduced conditions demonstrated that the variant fibrinogen was pure and that it was assembled into the 340-kDa molecule (data not shown). A reduced gel showed that all three chains were produced and were intact (Fig. 1). As previously demonstrated with the plasma fibrino-gen V/F IV (19), the variant ␥ chain (lane 2) had a different mobility than the normal ␥ chain (lane 1). The difference in mobility was clearly seen in lane 3, with a 1:1 mix of normal and variant ␥ chain.
"A-a" Interactions Are Missing in ␥⌬319,320 Fibrin Turbidity Studies-Thrombin-catalyzed polymerization was monitored as turbidity at 350 nm. Under conditions where normal fibrinogen polymerized rapidly, ␥⌬319,320 fibrinogen did not show a rise in turbidity (Fig. 2). Additional experiments (data not shown) were performed under varied conditions that augment polymerization. Nevertheless, no turbidity change was detected with ␥⌬319,320 fibrinogen, even when the thrombin concentration was increased to 10 units/ml, the fibrinogen concentration was increased to 1.3 mg/ml, the temperature was lowered to 14°C, the duration of the experiment was extended to 14 h, or the calcium concentration was raised to 10 mM.
Fibrinopeptide Release-To test whether the polymerization curves resulted from a blocked "A" site, we measured fibrinopeptide A and B release from normal and ␥⌬319,320 fibrinogen over a 2-h time course. The data were fit to two successive first-order rate equations under the assumption that FpA was released before FpB (37). Fig. 3 shows that indeed both FpA and FpB were released from ␥⌬319,320 fibrinogen, but the release of FpB was delayed relative to normal fibrinogen. The kinetic data presented in Table I show that FpA was released at a normal rate, whereas FpB was released at a rate about 3-fold slower than normal. The delayed release of FpB indicates that protofibril formation was impaired, as the release of FpB is reported to be accelerated by protofibril formation (5).
GPRP Protection-To determine whether the "a" site was functional in ␥⌬319,320 fibrinogen, we examined plasmin cleavage in the presence of GPRP, which mimics the "A" site (38). After incubating ␥⌬319,320 or normal fibrinogen with GPRP, we performed a plasmin digest as described under "Experimental Procedures." As shown in Fig. 4, normal fibrinogen was protected from plasmin cleavage by the presence of GPRP (lane 3), as the product labeled D1 remained. Normal fibrinogen was also protected from cleavage in the presence of 5 mM CaCl 2 (lane 4), as described previously (39). In the absence of calcium or GPRP (lane 1), plasmin cleavage was demonstrated by the loss of the D1 product and the appearance of D2 and D3, as described previously (39,40). We examined plasmin cleavage of ␥⌬319,320 fibrinogen under all three conditions and found no protection from plasmin cleavage as evidenced by the D2 and D3 products (lanes 6 -8). Because the two-residue de-letion removed part of the high affinity calcium-binding site (13,16), it was not surprising that calcium provided no protection to ␥⌬319,320 fibrinogen (lane 8), similar to data obtained with plasma V/F IV (19). We conclude from the plasmin cleavage profile that GPRP did not bind to ␥⌬319,320 fibrinogen.
␥⌬319,320 Fibrin Monomers Interact to Form Small Aggregates Size Exclusion Chromatography-We further examined the polymerization of ␥⌬319,320 fibrin by size exclusion chromatography to determine whether products larger than fibrin monomers were formed (Fig. 5). At various time points, thrombin was inhibited, and the reaction products were immediately run on a size exclusion column.
When thrombin was added to normal fibrinogen, polymerization was evident almost instantly. A peak at the void volume was seen as early as 28 s (data not shown), and a small monomer peak was still evident at 1 min (Fig. 5, panel B). After 3 min, a monomer peak was not detected (panel C), and only a small peak at the void volume was seen. With ␥⌬319,320 fibrinogen, the result was quite different. At 3 min (panel E), most of the protein was still in the monomer peak. A peak at the void volume was not seen until 30 min (data not shown), and after 240 min (panel F), the monomer peak was still evident. The void volume peak was not present when ␥⌬319,320 fibrinogen was incubated with inhibited thrombin for 240 min (data not shown), ensuring that the formation of high molecular components was thrombin-dependent.
Finally, we observed that the size distribution of larger species was different between the normal and ␥⌬319,320 fibrin. For example, at 1 min, normal fibrinogen had two distinctive fractions represented by two peaks with little material between the two peaks. In contrast, at 240 min, ␥⌬319,320 fibrin had two major peaks, but there was also evidence of a broad spectrum of material between these two peaks. We concluded that ␥⌬319,320 fibrin is capable of forming higher molecular weight species, although the nature of this material could not be determined by size exclusion chromatography.
Transmission Electron Microscopy-To characterize this high molecular weight material, we used transmission electron microscopy. We performed reactions as for size exclusion chromatography, inhibited thrombin at specific times, and immediately diluted and sprayed the solutions onto mica. The samples were visualized by rotary shadowing with tungsten ( Fig.  6). Under these conditions, normal fibrin became a gel within 1 min, such that at 1 min and thereafter only the unclottable/ soluble fraction of normal fibrin was visible by transmission electron microscopy, and the microscopic fields were sparsely populated with molecules (Fig. 6, panel C). We limited our analysis of normal fibrin to time points Յ1 min, and we assumed that the discernible fibrin monomers and aggregates were those that did not incorporate into the clot. In contrast, under these conditions ␥⌬319,320 fibrin did not form a gel, and over the 4-h incubation, microscopic fields remained densely populated with molecules (Fig. 6, panel D).
As can be seen in Fig. 6, the ␥⌬319,320 fibrinogen molecules (panel B) appeared the same as normal fibrinogen molecules (panel A); most ␥⌬319,320 fibrinogen molecules appeared trinodular, and the majority of molecules were monomeric, and there were no obvious abnormal structures. Following incubation with thrombin, both normal and ␥⌬319,320 fibrin demonstrated a shift in the population of molecules seen. We analyzed many fields, counted the number of molecules, and scored these as monomers, pairs, or aggregates (Table II). Paired molecules were defined as two molecules touching each other in any orientation, whereas aggregates were defined as a group of three molecules or more. The data were expressed as a percentage of molecules (of total counted) in monomers, pairs, or aggregates, where, for example, the number of molecules counted as pairs was twice the number of pairs. Aggregates from both normal and ␥⌬319,320 fibrin were scored as being either elongated, indicated by an arrowhead, or irregular-shaped, indicated by an asterisk, in Fig. 6, panels E and F. Often, it was difficult to discern if a nodule belonged to one molecule or another, so the number of monomers per aggregate was based on an estimate of the total number of nodules seen (one molecule is tri-nodular). Consistent with the chromatography data, after a 4-h incubation, the fraction of ␥⌬319,320 fibrin monomers decreased and the fraction of oligomers increased. Although there were no striking differences between normal and ␥⌬319,320 aggregates, analysis of the electron microscopy data showed that more elongated aggregates were seen with ␥⌬319,320 fibrin than in normal fibrin (Table II). From the microscopy data, we conclude that after prolonged incubation ␥⌬319,320 fibrin formed aggregates, that these aggregates were small (on average, four or fewer monomers), and that a significant fraction of aggregates (about 40% at 4 h) were elongated as might be expected for early polymerization.
Dynamic Light Scattering-Because both size exclusion chromatography and transmission electron microscopy only examined the molecules/aggregates that remained soluble, we also examined polymerization by dynamic light scattering which analyzed all products in a closed system. This method provided time-correlated molecular mass and diffusion coefficient data, which we compared with corresponding data for normal fibrinogen (41). Under the conditions of our assay, within 20 s, analysis of the data with normal fibrin would show a 6 -10-fold increase in molecular mass and a 10-fold drop in the diffusion coefficient, consistent with formation of elongated, rod-shaped structures, that is, protofibrils (41). When we examined ␥⌬319,320 fibrin by dynamic light scattering (Fig. 7), the results were significantly different from normal. The molecular mass increased only 2-fold, on average, and the diffusion coefficient dropped about 1.5-fold after 4 h. These data were consistent with the formation of small aggregates, consisting on average of 3.9 Ϯ 2.5 fibrin monomers per aggregate. We conclude that the light scattering data and the results seen by electron microscopy demonstrated that only small aggregates of fibrin were formed.

Distant Functional Sites in Fibrinogen Are Affected by the ␥⌬319,320 Deletion
Fibrin Cross-linking-We also investigated other ␥ chain coagulation functions, including fibrin cross-linking by the transglutaminase, FXIII. During polymerization, factor XIII is activated by thrombin to FXIIIa, which catalyzes the formation of a covalent bond between Gln and Lys of two ␥ chains (17) or two or more ␣ chains (42). We examined cross-link formation by incubating fibrinogen, thrombin, and FXIII, and analyzing the products on immunoblots from SDS gels run under reducing conditions. As shown in Fig. 8, with normal fibrinogen, the formation of ␥-␥ dimers was essentially complete after a 5-min incubation (lane 2). With longer incubation, the ␥ chain completely disappeared as ␥-␥ dimers formed, and the ␣ chain disappeared, indicating the formation of ␣-polymers. With ␥⌬319,320 fibrinogen, cross-linking was clearly impaired. After a 4-h incubation, a faint ␥-␥ dimer band was obvious as was a decrease in both the ␣ and ␥ chains (lane 10). After 24 h (lane 16), the ␥-␥ dimer band was still faint, and the ␥ chain band had faded, but the ␣ chain had completely disappeared. Because we might expect fibrin cross-linking to be impaired with monomers that did not polymerize, we also examined crosslinking after ␥⌬319,320 fibrinogen was incubated first with thrombin for 4 h, followed by addition of FXIII (lanes 11-15). The immunoblot demonstrated that cross-linking of ␥⌬319,320 fibrin was somewhat enhanced by prior incubation with thrombin, but the rate and extent of cross-link formation was still markedly impaired. Fibrinogen Cross-linking-Although polymerizing fibrin is a better substrate, FXIIIa will also cross-link fibrinogen. In these experiments, we first activated FXIII and initiated the reaction by adding FXIIIa to fibrinogen, as described under "Experimental Procedures," and analyzed the products by Western blot (Fig. 9). After 1 h, with normal fibrinogen, the blot demonstrated complete loss of the A␣ chain, indicative of ␣-polymer formation, and a strong ␥-␥ dimer band. After 1 h, with ␥⌬319,320 fibrinogen, there was also complete loss of the A␣ chain, yet the ␥ chain band remained essentially unchanged, and no band representing ␥-␥ dimers was seen. Even after 24 h, there was no evidence for ␥-␥ dimer formation with ␥⌬319,320 fibrinogen (data not shown). We conclude that, specifically, the ␥ chain of ␥⌬319,320 fibrinogen was a very poor substrate for FXIIIa.
FIG. 5. Size exclusion chromatography. Thrombin was mixed with fibrinogen and incubated. After the addition of hirudin to inhibit thrombin, the sample was immediately run over an Amersham Pharmacia Biotech HR 10/30 column packed with Superose 6B. Panels A-C, normal fibrinogen after incubation for 0, 1, and 3 min, respectively. Panels D-F, ␥⌬319,320 fibrinogen after incubation for 0 and 3 min and 4 h, respectively. The arrows delineate the void volume. ated platelet aggregation, the hemostatic function that requires the last four residues (␥408 -411) of the ␥ chain (18). We monitored ADP-induced platelet aggregation as a change in light transmission, as described under "Experimental Proce-dures," and the results are shown in Fig. 10. After ADP addition (indicated by arrows), we saw in both samples the expected decrease in light transmission that reflects the platelet shape change indicative of platelet activation. In the presence of  normal fibrinogen, subsequent light transmission increased rapidly as the platelets aggregated and the solution became more transparent. In contrast, in the presence of ␥⌬319,320 fibrinogen, light transmission did not change after the initial decrease. We conclude that ␥⌬319,320 fibrinogen did not support normal platelet aggregation. DISCUSSION We have synthesized the recombinant ␥⌬319,320 fibrinogen, modeled from the dysfibrinogen V/F IV, which was shown to be defective in calcium binding, fibrin polymerization, and platelet aggregation (19,43). As with many dysfibrinogens expressed in heterozygous patients, however, the consequence of the mutation cannot be fully appreciated in the presence of normal molecules. Therefore, we synthesized the variant recombinant fibrinogen to study the effect of this two-residue deletion in the absence of normal molecules.
The data presented here suggest that the deletion of residues 319 and 320 affected the structural integrity of the C-terminal domain of the ␥ chain. Based on the crystallography data, it was reasonable that a deletion in the calcium binding pocket would have profound effects on the adjacent polymerization site, "a" (12,13,16). However, it was unexpected that functions associated with distant residues were also affected. The FXIIIa-catalyzed formation of ␥-␥ dimers was severely impaired, with either ␥⌬319,320 fibrinogen or fibrin. Because ␥-␥ dimers were not seen with fibrinogen, we concluded that the loss of function was inherent in the ␥ chain structure per se and not simply associated with the impaired polymerization of fibrin. In addition, ADP-induced, fibrinogen-mediated platelet aggregation was nonexistent. Previous studies suggest that the ␥ chain residues needed to support FXIIIa-catalyzed crosslinking and platelet aggregation are independent from the rest of the molecule (44 -46). For example, a glutathione S-transferase chimeric protein containing residues ␥398 -411 was able to bind to GPIIbIIIa, the platelet receptor that mediates fibrinogen-dependent aggregation (46). This chimeric protein was also a substrate for FXIIIa, forming cross-linked dimers. The crystal structure of this chimeric protein suggested that this 14-residue segment of the ␥ chain has a flexible, "Z-shaped" structure, and when this structure was superimposed onto that from cross-linked fragment D, the residues involved in both platelet aggregation and FXIII cross-linking were available on the surface of the domain (46). We conclude, therefore, that in ␥⌬319,320 fibrinogen, these residues are either inaccessible or in an altered configuration. Perhaps these residues are buried in the ␥⌬319,320 fibrinogen molecule. Irrespective of the exact nature of the change, the data support the conclusion that the deletion of residues 319 and 320 altered the structure of these distant residues. We interpret these data as indicating that the overall structure of the C-terminal domain of the ␥ chain has been altered as a result of the two-residue deletion.
The experiments with ␥⌬319,320 demonstrated the value of studies on dysfibrinogens in a homogenous form. Polymerization of plasma fibrinogen V/F IV, purified from a heterozygous individual, was shown to be impaired (19), yet clotting of recombinant ␥⌬319,320 fibrinogen was undetectable. Importantly, we have identified a variant fibrinogen in which normal polymerization does not occur. Previous to these studies, every variant fibrinogen, whether recombinant or from plasma, has shown at least minimal polymerization by turbidity studies. Even fibrin with a substitution at a key residue in "a," ␥D364H, demonstrated a rise in turbidity by 60 min with 0.45 mg/ml fibrinogen and 0.5 units/ml thrombin (25). We saw no increase in turbidity under these same conditions.
The results from the size exclusion chromatography, electron microscopy, and light scattering support the conclusion that the very first step in polymerization was lost, and normal protofibrils were not formed. The elongated aggregates apparent with ␥⌬319,320 fibrin could be either short protofibrils that cannot laterally aggregate or aggregates that are coinciden- FIG. 9. Fibrinogen cross-linking in the absence of thrombin with activated FXIII. Normal or ␥⌬319,320 fibrinogen was incubated with pre-activated FXIII and the reaction was stopped after 1 h. A reduced 7.5% SDS-PAGE gel (26) was immunoblotted with a polyclonal anti-fibrinogen antibody. The blot was slightly overexposed to enhance the ␥-␥ band signal.
FIG. 10. ADP-induced platelet aggregation. 2 ϫ 10 8 platelets/ml were preincubated at 37°C with 250 nM fibrinogen, prior to activation with 10 M ADP (indicated by arrows). Aggregation of stirred solutions was monitored as an increase in light transmission. Thick line, normal fibrinogen; thin line, ␥⌬319,320 fibrinogen. tally elongated in shape, similar to protofibrils. We favor the latter based on their appearance in comparison to previously published transmission electron microscopy data of normal protofibrils (47,48). Because aggregate formation was thrombindependent, interactions that are thrombin-dependent yet weaker than "A-a" may have a role. One potential interaction would follow the release of FpB, exposing the "B" site that can interact with the "b" site. Based on the structure of fragment D (12), we would expect "B-b" interactions to form unusual polymers, such as the oligomers seen in ␥⌬319,320 fibrin. Previous biochemical studies indicate that an interaction between "A" and "b" might lead to abnormal polymer forms (49), although crystallography studies did not detect any "A" binding to "b" (8).
Another thrombin-dependent interaction may occur through the ␣C domains, which contact each other and the central E nodule in fibrinogen (30,50). In fibrin, the ␣C domains separate from the center of the molecule (30), possibly uncovering new sites for intermolecular interactions. In transmission electron microscopy studies, Veklich et al. (30) identified fibrin aggregates that interact via their ␣C domains. Similar aggregates were seen on occasion with ␥⌬319,320 fibrin (data not shown), so such interactions may support this abnormal polymer. Regardless of whether "B-b" or "A-b" or ␣C-␣C interactions mediate the formation of ␥⌬319,320 fibrin aggregates, the results provide evidence that without a functional "a", protofibrils, and consequently a fibrin clot, cannot form.
Our polymerization results also support one proposed model for the structure of ␤-fibrin, the product of cleavage by a protease that releases predominantly FpB. This fibrin polymer only forms at 14°C (51), but it is normal in that there are normal protofibrils, fibers, and clots (6,52). It has been proposed that ␤-fibrin is mediated through "B-a" interactions (8). As no turbidity changes were detected when ␥⌬319,320 fibrinogen was mixed with thrombin at 14°C (data not shown), our data are consistent with this model of ␤-fibrin that requires a functional "a" site.
In summary, we have synthesized a recombinant fibrinogen modeled after the dysfibrinogen V/F IV, a deletion of residues ␥319 and ␥320. The loss or impairment of several coagulation functions associated with the ␥ chain suggests that a significant change in structure has occurred in the molecule. Nonetheless, the deletion had no apparent effect on assembly or secretion, and transmission electron microscopy showed that the overall tri-nodular structure of the molecule was normal. Thus, we hypothesize that the deletion has induced a significant structural change only within the C-terminal domain of the ␥ chain.