Expression of Human Soluble Tissue Factor in Yeast and Enzymatic Properties of Its Complex with Factor VIIa*

The extracellular domain of human tissue factor (TF, amino acids 1-217) was expressed in Saccharomyces cerevisiae, using the inducible yeast acid phosphatase promoter and the yeast invertase signal sequence to direct its secretion into the culture broth. Two active soluble forms sTFa (high molecular weight form) and sTFB (low molecular weight form) were purified, the yield being approximately 10 and 1 mglliter of culture supernatant, respectively. sTFa had an apparent molecular mass of 150 kDa on SDS-polyacrylamide gel electrophoresis and contained more than 200 residues of mannose/mol of protein. sTF@ had an apparent molecular mass of 37 kDa and contained 22 residues of mannose/mol of protein. N-Glycosidase F treatments of both rTFs reduced the apparent molecular mass to 35 kDa. The amino-terminal sequences and amino acid compositions of sTFa and sTF@ were consistent with those deduced from the cDNA sequence, thereby indi-cating that the difference in molecular mass is caused by heterogeneity of oligosaccharide structures. Of these recombinant TFs, sTFB enhanced factor VIIa-amidolytic

The extracellular domain of human tissue factor (TF, amino acids 1-217) was expressed in Saccharomyces cerevisiae, using the inducible yeast acid phosphatase promoter and the yeast invertase signal sequence to direct its secretion into the culture broth. Two active soluble forms sTFa (high molecular weight form) and sTFB (low molecular weight form) were purified, the yield being approximately 10 and 1 mglliter of culture supernatant, respectively. sTFa had an apparent molecular mass of 150 kDa on SDS-polyacrylamide gel electrophoresis and contained more than 200 residues of mannose/mol of protein. sTF@ had an apparent molecular mass of 37 kDa and contained 22 residues of mannose/mol of protein. N-Glycosidase F treatments of both rTFs reduced the apparent molecular mass to 35 kDa. The amino-terminal sequences and amino acid compositions of sTFa and sTF@ were consistent with those deduced from the cDNA sequence, thereby indicating that the difference in molecular mass is caused by heterogeneity of oligosaccharide structures. Of these recombinant TFs, sTFB enhanced factor VIIaamidolytic activity 40-fold toward the chromogenic substrate and 147-fold toward the fluorogenic substrate, affecting mainly the kcat value. The enhancement was comparable with that of TF purified from human placenta. The TF-mediated enhancement of factor VIIa-amidolytic activity was inhibited by heparinactivated antithrombin 111, forming a high molecular weight complex. As treatment of sTFj3 with denaturants such as guanidine hydrochloride or urea led to a biphasic loss of the activity, the extracellular domain of TF probably consists of two discrete domains. This expression system provides a significant amount of the extracellular domain of TF so that studies of interactions with factor VI1 are feasible.
* This work was supported in part by a grant-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan and by Grant HLBI 14147 (Specialized Center for Research in Thrombosis) from the National Institutes of Health (to J. E. S.). 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. $$ To whom correspondence should be addressed.
Tissue factor (TF)' is a membrane-bound glycoprotein which forms a complex with factor VII(a) to initiate extrinsic blood coagulation. Factor VIIa bound to membrane surface TF triggers the extrinsic pathway by activating zymogen factor X to the active form, factor Xa and also activates zymogen factor IX to the active form, factor IXa (1). cDNA clones have been isolated for human (2)(3)(4)(5), mouse (6, 7), rabbit (8,9), and bovine TF (10). Human TF consists of 263 amino acids and is composed of three domains: an extracellular one (amino acids 1-219) containing three potential Nlinked glycosylation sites and four cysteine residues, a transmembrane region (amino acids 220-242) consisting of highly hydrophobic residues, and a cytoplasmic tail (amino acids 243-263) containing one N-linked potential glycosylation site and one cysteine residue linked covalently with palmityl or stearyl thioester. The cDNA cloning has facilitated expression of the recombinant T F in Escherichia coli, a human kidney cell line (ll), and a Chinese hamster ovary cell line (12).
Genetically engineered soluble TF (sTF) lacking the transmembrane region and the cytoplasmic tail has been expressed in a Chinese hamster ovary cell line (13). Using the recombinant TF, the extracellular domain of sTF was seen to bind to factor VIIa, the result being enhancement of the amidolytic activity of factor VIIa toward synthetic peptidyl substrates. However, there is a paucity of information on the structural and functional aspects of the extracellular domain of sTF and on the molecular interaction with factor VII(a) because of difficulties in preparing a large amount of homogeneous TF from native tissue and from cell culture systems. Here we report the purification, characterization, and expression of the extracellular domain of TF in Saccharomyces cereuisiae.
Yeast is a eukaryote and has an elaborate secretory pathway which mediates post-translational modifications and secretion of proteins. Recombinant sTF expressed in yeast enhances factor VIIa activity toward peptidyl substrates showing much the same efficiency as that of native TF; hence this expression system provides a functional sTF.
Portions of this paper (including "Materials and Methods," Tables IM-IIIM, and Figs. 1M and 2M) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

Human Tissue
Factor Expressed in Yeast

RESULTS
Purification of Secreted sTFn and sTFP-Three liters of yeast culture were centrifuged to remove cells, and the supernatant was applied to a CM-Sepharose CL-GB column equilibrated with 20 mM sodium acetate buffer, pH 4.0. After washing the column with the equilibration buffer, the adsorbed proteins were eluted with a linear gradient of sodium chloride. The T F activity measured was based on potentiation of factor VIIa-catalyzed hydrolysis toward a peptidyl chromogenic substrate, S-2288, in the presence of TF. As shown in Fig. 2M, the T F activities were eluted into two peaks; the former was designated as sTFa and the latter as sTFB.
SDS-PAGE analyses of the sTFn fractions from CM-Sepharose chromatography indicated that fraction 28 contained a 150-kDa protein band and fraction 36 a 150-kDa protein band with a small amount of contaminant proteins (data not shown). The sTFB fraction 48 contained 39-and 37-kDa proteins, both of which reacted with monoclonal and polyclonal antibodies against human TF. Fraction 52 had the highest cofactor activity in the sTF@ fractions and contained mainly 37-kDa protein (data not shown). For further purification, fractions were pooled as indicated by bars in Fig. 2M, and each pool was applied, separately, to an anti-human TF monoclonal IgG column. Two forms of sTF were purified using these two-step procedures. The purification of sTFn and sTFB from yeast culture medium is summarized in Table  IM. From 3 liters of yeast culture medium, 28 mg of sTFn and 3 mg of sTFP were obtained.
SDS-PAGE of Purified sTFn and sTFIj- Fig. 1 shows an SDS-PAGE pattern of purified sTFn and sTF/3. The sTFtv sample gave a broad band with a molecular mass of 150 kDa, under both nonreducing and reducing conditions (lanes I and 2). In cont.rast, sTFP gave a single band with an apparent molecular mass of 37 kDa under nonreducing (lane 3 ) and 39 kDa (lane 4 ) under reducing conditions. The predicted molecular mass of 24,672 for the sTF polypeptide chain was considerably lower than the 150-and 39-kDa bands estimated by SDS-PAGE; these differences may be due to glycosylation of the recombinant proteins. T o examine the post-translational modifications of the recombinant proteins, the purified sTFa and sTFP were digested with N-glycosidase F. As shown in  were indistinguishable and consistent with values deduced from the sequence of cDNA used for their construction (Table  IIM). Moreover, the amino-terminal sequences of sTFtr and sTFB up to 20 residues agreed with the predicted sequence from the cDNA except for Asn-1 1 (Table IIIM), which was a putative carbohydrate attachment site.
On the other hand, the carboxyl-terminal residues of sTFn and sTF/j determined by vapor-phase hydrazinolysis did not agree with the carboxyl-terminal Arg-218 predicted from the cDNA sequence of the construct; instead, they yielded phenylalanine (0.24 mol/mol of sTFa and 0.23 mol/mol of sTF& respectively (uncorrected values)). Furthermore, carboxypeptidase B treatment with 1 nmol of sTFn yielded only a basal level of arginine. These results suggest that the carboxylterminal Arg-218 had been removed by a carboxypeptidase during the secretion of sTF.
In addition, we examined the internal structure of recombinant sTFa by sequencing peptide fragments derived from a lysyl endopeptidase digest. A total of 198 out of 217 amino acid residues and two carbohydrate attachment sites, Asn-124 and Asn-137, were confirmed for sTFtr (data not shown).
Component sugar analyses showed that sTFtr contained 4.1 mol of glucosamine and 272.3 mol of mannose/mol of protein and sTFP contained 2.9 mol of glucosamine and 22.1 mol of mannose/mol of protein.
Functional Actiuity of sTFm-To examine the functional activity of the recombinant protein, the cofactor activity of T F for factor VIIa-catalyzed hydrolysis of a synthetic substrate was measured. As shown in Fig. 3, factor VIIa activity was enhanced by increasing the amount of sTFtr or sTFd. The enhancement reached a plateau at over 1000 nM sTF. The apparent dissociation constants of sTFrr and sTF/j with factor VIIa were 8.98 x lo-' and 4.74 X 10"' M , respectively, when concentration of the complexes was calculated from specific activities of 1.53 mmol of p-nitroaniline/min/nmol of complex and 1.60 mmol of p-nitroaniline/min/nmol of complex for sTFn/factor VIIa and sTF(j/factor VIIa, respectively. T o compare the recombinant proteins with human placental TF, cofactor activity was measured at various concentrations (0-1.8 nM) of human factor VIIa. As shown in Fig. 4. both sTFo (open squares) and sTF/f (closed circks) demonstrated cofactor activities similar to that of detergent-soluhilized T F (open circles). The effects of sTFtv and sTFIf on the kinetic parameters of factor VIIa were also determined using various concentrations (1.6-3.2 mM) of S-2288 (Table I). In this experiment, the reaction mixtures contained 28 nM human VIIa and a 53.6-fold excess amount of sTFa or sTFP. In the presence of recombinant proteins the K, values of factor VIIa toward S-2288 were decreased approximately 2-fold, and the kc,, values were increased 19-fold (sTFa) or 32-fold (sTFP), when compared with the values obtained with factor VIIa alone. These results suggest that the catalytic efficiency (kca,/Km) of factor VIIa in the presence of sTFa and sTFP was enhanced 33.4-and 39.5-fold, respectively, over that seen with factor VIIa alone.
Substrate Specificities of Factor VIIa and Its Complex with sTFP toward Peptidyl-MCA-The 25 peptidyl-MCA substrates were first screened at a fixed concentration of 0.2 mM, a s described under "Materials and Methods." Table I1 shows the rates of hydrolysis of different peptidyl-MCA substrates by factor VIIa-sTFP complex and factor VIIa alone. The sequence of the best substrate for factor VIIa, Boc-Leu-Thr-Arg-MCA, is the same as that close to the cleavage sites  Table I11 shows the kinetic parameters of factor VIIa in the presence of sTFP for selected fluorogenic substrates. The kc,,/ K, value for the best substrate Boc-Leu-Thr-Arg-MCA was the highest, and the catalytic efficiency of factor VIIa-sTFP complex toward this substrate was enhanced 146.6-fold over that seen with factor VIIa alone. Factor VIIa preferentially hydrolyzed substrates containing Thr, Ser, or Ala residues at the P2 site (35).
Effects of Various Proteinase Inhibitors on the Enhanced Factor VIIa Activity in the Presence of sTFP-The effects of naturally occurring proteinase inhibitors on the amidolytic activity of factor VIIa-sTFp complex were examined. The amidolytic activity was evidently inhibited by antithrombin I11 in the presence of heparin, but not by a,-plasmin inhibitor and C1' inactivator. Several other inhibitors of serine proteases such as soybean trypsin inhibitor (toward factor XIa and Xa), hirudin (a-thrombin), aprotinin (plasma kallikrein and plasmin), and corn inhibitor (factor XIIa) had no apparent inhibitory effects on factor VIIa-sTFP. Fig. 5 shows the inhibition of factor .\IIIa-sTFP complex  amidolytic activity by antithrombin I11 and heparin. From these data, the pseudo-first order rate constant (k2) was calculated to be 0.203 min". Moreover, the inhibition constant (Ki) of antithrombin I11 for factor VIIa-sTFP complex was estimated to be 4.44 X low7 M, based on the reciprocal plot shown in Fig. 6. Fig. 7 shows complex formation between factor VIIa and heparin-activated antithrombin 111. In the presence of sTFP (Fig. 7 A ) , the protein band of a factor VIIa-antithrombin I11 complex detected by Western blotting appeared within 2 min under the conditions used. The molecular weight of the complex corresponded to that calculated for the sum of factor VIIa and antithrombin 111. However, the amounts of factor VIIa-antithrombin I11 complex formed at 2 and 16 min were almost the same, and "free" factor VIIa remained after incubation. It seems likely that a denatured or inactive form of factor VIIa which does not form a complex with antithrombin I11 is present in the preparation. Further studies on this point are required. In contrast, factor VIIa alone formed a complex with antithrombin I11 very slowly even in the presence of heparin (Fig. 7B). These results suggest that the enhanced amidolytic activity of factor VIIa with sTFP is inhibited by antithrombin 111-heparin, forming an acyl complex between factor VIIa and antithrombin 111.
Stability of sTFP-The stability of recombinant sTFP was examined under various conditions. The cofactor activity of sTFP was linearly decreased by increasing the concentration of guanidine hydrochloride up to 1 M during incubation at 37 "C for 6 h. However, the cofactor activity was relatively resistant to concentrations of guanidine hydrochloride greater than 1 M (Fig. &I). The effect of 2 M urea on the cofactor activity was almost the same as that of 1 M guanidine hydrochloride (Fig. 8B). The biphasic loss of cofactor activity in the presence of guanidine hydrochloride or urea suggests that sTFp might consist of at least two discrete domains. On the other hand, treatment of sTFP with over 100 mM 2-mercaptoethanol abolished the cofactor activity (Fig. 8C); hence Flc. 7. Complex formation between factor VIIa and antithrombin 111. Human factor VIIa (192 pmol) plus sTFB (3.1 nmol) ( A ) and human factor VIla alone (192 pmol) (8) were incuhated. respectively, with antithromhin 111 (2.0 nmol) and heparin (1 u n i t ) in 300 pI of THS containing 5 mM CaCI, at 3'7 "C. At the times indicated, samples taken from the reaction mixture were hoiled with 2.5% SIX to terminate the reaction and subjected to SDS-PACE. After electrophoresis, proteins were transferred electrophoretically to nitrocellulose membranes. Then, factor VIIa-containing hands were detected hy a mouse anti-human Vlla monoclonal antihody (designated hVlla-M93), using peroxidase-conjugated goat-anti-mouse IgG and 4-chloro-l -naphthol for color development. Vlln. human factor Vlla; A7'111, antithromhin Ill. disulfide bonds probably play a critical role in stabilizing the conformation of functional sTFP. sTFP was stable for 6 h a t 37 "C in buffers of pH 6-10; it also retained 74% of initial activity at pH 2.5 (Fig. 8D). Almost full cofactor activity was retained after 6 h of incubation at 50 "C ( Fig. 8E).

DISCUSSION
The expression vector for production of the extracellular domain of human TF was constructed and expressed in S.
cereoisiae. The recombinant protein (sTF) was secreted into the culture broth. Two forms, sTFa (high molecular mass form) and sTFP (low molecular mass form), both of which show cofactor activity, were isolated by two-step procedures. T h e yields of sTFn and sTF(j, as estimated by amino acid analysis, were approximately 230 and 36 nmol/liter of culture.  (25). Both forms of s T F n a n d sTFD showed almost equal cofactor activities for factor VIIa-catalyzed hydrolysis of S-2288, as noted by enhancement of the factor VIIa activity. This indicates that glycosylations at Asn-11, Asn-124, and Asn-137 found in sTFtr. despite the large molecular mass, do not affect cofactor activity. Furthermore, the enzymatic removal of the cnrhohydrate moiety from sTFn did not alter the k,,,,/K,,, value (data not shown). In agreement with these results, it has been known that deglycosylated TF retains cofactor activity in clotting assays (26) and that a recomhinant TF produced in E:. coli is also active in both chromogenic substrate and one-stage prothrombin time assays (11).
Chemical analyses of the recombinant proteins indicated that both forms of STFR and sTF/j consist of sequences from Ser-1 to Phe-217, as determined by amino acid composition. partial amino-terminal sequences, and detection of the carboxyl-terminal residue. The expression construct contained a chimeric cDNA consisting of the yeast invertase signal sequence and the TF cDNA from Ser-1 to Arg-218. This means that a signal peptidase and the secretion machinery in the cells correctly recognized the invertase signal sequence and secreted sTFn and sTF/j after removal of the invertase signal sequence. Carboxyl-terminal analvses revealed that the recombinant protein has Phe-217 at the carhoxyl terminus; the predicted carboxyl-terminal Arg-218 may have been excised by a carboxypeptidase-t?e protease, such as carbox>peptidase Y or Kexl (27).
The apparent dissociation constants ( K c ) of recomhinant sTFn and sTFB with human factor Vlla were 89.8 and G . 4 nM, respectively, values 28 and I5 times higher than that obtained with a membrane-hound TF on .J82 cells ( 2 8  factor VIIa and the factor VIIa-sTF complex. In contrast to t h a t of factor VIIa alone, the k,,,, values of the factor VIIa-sTF complex increased considerably. Therefore, the binding of factor VIIa to sTF does not seem to facilitate suhstrate recognition by the factor VIIa-sTF complex; rather it enhances catalysis. Our ohservations are in good agreement with the data of Ruf ~t 01. ( 1 3 , who determined the kinetic parameters of factor X cleavage by factor VIIa, factor VIIa-TF complex, and factor VIIa-sTF complex and found that catalytic activity of factor VIIa was enhanced hy complex formation with TF. T h e K , value of factor S a toward S-2288 was reported to be 2 mM (30). Our data show that the K,; values of factor VIIa and factor Vlla-sTF complexes are within 1-3 mM; hence, factors S a and VIla have a similar affinity for this suhstrate once factor VIla and sTF have formed a complex.
In our systematic studies previously reported ( 3 1 ), no gnod fluorogenic peptide suhstrate for factor VIla could be found due to the low catalytic activity of factor V11a in the absence of TF. We now have developed the sensitive suhstrate Hoc-Leu-Thr-Arg-MCA for the factor VIIa-sTF/j complex which enahles a rapid and accurate assay of as little as 5 pmol o f factor VIIa. In general, blood clotting proteases. except for plasma kallikrein, exhihit relatively strict specificity toward peptidvl-MCA substrates containing an apolar (Ala and Pro) or Gly residue a t t h e P, site and bulkier residues at the I' site, although they show substantial differences in catalytic efficiency toward such substrates ( 3 1 ). Factor VIIa-s'T'F,f cornplex, however, exhihits specificity toward Thr and .4la resi-FIG. 8. Stability of recombinant sTFP against guanidine hydrochloride, urea, 2-mercaptoethanol, pH, and temperature. The effects of guanidine, urea, and 2-mercaptoethanol are shown in A, B , and C, respectively. The purified sTFB was incubated at a concentration of 50 g / m l for 6 h a t 37 "C in T B S containing either urea (0, 0.5, 1, 2, 4, and 6 M), guanidine hydrochloride (0, 0.25, 0.5, 1, 2, and 4 M), or 2-mercaptoethanol (0, 12.5, 25, 50, 100, and 500 mM), respectively. After dialysis against TBS, aliquots of the samples were subjected to the functional assays as described under "Materials and Methods." The effect of pH is shown in D. The purified sTFB at a concentration of 50 pg/ml was incubated in 50 mM glycine buffer, pH 2.5, 0.1 M sodium acetate buffer, pH 4.0, 0.1 M sodium acetate buffer, pH 6.0, 50 mM Tris buffer, pH 8.0, 50 mM glycine buffer, pH 10.0, and 50 mM CAPS buffer, pH 11.5, for 6 h a t 37 "C. After dialysis against TBS, aliquots of the samples were subjected to functional assays. The effect of temperature is shown in E. The purified sTFP was incubated in TBS at a concentration of 50 kg/ml for 6 h a t 4, 25, 37, 50, 65, and 80 "C, respectively. After cooling on ice, the samples were subjected to functional assays. Data were plotted as average of two assays. The relative activity was plotted as the percentage of the values without guanidine hydrochloride dues but not Gly residues at the P2 site. This result is in contrast to that of factor Xa, for which a P2-Gly is most favorable (31). On the other hand, the factor VIIa-sTF complex accommodates bulky apolar residues such as Leu and Glu(OBz1) at the PB site (Tables I1 and 111). It is of interest that there are some differences in specificity between factor VIIa alone and the factor VIIa-sTFP complex toward peptidyl-MCA substrates. This result suggests that the substrate binding site of factor VIIa may be altered by complex formation with sTF& However, the detailed mechanism with respect to the enhanced effect of sTFP on the factor VIIa activity is still unclear.
In 1980, Broze and Majerus (37) reported that factor VIIa activity measured by a coagulation assay using factor VIIdeficient plasma was inhibited by antithrombin I11 and heparin. Our present result is in good agreement with them, in that the enhanced amidolytic activity of factor VIIa with sTFP was inactivated by excess antithrombin I11 (Figs. 5 and  6). Moreover, factor VIIa in the presence of sTF@ formed a high molecular weight complex with heparin-activated antithrombin 111, consistent with acyl complex formation (Fig. 7). The second order rate constant ( k 2 ) for factor VIIa-sTFP inhibition by antithrombin I11 plus heparin is very small; however, under some conditions in vivo this reaction might be further accelerated. If so, it could contribute to the phys-iological regulation of factor VIIa activity.
Bazan (32) proposed a structural and evolutionary relationship of TF with the cytokine and interferon receptor families. A tandem repeat consisting of seven &sheet structures has been predicted. Indeed, the CD spectrum of the extracellular domain of TF suggests that it consists of mainly P-sheet structures (33). Our denaturation experiments on the extracellular domain of sTFP with guanidine hydrochloride or urea showed a biphasic loss of TF activity, suggesting that one of the tandem repeat domains is less stable, resulting in a rapid loss of activity.
As our yeast expression system provides large quantities of functional sTF, detailed structure-function relationships of sTF can be examined. Moreover, this system enables the production of a variety of mutant sTF proteins which will facilitate studies on the mechanism of initiation of the extrinsic blood clotting pathway.