Design of Constructs for the Expression of Biologically Active Recombinant Human Factors X and Xa KINETIC ANALYSIS OF THE EXPRESSED PROTEINS*

Activation of vitamin K-dependent plasma proteases occurs by specific interaction with components of the blood coagulation cascade. In this report, we describe the direct expression and enzymatic characterization of the human coagulation zymogen factor X and its activated form, factor Xa, from transformed Chinese hamster ovary fibroblast cell lines. Expression was achieved using either a full-length factor X cDNA or a unique mutant factor Xa cDNA. The functional factor Xa precursor contained a novel tripeptide bridge in place of the native 52-amino acid activation peptide. This mutation allowed for intracellular processing and secretion of the activated form of factor X. Secreted recombinant factors X (rX) and Xa (rXa) were purified by sequential anion-exchange and immunoaffinity chromatography. The enzymatic activities of factors rX and rXa were compared with those of plasma factors X and Xa in three independent assay systems. In comparison to human plasma factor X, the amidolytic, prothrombinase complex, and plasma clotting activities of factor rX were 50, 85, and 43%, respectively. The corresponding comparative activities for

Factor X, a vitamin K-dependent blood coagulation glycoprotein, is the precursor for factor Xa, the enzyme component of the prothrombinase complex (1). In plasma, factor X circulates as a zymogen in the form of a two-chain polypeptide consisting of a 17-kDa light chain joined by a disulfide bridge to a 45-kDa heavy chain. During biosynthesis, a number of post-translational processing events occur (1,2). These include endopeptidic cleavage, glycosylation, and conversion of select glutamic acid residues in the amino terminus of the light chain to y-carboxyglutamic acid by a vitamin K-depend-HL44782-01 (to D. L. W.), HL29807-10 (to C. T. E.), and HL35058 * This work was supported by National Institutes of Health Grants and HL040467 (to W. R. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
To whom correspondence should be addressed COR Therapeutics, Inc., 256 E. Grand Ave., Suite 80, S. San Francisco, CA 94080. ** Investigator of the Howard Hughes Medical Institute. ent carboxylase. Calcium ion and phospholipid binding by factor Xa is a direct consequence of light chain carboxylation (2). The extrinsic coagulation pathway, composed of tissue factor and factor VIIa, and the intrinsic pathway, consisting of factors IXa and VIIIa, catalyze the proteolytic activation of factor X to its active species, factor Xa (1). Both pathways require calcium-dependent complex assembly on phospholipid membranes. Factor X is also activated in vitro by a factor X activator isolated from Russell's viper venom (RVV)' (3). Proteolytic cleavage at Arg234-Ile235 and release of a 52-amino acid activation peptide from the amino terminus of the heavy chain lead to the formation of the active enzyme, factor Xa. The catalytic site of factor Xa is located on the heavy chain (3,4).
Factor Xa forms a macromolecular complex with factor Va on negatively charged phospholipid surfaces. This complex catalyzes the activation of prothrombin to thrombin, an important enzyme that elicits both pro-and anticoagulant responses during cardiovascular trauma (1). Factor Xa also forms a ternary complex with factor VIIa and a lipoproteinassociated coagulation inhibitor ( 5 , 6 ) . Formation of the factor Xa-lipoprotein-associated coagulation inhibitor-factor VIIa complex has been shown to inhibit the extrinsic coagulation pathway.
One approach for the structure/function characterization of topographical sites on factors X and Xa could be accomplished by site-directed mutagenesis. In this report, we describe the direct expression in Chinese hamster ovary (CHO) cells of recombinant human factors X (rX) and Xa (rXa) and compare their biological activities to the plasma factors. This approach demonstrates a strategy for intracellular zymogen activation and establishes the foundation for a more systematic analysis of this important coagulation enzyme.
Human Factor X and X a Expression X-deficient plasma, human antithrombin 111, and human prothrombin were obtained from American Bioproducts Co. Human factors X, Xa, Va, and human a-thrombin; and purified RVV factor X activator were purchased from Haematologic Technologies. Human and bovine factors X were also isolated as described (7). Lipofectin reagent and G418-neomycin were from Bethesda Research Laboratories-GIBCO.
Coagulation reagents thromboplastin C (composed of rabbit brainderived thromboplastin) and actin FS (composed of soy phosphatides in ellagic acid) were from Baxter. All reagents were of analytical grade. Construction of Human Factor X and Xu-expressing Cell Lines-T h e HindIII-XbaI fragment of plasmid Bluescript containing the fulllength human factor X cDNA (8) was subcloned into the HindIII-XbaI site of m13mp19. Oligonucleotide site-directed mutagenesis was performed as described by Kunkel et al. (9) utilizing the following oligomer: 5"ACCCTGGAACGCAGGAAGAGGCGGAAAAGAAT CGTGGGAGGCCAGGAATGC-3'. The mutagenesis was composed of deletion of the activation peptide and duplication of the tripeptide cleavage site (Fig. 1). Sequence verification of the mutagenesis was performed by dideoxy chain termination nucleotide sequencing (10). The SmaI-EcoRV fragments of the modified factor X cDNA was subcloned into the XbaI site of the expression vector pBN following treatment with Klenow polymerase (11). The expression vector pBN was derived by subcloning the SR-a promoter (12) obtained from p B J l (13) into the NruI-XbaI site of pRC/CMV, thus replacing the CMV promoter with the SR-a promoter. CHO-K1 cells (obtained from the University of California Cell Culture Facility) were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum, 50 mg/ml penicillin/streptomycin, 2 mM glutamine, and 4 pg/ml vitamin K,. The factor X expression vectors pBNX and pBNXa were transfected into CHO cells by the method of lipofectin reagent DNA uptake (14). Transfected cells were selected for G418-neomycin resistance, and individual clones were isolated and screened at equal cell density for high level expression in serum-free medium. Factor X antigen was assessed using an enzyme-linked immunosorbent assay for factor X (15) and by immunoblotting (15). Factor Xa activity was determined using Chromozym X hydrolysis as described below. Serum-free medium, bovine and human factors X, and the human factor Xa were used as controls.
Purification of Factors rX and rXa-Recombinant factor X-or Xaexpressing CHO cells were grown to confluency in roller bottles. Following multiple washing of the cells with serum-free medium, the cells were cultured for consecutive 24-h periods at 37 "C in serumfree medium supplemented with 4 pg/ml vitamin KB. Conditioned medium was centrifuged a t 15,000 X g; for 20 min, followed by filtration of the supernatant through a 0.2-pm filter. To the medium was added Tris-HC1 (pH 7.5) to 20 mM and NaEDTA to 10 mM, followed by chromatography on a Q-Sepharose Fast Flow column (Pharmacia LKB Biotechnology Inc.). All chromatographic steps were performed a t 4 "C. The column was washed extensively with 20 mM Tris-HC1 (pH 7.5), 10 mM EDTA, 0.15 M NaC1; and bound proteins were eluted with 20 mM Tris-HC1 (pH 7.5), 0.5 M NaC1, 5 mM CaC12. Peak fractions were pooled and either stored a t -20 "C or applied directly to an anti-factor X monoclonal antibody affinity column as described previously (16). The particular antibody used for isolation (aHFX-ld) is specific for human factor X, is not influenced by Ca", and binds both factors X and Xa.' Recombinant factor Xa was purified further on a benzamidine-Sepharose column (Pierce Chemical Co.) as described by Krishnaswamy et al. (7). The concentrations of the recombinant proteins were determined by quantitative enzyme-linked immunosorbent assay; colorimetric protein assay (bicinchoninic acid) (15); and absorbance measurements a t 280 nm using extinction coefficients and molecular masses of 11.6 and 58.9 kDa for factor X (17) and 11.6 and 46 kDa for factor Xa (18), respectively. SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was performed as described by Laemmli (19), and the proteins were visualized by silver staining (15). Amino-terminal protein sequence analysis was performed using an Applied Biosystems Model 473A protein sequenator as described by Matsudaira (20). Approximately 4-8 residues were determined for each peptide. Amino acid sequence numbering of factor X is based on the single-chain precursor form and begins with the initiation methionine of the light chain signal sequence (8).
Functional Analysis of Factors rX and rXa-Plasma factor X and factor rX were activated by incubation with RVV for 15 min at room temperature a t a protease/protein ratio of 1:lOO (w/w) in 20 mM Tris- HCl (pH 7.5), 0.15 M NaCI, 5 mM CaC12, 0.1% bovine serum albumin. Under these conditions, maximal factor X activation was found to occur. The samples were placed on ice and used within 1 h following activation. Amidolytic activity of factor Xa toward Chromozym X (50-500 p~) was determined by following increments of absorbance a t 405 nm due to peptide nitroanilide hydrolysis. Assays were carried out a t room temperature using a V,,, plate reader (Molecular Devices). Initial rates of substrate hydrolysis were determined in 20 mM Tris-HC1 (pH 8.0), 0.15 M NaCl, 5 mM CaC12, 0.1% bovine serum albumin. The kinetic constants (kc,, and K,) were determined by nonlinear regression analysis using Michaelis-Menten Equations (Enzfitter, Elsevier Scientific Publishing Co., Inc.). Duplicate measurements were averaged for each analysis, and the reported data (Table I) represent the average of three to eight separate determinations.
The ability of factor Xa to activate prothrombin in the presence of factor Va, Ca", and synthetic phospholipid vesicles was determined by measuring the product of the activation, a-thrombin (21). Both human plasma factor X and factor rX were activated by RVV prior to assaying in the prothrombinase complex. The reaction mixture consisted of 0.5-1.5 nM factor Xa, 7.5 nM factor Va, 20 p M phosphatidylcholine (75%)/phosphatidylserine (25%) in 20 mM Tris-HC1 (pH 7.5), 0.15 M NaCl, 5 mM CaCl,, 0.1% bovine serum albumin. Reactions were carried out a t room temperature and initiated by adding prothrombin to the reaction mixture. Aliquots were removed at various times and assayed directly for the production of thrombin. A standard curve of a-thrombin concentration as a function of Chromozym T H hydrolysis was used to estimate the amount of thrombin generated. For determination of steady-state kinetic constants, the concentration of prothrombin was varied between 0.2 and 1.5 p~. Kinetic constants were calculated from the data using a nonlinear regression analysis (Enzfitter) of the initial rates of thrombin formation and assuming Michaelis-Menten type kinetics. The reported data (Table   I) represent the average of three to five separate determinations.
Functional clotting activities of factors rX and rXa were determined using an MLA Electra 800 fibrometer. Three assays were performed (i) a two-stage prothrombin time assay, where the clotting time was determined after addition of bovine factors VII-and Xdeficient plasma, rabbit brain cephalin, and Ca'+ to dilutions of purified plasma or recombinant factors Xa or X preincubated with purified RVV; (ii) an activated partial thromboplastin time assay, which was performed using either bovine factors VII-and X-deficient plasma or human factor X-deficient plasma, and actin FS; and (iii) a factor X-dependent prothrombin time assay, which was performed using human factor X-deficient plasma and reagent thromboplastin C. Clotting assays were performed in duplicate and repeated two to three times, and the activities of factors rX and rXa were interpolated from standard clotting curves of purified human factors X and Xa accordingly.
Active enzyme content was quantified as follows. The concentration of human a-thrombin was determined by active-site titration employingp-nitropheny1p'-guanidinobenzoate HCl(22). This standard a-thrombin preparation was then used to calibrate a reference solution of human antithrombin 111 (23). Accordingly, human antithrombin 111 titrations of plasma and recombinant human factors X and Xa, based on inhibition of amidolytic activity (see above), were performed assuming a 1:l stoichiometry for protease inhibitor (24).

RESULTS
Plasma coagulation proteins are unique because of their requirement for several intracellular post-translational modifications including proteolysis of the signal peptide and propeptide (Fig. 1, arrows 1-2); y-carboxylation of glutamate residues in the light chain; and, as in the case of factor X and protein C, intramolecular cleavage of the single chain precursor to the two-chain form (1,2). During synthesis, proteolytic cleavage of the factor X precursor also occurs at Arg'79-Ser1R" (Fig. 1, arrows 3-4) (8, 18,26). Cleavage a t A~-p -I l e~~~ results in activation (Fig. 1, arrow 5 ) . Following zymogen activation, it is suggested that factor Xa undergoes autoproteolysis, possibly a t Arg464-Ser465 (Fig. 1, arrow 6) (3, 26). The net effect of this cleavage on factor Xa enzymatic activity is unknown at present. Signal sequence cleavage most likely occurs at LeuZ3-Serz4 (Fig. 1, arrow 1 ). To combine the intracellular and  vascular processing events that lead to factor X activation into one step, we attempted to express factor X cDNA constructs based on: 1) deletion of the activation peptide and 2) joining of the light and heavy chains by minor modifications of the Arg'"'-Ser'':' intrapeptide cleavage sites. This report details the expression and enzymatic characterization of factor rX and a novel factor, rXa, derived by cellular processing of a unique monomeric precursor.
Human factor X cDNA constructs were transfected into CHO cells, subcloned, and selected for G418-neomycin resistance (Fig. 1). Factor X antigen produced by the transfected stable producing clones was determined by enzyme-linked immunosorbent assay and immunoblotting. Antigen levels averaged -1 pg/ml for 24-h cultures for both factors rX and rXa. Factors rX and rXa were purified as described above (see "Experimental Procedures"). Homogeneity of purified factors rX and rXa was determined by SDS-PAGE (Fig. 2). Factor rX (Fig. 2, lune 3 ) under reducing conditions had three polypeptides, a t 75, 45, and 22 kDa. The 45-and 22-kDa species corresponded in migration distance to plasma factor X heavy and light chains (Fig. 2, lune I), and the 75-kDa species is similar in size to the putative single chain factor X observed previously in transfected COS-1 cells (8), HepG2 human hepatoma cells (27), and plasma (27). The factor rX precursor monomer content was estimated to be 26% by scanning autoradiographs of immunoblots (data not shown). Purified factor rXa had a more complex pattern, with polypeptides at 43, 35, 31, and 22 kDa. Additional silver-stained peptides were observed a t -17 kDa (Fig. 2, lune 4 ) . The 35and 31-kDa species corresponded to factor Xa,, and Xa,+ heavy chains (Fig. 2, lunes 2 and 4 ) , and the species at 22 kDa corresponds to the light chain of factor X (3,26).
Amino-terminal sequence analysis of the polypeptides was performed following electrotransfer to nylon filters (20). For factor rX, the sequence of the 45-and 22-kDa species corresponded to the expected factor X heavy and light chain amino acid sequences (8). The light chain sequence was heterogenous, with 27% initiating a t Val:" and 73% initiating at Ala"'. The 75-kDa species gave the expected sequence for the factor X light chain, confirming its identity as the uncleaved single chain factor X precursor, with 41% initiating a t Val'' and 59% initiating a t Ala"'. The data suggest that single chain factor rX is correctly cleaved a t Arg1x'2-Ser1x:' during processing. However, sequencing of the carboxyl terminus of the light chain was not performed, and the accuracy of cleavage at In the factor rXa preparation, sequences for the 35-and 31-kDa species corresponded to the activated plasma factor X heavy chain amino-terminal sequence, and the sequence of the 22-kDa species was comparable to that of the factor rX light chain sequence, with 29% initiating a t Val:'' and 71% initiating a t Ala"'. The 43-kDa species did not give reliable amino acid sequence by this procedure. The minor sequences (17 kDa) suggested that proteolytic cleavages had occurred at Ar~"-Gln:":{ and Lys:'7X-Met''i" (Fig. 1, arrow 7). Protease inhibitors such as benzamidine or soybean trypsin inhibitor were not used during purification. This could account, in part, for the observed endopeptidic cleavage of the heavy chain by factor Xa and possibly by secreted CHO proteases.
The enzymatic activity of factor rX following activation with RVV was compared with that of RVV-activated plasma factor X (Table I). SDS-PAGE and immunoblot analysis of RVV-activated factor rX demonstrated that the single chain precursor content did not change significantly (data not shown). The reduced RVV sensitivity of the factor rX single chain precursor suggests that its participation in the enzymatic activity of factor X in vitro as well as its possible role in vivo are minimal. The catalytic efficiency ( kcal/K,,,) of factor rX was 50% of that of plasma factor X in the amidolytic assay. This was due to a decrease in kc,, and only a slight difference in K,, for recombinant versus plasma factor X.
Factor rXa, generated by activation of factor rX with the venom activator and assayed directly without further purification, cleaved prothrombin in the presence of factor Va, phosphatidylcholine/phosphatidylserine vesicles, and Ca'+ ArgIi"Arg180 could not be confirmed. " 2PT, two-stage prothrombin time assay; APTT, activated partial thromboplastin time assay; PT, prothrombin time assay.
( Table I). The kinetic constants (k,,, and K,) for prothrombin activation were 43 and 45%, respectively, of the activity of plasma factor X activated in an identical manner. Comparison of k,,,/K,,, ratios demonstrated 85% activity for factor rX. The observed kinetic constants for plasma factor Xa were within experimental error compared with the values reported previously for human prothrombin activation (7).
In three plasma-based clotting assays, factor rX had 43% the specific activity of plasma factor X in a two-stage prothrombin time assay, 9% the specific activity of plasma factor X in an activated partial thromboplastin time assay, and 50% the activity of plasma factor X in a prothrombin time assay (Table I). The reduced activated partial thromboplastin activity could reflect unique processing or conformational differences between human factor X and factor rX that alter association with the factor IXa-VIIIa intrinsic pathway complex. These modifications may suggest possible differences in factor X association with the extrinsic and intrinsic factor X activating complexes.
The expressed factor rXa was catalytically active toward both small synthetic substrates and prothrombin (Table I).
Incubation with the venom activator had no effect on factor rXa activity. In amidolytic assays factor rXa had a catalytic efficiency that was 32% of purifiedplasma factor Xa activated by the venom activator. This difference in catalytic efficiency was due to a lower kc,, than that of RVV-derived plasma factor Xa. The K , values were comparable. In prothrombin activation assays, the kc,, values of factor rXa were 55%, and K, values were 85% of RVV-derived plasma factor Xa. Factor rXa in a two-stage prothrombin time clotting assay had 48% the activity of purified RVV-derived plasma factor Xa ( Table  I).
To better characterize the specific activities of the various plasma and recombinant factor X preparations, antithrombin 111-based active-site titration was performed (22-24). Active enzyme contents of RVV-treated plasma factor X and factor rX were 78 and 37%, respectively. Values of 92 and 59% for plasma factor Xa and factor rXa, respectively, were also derived. The lower content of active factors rX and rXa suggests that RVV activation was less efficient and that the preparations contain an inactive population of proteins composed of single chain precursors, proteolytically degraded species, and possibly improperly folded forms due to abnormal disulfide pairing ( Fig. 2 and Table I). Therefore, based only on estimates of active-site quantitation, the relative activities of factors rX and rXa are comparable to plasma factors X and Xa.

DISCUSSION
This report describes the expression and characterization of CHO-derived human factors X and Xa. The method chosen to express factor rXa entailed deletion of the activation peptide and duplication of the tripeptide cleavage site (Fig. 1). The modification did not significantly affect the biosynthesis or post-translational processing of the novel precursor. Since factor rXa is capable of being processed in other mammalian cell lines (data not shown), it seems likely that proteolytic cleavage sites, based on minor modifications of native sequences, can be engineered into proteins with some degree of specificity ( 2 8 ) .
The novel factor Xa hexapeptide cleavage site shares similarity with the intracellular peptidase endopeptidase Kex2p and carboxypeptidase Kexlp cleavage sites (30, 31) and the recently characterized mammalian FUR gene product (32), which cleaves after paired basic amino acids (PACE) ( 3 3 ) .
Although it is not known whether these particular proteases are directly involved in the processing of the novel factor Xa monomer precursor, the possibility remains that the unique processing site may be of some utility in other systems.
Attempts to secrete fully active vitamin K-dependent proteins in CHO cells have met with variable success (2,34). Although prothrombin is expressed in a fully y-carboxylated and biologically active form, factor IX, protein C, and protein S have been reported to be secreted with incomplete ycarboxylation (y-carboxyglutamic acid) and low functional activity. In agreement with CHO-based prothrombin expression, both factors rX and rXa were found to be secreted in biologically active forms (2). However, preliminary y-carboxylation estimations based on anion-exchange fractionation suggest a mixture of fully and partially y-carboxylated forms. Work is in progress to further characterize the process and extent of y-carboxylation in these particular cell lines (29).
Relative to the plasma factors, the lower active enzyme contents of the recombinant factors, 47% for factor rX and 64% for factor rXa, were partially due to a reduction in RVV activation and proteolysis. The reduced RVV conversion could be caused by differing conformations and possible post-translational modifications of factor rX, affecting both the rate and extent of RVV activation. As mentioned before, factors rX and rXa were not prepared in the presence of protease inhibitors. We have noted that unlike factor rX, factor rXa is less stable in CHO media and is prone to autoproteolysis in purified form. Accordingly, expression of proteolytic inactivated forms of factor rXa is also subject to heavy chain a / @ conversion (data not shown). Modifications in cell culture methodologies, purification, and RVV activation should improve the overall specific activities of the recombinant proteins. This is the first example of the direct expression and characterization of biologically active human factors X and Xa in CHO cells. The ability to express both zymogen and proteolytically active forms of factor X will greatly facilitate the characterization of functional domains important for activation, complex assembly, substrate association, and catalysis during coagulation. One outcome of this analysis may be the further development of novel coagulation inhibitors useful as antithrombotic agents ( 2 5 ) .