The _EC Domains of Human Fibrinogen₄₂₀Contain Calcium Binding Sites But Lack Polymerization Pockets

TWO SUBCLASSES OF fibrinogen molecules can be distinguished in normal blood based on their α chain isoforms. Recognition of the existence of a second subclass evolved from the discovery that the complete sequence of the α gene contains an additional exon (exon VI of the human gene) encoding 236 amino acids (VI-domain) homologous to the fibrinogen β- and γ-chain carboxyl termini.1 The α gene, in addition to giving rise to the conventional form of the α subunit that lacks the VI-domain, also yields a transcript that, when alternatively spliced, encodes the extended α chain (α_E isoform)2 containing the exon VI-encoded carboxy terminus (α_EC).

Fibrinogen containing α_E has subsequently been shown to be a symmetrical molecule of the structure (α_Eβγ)₂,3 ie, the common fibrinogen α chains have both been replaced by α_Echains against all stoichiometric odds, given the overwhelming 100:1 ratio of α mRNA:α_E mRNA in the hepatocyte.2Based on mass prediction, (α_Eβγ)₂ has been termed Fibrinogen₄₂₀ (Fib₄₂₀) to distinguish it from the common (αβγ)₂ subclass, similarly referred to as Fibrinogen₃₄₀(Fib₃₄₀).

In the normal adult population, on average only 1 of every 100 molecules of fibrinogen is a molecule of Fib₄₂₀.4 However, homologues of Fib₄₂₀’s α_E chains have been identified throughout the vertebrate kingdom and α_EC itself represents the largest conserved segment of the α gene,1,5 6 a startling observation that suggests that the domain imparts a critical and distinctive functionality.

Clot formation and lysis, the primary functions of the major form of fibrinogen (Fib₃₄₀), have been characterized in detail over a period of several decades.7 8 Briefly, thrombin cleavage of fibrinogen releases fibrinopeptides A and B from the amino termini of the α and β chains, and the resulting fibrin monomers spontaneously polymerize. Interactions between each newly exposed α-chain N-terminus and complementary polymerization pockets located in the carboxyl-terminal domains of the γ chains provide a main driving force for fibrin monomer association. This noncovalent association brings together each central amino terminal E-domain with the distal carboxyl terminal D-domains of two other fibrin molecules, forming a half-staggered, double array of protofibrils. This organization enables factor XIIIa-catalyzed covalent stabilization of the polymers by ε-amino-(γ-glutamyl) lysine cross-links. Calcium ions promote both the formation and cross-linking of fibrin polymers, and calcium binding limits fibrin(ogen)’s susceptibility to proteases such as plasmin.

Fib₄₂₀ also participates in clot formation3(Applegate et al, manuscript in preparation); however, the influence of its two α_EC domains on this process and on clot lysis is not known. In this context, the current study reports the generation of a soluble recombinant form of the human α_EC (rα_EC) and an evaluation of its physical and biochemical properties.

Human plasmin, thrombin, and fibrinogen fragment D₁9 were generous gifts of M. Mosesson (Sinai Samaritan Medical Center, Milwaukee, WI). Human factor XIII was graciously provided by P. Bishop (ZymoGenetics, Seattle, WA). A GPR peptide column was kindly provided by R. Doolittle (University of California at San Diego, La Jolla, CA).

A segment encoding the C-terminal domain of α_E(Val610-Gln847) was generated by polymerase chain reaction (PCR), inserted in pPIC9 at the Avr II site, and verified by DNA sequencing. The recombinant α_EC domain was expressed in the methylotrophic yeast strain Pichia pastoris (Mut−) according to the manufacturer’s protocol (Invitrogen, San Diego, CA). Throughout this report, residues in rα_EC are numbered according to their corresponding position in the α_E chain (GenBank accession no. M58569).

Production of rα_EC was scaled up from flask to fermentor culture in a 12-L Microferm fermentor (New Brunswick Scientific, Edison, NJ), based on the Invitrogen fermentation protocol. Briefly, growth phase was initiated with inoculation of 6 L of medium with a P pastoris clone expressing rα_EC and maintained as a glycerol batch culture for 24 hours, followed by continuous feeding of glycerol for another 24 hours. Feeding was interrupted for 0.5 hours (starvation) and the production phase was induced by the addition of methanol. The culture was stirred at 1,000 rpm and kept at 30°C throughout, with compressed air (5 psi) providing aeration. Dissolved oxygen, pH, and wet weight of cells were monitored. The antifoam agent Struktol KFO 673 (0.2%; Kabo Chemicals, Inc, Jackson Hole, WY) was present from the beginning. Because of fibrinogen’s sensitivity to low pH, a pH of 6 was maintained throughout the fermentation. Casamino acids (1%; Difco, Livonia, MI) were included during the production phase to suppress yeast proteases that may be active at this pH. Production of rα_EC was monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western analysis. After a total run of 100 hours (52-hour production phase), the yeast cells (wet weight, ∼200 g/L) were pelleted at 3,600 rpm. The recovered supernatant (5.4 L) was stored at 4°C overnight to dissolve the antifoam reagent. The conditioned medium was first filtered (0.45-μm Nalgene membrane, with a double layer of prefilters) and then concentrated approximately 10-fold with a Pellikon apparatus (Millipore, Bedford, MA; exclusion 10 kD). The crude fermentor culture supernatant was stored at −80°C.

Crude fermentor supernatant was dialyzed against 20 mmol/L Tris, pH 7.5, clarified by centrifugation at 5,000g, and subsequently applied to a Mono Q HR 5/5 anion exchange column (Pharmacia, Piscataway, NJ) pre-equilibrated with the same buffer. Bound protein was eluted with a 20-column volume linear salt gradient (20 mmol/L Tris, pH 7.5, to 1 mol/L NaCl, 20 mmol/L Tris, pH 7.5).

Samples were prepared for electrophoresis in Laemmli sample buffer in the absence or presence of 0.1 mol/L dithiothreitol10 and separated on SDS-PAGE using a Mini-Protean II Electrophoresis Cell (Bio-Rad, Hercules, CA). Protein was stained with Coomassie Blue R250. For Western blots, transfer onto 0.2-μm nitrocellulose membranes was performed with a Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad). Membranes were incubated with primary monoclonal antibodies followed by secondary antibody, horseradish peroxidase (HRPO)-labeled goat antimouse IgG (Pierce, Rockford, IL). To visualize enzyme activity, signals were developed by enhanced chemiluminescence (SuperSignal Substrate; Pierce) and filmed.

Linear 4-6.5 gradient IEF PhastGels were run on the PhastSystem (Pharmacia). Recombinant α_EC was diluted in 125 mmol/L NaCl, 25 mmol/L HEPES, pH 7.4, and compared with low isoelectric point calibration markers (Pharmacia Biotech, Piscatway, NJ) run under the same conditions.

The material was run on SDS-PAGE under reducing conditions, transferred to a polyvinylidene difluoride membrane, and stained with Coomassie Blue G250. The protein band was cut out and subjected to 20 cycles of microsequencing (Model 477A; Applied BioSystems, Inc, Foster City, CA).

N-linked glycosylation of rα_EC was evaluated by Glyko, Inc (Novato, CA). In short, protein-bound oligosaccharides were released with PNGase F endoglycosidase, and the sugars were labeled with fluorophore, separated using PAGE, and visualized and analyzed using Glyko’s FACE Imager and software. The exact nature of the bands was determined by fingerprinting using several exoglycosidases alone or in combination: neuraminidase, β-galactosidase, β-N-acetylhexosaminidase, and α-mannosidase.

rα_EC was cleaved with CNBr in 70% formic acid for 18 hours as described by Blomback et al.11 The fragments were separated by high-performance liquid chromatography (HPLC) using a reverse-phase C18 column for analysis with a MALDI TOF mass spectrometer (Voyager DE; PE Biosystems, Framingham, MA).

rα_EC (0.4 mg/mL) was digested with human plasmin (0.5 U/mL) at 37°C in a buffer containing 125 mmol/L NaCl, 25 mmol/L HEPES, pH 7.4, in the presence of either 5 mmol/L EDTA or 5 mmol/L CaCl₂. For plasmin protection experiments with peptides (5 mmol/L), the samples contained 5 mmol/L EDTA and either GPRPamide or GHRP (Sigma, St Louis, MO). In all cases, reactions were stopped by the addition of aprotinin (60 KIU/mL), followed by the addition of Laemmli sample buffer and boiling.

rα_EC (0.2 mg; ∼33 nmol) in 135 μL of buffer containing 125 mmol/L NaCl, 25 mmol/L HEPES, pH 7.4, and 5 mmol/L CaCl₂ was loaded on a 1.6-mL GPR peptide column pre-equilibrated with the same buffer. The column was then washed with 10 column volumes of loading buffer. Fractions (1 mL) were collected and absorbance at 280 nm was determined. After confirming that absorbance had returned to zero, bound protein was eluted with 3 column volumes of 1.0 mol/L NaBr, 0.05 mol/L sodium acetate, pH 5.3.12 13 

The reaction was performed in 5 mmol/L CaCl_2, 100 mmol/L NaCl, 50 mmol/L HEPES, pH 7.4, at room temperature. It was initiated by the addition of human thrombin (1 U/mL) to a solution containing recombinant human factor XIII (40 μg/mL) plus substrate (4 mg/mL), either fibrinogen fragment D₁ or rα_EC. Aliquots were removed and the reaction was stopped by boiling in Laemmli sample buffer.

A protein of about 38 kD was the predominant protein secreted upon methanol induction of the yeast P pastoris transformed with a pPIC9 vector carrying the exon VI sequence of the human fibrinogen α gene (Fig 1B, lane 1). Recognition by a polyclonal antibody highly specific for the VI-domain of α_E3 confirmed this band to be the recombinant α_EC domain. In the first 52 hours of the methanol induction phase, rα_EC accumulated to a level of about 200 μg/mL in the fermentation culture medium. After concentration, rα_EC was purified from the medium supernatant by anion exchange chromatography (Fig 1A). Gel analysis showed that the majority of protein contaminants appeared in the flow-through (Fig 1B, lane 2). The rα_EC domain eluted predominantly in five fractions between 350 and 400 mmol/L NaCl (peak c). These fractions were combined, dialyzed against 125 mmol/L NaCl, 25 mmol/L HEPES, pH 7.4, and concentrated for use in this study.

The purified rα_EC appears to be essentially intact based on N-terminal sequence analysis, on C-terminal epitope recognition, and on mass spectrometry of CNBr fragments. Microsequencing established the 20 N-terminal residues of rα_EC to be[YVEFPR]VRDXDDVLQTXPXG, representing an initial set of six residues generated by the multiple cloning region of the vector (in italics) and continuing with valine as the first bona fide fibrinogen residue (Val610), the final amino acid of the conventional α chain as it is found in circulating Fib₃₄₀. In rα_EC, this valine is followed, as in α_E, by Arg611 and then the 236 residues of the VI-domain proper (Asp612-Gln847). A product beginning at Ile631 of the α_E chain was also detected. Accounting for less than 10% of the purified rα_EC, it is evident as the fainter band migrating just ahead of intact rα_EC on SDS-PAGE (Fig 1B) and presumably represents partially degraded rα_EC.

On SDS-PAGE (Fig 2), reduction of rα_EC produces an upward mobility shift from an apparent size of 34 to 38 kD, indicating the presence of internal disulfide bridges. The α_EC domain has 4 cysteine residues available for such bonds at positions 613, 644, 780, and 793. Their positions, invariant among α_E homologues throughout vertebrate evolution, align precisely with cysteines in the βC and γC domains.2,5 6 By analogy with βC and γC, as well as from the results of mutational analysis of α_E,^13a we expected disulfide bonding in rα_EC to be restricted to the cysteine pairs α_E613/644 and α_E780/793.

To test the prediction, rα_EC was treated with CNBr. The three methionines in the sequence of the recombinant protein are distributed such that cleavage after these positions should generate four unequal fragments, with the predicted disulfide loops internally situated in two separate fragments (Table 1). Indeed, four major CNBr-generated fragments of rα_EC were observed by SDS-PAGE under both reducing and nonreducing conditions. Moreover, upward mobility shifts upon reduction provided positive, albeit indirect, evidence for internal disulfide bonds within fragments 1 and 3, consistent with the predicted assignments. In contrast, fragment 2, with no cysteines, migrated as a broad band with the same mobility whether reduced or not. The smallest fragment, containing the domain’s C-terminal amino acid residues, was barely detectable by SDS-PAGE, but was identified by mass spectrometry as having the predicted molecular mass of 954 Daltons.

The calculated mass of the 244 amino acids of the full recombinant fusion protein, TyrValGluPheProArgVal610-Gln847, is 27,653 Daltons. The sizable discrepancy between this number and 34 kD, its apparent molecular mass on SDS-PAGE under nonreducing conditions (Fig2), can be attributed to glycosylation. Whereas the conventional α chain of Fib₃₄₀ has no carbohydrate, the α_EC domains of Fib₄₂₀ are decorated, attachment occurring only at Asn667^13a even though a second potential site exists at Asn812.2 Carbohydrate analysis of the yeast-expressed rα_EC showed that roughly two thirds of the molecules contained high mannose N-linked oligosaccharides (Man_6-13GlcNAc₂) common to other glycoproteins expressed in P pastoris14; phosphorylated low mannose sugars (Man₂P GlcNAc₂) were deduced to be present on the remainder. This heterogeneity permits assignment of the attachment site in rα_EC to CNBr fragment 2, which migrated on SDS-PAGE not only as the broadest band, but also displayed an apparent mass considerably greater than that calculated from the amino acid sequence alone (Table 1). This finding indicates sole use of the site at Asn667 by yeast, consistent with glycosylation of the α_E chains of human Fib₄₂₀. The structure of the carbohydrate attached to the native α_E remains to be determined, although it is likely to be of the low mannose type ending predominantly in sialic acid, as in the β and γ chains,15 rather than the high mannose type found in the yeast-derived recombinant domain.

The recombinant domain’s predicted pI is 4.3, based on its amino acid sequence. Isoelectrofocusing of rα_EC yielded two closely migrating bands located between the pI markers at 4.15 and 4.55 (not shown), consistent with the prediction. It may be that these doublet bands reflect the heterogeneity of the carbo- hydrate side chain described above.

The primary structure of Fib₄₂₀ is identical to that of Fib₃₄₀ but for the α_EC domains.2 3 We used rα_EC to investigate whether these domains might contribute to Fib₄₂₀ additional cross-linking sites, calcium-binding sites, and/or polymerization pockets. This yeast recombinant closely approximates the conformation of the native domain as indicated by its use of the same carbohydrate attachment site and the deduced congruence of its internal disulfide bridges described above.

It has been suggested that the α_EC domains participate in factor XIIIa-catalyzed cross-linking based on studies in lamprey.13 To evaluate cross-linking between the human α_EC domains, rα_EC was incubated with factor XIIIa (Fig 3). Although the reaction was performed at high substrate and enzyme concentrations to increase its efficiency, no formation of rα_EC-dimers was observed, even after prolonged incubation. Under similar conditions, significant cross-linking of the γ-chain fragment of fragment D₁(γD) occurred, as evidenced by the steady accumulation of γD-dimer over the 15-hour period. In addition, reactions conducted with Fib₃₄₀ as substrate in a 10-fold molar excess of rα_EC yielded no evidence of the domain cross-linking either to itself or to fibrinogen (not shown). These data suggest that Fibrin(ogen)₄₂₀’s α_EC domains may not cross-link. Taken together with recent mutagenesis studies finding no evidence of disulfide bonding between α_EC domains and the rest of the Fib₄₂₀ molecule,^13a the findings support the notion that, in a clot, the α_EC domains may be tethered only via their upstream “αC” protein sequence.

In the presence of 5 mmol/L EDTA, rα_EC is highly susceptible to plasmin (0.5 U/mL), being degraded to species no larger than 18 kD within 30 minutes (Fig 4). However, calcium ions (5 mmol/L) largely protect rα_EC from plasmin for at least 15 hours. The calcium protection is reminiscent of that observed for the carboxyl terminal domain of γ chains (γC)16,17 and argues strongly for an active calcium binding site within α_EC. By analogy with the site elucidated in the γC domain,18-20 we predict that the α_EC calcium binding site involves residues 772-778: DADQWEE. Interestingly, the corresponding region of the βC domain has a five-residue insert that, according to a recently reported structural study, diminishes but does not obliterate its affinity for calcium.21 

Plasmin digestion of rα_EC in the presence of 5 mmol/L EDTA was evaluated by Western analysis with two different monoclonal antibodies, one that recognizes an epitope at the C-terminus (29-1) and another (3-10) that recogizes an epitope downstream in the N-terminal third of the VI-domain (Fig 5). The immunoblot shows generation of a relatively stable 18-kD degradation fragment containing the C-terminal epitope at the highest plasmin concentration, whereas lower levels of the enzyme produce a variety of fragments in the 20- to 25-kD range. The relative amounts of the latter, which bear either the N- or C-terminal epitope, are dependent on the time of exposure and the concentration of plasmin. In this context, it is noteworthy that both antibodies recognize the starting material’s major 38-kD band as well as the minor species migrating immediately below it (presumably the species beginning at α_E631), providing further evidence that the yeast-expressed rα_EC domain has a structurally intact C-terminus.

The synthetic peptide GPRPamide, resembling the amino terminus of the fibrin α chain, limits the plasmic digestion of fragment D and the γC domain.16 22 Analysis of rα_EC incubated with plasmin showed that the synthetic peptides GPRPamide or GHRP (the latter mimicking the amino terminus of the fibrin β chain) rendered no protective effect (not shown). To test whether analogues of the fibrin α chain amino terminus bind rα_EC without affecting its sensitivity to plasmin, we performed affinity chromatography by passing purified rα_EC over a GPR peptide column as described in Materials and Methods. Under conditions in which intact fibrinogen bound completely, no rα_EC bound to the column, all of it appearing in the flow-through. These results suggest that the α_EC domain lacks a polymerization pocket.

In conclusion, this study strongly suggests that the conformation of recombinant α_EC expressed in yeast closely approximates that of the native domain. Experiments with the recombinant domain indicate that Fib₄₂₀ is endowed with one extra calcium-binding site per α_EC but no additional polymerization pockets or cross-linking sites. The usefulness of the rα_EC generated in this study is underscored by our recent observation that, when Fibrin(ogen)₄₂₀ is treated with plasmin, the α_EC domains are readily clipped off and, under physiologic conditions, remain intact (Applegate et al, manuscript in preparation). Intriguingly, the freed domain is detectable in human plasma, particularly under pathophysiologic conditions. Thus, rα_EC may serve as an excellent model substance for testing the potential biological function(s) of this unique and stable degradation product generated from Fibrin(ogen)₄₂₀.

The authors thank Dr James Farmar of the Microchemistry Laboratory for performing the mass spectral analyses and Dr Bohdan Kudryk for many helpful discussions. Our gratitude goes to Dr Jack Goldstein and members of his laboratory, particularly Drs Alex Zhu, Zhanfan Zhang, and Lin Leng, for help with the fermentation process.

As this report was being processed for publication, the x-ray structure of the α_EC domain of Fib₄₂₀ was solved using rα_EC crystals, providing confirmation of the predictions of this biochemical study.23