Characterization of Novel Forms of Coagulation Factor XIa

Factor XI is the zymogen of a dimeric plasma protease, factor XIa, with two active sites. In solution, and during contact activation in plasma, conversion of factor XI to factor XIa proceeds through an intermediate with one active site (1/2-FXIa). Factor XIa and 1/2-FXIa activate the substrate factor IX, with similar kinetic parameters in purified and plasma systems. During hemostasis, factor IX is activated by factors XIa or VIIa, by cleavage of the peptide bonds after Arg145 and Arg180. Factor VIIa cleaves these bonds sequentially, with accumulation of factor IXα, an intermediate cleaved after Arg145. Factor XIa also cleaves factor IX preferentially after Arg145, but little intermediate is detected. It has been postulated that the two factor XIa active sites cleave both factor IX peptide bonds prior to releasing factor IXaβ. To test this, we examined cleavage of factor IX by four single active site factor XIa proteases. Little intermediate formation was detected with 1/2-FXIa, factor XIa with one inhibited active site, or a recombinant factor XIa monomer. However, factor IXα accumulated during activation by the factor XIa catalytic domain, demonstrating the importance of the factor XIa heavy chain. Fluorescence titration of active site-labeled factor XIa revealed a binding stoichiometry of 1.9 ± 0.4 mol of factor IX/mol of factor XIa (Kd = 70 ± 40 nm). The results indicate that two forms of activated factor XI are generated during coagulation, and that each half of a factor XIa dimer behaves as an independent enzyme with respect to factor IX.

During coagulation, FXIa converts factor IX (FIX) to the protease factor IXa␤ (FIXa␤), by cleaving the Arg 145 -Ala 146 and Arg 180 -Val 181 peptide bonds (8 -12). FIX is also activated by factor VIIa (FVIIa) by cleavage at these same bonds (12). FVIIa bound to the membrane protein tissue factor (TF) initially cleaves FIX after Arg 145 , generating the intermediate FIX␣, prior to cleavage after Arg 180 to generate FIXa␤ (9,(12)(13)(14). In contrast, little intermediate appears to be generated during activation by FXIa (8,15). It has been proposed that the dimeric structure of FXIa may account for its capacity to activate FIX without intermediate formation (15,16), with the two protease domains each cleaving one FIX activation site prior to releasing FIXa␤.
Here, we report that FXI activation by factor XIIa or ␣-thrombin proceeds through an intermediate in which only one subunit of the dimer is cleaved, and that this intermediate species is formed in plasma during contact activation-induced coagulation. We purified and characterized the intermediate and used it, along with other forms of FXIa with one active site per molecule, to address the importance of the dimeric structure of FXIa to FIX activation.
Purification of Plasma FXI-Frozen plasma (2 liters) collected in acid-citrate-dextrose was thawed at 4°C, and supplemented with benzamidine (20 mM). FXI was purified from the cryosupernatant by affinity chromatography using the anti-human FXI antibody 1G5.12 (17). After loading, the column was washed with 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 20 mM benzamidine and eluted with 2 M NaSCN in the same buffer.
The eluate was concentrated by ultrafiltration and dialyzed against 50 mM Hepes, pH 7.4, 125 mM NaCl, 20 mM benzamidine. Purity was assessed by SDS-PAGE and concentration by colorimetric assay (Bio-Rad).
Western Blots of FXI Activation in Plasma-Human plasmas with 0.38% sodium citrate (George King, Overland Park, KS) were mixed with equal volumes of PTT A reagent (Diagnostica Stago, Asnières-sur-Seine, France) at 37°C. At various times, 9 l of reactions were mixed with 6 l of non-reducing sample buffer (233 mM Tris-Cl, pH 6.8, 138 mM SDS, 19% glycerol, 0.01% bromphenol blue), fractionated on 6% polyacrylamide-SDS gels, and transferred to nitrocellulose. The primary antibody was goat anti-human FXI IgG (Enzyme Research Laboratories, South Bend, IN) and secondary antibody was horseradish peroxidase-conjugated anti-goat IgG. Detection was by chemiluminescence.
Preparation of FXI with a Single Catalytic Active Site (1/2-FXIa)-As will be shown, FXI activation proceeds through an intermediate with one activated subunit (1/2-FXIa). Plasma FXI (0.6 -12 M) in 50 mM Hepes, pH 7.4, 125 mM NaCl, underwent limited digestion by incubating with FXIIa (625 nM) or thrombin (860 nM) for 1 h at 24°C. Reactions were terminated by addition of corn trypsin inhibitor (8.6 M) or hirudin (20 M), respectively. The mixture was chromatographed on benzamidine-Sepharose. Elution was with 50 mM Hepes, pH 7.4, 125 mM NaCl, 50 mM benzamidine. FXI was found in the flow through and 1/2-FXIa in the eluate. 1/2-FXIa was also prepared by chromatography on soybean trypsin inhibitor-agarose. Elution was with 50 mM Hepes, pH 7.4, 1 M NaCl, 1 M benzamidine, 1 mM EDTA. The active site concentration for 1/2-FXIa was determined by complete inhibition with fluorescein-Phe-Pro-Arg-CH 2 Cl, followed by dialysis to remove free inhibitor. The protein concentration was determined by absorbance at 280 nm (corrected for absorbance of the fluorophore) with ⑀ of 214,400 M Ϫ1 cm Ϫ1 , and the fluorescein concentration was determined by absorbance at 491 nm with ⑀ of 79,000 M Ϫ1 cm Ϫ1 .
Preparation of FXIa Bound to Polyacrylamide Beads-FXIa bound to UltraLink Iodoacetyl Resin (Pierce) was prepared by a modification of the method of Dharmawardana and Bock (18). 1/2-FXIa (1 M) in 50 mM Hepes, 125 mM NaCl, 1 mM EDTA, 1 mg/ml polyethylene glycol 8000, pH 7.4 (coupling buffer), was incubated with 20 M ATA-Phe-Pro-Arg-CH 2 Cl (18,19) to inhibit active sites. After dialysis to remove free inhibitor, dialysate (500 l) was mixed with 100 l of packed UltraLink Resin, previously washed with coupling buffer. Protein coupling was initiated by addition of NH 2 OH to 0.1 M, and incubation at 24°C for 4 h with rocking. Beads with bound protein showed no activity in a chromogenic assay using S-2366, indicating the active sites of bound 1/2-FXIa were blocked. To generate an active site on the uncleaved 1/2-FXIa subunit, the beads were incubated with FXIIa (600 nM) at 37°C for 6 h with frequent mixing. Unreacted sites were blocked by rocking with monomeric bovine serum albumin (18) (20 mg/ml) for 12 h at 24°C. Specific activity of the reacted beads was determined by cleavage of S-2366 using FXIa as a standard (  (22), were expressed in HEK 293 cells as described. Protein from stably transfected clones was purified from conditioned media (Cellgro Complete, Mediatech, Herndon, VA) by chromatography using anti-human factor XI-IgG 1G5.12 (17). The column was eluted with 2 M sodium thiocyanate in 25 mM Tris-HCl, pH 7.5, 100 mM NaCl (Tris/NaCl). Protein containing fractions were pooled and concentrated by ultrafiltration, dialyzed against Tris/NaCl, and stored at Ϫ80°C. FXI/PKA4 and FXI-Ser 362,482 (100 -300 g/ml) were activated by incubation with 2 g/ml FXIIa at 37°C. Complete activation was confirmed by reducing SDS-PAGE. FXI-Ser 362,482 , which lacks the disulfide bond that connects the FXIa heavy chain to the catalytic domain, was reapplied to the 1G5.12 column. The catalytic domain (FXIa CD ) binds to the column, whereas the heavy chain is found in the flow through. FXIa CD was eluted as described above and dialyzed against Tris/NaCl.

Preparation of Recombinant FIX Cleavage Site Mutants-
The nucleotides coding for Arg 145 or Arg 180 in the human FIX cDNA were changed to Ala (GCG) using a Chameleon kit (Stratagene, La Jolla). The constructs encode proteins designated FIX-Ala 145 and FIX-Ala 180 . cDNAs were ligated into vector pJVCMV, and stably expressing HEK293 cell lines (ATCC CRL 1573) were prepared as described (23). Proteins were purified from conditioned media (Cellgro Complete supplemented with 10 g/ml vitamin K 1 ) by chromatography using a calciumdependent monoclonal IgG (SB 249417) that recognizes the properly ␥-carboxylated FIX Gla domain, as previously described (23).
FIX Activation by FXIa Followed by Chromogenic Substrate Cleavage (22)-FIX (25-2000 nM) in assay buffer was activated by FXIa (1-6 nM active sites), 1/2-FXIa (1-4 nM active sites), or FXIa-1/2i (2.5-10 nM active sites) at 24°C. At various time points between 0 and 120 min, 60-l aliquots were removed and mixed with 6 l of assay buffer containing 150 M aprotinin. Aprotinin completely inhibited FXIa without affecting FIXa␤ activity. Sixty-six microliters of 1 mM S299 in assay buffer with 66% ethylene glycol was added to the quenched sample, and substrate hydrolysis was followed by measuring the change in absorbance at 405 nm. Generation of FIXa␤ as a function of time was determined by interpolation of the linear dependence of the initial rate of S299 hydrolysis on known concentrations of FIXa␤.
Initial rates for progress curves of FIXa␤ generation were obtained by analyzing the first 5 min of each curve with a second order polynomial equation. Resulting v o values were fit by the Michaelis-Menten equation, and values for K m and k cat were obtained from direct non-linear least squares analysis. The initial 5 min of FIX activation are minimally influenced by product inhibition, and defined the K m and k cat values adequately. The values for K m and k cat were used to analyze complete progress curves by the integrated Michaelis-Menten equation. With values for K m and k cat fixed, full progress curves were fitted simultaneously by the integrated rate equation with product inhibition to obtain estimates of K i for FIXa␤.
Activity of FXIa in Plasma Clotting Assays-FXIa enzymes were diluted to 5 g/ml in 20 mM Tris-Cl, 100 mM NaCl, 1 mg/ml bovine serum albumin, pH 7.4, and serial 1:2 dilutions were prepared in the same buffer. Sixty l of each dilution was mixed with an equal volume of FXI-deficient plasma, and rabbit brain cephalin, followed by incubation for 30 s at 37°C. Sixty l of 25 mM CaCl 2 was added and the time to clot formation was determined on a Dataclot 2 fibrometer (Helena Laboratories, Beaumont, TX). Clotting times were plotted against enzyme concentration on a log-log plot, and FXIa activity was determined as a percent of control by comparison to a control curve constructed with plasma FXIa.
FIX Activation Followed by Western Blot-Plasma or recombinant FIX (100 nM) were incubated in assay buffer at 24°C with various FXIa species, or with FVIIa in the presence of saturating human TF (Innovin, Dade-Behring, Miami, FL). In some reactions, FXIa inhibited by a tripeptide chloromethyl ketone (FXIai) was included. At various times, 7-l samples were mixed with 7 l of reducing sample buffer (233 mM Tris-Cl, 138 mM SDS, 19% glycerol, 10% 2-mercaptoethanol, 0.01% bromphenol blue, pH 6.8), fractionated on 12% polyacrylamide-SDS gels, and then transferred to nitrocellulose. The primary antibody was goat anti-human FIX polyclonal IgG (Enzyme Research Laboratories), and the secondary antibody was horseradish peroxidase-conjugated anti-goat IgG. Detection was by chemiluminescence. The relative positions of bands representing FIX, FIX␣, FIXa␣, and FIXa␤ were determined by Western blots of standards for each protein.
Titration of Active Site-labeled FXIa with FIX and FIXa␤-FXIa (6.25 M) was diluted in titration buffer (50 mM Hepes, 125 mM CaCl 2 , 1 mg/ml polyethylene glycol 8000, pH 7.4) and inhibited with a 10-fold molar excess of 1,5-dansyl-Glu-Gly-Arg-CH 2 Cl at 24°C. Residual FXIa activity was determined by diluting aliquots to 20 nM FXIa in 120 l of titration buffer with 500 M S-2366, and monitoring changes in absorbance at 405 nm until FXIa activity was reduced Ն99.9%. The concentration of inhibited active sites per mol of FXIa was determined by measuring dansyl concentration by absorbance at 335 nm with ⑀ of 8,000 M Ϫ1 cm Ϫ1 , and FXIa concentration by absorbance at 280 nm (corrected for absorbance of the fluorophore) with ⑀ of 214,400 M Ϫ1 cm Ϫ1 . Probe incorporation was 1.98 moles per mole of FXIa. Fluorescence titrations were performed with an SLM 8100 fluorometer, using acrylic cuvettes coated with polyethylene glycol 20,000. Fluorescence intensity titrations of dansyllabeled FXIa were performed with 335 nm excitation (16-nm band pass) and 552 nm emission (16-nm band pass) in titration buffer supplemented with 2 M D-Phe-Pro-Arg-CH 2 Cl at 24°C. The quadratic binding equation was fit to fluorescence changes ((F obs Ϫ F o )/F o ϭ ⌬F/F o ) as a function of total FIX concentration, to determine the maximum change in fluorescence (⌬F max /F o ), dissociation constant (K d ), and stoichiometry (n) using SCIEN-TIST software (MicroMath Scientific Software, Salt Lake City, UT). Parameter errors represent 95% confidence intervals.

Activation of FXI by Factor XIIa and Thrombin-
The conversion of the 80-kDa subunits of FXI to the 50-kDa heavy chains and 30-kDa catalytic domains of FXIa is evident on reducing polyacrylamide gels (Fig. 1A). On non-reducing gels, the 160-kDa FXI dimer migrates slightly more rapidly than FXIa (Fig. 1B). During time course experiments, a species migrating in an intermediate position between FXI and FXIa was observed on non-reducing gels (Fig. 1B). The PTT assay is used in clinical laboratories to assess plasma coagulation initiated by surface-dependent activation of FXII (contact activation). Western blots of human plasma exposed to a silica-containing PTT reagent (Fig. 2) demonstrated a band migrating between FXI and FXIa in normal plasma ( Fig. 2A). Generation of FXIa and the intermediate were dependent on FXIIa (Fig. 2B) and the plasma protein HK (Fig. 2C), which is required for FXI binding to the contact surface.
An intermediate was also observed during FXI activation by ␣-thrombin (Fig. 3A). When a mixture of FXI and the intermediate was chromatographed on benzamidine-Sepharose, the intermediate bound (  FXIa with One Inhibited Active Site (FXIa-1/2i)-We prepared an additional single active site FXIa species, FXIa-1/2i, using the procedure shown in Fig. 3A. 1/2-FXIa was incubated with a tripeptide chloromethyl ketone to irreversibly inhibit the active sites. After dialysis, the uncleaved subunit of 1/2-FXIa was activated with FXIIa. The only new active sites formed in this step are due to cleavage of this subunit. No new fully active FXIa was generated because zymogen FXI was removed during preparation of 1/2-FXIa, and any FXIa in the 1/2-FXIa preparation was inactivated by the chloromethyl ketone. The product, FXIa-1/2i, has two active sites per molecule, one of which is blocked by the inhibitor, and was separated from traces of inhibited FXIa, by benzamidine-Sepharose chromatography, as shown (Fig. 3A).
Activities of Single Active Site FXIa Species in Chromogenic Substrate Assays-The kinetic parameters for cleavage of the tripeptide substrate S-2366 by FXIa, 1/2-FXIa, and FXIa-1/2i were determined, and the Michaelis-Menten equation was fit to the data (Table 1), treating each active FXIa subunit as an independent enzyme. Substrate affinity (K m ) and catalytic efficiency (k cat ) were similar for the three proteases. Activation of FIX was studied by progress curve analysis and from the FIX dependence of the initial rate of FIXa␤ formation, using a chromogenic assay (Fig. 4) (22). The results, summarized in Table 1, indicate that FXIa and its singly active derivatives have similar apparent affinities (K m ) and turnover numbers (k cat ) for FIX. The results for 1/2-FXIa are unlikely to be due to contaminating FXIa. If 1/2-FXIa (the vast majority of protease in the preparation) did not cleave FIX, and all FIX activation was due to traces of FXIa, the turnover number (k cat ) would be very low relative to the FXIa control. Previously, we demonstrated that FIX and FIXa␤ bind to FXIa in a mutually exclusive manner, and that the K i for product inhibition and the K m for FIX activation were similar (22). In the present studies, the K i values for product inhibition  were similar for FXIa, 1/2-FXIa, and FXIa-1/2i (Table 1), and were reasonably similar to K m values, given the techniques used.
Activities of Single Active Site FXIa Species in Plasma Coagulation Assays-To determine whether 1/2-FXIa and FXIa-1/2i activate FIX in plasma, the enzymes were compared with FXIa in an assay that requires FXIa to activate FIX during generation of a fibrin clot. The relative activities of 1/2-FXIa (102%) and FXIa-1/2i (130%) were comparable with fully active FXIa (activity arbitrarily set at 100%) when corrected for the number of active sites per molecule.
Cleavage of FIX by Factor VIIa/TF and FXIa-During FIX activation by FVIIa/TF, the Arg 145 -Ala 146 bond is cleaved initially, resulting in formation of the intermediate FIX␣ (Fig. 5, A  and B) (13)(14)(15). In contrast, FIX cleavage by FXIa generates little intermediate (Fig. 5C). In our hands (Fig. 5D), and others (15), traces of intermediate migrating in the position of FIX␣ are seen in some time courses with FXIa. FXI circulates in plasma as a complex with HK (24). The apparent rate and pattern of FIX cleavage by FXIa was not changed by a saturating concentration (500 nM) of HK under identical conditions to those in Fig. 5 (data not shown).
Cleavage of FIX by FXIa in the Presence of FXIai-The initial interaction of FIX with FXIa involves exosites on the FXIa heavy chain that are available on active site inhibited FXIa (FXIai, see below) (22), and FXIai is expected to behave as a competitive inhibitor of FIX activation by FXIa. The rate of FIX cleavage is significantly reduced when FIX is activated by FXIa in the presence of a 1000-fold molar excess of FXIai (Fig. 6). Interestingly, accumulation of FIX␣ is also observed, suggesting that 1) FXIa initially cleaves the FIX Arg 145 -Ala 146 bond to form FIX␣, and 2) that FIX␣ is released from FXIa and is avail-

TABLE 1 Kinetic parameters for cleavage of S-2366 and activation of FIX by FXIa, 1/2-FXIa, and FXIa-1/2i
Values for K m and k cat for S-2366 cleavage were determined by fitting the Michaelis-Menten equation to substrate dependence curves using eight concentrations of S-2366. K m and k cat for FIX activation were based on initial rates for FIX activation from full progress curves (Fig. 4) as described under "Experimental Procedures." The resulting v o values were analyzed by fitting the Michaelis-Menten equation, and values for K m and k cat were obtained from direct non-linear least squares analysis. Values for K m and k cat were used to analyze the complete progress curves by the integrated Michaelis-Menten equation. With values for K m and k cat fixed, full progress curves were fitted simultaneously by the integrated rate equation with product inhibition to obtain estimates of K i for FIXa␤. Errors in parameters represent 95% confidence intervals.  (Fig. 7). FIX-Ala 145 and FIX-Ala 180 can only be cleaved to FIXa␣ and FIX␣, respectively. FXIa clearly shows a preference for cleavage after Arg 145 (Fig. 7B) Fig. 8A, and was not observed in the kinetic studies (Fig. 4).

Substrate
FXIa-1/2i is a dimer with one functional and one blocked active site. The method used to prepare this enzyme makes contamination with FXI or FXIa unlikely. Activation of FIX by FXIa-1/2i generated FIXa␤ with little intermediate accumulation (Fig. 8B). We also examined FIX activation by FXIa bound to polyacrylamide beads through an inhibitor occupying one of the FXIa active sites. Functional FXIa active sites are generated after protein was bound to the bead, and it is not possible for fully active FXIa to be present. Intermediate accumulation was not observed when bound FXIa activated FIX (Fig. 9), showing that FXIa with a single active subunit cleaves the FIX activation sites normally when the other subunit is tethered to a surface.
Activation of FIX by FXIa/PKA4-Availability of monomeric FXIa would offer another approach to address the importance   of the dimeric structure to FIX activation. Removing the interchain disulfide bond involving Cys 321 in the FXI fourth apple (A4) domain does not produce a pure monomeric protein because of non-covalent interactions between the two FXI subunits (21,25). FXI has a high degree of structural homology to plasma prekallikrein (PK) (4, 26), which is a monomer. Previously, we described recombinant FXIa in which the A4 domain is replaced with the PK A4 domain (20). The chimera, FXIa/ PKA4, is a monomer on size exclusion chromatography (21), and cleaves S-2366 and FIX with similar kinetic parameters to FXIa (20). As with the single active site plasma FXIa species, little intermediate was formed during FIX activation by FXIa/ PKA4 (Fig. 8C).
Activation of FIX by FXIa CD -Previously we described recombinant FXIa catalytic domain (FXIa CD ), prepared by activating a FXI variant lacking the Cys 362 -Cys 482 disulfide bond that connects the heavy chain and catalytic domain in FXIa (22). FXIa CD cleaves a chromogenic substrate similarly to FXIa, but is a poor activator of FIX (Fig. 10A), likely due to the loss of substrate binding exosites on the heavy chain (22). Sinha et al. (27) showed that FXIa CD cleaved FIX with accumulation of FIX␣. In these studies, which used stained gels and high substrate concentrations, FIX␣ was also evident during FIX activation by wild type FXIa. We activated a physiologic concentration of FIX (100 nM) with a high concentration of FXIa CD and also observed significant accumulation of FIX␣ (Fig. 10B). The result is distinctly different from those for other FXIa single active site species, and indicates that activation of FIX with limited intermediate accumulation requires the FXIa heavy chain.
The Stoichiometry of FIX and FIXa␤ Binding to FXIa-The data presented so far strongly indicate that each FXIa subunit behaved as a complete enzyme toward FIX. Each FXIa dimer, therefore, should bind two FIX molecules. FXIa was inhibited with a tripeptide chloromethyl ketone linked to a fluorescent dansyl group that functions as a reporter of change in the microenvironment around the FXIa active site (19,28). The fluorescence probe is covalently linked via the chloromethyl ketone to the active site catalytic histidine and serine residues (Fig. 11A). Because initial binding of FIX to FXIa involves interactions with exosites remote from the active site (20,22), blocking the active site does not significantly affect FIX binding.
Binding of FIX to labeled FXIa increased the dansyl fluorescence (Fig. 11B), with a maximum enhancement at saturation of 43 Ϯ 3%, and a stoichiometry of 1.9 Ϯ 0.4 mol of FIX per mol of FXIa. The K d for the interaction (70 nM Ϯ 40 nM) was consistent with published results (22,23). Earlier work showed that FIX and FIXa␤ have similar affinities for FXIa (22). Binding of FIXa␤ to labeled FXIa increased the fluorescence to a maximum of 23 Ϯ 1% (Fig. 11C). The stoichiometry of FIXa␤ binding to FXIa was ϳ2:1 (2.2 Ϯ 0.4), with a K d of 100 Ϯ 50 nM. The results support the conclusion that each half of the FXIa dimer functions as a complete enzyme toward FIX.

DISCUSSION
In 1977, Bouma and Griffin (3) reported that human FXI was comprised of two disulfide bond-linked polypeptides. FXI is the only coagulation protease that is a dimer (1,2). An interchain bond involving Cys 321 links the identical 80-kDa FXI subunits in all mammalian species studied, with the exception of the rabbit (His 321 ) (29). Rabbit FXI is, however, a non-covalently associated dimer (29). Activation of each FXI subunit requires cleavage of the Arg 369 -Ile 370 bond (3)(4)(5). It has been assumed that FXIa formed during coagulation has two cleaved subunits, and essentially all studies of the kinetic and binding properties of plasma FXIa have been conducted with this type of protease. Bouma and Griffin (3) proposed the existence of a form of FXIa with one cleaved subunit, but to date it has not been described in either purified or plasma systems. Indeed, the standard practice of following FXI activation with reducing gels will not allow fully and partially activated species to be distinguished.  Activation of FXI results in formation of functional exosites and active sites that facilitate FIX binding and cleavage (3,4,22). The accompanying conformational changes cause FXIa to migrate slower than FXI on non-reducing polyacrylamide gels. When FXI is activated in solution by FXIIa or ␣-thrombin, or in plasma by contact activation, a species migrating in an intermediate position between FXI and FXIa is evident.  , is an active protease as demonstrated by its capacity to 1) bind to benzamidine and soybean trypsin inhibitor, 2) cleave the substrates S-2366 and FIX, and 3) promote clot formation in plasma.
As discussed, it is unlikely that the activities of 1/2-FXIa are attributable to contaminating FXIa. However, to address this issue we prepared FXIa-1/2i, which has one functional and one inhibited active site per dimer, using techniques that remove FXI and FXIa. FXIa-1/2i in solution, or bound to the surface of a bead performs similarly to FXIa and 1/2-FXIa, providing further support for the concept that one active site is sufficient for FXIa activation of FIX.
These data raise questions concerning the predominant FXIa species during coagulation. While fully activated FXIa forms in plasma during contact activation, a relatively large amount of FXIa is required to initiate coagulation through this mechanism. It is postulated, however, that fibrin formation in vivo is initiated by factor VIIa/TF, with FXIa serving a secondary role in clot maintenance (30). Models of this process indicate that very low concentrations of FXIa activity (subpicomolar) can affect fibrin stability (31)(32)(33). In the absence of an artificial surface, both FXIIa and ␣-thrombin activate FXI to fully activated FXIa slowly, and 1/2-FXIa may be a major active FXI species in plasma.
In current models of hemostasis, FIX activation by FVIIa/TF is required for sustained generation of factor Xa (1, 2, 30), whereas activation by FXIa contributes to consolidation of coagulation in tissues with high fibrinolytic activity (34). In both cases, FIX is cleaved after Arg 145 and Arg 180 to generate FIXa␤ (9 -11). FVIIa/TF initially cleaves FIX after Arg 145 , forming FIX␣, which accumulates prior to formation of FIXa␤ (9,(12)(13)(14). This suggests that cleavage at Arg 180 is rate-limiting, and that FIX␣ dissociates from FVIIa/TF and is reacquired prior to cleavage after Arg 180 . Interestingly, FXIa also cleaves FIX preferentially after Arg 145 , as shown in studies with FIX-Ala 145 and FIX-Ala 180 , however, an intermediate does not accumulate appreciably (9 -11, 15). Two mechanisms could explain these findings. FXIa may activate FIX by sequentially cleaving Arg 145 and Arg 180 prior to releasing FIXa␤, and without release of an intermediate (a processive mechanism). Alternatively, FIX␣ may dissociate from FXIa, and rebind to facilitate cleavage at Arg 180 . In this case, the rate of conversion of FIX␣ to FIXa␤ would be faster than for FIX to FIX␣ to explain the lack of intermediate accumulation. Activation of prothrombin to ␣-thrombin by factor Xa in the prothrombinase complex is a well characterized example of the latter mechanism (35-37).
Wolberg et al. (15) proposed a processive mechanism for FIX activation by FXIa, based on the absence of intermediate on Western blots accumulation, and the observation that FIX and the intermediates FIX␣ and FIXa␣ are converted to FIXa␤ by FXIa at approximately similar rates. They postulated that the two protease domains of FXIa may cleave the activation sites of one FIX molecule, either simultaneously or sequentially, prior to releasing FIXa␤ (15). In the crystal structure of zymogen FXI, the catalytic domains are at opposite ends of the molecule, and would be unable to interact simultaneously with one FIX mol- ecule (38). However, work on the structure of the FXI A4 domain by Samuel et al. (16) suggests that conformational changes occur during FXI activation that bring the catalytic domains into closer proximity. In the structure for FXI A4 dimer, an ␣-helix not present in the zymogen crystal structure is observed. It is postulated that this ␣-helix forms after cleavage of Arg 369 -Ile 370 , and alters interdomain contacts with the opposite A4 domain. This reorients the two halves of the dimer so that the catalytic domains are closer together, permitting them to interact with a single FIX molecule. The availability of the FXIa species with single active sites provided us with a means to address these intriguing proposals.
The results presented here demonstrate that the mechanism involved in FIX activation by FXIa applies to catalysis by individual subunits of the FXIa dimer, and do not support a model in which two catalytic domains are required for normal FIX cleavage. Each FXIa subunit, therefore, can be considered a separate enzyme. Of the more than one hundred trypsin-like proteases described, only FXIa (3)(4)(5)38) and the T-lymphocyte apoptotic protease granzyme A (39,40) are homodimers. Granzyme A must be a dimer for proper cleavage of macromolecular substrates. The substrate binding pockets of granzyme A are near the dimer interface, and substrate binding exosites on the adjacent subunit extend the active site clefts across the interface (39,40). The exosites are not available in the protease monomer, resulting in poor substrate recognition. Recognition of FIX by FXIa also involves exosite interactions (17,20,22,27,41), however, the current results indicate that the active site and exosites required for FIX activation reside on the same FXIa subunit.
Furthermore, the results are not supportive of a processive mechanism in which both factor IX activation sites are cleaved without release of an intermediate. Instead, the data strongly indicate that FIX is cleaved initially at the Arg 145 -Ala 146 bond, forming FIX␣, which is released from FXIa. FIX␣ must then rebind to FXIa, probably in a new conformation facilitating cleavage of the Arg 180 -Val 181 bond, to complete conversion to FIXa␤. It is clear that the FXIa heavy chain is crucial for this process. In addition to a marked decrease in rate of FIX activation (27,42,43), substantial accumulation of FIX␣ prior to formation of FIXa␤ is observed when FIX is incubated with the single active site species FXIa CD , which lacks a heavy chain. The loss of FIX binding exosites on the heavy chain appears to have a greater effect on cleavage after Arg 180 than Arg 145 (27). One FXIa heavy chain exosite likely interacts with the FIX-Gla domain (23). Sinha et al. (27) showed that FIX␣ accumulated during FIX activation by FXIa in the absence of calcium, supporting the notion that loss of the Gla-FXIa heavy chain interaction disproportionately affects cleavage at Arg 180 .
The observation that FXI is a dimer in all species examined strongly suggests that this is important for some aspect of protease function. Other coagulation serine proteases have vitamin K-dependent modifications to the Gla-domain that facilitate binding to platelets and phospholipid (1,2,44). FXI has no Gla-domain, but binds to platelets through platelet glycoprotein 1b (45)(46)(47). Previously, we proposed a model where one FXIa subunit tethers the molecule to a platelet, whereas the other interacts with FIX (48). The observation that 1/2-FXIa has full activity toward FIX despite having only one cleaved subunit raises the possibility that FXI can be activated to 1/2-FXIa on the platelet, with the unactivated subunit remaining bound to glycoprotein 1b. This hypothesis is supported by the finding that FXIa linked to beads through one active site activates FIX with minimal intermediate generation.