In vitro assembly of the component chains of fibrinogen requires endoplasmic reticulum factors.

Human fibrinogen (340 kDa) is a dimer, with each identical half-molecule composed of three different polypeptides (Aα, 66 kDa; Bβ, 55 kDa; and γ, 48 kDa). To understand the mechanisms of chain assembly, a coupled in vitro transcription translation system capable of assembling fibrinogen chains was developed. Fibrinogen chain assembly was assayed in an expression system coupled to rabbit reticulocyte lysate in the presence or absence of dog pancreas microsomal membranes. Fibrinogen chain assembly required microsomal membranes and oxidized glutathione. Co-expression of two of the chains, Bβ and γ or Aα and γ, yielded free chains and two-chain complexes. Unlike combinations of Aα with γ and Bβ with γ, co-expression of Aα and Bβ did not form a single two-chain complex but produced a mixture of two-chain complexes. Co-expression of all three chains yielded free chains, two-chain complexes, and higher molecular weight complexes that corresponded to a half-molecule and to fully formed fibrinogen. Upon treatment of this mixture with thrombin and factor XIIIa, a γ·γ dimer, similar to that obtained from cross-linked human fibrin, was produced, indicating that properly folded fibrinogen was formed in vitro. Molecular chaperones may participate in fibrinogen assembly, since antibodies to resident proteins of the endoplasmic reticulum (BiP, Hsp90, protein disulfide isomerase, and calnexin) co-precipitated the chaperones together with nascent fibrinogen chains and complexes.

Human fibrinogen is a dimer with each half-molecule containing three nonidentical polypeptide chains (A␣, B␤, and ␥). Two of the chains, B␤ and ␥, are glycoproteins. Fibrinogen is an elongated, trinodal molecule composed of a central domain (E domain), which contains the NH 2 termini of the six chains, and two terminal D domains, formed by the carboxyl-terminal globular regions of B␤ and ␥ chains and a segment of the A␣ chain. The central E domain and the terminal D domains are connected by a "coiled-coil" ␣-helical region composed of the three chains. The coiled region is flanked at both ends by a set of interchain "disulfide rings." In the central E domain the 2 half-molecules are joined at their NH 2 -terminal regions by three symmetrical disulfide bonds between adjacent A␣ Cys-28 and ␥ Cys-8 and ␥ Cys-9 residues (for reviews see Refs. [1][2][3][4][5]. Site-directed mutagenesis, changing NH 2 -terminal cysteine residues to serine, also indicates that A␣ Cys-36 of one halfmolecule is disulfide-linked to B␤ Cys-65 of the other halfmolecule (6,7). Fibrinogen has a total of 29 inter-and intrachain disulfide bonds with no free cysteine residues (8 -13). In addition, human plasma contains a small amount of fibrinogen, which has an extended A␣ chain with a COOH-terminal globular domain similar to that of the B␤ and ␥ chains. (14,15).
Chain assembly into six-chain fibrinogen occurs in the endoplasmic reticulum (ER) 1 of hepatocytes (16). As with other multichain proteins, proper folding takes place in the ER before the nascent protein progresses to the Golgi complex, where further processing occurs, and before it is secreted. For most proteins initial assembly and folding is facilitated by transient interactions with a number of molecular chaperones that reside in the ER (for reviews see Refs. [17][18][19]. Pulse-chase studies with Hep G2 cells demonstrated the step-wise assembly of fibrinogen chains, progressing from twochain complexes to three-chain half-molecules, which are finally coupled together to form six-chain fibrinogen (20,21). Several recombinant fibrinogen systems have been described that are capable of assembling and secreting fibrinogen (22)(23)(24)(25). The intracellular complexes found in some of these cells are similar to those in Hep G2 cells, suggesting similar paths of assembly. Early pulse-chase studies with Hep G2 cells showed accumulation of an A␣⅐␥ complex and postulated that A␣⅐B␤ and B␤⅐␥ are also precursors in the formation of half-molecules and six-chain fibrinogen (21,26). The accumulation of precursor and intermediate fibrinogen complexes under steady-state conditions have been characterized in Hep G2 cells and in transfected COS and baby hamster kidney cells. In these cells A␣⅐␥, B␤⅐␥, half-molecules, and six-chain fibrinogen were identified, but A␣⅐B␤ complexes were not detected (27). A␣⅐B␤ complexes were also not found in transiently transfected COS cells (23). However, all of the studies indicate a stepwise assembly, from single chains to two-chain complexes, addition of a third chain to form the half-molecule, and finally dimerization to form the final six-chain complex. Studies with transfected cells, using deletion and substitution mutants of the fibrinogen chains, indicate that formation of the coiled-coil structure and its disulfide rings play important roles in dimer formation (7).
To further understand the mechanisms of fibrinogen assembly, an in vitro system that couples transcription, translation, and events that normally occur in the ER has been developed. tase were purchased from Boehringer Mannheim. Endoglycosidase-H was obtained from Genzyme Corp. (Cambridge, MA), and T4 DNA ligase was from New England Biolabs (Beverley, MA). Rabbit polyclonal antibody to GRP78 (BiP) and mouse monoclonal antibody to GRP90 were purchased from Affinity BioReagents, Inc. (Neshanic Station, NJ). Rabbit polyclonal antibody to bovine protein disulfide isomerase was a gift from Birger Blomback (Karolinska Institute, Stockholm, Sweden). Polyclonal antibody to rat calnexin was also a gift from John Bergeron (McGill University, Montreal, Quebec, Canada). Other reagents used, including the antibodies to fibrinogen, have been described previously (22,28).
Construction of Expression Vectors-Expression vectors containing fibrinogen cDNAs for single chains, combinations of two chains, and all three chains, were inserted in multiple cloning sites at the 3Ј end of a T7 promoter in pYES2 plasmid (Invitrogen Corp., San Diego, CA) (Fig. 1). In the construction of expression vectors containing single fibrinogen chain cDNAs, full-length cDNAs were released by appropriate restriction enzymes from previously described constructs (22,30) and ligated at the 3Ј end of the T7 promoter. The single chain constructs were termed pYES2A␣, pYES2B␤, and pYES2␥. Other constructs had two fibrinogen chain cDNAs, in various combinations and all three cDNAs chains in tandem, each under the control of T7 promoter.
To make the two-chain construct pYES2A␣,B␤, a 2.5-kb fragment (GalT7A␣) was released from pYES2A␣ by BamHI and partial SpeI digestion. The 2.5-kb fragment was ligated to SpeI-cut, dephosphorylated pYES2B␤ plasmid followed by a fill-in reaction and blunt end ligation. To construct pYES2A␣,␥, the GalT7A␣ fragment was ligated to BamHI-cut, dephosphorylated pYES2␥ plasmid that had been treated with BamHI, followed by a fill-in reaction and blunt end ligation. To obtain pYES2B␤,␥, a GalT7␥ (2.0-kb) fragment was released from pYES2␥ plasmid by SpeI and ligated to SpeI-cut, dephosphorylated pYES2B␤ plasmid.
To obtain the construct with all three cDNAs (pYES2A␣,B␤,␥) the GalT7␥GalT7A␣ fragment (4.5 kb) was released from pYES2A␣,␥ by BamHI and partial SpeI treatment and was ligated to SpeI-cut, dephosphorylated pYES2B␤ plasmid, followed by a fill-in reaction and blunt end ligation. The orientation of the cDNAs in the above constructs was determined by treatment with the appropriate restriction enzymes. The procedures for elution of DNA fragments from agarose gel, dephospho-rylation of plasmids by calf intestinal phosphatase, the fill-in reaction by Klenow fragment, and ligation were performed by standard procedures (29).
Cell-free Transcription and Translation of Fibrinogen Chains-Transcriptions and translations were performed using the TNT T7 transcription system coupled with a rabbit reticulocyte system according to the manufacturer's directions. A typical reaction mixture (12.5 l) contained 55% TNT T7 reticulocyte lysate, 0.25 units of T7 RNA polymerase, 80 mM amino acid mixture without L-methionine, 10 Ci of L-[ 35 S]methionine, 10 units of RNasin, 0.5 mg of DNA substrate in TNT buffer supplied by the manufacturer. Incubations were at 30°C for 90 min. In some reactions, two equivalents of canine pancreatic microsomal membranes and oxidized glutathione (GSSG) (ranging from 1 to 10 mM) were included. Aliquots (2 l) of translated products were either directly subjected to SDS-PAGE analysis or immunoprecipitated with a polyclonal antibody to human fibrinogen or with mouse monoclonal antibodies to A␣ or B␤ chains followed by one-dimensional or twodimensional SDS-PAGE analysis and autoradiography. Two-dimensional SDS-PAGE was performed in 7.5% polyacrylamide gels, first in nonreducing conditions followed by reelectrophoresis in the second dimension in reducing conditions. The radioactive products were quantitated either by densitometry or by excising the radioactive areas and measuring radioactivity.
Co-precipitation of Nascent Fibrinogen with Antibodies to Molecular Chaperones-Following in vitro translation of fibrinogen complexes in the presence of oxidized glutathione and microsomal membranes, the reaction mixture (100 l) was divided into six equal parts. To each was added 500 l of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, and 0.5% Nonidet P-50 followed by antibodies to one of the following: fibrinogen, GRP78 (Bip), GRP90, protein disulfide isomerase, or calnexin. As a control, equivalent amounts of nonimmune rabbit serum was added to one of the samples. The mixtures were incubated for 16 h at 4°C. Immunoprecipitation was completed by the addition of a second immobilized antibody. Protein A-Sepharose was added to the control and to the samples containing rabbit antibody to fibrinogen, GRP78 (BiP), protein disulfide isomerase, and calnexin, and anti-mouse IgM-agarose was added to the sample containing mouse monoclonal antibody to GRP90. The samples were further incubated for 2 h at 4°C, and the immunocomplexes were collected by brief centrif- ugation (1 min at 12,000 rpm). The pellets were washed three times with 1 ml of 50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 0.5% sodium deoxycholate, 0.5% Nonidet P-50 and once with 50 mM Tris-HCl, pH 6.8, and 150 mM NaCl. The proteins were solubilized by heating for 5 min at 100°C with 40 l of SDS-PAGE loading buffer containing 1% SDS. The eluted proteins were separated by one-or-two-dimensional SDS-PAGE.
Other Methods-Endoglycosidase-H treatment of translated products, thrombin treatment, and cross-linking with factor XIIIa were performed as described previously (28).

Synthesis and Processing of Fibrinogen
Chains-In vitro expression of polypeptides in the presence of stripped dog pancreas microsomes leads to initial proteolytic processing, translocation of the nascent polypeptides into the lumen of microsomes, and glycosylation of the B␤ and ␥ chains. The addition of microsomal membranes to the translation mixture containing pYES2A␣, pYESB␤, and pYES2␥ resulted in a mobility shift as determined by SDS-PAGE of B␤ and ␥ chains, which are glycoproteins, but not of A␣ (Fig. 2). Similarly, when combinations of two chains or all three chains were expressed, the B␤ and ␥ chains, but not A␣, had slower mobilities on SDS-PAGE in the presence of microsomal membrane (Fig. 2). Multiple smaller fragments of all fibrinogen chains were observed. These small polypeptides are probably due to incomplete translation products and also to degradation.
To determine whether N-glycosylation occurred, the translation products of pYES2A␣,B␤,␥ with microsomal membranes were immunoprecipitated and treated with endoglycosidase H. Endoglycosidase H treatment did not affect the mobility of the chains expressed in the absence of microsomal membranes but increased the mobilities of B␤ and ␥ chains, translated together with microsomal membranes. This indicates that the B␤ and ␥ chains were N-glycosylated and contain mannose-rich oligosaccharides. The electrophoretic migration of A␣ chain, which normally is not N-glycosylated, was not affected by treatment with endoglycosidase H (data not shown).
Chain Assembly Requires Oxidized Glutathione-Oxidized glutathione, added to cell-free systems, has been shown to result in the correct disulfide-bonded assembly of biologically active proteins (30 -33). To determine whether oxidized glutathione was necessary for fibrinogen chain assembly, pYES2A␣,B␤,␥ was transcribed and translated in the presence of microsomal membranes and varying concentrations of GSSG. Analysis of nascent proteins, in nonreduced SDS-PAGE showed that in the absence of GSSG free chains and high molecular weight aggregate were the principal products. In the presence of 1 mM GSSG two-chain complexes (A␣⅐␥ and B␤⅐␥) and some higher molecular weight products were formed. With increased concentrations of GSSG (2-15 mM), more of these assembled fibrinogen complexes appeared. The optimum concentration of GSSG was about 5-10 mM.
Kinetic studies showed that at the earliest time measured (15 min) mostly free A␣, B␤, and ␥ chains were present, with larger amounts of ␥ chains. Intermediate two-and three-chain complexes and material, similar in size to fibrinogen, were detected at 30 min and continued to accumulate until 120 min of incubation. Scanning densitometric analyses showed that the percentage of radioactivity in free chains decreased with time and that the radioactivity in fibrinogen and its intermediates complexes increased. At the end of 120 min of incubation, approximately 20% of the total protein radioactivity was accounted for in a location where six-chain fibrinogen migrates and 15-20% was in intermediate two-chain and three-chain complexes (Fig. 3).
Assembly of Two-and Three-chain Complexes Requires Microsomal Membranes-Two-and three-chain complex formation required that the chains be translated in the presence of microsomal membranes. In the absence of microsomal membranes only free chains, large molecular weight aggregates, and proteins that migrated as a streak on SDS-PAGE were noted (data not shown). In the presence of microsomal membranes, co-translation of A␣ and ␥ and of B␤ and ␥ led to the assembly of discrete bands that migrated as expected for twochain complexes (Fig. 4, B and C). However, co-translation of  A␣ and B␤ did not yield a distinct two-chain product, although a darker diffuse area was noted in the region where two-chain complexes are expected to migrate (see Fig. 4A). Co-translation of all three chains, in the absence of microsomal membranes, led to the formation of a high molecular weight aggregate and free chains. In the presence of microsomal membranes, in addition to free chains, two bands were observed in the location where two-chain complexes migrate. Higher molecular weight products, corresponding to half-molecules (three-chain complexes) and to six-chain fibrinogen were also formed (Fig. 4D).
Two-dimensional SDS-PAGE Analysis-To characterize the complexes formed when different chains are co-translated in the presence of microsomal membranes, the translation products, isolated by immunoprecipitation with polyclonal antibody to fibrinogen, were analyzed by two-dimensional SDS-PAGE first in nonreduced conditions and then in reducing conditions. Co-expression of A␣ and B␤ yielded, in nonreducing conditions, aggregated material at the top of the gel, a diffuse band at the region where two-chain complexes are expected to migrate, and free chains. Upon reduction and electrophoresis in a second dimension, both the high molecular weight aggregate and the area that may contain two-chain complexes yielded A␣ and B␤ chains. Greater amounts of A␣ chain than B␤ chains were present in these areas. The free chains, as expected, gave single bands (Fig. 4A).
Co-expression of A␣ and ␥ yielded two bands that, because of their electrophoretic migration in the first dimension, are putative two-chain complexes. One higher molecular weight band was composed of A␣ and ␥ chains, and there was indication of a ␥⅐␥ dimer, which migrated just below the A␣⅐␥ complex. Free A␣ and ␥ chains were also present (Fig. 4B). Co-expression of B␤ and ␥ chains gave a single two-chain complex and free chains. There were many more free ␥ than free B␤ chains. There were few, if any, higher molecular weight complexes formed. Upon reduction, in the second dimension, the two-chain complex yielded B␤ and ␥ chains (Fig. 4C).
Co-translation of all three chains showed the presence of several complexes (Fig. 4D). In the areas where six-chain fibrinogen and the half-molecules are expected to migrate, A␣, B␤, and ␥ chains were obtained. There was greater recovery of A␣ and B␤ than of ␥ chains. Similarly, in the area where two-chain complexes migrate, A␣ and ␥ chains were detected in one area and B␤ and ␥ from another area. In all cases there was much lower recovery of ␥ chains in the second dimension. The reasons for this are not clear, although partly it is due to lower methionine content and, therefore, less radioactivity, in ␥ as compared with A␣ and B␤ chains. Fewer ␥ than A␣ and B␤ chains were also noted in the regions where fibrinogen, halfmolecules, and the two-chain complexes migrate.
Characterization of A␣⅐B␤ Complexes with Chain-specific Antibodies-To further characterize the complex formed when A␣ and B␤ are co-translated in the presence of microsomal membranes, the radioactive products were isolated with antibodies specific to A␣ and B␤ chains. In nonreducing conditions (first dimension) the pattern obtained with the chain-specific antibodies was slightly different from that obtained with the polyclonal antibody to fibrinogen. In addition to the free chains and the diffuse streaking in the high molecular weight region, a distinct band was obtained just above the 200-kDa marker. Upon reduction (second dimension), this band, isolated with anti-A␣, yielded mostly A␣ chain with a trace of radioactive material in the region where B␤ migrates. The "diffuse" streak between the 200-and 100-kDa markers was composed mostly of A␣ chains. Free A␣ chain, and lower molecular weight bands, which could be degradation products or incomplete A␣ chains, were also detected (Fig. 5A). Isolation with anti-B␤ showed that the radioactive product, which migrated just above the 200-kDa marker, contained mostly B␤ but also contained a trace of A␣ and proteins with lower molecular weights than B␤. The diffuse streak between the 200-and 100-kDa markers was mostly composed of B␤ (Fig. 5B). Free B␤ chains and proteins of lower size than B␤ were also isolated.
Taken together, these results indicate that co-expression of A␣ and B␤ leads mainly to the formation of high molecular weight homocomplexes. Very few heterocomplexes of A␣ and B␤ are formed, although there is an indication that some A␣ and B␤ chains are assembled together, since anti-B␤ isolates small amounts of A␣ (Fig. 5B) and anti-A␣ isolated small amounts of B␤ (Fig. 5A).
Co-precipitation of Nascent Fibrinogen and Molecular Chaperones-The three fibrinogen chains were co-expressed in the presence of microsomal membranes and GSSG, and the radioactive proteins formed were treated with nonimmune rabbit serum as a control or with rabbit antiserum to fibrinogen, GRP78 (BiP), protein disulfide isomerase, or calnexin or with mouse monoclonal antibody to GRP90. The radioactive proteins, present in the immunoprecipitate, were separated by one-or two-dimensional SDS-PAGE and detected by autoradiography.
No radioactive proteins were isolated with nonimmune rabbit serum. The other antibodies precipitated fewer radioactive proteins than anti-fibrinogen, but the proteins isolated with the antibodies to the different chaperones had the same electrophoretic properties as those isolated by anti-fibrinogen. The antibodies to chaperones isolated less radioactive proteins than anti-fibrinogen. Anti-calnexin isolated 15%, and the other an-tibodies isolated 25-30% as compared with anti-fibrinogen, (Fig. 6A).
Two-dimensional SDS-PAGE was performed to identify the precipitated radioactive proteins. All of the antibodies to the molecular chaperones isolated free A␣ and ␥ chains (Fig. 6, C,  D, E, and F). Very little free B␤ chain was present, even in the immunoprecipitate isolated with anti-fibrinogen. An A␣⅐␥ complex (which migrated below the 200-kDa marker) was also isolated by all of the antibodies tested. The half-molecule, which is detected upon isolation with anti-fibrinogen (it migrated below fibrinogen) (Fig. 6B) was only clearly identified with anti-GRP90 (Fig. 6D) and anti-protein disulfide isomerase (Fig. 6E). Very few half-molecules were detected in the precipitate obtained with anti-BiP (Fig. 6C) and anti-calnexin (Fig.  6F). Fibrinogen, which migrates at the top of the band and is identified by yielding the three component chains upon reduction (Fig. 6B), was isolated by anti-BiP (Fig. 6C), anti-GRP90 (Fig. 6D), and anti-protein disulfide isomerase (Fig. 6E). Although anti-calnexin co-isolated, on the first dimension, a protein that migrated similarly to fibrinogen, this compound contained very little of the three component chains of fibrinogen (Fig. 6F). The high molecular weight proteins isolated by anticalnexin could be aggregated forms of fibrinogen polypeptides.
Factor XIIIa-catalyzed Cross-linking of in Vitro Fibrinogen-A functional test for fibrinogen is its ability to undergo thrombin-induced polymerization and factor XIIIa-catalyzed cross-linking. To determine whether the in vitro assembled fibrinogen met this criterion, the total radioactive proteins, obtained by co-expression of all three chains in the presence of GSSG and microsomal membranes were treated with thrombin and factor XIIIa. Plasma fibrinogen treated in this manner yields high molecular weight cross-linked A␣ chains, ␥⅐␥ dimers, and unreacted B␤ chains. In vitro translated fibrinogen chains gave ␥⅐␥ dimer, indicating the presence of some properly assembled fibrinogen, but a substantial amount of free A␣ and ␥ chains remained, indicating that a number of translated chains were not assembled into six-chain fibrinogen capable of undergoing polymerization and cross-linking (data not shown).

DISCUSSION
In hepatocytes, assembly of the component chains of fibrinogen into the final six-chain molecule occurs in a stepwise fashion. First, single chains are linked to form two-chain complexes, and later these duplexes acquire a third chain to form three-chain half-molecules. In the final step two half-molecules are joined to form the six-chain fibrinogen molecule (20,21,26,27). In the in vitro system this cellular assembly process appears to be duplicated, since similar precursor and intermediate products, are formed. Assembly of the chains requires GSSG and microsomal vesicles, derived from the ER. GSSG is needed for the in vitro formation of disulfide bonds in other proteins (30 -33). The ER, in contrast to the cytoplasm, maintains an oxidizing environment, with ratios of GSSG and GSH primarily contributing to its redox condition (34). Although this oxidizing environment ensures disulfide formation of most proteins, the process is facilitated by protein disulfide isomerase, an abundant ER enzyme that catalyzes thiol/disulfide interactions and promotes disulfide bond formation and isomerization (35,36). Therefore, it is not surprising that nascent fibrinogen, which contains 29 disulfide bonds, may interact transiently with protein disulfide isomerase and is co-precipitated with an antibody specific for this chaperone.
Fibrinogen chain assembly also requires microsomal membranes, and in their absence the expressed chains do not form recognizable chain complexes. The ER contains, in addition to protein disulfide isomerase, a number of resident proteins that function to recognize nascent proteins, prevent incorrect association or folding, and assist in proper folding and stabilization. These early co-and post-translational steps are required before the nascent proteins exit the ER and travel to the Golgi, where further processing occurs (17)(18)(19). In keeping with this view, antibodies to several ER chaperones co-precipitated free chains, intermediate fibrinogen complexes, and nascent fibrinogen. Previously it was shown, in both hepatocytes and recombinant systems, that BiP, a resident protein of the ER, is associated with fibrinogen (37). The association of nascent fibrinogen and its precursors with different chaperones varied. In all cases free A␣ and ␥ chains and A␣⅐␥ duplex, appeared to be complexed to all the chaperones. On the other hand, halfmolecules were not detected with BiP or calnexin, and six-chain fibrinogen was not noted together with calnexin. Failure to detect these associations does not necessarily mean that the nascent fibrinogen chains are not complexed to the chaperones, but it may reflect low recovery, perhaps due to a loose transient association, which could be disrupted by the experimental procedures employed. It is interesting to note that calnexin, which is an integral membrane protein of the ER and is thought to interact with nascent proteins during initial N-glycosylation FIG. 6. Antibodies to molecular chaperones co-isolate nascent fibrinogen. Antibodies to molecular chaperones or to fibrinogen were added to a detergent-solubilized mixture of nascent radioactive fibrinogen complexes produced by co-expression of all three chains in the presence of GSSG and microsomal membranes. The isolated proteins were separated by SDS-PAGE in nonreduced conditions (A) or by two-dimensional SDS-PAGE (B, C, D, E, and F). Autoradiograms are shown. A, lane 1, control, nonimmune rabbit serum; lane 2, antiserum to fibrinogen; lane 3, antibody to BiP; lane 4, antibody to Hsp90; lane 5, antibody to protein disulfide isomerase; lane 6, antibody to calnexin. B, two-dimensional SDS-PAGE of fibrinogen complexes isolated with anti-fibrinogen. C, proteins isolated with anti-BiP. D, proteins isolated with anti-Hsp90. E, proteins isolated with anti-protein disulfide isomerase. F, proteins isolated with anti-calnexin. (38 -42), appears to interact with the A␣ chain of fibrinogen, which is not normally N-glycosylated. The A␣ chain of human fibrinogen has two potential N-glycosylation sites at asparagine residues 269 and 400, but these residues are not N-glycosylated in vivo. Neither A␣ glycosylated in vitro, since treatment with endoglycosidase H did not affect nascent A␣ expressed in vitro but did affect B␤ and ␥. This is in agreement with other evidence that calnexin may not only interact with N-glycosylated proteins but that nonglycosylated domains may also be recognized (39,(42)(43)(44)(45)(46).
In vitro fibrinogen chain assembly resembles that occurring in hepatocytes and in recombinant systems. In all cases more free A␣ and ␥ chains than B␤ chains accumulate in the system. This could be due to rapid assembly of B␤ chains into larger complexes, coupled with degradation of free B␤ chains. Earlier studies with Hep G2 cells indicated that B␤ is tagged by BiP and is degraded in the ER (37). In the in vitro system there are also noticeable accumulations of several intermediates, including A␣⅐␥ and B␤⅐␥ duplexes, and half-molecules. In the in vitro system, as in steady-state labeled Hep G2 cells (27), there was little A␣⅐B␤ duplex formation. Co-expression of only A␣ and B␤ chains led to the formation of A␣ and B␤ homopolymers with only minor amounts of A␣⅐B␤ duplex assembly. Thus, both in vitro and in Hep G2 cells (20,21,27) fibrinogen chain assembly appears to commence with the formation of two-chain complexes, primarily A␣⅐␥ and B␤⅐␥, followed by the addition of a third chain to form half-molecules, which dimerize to form the six-chain final product. Covalent disulfide bond interactions, especially those that flank the coiled-coil region, are important in stabilizing the complexes and allowing chain assembly (7). Final linkage of the two half-molecules involves both noncovalent and disulfide bond interactions (47). Some of the halfmolecules noted, both in the in vitro system and in Hep G2 cells, could be noncovalently associated but may be separated in the denaturing SDS-PAGE procedure.
In contrast to intact cells in which the assembly process goes to completion, in vitro only about 20% of radioactive methionine was incorporated into fully formed fibrinogen. The remainder of the radioactivity was in free chains and intermediate products. It is difficult to determine how much of the six-chain molecule formed in vitro is properly folded. However, some of it appears to be properly assembled, since it underwent factor XIIIa-catalyzed cross-linking of ␥ fibrin chains. For ␥ chain cross-linking to be accomplished the in vitro product should be properly folded in order to allow correct alignment of the chains so that factor XIIIa-catalyzed formation of ⑀-(␥glutamyl) lysine isopeptide bridges between neighboring ␥ chains can occur.