Assembly and Secretion of Recombinant Human Fibrinogen*

Expression vectors containing full-length cDNAs for each of the human fibrinogen chains were constructed. COS-1 cells were transfected with single vectors, mix- tures of two, or with all three vectors and stable cell lines selected. Cells transfected with single vectors, or with mixtures of any two vectors, expressed the appropriate fibrinogen chains but did not secrete them. COS cells transfected with three vectors expressed all of the chains and secreted fibrinogen. COS cells transfected with three vectors contained, intracellularly, a mixture of fibrinogen-related proteins. The four main in- tracellular products were nascent fibrinogen, an Act . y complex, free Act chains, and free y chains. This is a similar pattern to that noted in Hep G2 cells. The intracellular forms of fibrinogen were sensitive to en- doglycosidase H, indicating that they reside in a pre-Golgi compartment. Secreted fibrinogen was endogly- cosidase H-insensitive, suggesting that the secreted glycoprotein moieties were processed in the normal manner. When mixed with plasma fibrinogen, radiolabeled recombinant fibrinogen was incorporated into a thrombin-induced clot. These studies demonstrate that COS cells transfected with all three fibrinogen chain cDNAs are capable of assembling and secreting a functional fibrinogen molecule. The incubation medium of COS-N,~,~ cells and of Hep G2 cells, incubated with L-["SJmethionine for 2 h at 37 "C, was collected. An aliquot (0.75 ml) was treated with 220 units/ml, final concentration, of Trasylol, and mixed with purified human plasma fibrinogen (1.4 mg/ml) and CaC12 (0.024 M). Some samples also contained 5 mM iodoacetamide to inhibit Factor XI11 that is usually found in most plasma fibrinogen preparations. This treatment blocks cross-linking that occurs with fibrinogen and other proteins. Clotting was initiated by the addition of 3 units/ml thrombin. The clot was allowed to form overnight in a sealed Centrex tube (Schleicher & Schuell) housing a 0.45-pm cellulose acetate filter in which the bottom of the upper stage was sealed. The next day the clot was percolated with 0.05 M Tris, 0.1 M NaCI, 1 mM EDTA, pH 7.4, until the radioactive background of the eluate had stabilized at its lowest level. The tube was then centrifuged to remove all liquid from the clot and the clot was hydrolyzed in 0.2 M NaOH containing 40% urea. The hydrolyzed clot was neutralized with HCI and radioactivity determined. As a control, radioactive medium from COS cells transfected with an expression vector that did not contain fibrinogen chain cDNAs was treated in the same manner.

Expression vectors containing full-length cDNAs for each of the human fibrinogen chains were constructed. COS-1 cells were transfected with single vectors, mixtures of two, or with all three vectors and stable cell lines selected. Cells transfected with single vectors, or with mixtures of any two vectors, expressed the appropriate fibrinogen chains but did not secrete them. COS cells transfected with three vectors expressed all of the chains and secreted fibrinogen. COS cells transfected with three vectors contained, intracellularly, a mixture of fibrinogen-related proteins. The four main intracellular products were nascent fibrinogen, an Act . y complex, free Act chains, and free y chains. This is a similar pattern to that noted in Hep G 2 cells. The intracellular forms of fibrinogen were sensitive to endoglycosidase H, indicating that they reside in a pre-Golgi compartment. Secreted fibrinogen was endoglycosidase H-insensitive, suggesting that the secreted glycoprotein moieties were processed in the normal manner. When mixed with plasma fibrinogen, radiolabeled recombinant fibrinogen was incorporated into a thrombin-induced clot. These studies demonstrate that COS cells transfected with all three fibrinogen chain cDNAs are capable of assembling and secreting a functional fibrinogen molecule.
Fibrinogen is composed of three different polypeptides (Aa, BP, and y), arranged as a dimer with each half-molecule containing a set of each of the chains. The two half-molecules are linked together by three disulfide bonds at the NH2terminal portions of the polypeptides. Two of the symmetrical bonds are between adjacent y chaics and one is between Aa chains. In addition a complex set of inter-and intrachain disulfide bonds (there are 29 disulfide bonds with no free sulfhydyl groups) are involved in maintaining proper structure Our studies are aimed at determining how this multichain protein is synthesized, assembled, and secreted. Hepatocytes are the principal site of synthesis and each of the component chains of fibrinogen is encoded by a separate gene (5-8).
Previously we demonstrated that Hep G2 cells have surplus pools of Aa and y chains that occur either as free chains or complexed to each other, primarily as an Aa. y complex (9)(10)(11). Hep G2 cells maintain these surplus amounts of Aa and y chains even when fibrinogen synthesis and secretion is stimulated by production of enhanced amounts of BO chain * This study was supported by National Institutes of Health Grant HL-37457. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

(12).
Pulse-chase experiments demonstrated that chain assembly commences by the attachment of preformed Aa and y chains to nascent B/3 chains. On completion of Be chain elongation, the B@. y and BP. Aa complexes are released into the lumen of the rough endoplasmic reticulum and acquire the third chain to form half-molecules. The two half-molecules are then joined to form dimeric fibrinogen. Chain assembly occurs in the rough endoplasmic reticulum (13).
To obtain further information on the mechanisms which govern chain assembly we prepared a set of stable transfected COS-1' cells which express either the individual fibrinogen chains, mixtures of two of the chains, or all three chains and studied the assembly and secretion of fibrinogen.

Materials
~-["'S]Methionine, approximately 1.1 Ci/mmol, was purchased from Du Pont-New England Nuclear, fetal calf serum from Hyclone, endoglycosidase H from Genzyme, geneticin from Sigma, restriction and modifying enzymes from Boehringer Mannheim, and T4 DNA ligase from New England Biolabs. Human fibrinogen (Imco, Stockholm), prepared as previously described (141, was stored at -70 "C as a stock solution of about 14 mg/ml in 50 mM Tris-HC1 (pH 7.4) buffer containing 100 mM NaCl and 1 mM EDTA. The fibronectin present in this preparation was removed by affinity chromatography on gelatin-Sepharose (15). Fibrinogen concentration was measured spectrophotometrically in alkaline urea using Et%, = 16.5 at 282 nm.

Construction of Expression Vectors
Full-length Aa and y fibrinogen chain cDNAs were cloned into the PstI site of BR322 (17,18) and were kind gifts from Dr. Dominic Chung, University of Washington, Seattle, WA. Both Aa and y cDNA have internal PstI sites and both also have stop codons at the 5' end. Therefore, to obtain full length Aa and y chain cDNAs, capable of being expressed, the following procedures were used to construct the expression vectors.
pBC1PBI-Aa-The Aa cDNA was released from pBR322 by treatment with MstI. The resulting 3.2-kb fragment (200 ng) was then digested with nuclease Bal 31 for 4 min to remove 50 bp from both ends so that the stop codon a t position -28 together with 22 bp of pBR322 sequence was removed. The resulting DNA fragment, which The abbreviations used are: COS, monkey kidney fibroblasts; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; RSV, Rous sarcoma virus; kb, kilobase(s); Neo, neomycinkanamycin-resistant gene; bp, base pair(s).
contains the full length coding region for Aa fibrinogen chain, was purified and blunt-ended with Klenow fragment followed by ligation with PhosphorylatedHindIII linker (10-mer, Boehringer Mannheim). This material was digested with HindIII/NcoI to create a HindIII site at the 5' end of the Aa cDNA and a NcoI site at the 3' end.
To prepare pBC12BI to receive the above fragment, pBC12BI was digested with BamHI/HindIII. The linear plasmid DNA was then ligated with phosphorylated BamHIIBlunt adaptor (ds24/20-mer) from Boehringer Mannheim that contained a NcoI site at the other end. The plasmid DNA (100 ng) containing the adaptor was digested with NcoI and then ligated with A a cDNA (described above). The resulting circular expression vector, having the Aa cDNA insert (Fig.  lA), was screened by digestion of a DNA minipreparation with BamHIIHindIII, which showed the presence of an insert of approximately 2 kb in the vector.
pRSVNeo-Aa-Aa cDNA was released from pBC12BIAa by digestion with BamHI/HindIII. The released Aa cDNA (200 ng) was ligated to 100 ng of pRSV Neo (19) that had been cut with HindIII and dephosphorylated with calf intestinal phosphatase. The linear vector containing An cDNA, was filled-in with Klenow fragment and then self-ligated to form the circular expression vector (Fig. l e ) . The correct orientation was determined by digesting the plasmid DNA with BamHI/HindIII and selecting the clones that yielded fragments pBCI2BI-y-The y fibrinogen chain cDNA, which had been cloned in pBR322 (17) contains stop codons at position -39 and -42 of the 5' end. To obtain the full length coding region without these stop codons, the y chain cDNA was released from pBR322 by digestion with Sac1 and HindIII. The released y chain cDNA (200 ng) was ligated to pBC12BI plasmid DNA (100 ng), that had been cut with HindIII, and then dephosphorylated. The linear vector containing y chain cDNA was filled-in with Klenow fragment and then self-ligated to form the circular expression vector (Fig. IC). The correct orientation was determined by digesting the plasmid DNA with BamHI/ HindIII and selecting the clones that yielded fragments of 5.3 kb and 200 bp. The 5.3-kb fragment is composed of 1.4 kb of y cDNA insert and 3.9 kb from the vector DNA.
pRSVNeo-y-The full-length y cDNA (200 ng), prepared as described above, was ligated with 100 ng of pRSVNeo plasmid DNA that had been cut with HindIII and then dephosphorylated. As described above the linear vector containing y cDNA was processed to form the circular expression vector (Fig. ID). The correct orientation was determined by digesting the plasmid DNA with BamHI/ HindIII that yielded a 4.8-kb fragment, containing 1.4 kb of y cDNA and 3.4 kb of vector DNA, and a 2.25-kb fragment of vector DNA.
General Methods All DNA fragments, obtained after restriction enzyme digestions, were purified by 1% agarose gel electrophoresis, electroelution, phenol-CHC13 extraction, and alcohol precipitation. Restriction enzyme digestions, dephosphorylation of vector DNA, fill-in reactions with Klenow fragment, ligation of cDNAs with vector DNAs, and phosphorylation of linker and adaptor were performed as described by Sambrook et al. (20).
Transformations with the constructed vectors were performed in Escherichia coli RRI competent cells on LB-agar plates containing 100 pg/ml ampicillin. Minipreparations of plasmid DNA, from selected bacterial colonies, were done by the alkaline-lysis method. To determine whether the cDNA inserts occurred in the correct orientation in the constructed vectors, the DNA was treated with appropriate restriction enzymes and the mobilities of the DNA fragments determined by electrophoresis on 1% agarose gels. Large scale preparations of plasmid DNA were performed by alkaline lysis of bacterial preparations followed by cesium chloride-ethidium bromide equilibrium gradient centrifugation. All of the above general methods were performed by standard procedures (20).

Transfection and Selection of Stable Cell Lines
To determine whether the Aor and y cDNAs were expressed in COS-1 cells, the cells were first transiently transfected with either pBC12BI-Aa or pBC12BI-y by the calcium phosphate method (21) using 5 pg of DNA/ml per 2 X IO5 cells in 60-mm culture dishes. The
To obtain stable cell lines COS-1 cells were transfected by the calcium phosphate method (21) with either pRSVNeo-BO, pRSVNeo-Aa, or pRSVNeo-y or with combinations of equal amounts of two of these expression vectors, or with equal amounts of all three expression vectors. In all cases 5 pg of DNA/ml of each vector was used. When only one, or a mixture of two vectors were used pRSVNeo DNA was added to increase the DNA concentration to 15 pg/ml. As a control, COS-I cells were also transfected with pRSVNeo (15 pg/ml) that did not contain fibrinogen cDNA inserts. The transfected cells were selected by resistance to 0.4 mg/ml geneticin for 5 weeks as previously described (12). After 2 weeks, two or three colonies remained in each of the transfected cell lines. After another 3 weeks the colonies were treated with trypsin, transferred to fresh plates, and allowed to grow to confluency in 60-mm plates in the presence of 0.4 mg/ml geneticin.' Incubation of Cells with L-T"SJMethionine Before labeling with ~-[~~S]methionine the cells were kept for 24 h without geneticin in Iscove's medium supplemented with 10% fetal calf serum and 1% glutamine. The 90% confluent cells were then labeled for 2 h at 37 "C in methionine-free Dulbecco's minimal essential medium (GIBCO) containing 200 pCi of ~-[~'S]methionine, 0.1 mg/ml heparin, and 1% glutamine (12).

Immunoprecipitation of Nascent Fibrinogen Chains
Radioactive fibrinogen chains were isolated by immunoprecipitation from cell lysates and from the cell medium. A rabbit polyclonal antibody that reacts with fibrinogen and its component chains was used (9, IO). Cell lysates were treated with iodoacetamide, detergents, and proteolytic inhibitors prior to immunoprecipitation. The chains were separated by SDS-PAGE under reduced and nonreduced conditions and detected by autoradiography. The nonreduced polypeptides were excised from the gels, reduced with mercaptoethanol, and re-electrophoresed on SDS-PAGE to identify the component chains. These procedures have previously been described (9)(10)(11)(12).
Protein radioactivity was determined by cutting out the radioactive areas from the polyacrylamide gels and counting by liquid scintillation spectrometry (22). In some cases relative amounts of radioactivity were determined by scanning autoradiograms with a Shimadzu densitometer.

Clotting of Recombinant Fibrinogen
The incubation medium of C O S -N ,~,~ cells and of Hep G2 cells, incubated with L-["SJmethionine for 2 h at 37 "C, was collected. An aliquot (0.75 ml) was treated with 220 units/ml, final concentration, of Trasylol, and mixed with purified human plasma fibrinogen (1.4 mg/ml) and CaC12 (0.024 M). Some samples also contained 5 mM iodoacetamide to inhibit Factor XI11 that is usually found in most plasma fibrinogen preparations. This treatment blocks cross-linking that occurs with fibrinogen and other proteins. Clotting was initiated by the addition of 3 units/ml thrombin. The clot was allowed to form overnight in a sealed Centrex tube (Schleicher & Schuell) housing a 0.45-pm cellulose acetate filter in which the bottom of the upper stage was sealed. The next day the clot was percolated with 0.05 M Tris, 0.1 M NaCI, 1 mM EDTA, pH 7.4, until the radioactive background of the eluate had stabilized at its lowest level. The tube was then centrifuged to remove all liquid from the clot and the clot was hydrolyzed in 0.2 M NaOH containing 40% urea. The hydrolyzed clot was neutralized with HCI and radioactivity determined. As a control, radioactive medium from COS cells transfected with an expression vector that did not contain fibrinogen chain cDNAs was treated in the same manner.

Quantitation of Secreted Fibrinogen
The amount of secreted fibrinogen present in the medium of cells incubated for 24 h at 37 "C was determined by an indirect competition enzyme-linked immunosorbant assay procedure using a monoclonal antibody (Fd4-7B3) that is specific for an epitope in the y chain of human fibrinogen fragment D (23). In brief, the assay procedure was as follows. Polyvinyl microtiter plates (Costar) were coated with pure human fibrinogen. An appropriate dilution of antibody was mixed with an equal volume of either buffer, pure human fibrinogen (concentration range of standard curve: 0.25-4.0 pglml), or media (from C O S -N ,~,~ or Hep G2 cells). After mixing, each sample was added to the fibrinogen-coated enzyme-linked immunosorbent assay plate. Following incubation and subsequent wash cycles, an appropriate dilution of peroxidase-conjugated rabbit immunoglobulin to mouse immunoglobulin was added. Enzyme-linked IgG binding was detected using a H202 and o-dianisidine solution. pBC12BI-Aa produced a radioactive protein which comigrated with authentic A a chains and COS cells transfected with pBC12BI-y expressed y chains (data not shown). Cells transiently transfected with pBC12BI-BP have previously been shown to express fibrinogen BP chains (16).

Expression
Knowing that COS cells are capable of expressing Aa, BP, and y chains of fibrinogen, we then transfected COS cells with pRSVNeo-Aa, pRSVNeo-BP, and pRSVNeo-y and selected stable transfected cell lines which were resistant to geneticin. Geneticin-resistant COS cells transfected with any one of the expression vectors expressed proteins which reacted with rabbit antibody to human fibrinogen and were of similar size to appropriate authentic human plasma fibrinogen chains. Cells transfected with combinations of two vectors containing different fibrinogen chain cDNAs (Aa and BP, Aa and y, and BP and y) synthesized both chains and cells transfected with all three vectors expressed the three fibrinogen chains (Fig. 2).
Analyses, in nonreducing conditions, of the fibrinogen chains produced by cells transfected with combinations of two vectors showed that Aa and BP, Aa and y, and BP and y  (Fig. 3). The size of the Aa.7 complex, as calculated from its electrophoretic mobility on SDS-PAGE, is larger than predicted but not large enough to suggest the presence of a third chain. In addition to these principal products, small amounts of larger size complexes were also noted, but no free chains were detected (Fig. 3 A ) . The chain composition of the complexes was determined by reduction with mercaptoethanol and re-electrophoresis of the products. The complexes yielded a mixture of the expected two chains (Fig. 3B).
Synthesis of Fibrinogen by COS-a,P,y Cells-COS-a,/3,y cells synthesized several fibrinogen-related proteins when analyzed under nonreducing conditions. The pattern noted is similar to that seen in Hep G2 cells (Fig. 4A). In COS-a,P,y cells, after 2 h of metabolic labeling with ~-["SS]methionine, 24.5% of the immunoprecipitable radioactivity was in fibrin- ogen, 22% in Aa. y complex, 6% in free Aa chains, and 41% in free y chains. A parallel experiment with Hep G2 cells showed 31% in fibrinogen, 20% in Aa.y,9% in free Aa, and 28% in free y chains. Thus both stable transfected COS cells and Hep G2 cells develop surplus amounts of Aa and y chains which reside intracellularly mainly as free y chain and as an Aa y complex.
The major intracellular forms of fibrinogen in COS-a,P,y cells were characterized, as had been done previously for Hep G2 cells, by excision of the radioactive bands from the polyacrylamide gel, reduction, and re-electrophoresis (10). Their chain compositions and estimated molecular weights allowed us to identify these as fibrinogen, Aa. y complex, and free Aa and y chains (Fig. 4B). Secretion-COS cells which expressed single fibrinogen chains, and those which expressed two of the chains, in any combination, did not secrete these proteins into the medium (Fig. 5A). These single and duplex radioactive fibrinogen chains were only detected in the cell lysate (Figs. 2 and 3). However, COS-a,P,y cells secreted the expressed proteins into the medium (Fig. 5A). When analyzed under nonreducing conditions the secreted fibrinogen chains were components of a high molecular weight disulfide-linked complex, with an apparent M, of 340,000 which is similar to that of plasma fibrinogen. This M, 340,000 complex accounts for 99.4% of the immunoprecipitable protein radioactivity secreted. No free fibrinogen chains, nor intermediate products of assembly were detected in the medium (Fig. 5B). A small amount of protein radioactivity (less than 1%) was sometimes noted at about 130 kDa and this may be due to leakage from the cell or may be a degradative product of fibrinogen. A similar pattern was noted in the secretion of fibrinogen by Hep G2 cells (Fig. 5B). In the case of fibrinogen secreted by Hep G2 cells 90.4% of the immunoprecipitable radioactivity occurred as fibrinogen ( M I 340,000) and a small amount, (-5%) was noted in a wide area, between 130 and 115 kDa (Fig. 5B).
Excision, reduction, and re-electrophoresis of the M, 340,000 radioactive protein secreted by COS-a,B,y cells showed that this large protein was composed of Aa, BP, and y chains. A similar pattern was noted in the fibrinogen secreted by Hep G2 cells (Fig. 5).
Endoglycosidase H Treatment of Nonsecreted Chains-Aa. y complex and free y chains expressed by transfected COS cells were treated with endoglycosidase H to determine whether nonsecreted y chain contains mannose-rich oligosaccharides which are cleaved by the enzyme, or whether the carbohydrates of the y chain had been further processed, making it insensitive to the enzyme (24-26). The y chain of the Aa. y complex on reduction, migrated on SDS-PAGE as a 48-kDa protein. On treatment with endoglycosidase H it migrated as a smaller 42-kDa protein (Fig. 6, lunes 1 and 2).
The free y chain, under nonreducing conditions, migrated as a 48-kDa protein and on treatment with endoglycosidase H it migrated faster, as a 42-kDa protein (Fig. 6, lunes 3 and 4 ) .
The BP and y chain components of intracellular fibrinogen, synthesized by COS-a,P,y cells were also both endoglycosidase H-sensitive (Fig. 6, lanes 5 and 6). Thus, the major intracellular forms of fibrinogen, Aa . y complex, the free y chain, and nascent fibrinogen, are all endoglycosidase H-sensitive, indicating that they accumulate or are retained in a pre-truns-Golgi membrane compartment. Similar results were obtained when nascent intracellular fibrinogen, synthesized by Hep G2 cells, was analyzed (data not shown).
By contrast, the glycoprotein chains of secreted fibrinogen, produced by either COS-a,P,r (Fig. 6, lanes 7 and 8) or Hep G2 cells (not shown), are endoglycosidase H-insensitive. This suggests that recombinant fibrinogen follows the conventional secretory pathway with normal glycosylation.
Clotting of Recombinant Fibrinogen-To determine whether secreted recombinant fibrinogen is capable of clotting, the incubation medium of control (nontransfected) and COSa,P,y cells, incubated for 24 h with ~-[~~S ] m e t h i o n i n e , was mixed with human plasma fibrinogen and induced to clot by the addition of thrombin. The clotting ability of recombinant fibrinogen was compared to that of fibrinogen secreted by Hep G2 cells. Clots formed in the presence of radiolabeled media from COS cells that were not transfected with the fibrinogen chain cDNAs had only background amounts of radioactivity associated with the clot matrix (2-3% of total trichloroacetic acid-precipitable radioactivity from the media). By contrast, clots formed in the presence of radiolabeled media from COS-a,/3,y cells, or from Hep G2 cells, had 30 to 45 times background levels of radioactivity associated with the clot. Clotting in the absence of Factor XI11 cross-linking (i.e. in the presence of iodoacetamide) also produced highly radiolabeled clots, 18 to 24 times the background level. This indicates that the radiolabeled secreted fibrinogen became associated with the clot matrix through a thrombin-dependent polymerization mechanism.
Amount of Recombinant Fibrinogen Secreted-COS-a,P,y cells secreted comparable amounts of fibrinogen as compared to Hep G2 cells. In two experiments COS-a,P,y cells (2 x lo6 cells) secreted an average of 2.08 pg of fibrinogen in 24 h and the same number of Hep G2 cells secreted 1.94 pg of fibrinogen.

DISCUSSION
Fibrinogen is a multichain protein with a well ordered structure. It is sensitive to thrombin and acts in the final stages of blood clotting. Fibrinogen assembly which involves the arrangement of three different polypeptides into a symmetrical dimer probably occurs on structures within the endoplasmic reticulum which mediate proper alignment of the chains and also specific disulfide interactions. As such, a group of proteins known to be present in the lumen of the endoplasmic reticulum which probably include the immunoglobulin binding protein and protein disulfide isomerase are likely to be involved in a concerted effort to assemble the various chains into a functional molecule. (For reviews see Refs. 27,28). Previously, the individual chains of fibrinogen have been expressed in surrogate cells, either E. coli (29)(30)(31) or in COS cells (16). However, expression, assembly, and secretion of fully formed, functional recombinant fibrinogen has not been reported. We show that COS cells, transfected with single fibrinogen chain cDNAs or with any combination of two fibrinogen chain cDNAs, express the appropriate fibrinogen chains but cannot secrete them. In contrast, COS cells containing all three fibrinogen chain cDNAs express, assemble, and secrete the chains in a form which is capable of forming a thrombin-induced clot. This indicates that factors needed for proper assembly of fibrinogen chains are not restricted to the two tissues, hepatocytes and megakaryocytes, which normally express fibrinogen (32,33). This further demonstrates that for fibrinogen chains to be properly transported and secreted they must exist as part of fully formed dimeric fibrinogen. This suggests that intact fibrinogen contains a signal which allows intracellular transport and secretion to occur and that individual chains are recognized as products not to be secreted. I n vivo free fibrinogen chains have not been detected in the circulation. In dogs, injected with radioactive amino acids, nearly all of the secreted fibrinogen chain radioactivity is accounted for in fibrinogen (34) and studies with cells in culture have indicated that fully formed fibrinogen is the main, if not the only form, of secreted fibrinogen chains (35,36). This occurs in spite of the fact that in hepatocytes of several species studied, there is a surplus of two of the component chains of fibrinogen (9,10,(35)(36)(37). In dogs (34) and rabbits (38) surplus chains have not been detected intracellularly, but different specific radioactivities of the component chains of secreted fibrinogen indicate that pools of Aa and y chains may occur. Thus, hepatocytes have a mechanism for distinguishing the surplus forms of fibrinogen chains from fully formed fibrinogen. A similar mechanism occurs in transfected COS cells. The stable transfected COS cell lines only secrete fully formed fibrinogen. Under nonreducing conditions 99.4% of the secreted immunoprecipitable protein is fibrinogen. There is less than 1% protein radioactivity in lower molecular weight proteins. As evidenced by sensitivity to endoglycosidase H treatment, the nonsecreted fibrinogen chains, and also nascent fibrinogen which is not yet fully processed, are retained in a pre-Golgi compartment. This is similar to the assembly and degradation of other heteroligomeric proteins. In both the human asialoglycoprotein receptor (39) and the T-cell receptor proteins (40,41) surplus chains are produced and some of these excess chains are degraded in a nonlysosomal pre-Golgi compartment.
Previous pulse-chase experiments, which carefully measured kinetic precursor-product relationships in Hep G2 cells, showed that surplus Aa and y fibrinogen chains participate in fibrinogen synthesis and assembly and that unused chains are retained and degraded intracellularly (9,10). In COSa,P,y cells incubated for 2 h with ~-[~~S ] m e t h i o n i n e , which is near steady-state conditions, most of the radioactivity in fibrinogen chains occurs in three forms; as fully assembled fibrinogen whose carbohydrates have not been completely processed, as an A a . y complex and as free y chains (Fig. 4). In addition, some free Aa chains, and other intermediate forms, account for a small percentage of intracellular fibrinogen chains. This pattern is similar to that noted in Hep G2 cells and suggests that COS-a,P,y cells assemble chains in a similar manner to Hep G2 cells. Kinetic pulse-chase experiments have not yet been performed with transfected COS cells and it is not clear whether all the intracellular precursor fibrinogen forms detected in Hep G2 cells are also present in transfected COS cells; or whether all of the intermediate forms present in COS cells participate in fibrinogen assembly. However, it is apparent that, as in hepatocytes of all species studied, surplus y chains are generated in COS cells during fibrinogen assembly.
The mechanism by which surplus y chains are generated is not understood. In Hep G2 cells, the initial rates of synthesis of the three chains are unequal with that of BP being less than that of Aa and y (10). However, unequal degradative rates have not been ruled out. In transfected COS cells the expression of the three fibrinogen chains is driven by the same viral promoter present in the expression vector pRSVNeo and thus regulation is unlikely to occur at the nuclear level; although we cannot rule out that different mRNAs are exported from the nucleus at different rates or that they have different stabilities. More likely the generation of surplus y chains in transfected COS cells is a consequence of the chain assembly process and is probably a posttranslational event.