Isolation and characterization of thrombin-activated human factor VIII.

Recombinant human factor VIII (fVIII) was activated by thrombin at pH 7.4, followed by CM-Sepharose chromatography at pH values ranging from 3.5 to 7.4. Optimal coagulant activity was recovered at pH 5.5 and was associated with the isolation of an A1/A2/A3-C1-C2 heterotrimer. The activity was stable at -80 degrees C, but decayed slowly (t1/2 approximately 1 week) and nonproteolytically at room temperature or 4 degrees C. The coagulant activity of the pH 5.5 fVIIIa preparation assayed in human hemophilia A plasma was only 20% that of porcine factor VIIIa. However, its activity was approximately 75% that of porcine fVIIIa in a plasma-free assay, indicating that human fVIIIa is unstable relative to porcine fVIIIa during the coagulation assay. The first-order rate constant for spontaneous, nonproteolytic loss of activity of human fVIIIa at pH 7.4 was decreased 8-fold by fIXa and phospholipid, indicating that human fVIIIa is stabilized when incorporated into the intrinsic pathway factor X activation complex.

its isolation. Porcine fVIIIa has been isolated by m o n o 3 f a s t protein liquid chromatography at pH 6.0 and is indefinitely stable when stored at concentrations exceeding 0.2 VM (1). However, several attempts to isolate human fVIIIa by this or other procedures have not been successful because of dissociation of the A2 subunit and recovery of little or no heterotrimeric fVIIIa (2,3,12,13). Whether  fVIIIa are shared by human WIII and MIIa has been questioned (14,151. For example, the decay rate of porcine fVIIIa is decreased by approximately 10-fold when it is bound to f E a on a phospholipid surface (7)(8)(9). However, f E a cleaves and inactivates human fVIII (15,16), and the notion that human fVIIIa is stabilized by fIXa, while supported by one study (17), has been challenged (15). In none of these studies has the stabilization of fVIIIa by M a been directly tested by using purified fVIIIa. Additionally, it has been proposed that human WIII has less coagulant activity than porcine fVIII because of more rapid loss of its A2 subunit (2, lo), but comparison of purified human and porcine fVIIIa in coagulation assays has not been possible. In this study, the isolation of active, heterotrimeric human fVIIIa and its comparison to porcine fVIIIa is described.
EXPERIMENTAL PROCEDURES Materials-Human thrombin (1800 units/mg) was obtained from Baxter Biotech, Hyland Division (Glendale, CA). Normal human plasma and hemophilia A plasma were purchased from George King Biomedical. D-Phenylalanyl-prolyl-arginyl chloromethyl ketone was purchased from Calbiochem.
Isolation of Human FVIIIa-All procedures were carried out at room temperature unless otherwise indicated. The starting material for the purification of human M I I a was bulk human recombinant M I 1 obtained from Baxter Biotech Group, Hyland Division. FVIII (0.4 mg/ml) in 0.4 M NaCl, 20 m M Tris-C1,5 m M CaC12, pH 7.4,0.01% (v/v) Tween 20, pH 7.4, was first dialyzed into 0.15 M NaCl, 20 m M Hepes, 5 m~ CaCl, overnight at 4 "C and then stored at -20 "C until further use. FVIII (3.1 p~, 10 ml) was activated by human thrombin (0.1 PM) for 10 min at 37 "C. Activation was stopped by addition of the thrombin inhibitor FPR-CH,Cl(O.3 p~) along with 10 ml of 100 m~ sodium acetate buffer, at pH 5.5. The sample was loaded at 2 ml/min onto a CM-Sepharose column (1 x 10 cm) equilibrated in 0.1 M NaCI, 25 m M sodium acetate, 5 m~ CaClZ, 0.01% Tween 20, pH 5.5. FVIIIa was eluted at 0.5 ml/min using a 30-ml 0.1-1.0 M NaCl gradient in the same buffer. Fractions containing M I I a were collected and immediately adjusted to pH 6.0 by addition of 0.5 M Mes.
Extinction Coefficients and Molecular Weights-Extinction coefficients (EY:Z) at 280 nm and molecular masses used were: porcine MIIa, 1.60, 160 kDa (1); porcine m a 1.52,45 kDa (7); porcine M 1.04, 57 kDa (7). The extinction coefficients of human M I 1 and M I I a were assumed to be the same as porcine MIIa.
Coagulation Assays-FVIIIa was measured by using human M I Ideficient plasma as a substrate as described previously (22). The coagulant activity was determined in reference to the standard one-stage too TIME (MINI  Circular Dichroism Spectroscopy-Far-ultraviolet CD spectra were obtained a t 25 "C using a Jasco 710 spectropolarimeter. Samples were dialyzed overnight a t 4 "C into 0.1 M NaCI, 0.01 M Mes, 5 mM CaCI,, pH 6.0. The scans were performed by using a 0.05-cm path length cylindrical cell, 8-s response time, and a 5-ndmin scan speed. All CD scans of M I I a preparations were performed within 24 h of thrombin activation. Four scans were averaged for each sample and corrected by subtraction of the average of four buffer scans. ElectrophoresisSDS-8% polyacrylamide gel electrophoresis was done using the Laemmli buffer system (23) and the protein bands visualized by silver staining (24) or Coomassie Blue staining.

RESULTS
Isolation of Human fVZZZa-Attempts to isolate active human M I I a by mono-S chromatography at pH ranging from 4.5 to 7.4 were not successful. We then attempted to purify human M I I a by using CM-Sepharose to lower the charge density of the column matrix. Several pH conditions ranging from 3.5 to 7.4 were tested ( Fig. 1). At pH 3.5, very little protein was recovered. At pH 4.5-4.7, near the pK, of the CM-Sepharose carboxymethyl group, the A1 subunit was eluted as a separate peak prior to a A21A3-Cl-C2 dimer, as judged by SDS-PAGE analysis ( Fig. 2). At the other pH extreme, pH 7.4, the A2 subunit dissociated from the AlIA3-Cl-C2 dimer. Isolation of the AllA21A3-Cl-C2 heterotrimer was optimal between pH 5.0 and 6.0 as judged by the amount ofAl andA2 subunits detected by SDS-PAGE (Fig. 2).
The specific coagulant activity of fVIIIa isolated at the various pH conditions tested was maximal at pH 5.0-5.5 and decreased markedly at lower and higher pH values (not shown). The activity of fVIIIa isolated at all pH values decreased very slowly relative to the decay of activity of M I 1 activated in situ at the normal plasma concentration of M I 1 (-1 nM) at pH 7.4 (tM = 2 min) (10). However, these results are in contrast to the behavior of isolated porcine M I I a , which is indefinitely stable a t 4 "C or room temperature at pH 6.0 a t concentrations exceeding 0.2 p (1).
The slow loss of activity of the pH 5.5 preparation (tth = 6 days) was not decreased by D-phenylalanyl-prolyl-arginyl chloromethyl ketone or hirudin. After several weeks a t 4 "C, a minor band at 25 kDa was observed whose formation was inhibited by benzamidine. Thus, the slow loss of activity appears to be predominantly nonproteolytic with an additional minor loss due to proteolysis. Since many oligomeric proteins undergo denaturation a t 4 "C but not room temperature (251, the pH 5.5 preparation was also evaluated at room temperature, but was not more stable. FVIIIa was indefinitely stable when stored at -80 "C. Circular Dichroism Spectra of Human FVZZZ and FVZZZa-The far-ultraviolet CD spectra of human M I 1 and several preparations of human M I I a are shown in Fig. 3. The CD spectra of the different pH preparations of human fVIIIa are indistinguishable, suggesting that a pH-dependent loss in coagulant activity is not due to gross differences in the secondary structure of fVIIIa. Significant differences between the CD spectra of human M I 1 and human M I I a were observed below 235 nm. The observed difference is consistent with less random MIIa, pH 5.0 (0); MIIa, pH 5.5 (0); MIIa, pH 6.0 (0). coil structure in human fVIIIa, presumably due to the removal of the thrombin-cleaved B-domain by CM-Sepharose chromatography. The increase in the negative CD band at 218 nm may also indicate an increase of P-sheet structure in fVIIIa formed from fVIII.
Coagulant Activity of Human fVZZZa-The coagulant activities of human fVIIIa preparations isolated at pH 4.7 and 5.5 were compared to porcine fVIIIa (Table I). The specific activity of human M I I a was constant over at least a 10-fold concentration range, corresponding to clotting times of 55-65 s. The determinations made to construct Table I correspond to clotting t h e s in this range.
The specific activity of h u m a n M I I a isolated at pH 5.5 (130,000 unitdmg, 21,000 unitdnmol) is similar to the specific coagulant activity produced at the peak of thrombin activation of human fVIII (80,000-110,000 units/mg, 17,000-23,000 unitd nmol) (2, 31, suggesting that the isolated fVIIIa is fully h ctional. However, the coagulant activity of the human pH 5.5 preparation was approximately 6-fold lower than that of porcine fVIIIa. Possibly, human fVIIIa decays faster than porcine M I I a during the performance of coagulation assay, i.e. from the time that fVIIIa is diluted to the time of visible clot formation. At the concentrations used in the coagulation assays, the half-lives of human fVIIIa and porcine fVIIIa activities under these conditions are 2 and 6 min, respectively (10). These decay times are on the same scale as the time necessary to produce the end point of the coagulation assay, about 1 min. Thus, the relatively greater loss of activity of human fVIIIa could lead to a lower observed specific activity, even if both human and porcine fVIIIa were fully functional prior to dilution into the assay system.

Stabilization of Human fVZZZa by frXa and Phospholipid-
The decay of activity of M I I a can be measured using a defined plasma-free assay (10). The relative activities of human and porcine fVIIIa can then be compared without the potential complications of undetermined loss of activity during the assay. Before comparing human and porcine fVIIIa in a defined system, the effect of fKa on fVIIIa was evaluated, since it has been reported that fKa proteolytically inactivates human fVIIIa (15, 16), whereas fKa stabilizes porcine fVIIIa (7). In the absence of fKa and phospholipid, the half-lives of human and porcine M I I a were approximately 2 and 6 min, respectively (Fig. 4).
Addition of fMa and phospholipid increased the half-life of human fVIIIa to 12 min and porcine fKa to 60 min.
The stabilization of porcine fVIIIa by fKa or D E G R -f f i was also compared. We examined the possibility that D E G R -a , * Data are expressed as the mean 1 S.D. for four to eight determinations. Human or porcine MIIa samples were diluted 20,00~100,000-fold to adjust the measured clotting times to 55-65 s. Percent activity was determined by extrapolation to zero time of initial velocity curves.  (20 p~) was evaluated exactly as for porcine fMa (W. The data are expressed as the fraction remaining relative to initial MIIa (fVIIIa,), which was calculated by extrapolation of a semi-log plot to zero time.
which is active site-blocked and has no detectable enzymatic activity, might stabilize fVIIIa to a greater extent than fKa. Fig. 4 shows that DEGR-fKa produced a 1.7-fold longer halflife than = (100 versus 60 rnin). SDS-PAGE analysis of the porcine fVIIIa after incubation revealed the appearance of a minor band migrating between the A1 and A2 fragments when fIXa but not DEGR-fKa was present (not shown). This band has similar electrophoretic behavior to that described during cleavage of human MI1 and fVIIIa by f E a (15, 16) and identified as derived from the A1 domain, presumably due to cleavage at Ar236. Thus, stabilization of factor VIIIa by fKa and inactivation of factor VIIIa by fKa are competing processes. At the concentration of fIXa used in Fig. 4 (10 nM), fVIIIa is nearly saturated with fKa (20), representing optimal stabilization. As the concentration of fKa increases, the inactivation reaction would become more prominent.
Comparison of Human and Porcine FvIZIa Activity in a Defined System-In contrast to the coagulation assay, where possible loss of fVIIIa activity cannot be evaluated, fVIIIa activity can be followed directly in the plasma-free assay. In this assay, the activity of the pH 5.5 human fVIIIa preparation was 75% that of porcine fVIIIa (not shown). This result is in contrast to a relative activity of approximately 20% when compared to the coagulation assay (Table I). When human fVIIIa was isolated at pH 4.7, the specific activity of the preparation in the plasmafree assay was considerably lower than the pH 5.5 preparation (not shown), consistent with results obtained with the coagulation assay (Table I).
Human and porcine fVIIIa were also compared by activating highly purified fVIII with thrombin and comparing initial velocities of M activation in the plasma-free assay. At the concentration of thrombin used M I 1 was activated within 1 s, which allows comparison of peak activation of fVIII and subsequent decay rates. Both plasma-derived and recombinant human fVIII were compared to porcine fVIII and gave similar results (Fig. 5). In the porcine system, the initial rates of M activation when fVIII is activated in situ at 200 PM are very similar to the values obtained by using purified fVIIIa at the same concentration, indicating that no major changes in activity of fVIIIa occur during the process of separating fVIIIa from activation peptides and thrombin. Human fVIII was activated to levels that were indistinguishable from porcine fVIII. Comparison of the results shown in Fig. 3 to those obtained with fVIIIa preparations (Table I) indicates isolation of human fVIIIa at pH 5.5 yields a preparation that approaches, but may not achieve, complete activity. of Fl-M-FPR-DZa as a function of fVIIIa concentration can be used to determine the stoichiometry and the apparent dissociation constant of the porcine fVIIIa/fMa interaction. The titration of porcine F1-M-FPR-fMa by the pH 4.7 and 5.5 human fVIIIa preparations was studied fluorometrically and compared to porcine fVIIIa (Fig. 6). The stoichiometry and apparent dissociation constant for porcine fVIIIa (Table 11) was similar to values previously reported (20). The anisotropy increase produced by the two human fVIIIa preparations correlated with the measured activity. Both human fVIIIa preparations produced lower anisotropy increases than porcine fVIIIa. The human preparations yielded apparent stoichiometry val-  ues greater than 1. If these high stoichiometry values are due to the presence of inactive forms, this would lead to an underestimation of the true concentration of active fVIIIa and an overestimation of the dissociation constant, The higher dissociation constants could also arise from weaker binding of human fVIIIa to porcine Fl-M-FPR-ma. DISCUSSION The assembly of the intrinsic pathway fX activator complex is essential for normal hemostasis. Difficulties associated with the isolation of a stable form of fVIIIa have hampered the analysis of this complex. Activation of porcine fVIII by thrombin followed by mono-S high pressure liquid chromatography a t pH 6 results in the isolation of a Al/A2/A3-Cl-C2 heterotrimer with indefinitely stable activity at 4 "C (1). Porcine fVIIIa isolated in this way appears completely active because it binds fIXa with 1:l stoichiometry (20) and molecular weight analysis by analytical equilibrium ultracentrifugation is consistent with a homogeneous population of heterotrimers (1). However, attempts to isolate human fVIIIa by mono-S HPLC in this study and others (2,3,13) were unsuccessful, instead yielding mainly the inactive Al/A3-Cl-C2 dimer. We now find that active, heterotrimeric human M I I a can be isolated by CM-Sepharose chromatography at pH 5.5. This procedure constitutes the first report in which the A2 subunit of human fVIIIa is recovered in high yield.
The properties of this human fVIIIa preparation were characterized in several ways and used to address controversial or unresolved issues in the analysis of fVIIIa structure and function. Compared to porcine fVIIIa, the coagulant activity of human fVIIIa preparations is low (Table I). However, this appears to be at least partly due to the relative instability of human fVIIIa during the assay itself ( i e . in plasma during the development of the fibrin clot) since the discrepancy is much lower when a plasma-free assay is used for comparison (Table I). Inactivation of fVIIIa during the plasma-free assay can be monitored and is slow relative to the assay period. The activity of the pH 5.5 human fVIIIa preparation approaches that of porcine fVIIIa in this assay and is similar to that obtained by the peak activation of human M I 1 by thrombin. In the coagulation assay, the activity of this preparation (130,000 unitdmg) is similar to the peak coagulant activity generated by the activation of fVIII by thrombin (2,3). The increase in fluorescence anisotropy of Fl-M-FPR-fIXa produced by this preparation of human fVIIIa yields an apparent stoichiometry for the fVIIIa-fKa interaction slightly greater than 1 (Table 11). These results suggest that human fVIIIa isolated in this way is nearly completely active, and provides a suitable reagent for analysis of intrinsic pathway activation of M in the human system.
In contrast to porcine fVIIIa, human fVIIIa was not indefinitely stable at 4 "C or room temperature. Although benzamidine slightly decreased the rate of decay and prevented the appearance of a 25-kDa proteolytic fragment, most of the loss of activity was not correlated with a proteolytic event. This nonproteolytic loss of activity is distinct from the nonproteolytic decay in activity that occurs over a period of a few minutes a t pH 7.4 at nanomolar concentrations of fVIIIa due to dissociation of the A2 subunit. This slow loss of activity occurred a t M I I a concentrations much greater than the dissociation constant for the binding of the A2 subunit to the Al/A3-Cl-C2 dimer at pH 6.0 (28 nM) (26). Thus, the loss of activity appears to be due to a secondary pathway accessible to human fVIIIa, possibly irreversible denaturation of the heterotrimer or of the A2 subunit while free in solution. Storage at -80 "C is a practical way to maintain active fVIIIa preparations for further characterization.
The pH 5.5 human fVIIIa preparation was used to address a controversial issue regarding the lack of stability of fVIIIa. Although porcine fVIIIa is stabilized by f K a and phospholipid, it has been proposed that this is not the case with human fVIIIa because fIXa cleaves and inactivates human fVIIIa (15). However, the first-order decay of human fVIIIa activity is decreased 8-fold by nanomolar concentrations of fIXa and saturating concentrations of phospholipid (Fig. 4). This result was compared to the stabilization of porcine fVIIIa, since previous studies measured porcine fVIIIa in activation mixtures that contained thrombin and fVIII activation peptides. Purified porcine fVIIIa, like purified human fVIIIa, is stabilized by fIXa and phospholipid (Fig. 41, indicating that neither the activation peptides nor thrombin contribute to the stabilization process. As in the case of human fVIIIa, fIXa cleaves porcine fVIIIa, apparently in the A1 domain at Ar$36. This cleavage either inactivates or reduces the activity of porcine fVIIIa, because the stabilization of fVIIIa by active site-blocked f K a is greater than with active fIXa (Fig. 4). However, the dominant effect of fIXa at nanomolar concentrations is the stabilization of fVIIIa.
The behavior of human fVIIIa in the coagulation assay has implications regarding the measurement of fVIII in the clinical setting. The fVIIIa assay used in this study utilizes the same reagents as the standard one-stage fVIII clinical assay. In the latter assay, fVIII is activated endogenously by feedback from thrombin (27). Porcine fVIII has higher specific coagulant activity than human fVIII (lo), suggesting either a difference in the rate of fVIII activation or in the stability of fVIIIa. By studying the coagulant activity of purified human fVIIIa and thus eliminating consideration of the activation process, the stability of fVIIIa appears to be a critical determinant of coagulant activity.
The simplest mechanism consistent with this result is that after fVIIIa is diluted into plasma, substantial dissociation of the A2 subunit and loss of fVIIIa activity occurs during the 55-85 s required for the formation of a visible fibrin clot. The A2 subunit is in reversible equilibrium with the Al/A3-Cl-C2 dimer (10,17,26). The dissociation constants for human and porcine fVIIIa are similar at pH 7.4, approximately 0.2-0.3 PM, which is much higher than the concentration of fVIII in plasma or in clotting assays. Thus, in dilute solutions of fVIIIa, activity is governed entirely by the dissociation rate constant. The dissociation rate constant of human fVIIIa at pH 7.4 is 3-fold higher than porcine M I I a (0.35 min-l uersus 0.12 min-'1 (10) which may explain the difference in behavior between human and porcine M I 1 and fVIIIa in coagulation assays. Additionally, unidentified factors in human plasma may exist that further destablize human fVIIIa relative to porcine fVIIIa.