The Functional Domains of Coagulation Factor VIII:C*

A lack of factor VIII:C, manifested as a bleeding disorder due to the absence of clot formation, is known as hemophilia A, an X chromosome-linked inherited disease afflicting 1-2 males/10,000. To determine the minimum functional domain(s) essential for factor VIII:C activity, we have expressed the amino-terminal (92-kDa) and carboxyl-terminal (80-kDa) proteolytic cleavage products as individual, secreted polypeptides in monkey cells without the 909-residue central region. We have found that neither terminal domain alone is able to promote coagulation in factor VIII:C-deficient plasma. However, when the 92- and 80-kDa peptides are co-expressed, clotting activity is readily detected. Thus, these two chains alone constitute an active or activatable complex. The central domain is required neither for activity nor for the assembly of an active complex from two chains expressed in trans. These results suggest that a truncated derivative of factor VIII:C may be useful in coagulation therapy.

In addition to the proteolytic processing, an examination of the sequence of factor V1II:C suggests the existence of three domains. The amino and carboxyl termini contain the majority of the cysteine residues, but a paucity of potential Nlinked glycosylation sites, whereas the interior linker region has the opposite distribution. The homology of factor VIIIC to another protein in the coagulation cascade, factor V, also suggests a multidomain structure. Activated bovine factor V has been shown to consist of two polypeptide chains, the amino-and carboxyl-terminal species, linked by a calcium bridge (14)(15)(16)(17). Both chains seem to be required for factor V activity. The presence of a similar calcium bridge linking the amino-terminal domain to the carboxyl-terminal domain of factor V1II:C has been demonstrated by us and others (8,13,18).
By engineering the factor VII1:C cDNA, we have expressed the amino-terminal 92-kDa chain and the carboxyl-terminal 80-kDa chain as separate proteins, Co-expression of these individual 92-and 80-kDa regions results in a level of factor V1II:C activity comparable to that obtained from the expression of the whole factor VII1:C coding region. These proteins assemble together to form a thrombin-activatable complex. The presence of the interior linker region, containing 909 amino acids or about 40% of the total for the intact protein, is not required for factor VII1:C activity assayed in uitro. Moreover, assembly of the complex does not require the presence of the central region and occurs efficiently for the two chains expressed in trans. Factor Vll1:C Assays Coagulation-Aliquota of 75 pl of medium conditioned by the growth of COS cells transfected with the indicated plasmids or mocktransfected were assayed for their ability to decrease the prolonged partial thromboplastin time of factor VIIIC-deficient plasma in the one-stage assay (23). Briefly, 75 pl of Platelin (General Diagnostics) was incubated for 3 min at 37 "C followed by the addition of 75 rl of factor VIIIC-deficient plasma plus 75 p1 of the test sample for an additional 5-min incubation at 37 "C. A 75-pl aliquot of prewarmed 0.025 M CaClz was added, and the clotting time was measured with a Becton Dickinson fibrometer. Normal human plasma diluted in COS Functional Domains of Factor VI1I:C cell medium was used as a standard and is assumed to contain 1 milliunit of factor VII1:C activity per pl, corresponding to approximately 100 pg of factor V1II:C protein (4). For the coagulation inhibition assay, 160 p1 of the indicated conditioned medium was incubated with 20 p1 of a 100-fold dilution of a human factor VII1:C inhibitory serum (24) (Bethesda titer 1500 units/ml) or a similar dilution of pooled normal human serum or diluent alone (50 m M imidazole, 0.1 M NaCl, 100 pg/ml bovine serum albumin, pH 7.3) for 2 min at 37 "C. These samples were then assayed for residual coagulation activity.

DNA
Chromogenic-The Coatest assay (obtained from Kabi) measures the generation of activated factor X (Xa) as a linear function of the concentration of factor VIIIC (25). The concentration of factor Xa is measured by the proteolytic cleavage of p-nitroaniline from a synthetic peptide substrate for Xa. Normal human plasma diluted in COS cell medium was used as the standard. Aliquots of 20 pl of conditioned medium were measured in the end point assay as outlined by the manufacturer except that the assay volumes were uniformly The VII1:C 92-kDa protein was isolated from plasma as described (8) and was greater than 50 percent homogenous as estimated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and silver staining. After incubation for 16 h at room temperature, the plates were washed, and the amount of '=I in the individual wells was measured in a y counter. An intermediate purified commercial factor V1II:C preparation (factor VIII, NORDISK) with a specific activity of 0.5 unit of coagulation activity per mg was used as the standard. This standard was calibrated against the World Health Organization Third International factor VI1I:C standard. We defined our intermediate purified standard to contain a 92-kDa RIA activity/ factor VI1I:C coagulation activity ratio of 1. Enzyme-linked Immunosorbent Assay (ELISA) Specific for the 80-kDa Polypeptide-The ELISA assay was performed as described (24). Normal human plasma was used as a standard.

Partial Purification of a Coagulation Active Complex
One hundred pg of an anti-80-kDa monoclonal antibody (56 I&) (26) coupled to Sepharose CL-4B was incubated overnight at 20 "C with 1.4 ml of medium obtained from COS cells co-transfected with plasmids pSVF8-92 and pSVF8-80 containing a total of 6.2 milliunits of activity (measured by the Coatest assay). After incubation, the slurry was loaded into a column, and the unretarded fraction was collected. The column was washed with 300 #I of 50 mM imidazole, 0.1 M NaCl, 0.1% sodium insulin, 0.2% NaN3, pH 7.3, and then eluted with 300 rl of 2.

RESULTS AND DISCUSSION
To determine the functional significance of the apparent domain structure of factor VIII:C, a series of three plasmids was constructed to express and secrete the complete factor VII1:C protein, as well as both terminal regions as separate proteins. The amino-terminal 92-kDa domain is encoded by the expression vector pSVF8-92 (Fig. lC), a plasmid which specifies the first 759 amino acids of factor VIIIC ( Fig. 1A) including the 19-residue signal sequence and which terminates at a thrombin site ( A~g~~-S e r '~l ) present in the intact protein. The carboxyl-terminal 80-kDa domain was expressed as a 684-residue polypeptide by plasmid pSVF8-80 (Fig. ID). A 'The abbreviations used are: RIA, radioimmunoassay; ELISA, enzyme-linked immunosorbent assay; bp, base pairs. heterologous signal sequence was provided at the amino terminus of this protein by a precise fusion to the first 35 amino acids of tissue type plasminogen activator (28,29). This fusion generates an Arg-Ser cleavage site at the tissue plasminogen activator-factor VII1:C junction to substitute for the Arg"-Glula9 proteolytic site used in the release of the 80-kDa domain from intact factor V1II:C. For purposes of comparison, the entire factor VIIIC coding sequence was placed into the vector pSV7d to generate the plasmid pSVF8-200 (Fig. 1E)

(8).
Plasmids pSVF8-92, pSVF8-80, and pSVF8-200 were transfected into COS 7 cells both singly and together as noted in Table I to achieve transient expression of the cloned genes. At specific intervals post-transfection, conditioned medium was removed from the cells and tested for the ability to decrease the prolonged partial thromboplastin time of factor VI1I:C-deficient plasma in a standard coagulation assay (23). The more specific Coatest assay (25), which measures the generation of activated factor X (Xa) as a linear function of the concentration of exogenously supplied factor VIII:C, was used to verify the results of the coagulation assay. The concentration of immunologically reactive factor VI1I:C protein in the medium was determined by the use of a radioimmunoassay developed to detect the 92-kDa polypeptide and by an ELISA specific for the 80-kDa polypeptide (24). As shown in Table I, expression of the 92-kDa polypeptide or of the 80-kDa polypeptide alone produced no detectable activity even though high levels of each of the individual proteins were present in the conditioned medium. However, the medium obtained from cells co-transfected with both plasmids contained about 20 milliunitsjml of coagulation activity. The same relative level of coagulation activity was secreted by cells transfected with the plasmid pSVF8-200 encoding the complete factor V1II:C protein. These results indicate that a complex of the amino-and carboxyl-terminal domains of factor VIIIC retains intrinsic coagulation activity and that the interior domain is not essential for activity nor for the assembly of an active complex from separate chains.
The results from Table I would suggest that the level of coagulation activity of the two-chain complex is limited by the expression of the 92-kDa chain since the ratio of factor V1II:C chromogenic activity/92-kDa RIA activity is 0.4, whereas the ratio of factor V1II:C chromogenic activity/80-kDa ELISA activity is 0.018. Also, it appears that the ratio of Coatest activity/92-kDa RIA activity is approximately the same for the complex as for the entire protein, 0.4 and 0.36, respectively, suggesting that assembly of the complex is a relatively efficient process. However, we note that the absolute amounts of the 92-kDa protein, the 80-kDa protein, and the 330-kDa protein are not strictly comparable, as the immunoassays and the coagulation assays may be dependent on the exact molecular form of the molecule(s). Furthermore, different standards are used in the assays. It is, however, obvious that there is a large excess of coagulation-inactive 80-kDa protein in the medium from the co-transfectants.
When conditioned media from the pSVF8-92 and the pSVF8-80 single transfectants were mixed together (using several different conditions as outlined in Table I), no activity was measurable. The failure to reconstitute the activit9 of the complex in uitro could result from the instability of one (or both) polypeptides when present individually. Alternatively, the concentration of one (or both) of the individual peptides in the media may be too low to support complex formation. These results contrast with those obtained with factor v where the analogous polypeptides can be reassembled into en active complex from the isolated chains in vitro ( 16, 17). The 2267-bp SacI to HindIII fragment from the factor VI1I:C cDNA, and pSV7d from HindIII to BamHI. D, the 80-kDa polypeptide expression vector. Starting from the SalI site in the polylinker of pSVITd, pSVF8-80 consists of a 201-bp fragment of a tissue plasminogen activator cDNA from nucleotides -98 to +lo3 terminating at a BglII site (28,29), a 29-bp synthetic BglII to BclI linker-adapter encoding nucleotides +5002 to +5031 of factor V1II:C (8) ligated to a 2464-bp BclI fragment of factor VIIIC spanning from a BclI site created at nucleotide 5028 of the factor V1II:C cDNA through in vitro mutagenesis (33), to a BclI site in the 3' untranslated region, at nucleotide 7492, and a 400-bp fragment of tissue plasminogen activator 3' untranslated sequence spanning from a BglII site at nucleotide 2076 to a synthetic PstI site generated from cDNA cloning, followed by the polylinker from the vector M13mp9 (34) and then pSV7d. E, the complete factor VII1:C protein expression vector pSVF8-200. This plasmid contains the entire factor VII1:C cDNA coding and 3' untranslated sequences (8). The 5' untranslated sequences are the same as described above for pSVF8-92.
sensitivity limits of the assays correspond to about 0.9 milliunit/ml for the coagulation test and 1.0 milliunit/ml for the Coatest assay. Thus, if active individually, the separate polypeptide chains must be at least 1000-fold less active than the complex under the conditions tested.
To verify that the observed coagulation activity was due to factor VIIIC, the sensitivity of the coagulation to inhibition by an antibody specific for factor VI1I:C was determined. Prior to assay, aliquots of conditioned medium was preincubated with normal human serum or serum from a hemophiliac who had developed a high titer of inhibitory antibodies to factor VIIIC (24). The coagulation activity of the complete molecule and that of the 92-80-kDa complex were both reduced specifically by the inhibitory serum (Table 11). The same results were obtained using three different inhibitory mouse monoclonal antibodies which bind to the 80-kDa species (data not shown).
To demonstrate more clearly the existence of a two-chain complex, we have partially purified the active species from the COS cell medium by passage over a monoclonal antibody column directed against the 80-kDa portion. Approximately 65% of the applied activity was retained, and 50% of this bound material was eluted in an active and 5-fold concentrated form (Table 111). Thus, an active complex can be  4 . 0 ndb nd a A variety of mixing conditions were tested including preincubation for various times up to 2 h at 37 'C, 20 "C, or 4 "C in the presence or absence of 10 mM CaCL The value reported in this table is representative of the data obtained. Some transfections have produced as much as 950 milliunits/ml of 92-kDa RIA activity and up to 4350 milliunits/ml of 80-kDa ELISA activity. Even at these much higher expression levels, neither individual subunit has coarmlation activitv nor is an active comulex uroduced by in vitro mixing.

TABLE I1
Coagulation inhibition assay A 160-pl aliquot of conditioned medium from COS7 cells transfected with the indicated plasmids was preincubated with 20 pl of a 100-fold dilution of a human factor VIIIC inhibitory serum (Bethesda titer 1500 units) or a similar dilution of pooled normal human serum or diluent alone for 2 h at 37 "C. These samples were then assayed using the one-stage coagulation assay (23).

I11
Partial purification of 92-80-kDa coagulation active complex Conditioned medium from COS7 cells co-transfected with the plasmids pSVF8-92 plus pSVF8-80, 1.4 ml in total, containing 6.2 milliunits of Coatest activity, was absorbed batchwise to an 80-kDa chain-specific monoclonal antibody coupled to Sepharose CL-4B. The slurry was loaded into a column, and the unbound fraction was collected. The column was washed with 50 mM imidazole, 0.1 M NaCl, 0.1% sodium insulin, 0.2% NaN3, pH 7.3, and then eluted with the same buffer supplemented with 2.5 M NaCl, 50% ethylene glycol, and 0. We have also tried to purify the complex by chromatography on wheat germ lectin-Sepharose. Earlier experiments had shown that human factor VII1:C purified from cryopreciptate binds quantitatively to this matrix, and that 50% of the activity is readily eluted by a buffer containing 0.5 M Nacetylglucosamine.' However, when the recombinant 92-kDa-80-kDa complex was applied to the same column, no activity was bound. This result suggests that the presence of the internal domain, which contains the vast majority of potential R. L. Burke and M. Truett, unpublished results. N-linked glycosylation sites, is necessary for binding to wheat germ lectin.
We have shown here by expression of specific domains of the factor VII1:C protein that the amino-and carboxyl-terminal two-chain complex is sufficient to generate activity in an in vitro coagulation assay. This conclusion has also been reached by other workers in the field based on very different experimental approaches. For example, monoclonal antibodies which inhibit the coagulation activity of factor VII1:C in vitro have been found to bind to epitopes within the 80-kDa domain, as well as within the 92-kDa domain, whereas monoclonal antibodies directed against epitopes within the central region showed no inhibition (2, 3, 25). It has recently been shown that a M, = 170,000 complex consisting of two polypeptide chains of M, = 90,000 and M, = 80,000 can be purified from human plasma (10). Although this material is a minor fraction of the factor V1II:C activity, it has the same high specific activity as the major peak (13). Also, Toole and coworkers (35) reported that a factor VII1:C molecule with a large deletion in the heavily glycosylated region produced in vitro procoagulant activity when transfected into COS cells.
Although it is now accepted that the 92-and 80-kDa chains are the active domains of factor VIII:C, there remains some controversy concerning the molecular profile of the thrombinactivated species. Do the intact 92-and 80-kDa chains retain some level of intrinsic activity or, instead, are they a procoagulant complex which is activated in the coagulation assays? Based on SDS-polyacrylamide gel profiles of purified factor VIIIC protein during the time course of thrombin activation, Rotblat et al. (2) and Fulcher et al. (12) have both suggested that the intact 92-80-kDa chain complex is the active species, and, thus, that thrombin activation involves the release of the 92-kDa chain from higher molecular weight species. However, using the same technique, Eaton et al. (9) concluded that the generation of chains of M, = 50,000, 43,000, and 73,000 correlated with the maximum activity of human factor VI1I:C and Fass et al. (18) obtained analogous results for the porcine molecule. These discrepancies may be due to the particular conditions used for activation; in particular, the Ca'+ concentration has been shown to affect the species generated by thrombin (9). Recently, Anderson et al. (13) showed that the purified 90-80-kDa complex was activatable by thrombin yielding chains of 52 and 43 kDa, and a doublet at 70 kDa. We, too, find that the 92-80-kDa complex material produced by COS cells is a~tivatable,~ suggesting that maximum activity is produced by cleavage of the 92-and 80-kDa chains. HOWever, the molecular steps of the activation cascade are still R. L. Burke, unpublished results. not clear. Are the intact 92-and 80-kDa chains partially active? Is it cleavage of the 92-kDa chain or of the 80-kDa chain or both which is required for activation? These types of questions may be answered by altering the molecules by in vitro mutagenesis so as to eliminate the possibility of thrombin cleavage at particular sites. Our results shown here, that the interior domain is not essential for the assembly of an active complex, and, moreover, that each terminal chain may be made as a separate protein, make a functional dissection of the protein simpler.
These results also point to the question of what the central domain of factor VII1:C does in uiuo. The work of Brinkhous et al. (10) showed that a 170-kDa complex of the two terminal chains could combine with von Willebrand's factor and restore hemostasis in hemophiliac dogs lacking factor VII1:C and that this preparation had a slightly longer half-life than larger complexes with active components of M, = 185,000-280,000.
Thus, it is clear that a 92-80-kDa complex is sufficient to restore hemostasis, and that a definition of the role of the linker region requires further analysis.