Productive Recognition of Factor IX by Factor XIa Exosites Requires Disulfide Linkage between Heavy and Light Chains of Factor XIa*

Background: The heavy chain of FXIa (FXIa-HC) is essential for efficient FIX activation, although the light chain (FXIa-LC) contains the active site. Results: Efficient FIX activation by FXIa occurred only when FXIa-HC and FXIa-LC were disulfide-linked. Conclusion: Substrate recognition by FXIa exosites requires disulfide-linked FXIa-HC and FXIa-LC. Significance: Covalent linkage between FXIa-HC and FXIa-LC is required to form the substrate binding exosite. In the intrinsic pathway of blood coagulation factor XIa (FXIa) activates factor IX (FIX) by cleaving the zymogen at Arg145-Ala146 and Arg180-Val181 bonds releasing an 11-kDa activation peptide. FXIa and its isolated light chain (FXIa-LC) cleave S-2366 at comparable rates, but FXIa-LC is a very poor activator of FIX, possibly because FIX undergoes allosteric modification on binding to an exosite on the heavy chain of FXIa (FXIa-HC) required for optimal cleavage rates of the two scissile bonds of FIX. However preincubation of FIX with a saturating concentration of isolated FXIa-HC did not result in any potentiation in the rate of FIX cleavage by FXIa-LC. Furthermore, if FIX binding via the heavy chain exosite of FXIa determines the affinity of the enzyme-substrate interaction, then the isolated FXIa-HC should inhibit the rate of FIX activation by depleting the substrate. However, whereas FXIa/S557A inhibited FIX activation of by FXIa, FXIa-HC did not. Therefore, we examined FIX binding to FXIa/S557A, FXIa-HC, FXIa-LC, FXIa/C362S/C482S, and FXIa/S557A/C362S/C482S. The heavy and light chains are disulfide-linked in FXIa/S557A but not in FXIa/C362S/C482S and FXIa/S557A/C362S/C482S. In an ELISA assay only FXI/S557A ligated FIX with high affinity. Partial reduction of FXIa/S557A to produce heavy and light chains resulted in decreased FIX binding, and this function was regained upon reformation of the disulfide linkage between the heavy and the light chains. We therefore conclude that substrate recognition by the FXIa exosite(s) requires disulfide-linked heavy and light chains.

Factor IX (FIX) 3 , an essential component of the blood coagulation cascade, can be activated by the factor VIIa (FVIIa)/ tissue factor (TF) complex in the extrinsic pathway (1)(2)(3)(4) and by factor XIa (FXIa) in the intrinsic pathway of blood coagulation (5)(6)(7). Conversion of zymogen FIX to its activated form (FIXa) is achieved by cleavages at two peptide bonds one at Arg 145 -Ala 146 and the other at Arg 180 -Val 181 . FXIa, generated by proteolytic cleavage of zymogen FXI by FXIIa, thrombin, or FXIa (8,9), is an essential component of the consolidation phase of blood coagulation because tissue factor pathway inhibitor in complex with factor Xa shuts down the extrinsic pathway by inhibiting FVIIa/TF (10,11). FXIa exists in human plasma at a concentration of ϳ30 nM as a homodimer consisting of two identical subunits linked by a disulfide bond formed by Cys 321 in each heavy chain (12,13). Thus, each subunit consists of one N-terminal heavy chain joined to the C-terminal light chain (or catalytic domain) by a disulfide bond formed between Cys 362 in the heavy chain and Cys 482 in the light chain. It was originally established in our laboratory (14,15) and later confirmed by other investigators (16 -18) that binding of FIX to an exosite on the heavy chain of FXIa is essential for optimal activation of FIX. In addition, a second exosite on the light chain of FXIa has recently been identified from kinetic analyses of FIX activation by FXIa-LC (19).
This investigation is aimed at understanding the structural and functional features within FXIa required for the interaction of FIX with these two FXIa exosites in the generation of FIXa. We and others have shown earlier that the catalytic domain of FXIa (FXIa-LC) is a poor activator of the macromolecular substrate FIX, although both intact FXIa and the FXIa-LC hydrolyze the chromogenic substrate S-2366 at comparable rates (14 -16, 20). Thus, the isolated catalytic domain of FXIa possesses normal enzymatic activity against small peptide substrates but deficient activity against the normal macromolecular substrate FIX. There are several plausible explanations for this discrepancy, one of which is that FIX undergoes an allos-teric modification upon binding to the heavy chain of FXIa that facilitates access of the two FIX scissile bonds for cleavages by FXIa. Validation of this mechanism would predict that preincubation of FIX with the isolated heavy chain of FIXa would potentiate rates of FIX activation by the isolated catalytic domain of FXIa. Alternatively, if the affinity of the enzymesubstrate interaction is determined solely by the binding of FIX to the heavy chain of FXIa, then FIX activation by FXIa should be inhibited in the presence of excess of FXIa-HC because of substrate depletion. These two hypotheses were examined in experiments, the results of which are reported in this manuscript.
Preparation of Constructs for FXI/S557A, FXI/C362S/C482S, and FXI/S557A/C362S/C482S-The cDNA constructs of FXI mutants were made using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), with wild-type FXI cDNA inserted into the expression vector pJVCMV (12,18,19) serving as the template. Each mutant was sequenced using both reverse and forward primers to confirm the presence of the desired mutation and to ensure the absence of any polymeraseinduced errors.
Expression and Purification of FXI Mutants-Stable cell lines expressing the wild type and mutants of FXI were prepared by transfection of HEK293 cells as described earlier (19,21,22). Cells were grown in DMEM with 10% fetal bovine serum for 48 h and then switched to the same medium supplemented with 500 g/ml of G418. Expression of FXI in the medium was determined by enzyme-linked immunosorbent assay as described previously (19,21,22). Cell lines exhibiting maximal expression of FXI proteins were expanded in roller bottles (Nunc, Naperville, IL). After reaching confluence, cells were washed, and the media were replaced by serum-free DMEM supplemented with ITS (insulin-transferrin-selenium-A, Invitrogen), soybean trypsin inhibitor (10 g/ml), and lima bean trypsin inhibitor (10 g/ml). Conditioned media were removed every 48 h and kept frozen after supplementing with benzamidine (5 mM) and EDTA (5 mM) until ready to be purified.
Expressed wtFXI, FXI/S557A, FXI/C362S/C482S and FXI/ S557A/C362S/C482S proteins in the conditioned medium were purified using a monoclonal antibody (5F7) affinity column according to the protocol described earlier (19,21,22). Briefly, the cell supernatant was passed through the affinity column at a rate of 0.5 ml/min. The column was washed with 10 column volumes of TBS followed by elution of the protein using 2 M potassium thiocyanate. The eluted protein was dialyzed and concentrated using a Centricon Plus-20, 30,000 NMWL ultrafiltration unit (Millipore, Bedford, MA). To prepare FXI-HC and FXI-LC, the concentrated and purified FXI/C362S/C482S and FXI/S557A/C362S/C482S were activated with FXIIa at a substrate to enzyme ratio of 20:1 at 37°C for 24 h. The incubation mixture was passed through a 5F7 antibody column, the unbound LC was purified from FXIIa using corn trypsin inhibitor-coupled agarose beads, and the HC bound to the 5F7 column was eluted and purified as described for FXI/C362S/ C462S (21).
Protein concentration was estimated using the BCA assay (Pierce). The purified proteins were size-fractionated by 4 -15% SDS-PAGE under reducing and non-reducing conditions and stained using GelCode blue stain (Pierce). An Odyssey infrared imager (LI-COR, Inc., Lincoln, NE) was used for visualization and densitometric analysis of the bands.
Kinetic Assay of FIX Activation by FXIa-LC after Preincubation of FIX with the Heavy Chain of FXIa-FIX activation by FXIa-LC was carried out using a modification (19) of a coupled assay described originally by Wagenvoord et al. (23). In this assay, FIX is first activated by FXIa-LC to FIXa, which is then used to activate FX in the presence of FVIIIa, phospholipids, and calcium. FXa thus generated was estimated using the chromogenic substrate S-2765. All reagents were diluted in TBS containing 0.1% BSA (TBSA). Varying concentrations of FIX (0.2-2.2 M) were incubated with a fixed concentration (2.88 M) of FXIa-HC or buffer for 20 min, followed by activation by FXIa-LC (15-20 nM) for 5-10 min in a microtiter plate at 37°C in a total volume of 50 l of TBSA supplemented with 5 mM CaCl 2 . Adding EDTA to 25 mM and chilling on ice stopped the reactions. 5 l of the reaction mixture was then added to chilled TBSA containing 25 mM EDTA to make a total volume of 200 l. 5 l from the diluted reaction mixture was added to 60 l of TBSA containing 10 mM CaCl 2 , followed by addition of 10 l of 10 M phospholipid vesicles (phosphatidylcholine/phosphatidylserine) and incubation for 2 min at 37°C. 10 l of freshly activated FVIIIa (66 units/ml) followed by 10 l of FX (2 M) were then added, and the reaction mixture was incubated for 3 min at 37°C. Final concentrations of FVIIIa, FX, and phosphatidylcholine/phosphatidylserine in the reaction mixture were 6.6 units/ml, 0.2 M, and 1 M, respectively. Adding EDTA to 25 mM and placing it on ice stopped the activation of FX by FIXa. 50 l of each reaction was mixed with 50 l of 1 mM S-2765, and the change in absorbance at 405 nm was followed on a Ther-moMax microtiter plate reader. The concentration of FIXa generated was obtained from a standard curve constructed using known concentrations of purified FIXa.

Western Blot Analyses of FIX Activation by FXIa-LC after Preincubation of FIX with the Heavy Chain of FXIa-FIX (1 M)
was incubated with buffer or FXIa-HC (2 M) for 15 min at room temperature and then activated using FXIa-LC (15 nM) as described above. In a separate experiment, we examined the activation of FIX using a higher concentration of FXIa-LC (50 nM) after preincubating FIX (1 M) with varying concentrations of FXIa-HC (1-5 M). The reaction products were fractionated by SDS-PAGE. The protein bands were transferred to a nitrocellulose membrane using a semidry transfer apparatus (Bio-Rad). The membrane, after thorough washing, was incubated with goat anti-human FIX polyclonal antibody (ERL, South Bend, IN), and the bound antibody was detected by HRP-conjugated anti-goat IgG (Sigma-Aldrich, St. Louis, MO) and visualized by chemiluminescence. . FIX cleavage products were analyzed by Western blot analysis as described above.

Factor IX Binding to the Active Site Inhibited FXIa (FXIai), Isolated Heavy Chain (FXI-HC), Isolated Light Chain (FXIa-LCi), and to Zymogen FXI-To prevent activation of FIX by
FXIa or by FXIa-LC during prolonged incubation in the ELISA assay, the two proteins were irreversibly inhibited by incubation with 4-amidinophenylmethanesulfonyl fluoride by making five sequential additions of 5 l of 10 mM 4-amidinophenylmethanesulfonyl fluoride to 500 l of the protein solution (2-5 M) and allowing 5 min after each addition for the inhibitor to interact with the protein. The inactivated proteins were then dialyzed extensively before being used in the ELISA assay. The proteins were inactivated by Ͼ 97%, as judged by their ability to hydrolyze S-2366. 100 l of FIX (10 g/ml) in 50 mM carbonate buffer (pH 9.6) was added to wells of a microtiter plate and incubated overnight at 4°C. Uncoated sites of the wells were blocked by incubation with 200 l of BSA at a concentration of 1 mg/ml. FIX coated to the wells was allowed to interact with varying concentrations (as indicated in the figures) of FXIai, FXIai/LC, FXI-HC, or FXI in TBS containing Ca 2ϩ (5 mM) for 90 min at room temperature. The wells were thoroughly washed with TBS containing Ca 2ϩ (5 mM) and Tween 20 (0.02%) to remove the unbound proteins, followed by detection of FIX-bound FXIai, FXIai/LC, FXI-HC, or FXI using HRPconjugated polyclonal antibody made against FXI (Enzyme Research Laboratories) according to the protocol provided by the supplier.
Comparison of FIX Binding to FXIa/S557A and FXIa/S557A/ C362S/C482S-Like plasma FXIa, the mutant FXIa/S557A is a dimer of two identical subunits joined by a disulfide linkage through Cys 321 in the heavy chain, and each subunit of the enzyme is a two-chain molecule joined by Cys 362 in the heavy chain and Cys 482 in the light chain. Because both Cys 362 and Cys 482 are mutated in FXI/S557A/C362S/C482S, the activated triple mutant produces a dimeric heavy chain (one heavy chain from each subunit joined by the interchain disulfide bond at Cys 321 ) of ϳ90 kDa and a free light chain of ϳ35 kDa. ELISA assays of FXIa/S557A and FXIa/S557A/C362S/C482S for FIX binding were performed as described above.
In a separate experiment a fixed concentration of FIX (5 nM) was incubated with varying concentrations (15 nM-1 M) of either FXIa/S557A or FXIa/S557A/C362S/C482S in solution for 15 min at room temperature. These preincubated samples (100 l) were then added to the wells of a microtiter plate coated with FXIa/S557A. FIX bound to the wells was detected using HRP-conjugated polyclonal antibody to FIX.
Comparison of HK Binding to FXIa/S557A and FXIa/ S557A/C362S/C482S-FXI/FXIa binds to HK through its heavy chain (24,25). To examine whether FXIa-HC retains its capacity to bind to HK we compared binding of HK to FXIa/S557A and FXIa/ S557A/C362S/C482S. For this purpose, the wells of a microtiter plate were coated with HK by incubation of 100 l of HK at a concentration of 10 g/ml in 50 mM carbonate buffer (pH 9.6) as for the ELISA assays described above. After blocking free sites with BSA, either FXIa/S557A or FXIa/S557A/C362S/C482S at concentrations designated in the figure were added and incubated for 90 min followed by extensive washing. HK-bound proteins were then detected using peroxidase-conjugated polyclonal antibody to FXI as described above.
Comparison of FXIa and FXIa/C362S/C482S in the S-2366 Hydrolysis Assay-To assess whether the light chain of FXIa free from the heavy chain retains its active site integrity, we compared FXIa and FXIa/C362S/C482S in their ability to hydrolyze the small peptidyl substrate S-2366. FXIa or FXIa/ C362S/C482S at a concentration of 3 nM in TBS containing 1 mg/ml BSA was added to varying concentrations (0 -1.5 mM) of S-2366 in TBS, and the rate of hydrolysis was measured by monitoring generation of p-nitroaniline at 405 nm in a plate reader.
Comparison of Plasma FXI (pFXI), FXI/S557A, and FXI/ S557A/C362S/C482S Binding to Monospecific Polyclonal Antibody Raised against FXI-To examine whether any major structural alteration resulted from mutations at Cys 362 and Cys 482 , we compared plasma FXI, FXI/S557A, and FXI/S557A/ C362S/C482S in their capacity to bind to a monospecific polyclonal antibody raised against plasma FXI (FXI ELISA kit, Enzyme Research Laboratories) using the protocol provided by the manufacturer.
Partial Reduction of FXIa/S557A Followed by Reoxidation-200 l of FXIa/S557A (3.75 M) was partially reduced using 150 M DTT by incubation for 20 min at room temperature as described previously (14). Half of the reduced product was alkylated by further incubation for 30 min with 300 l of iodoacetamide, and both the alkylated and non-alkylated samples were dialyzed for 48 h with two changes of buffer containing 0.02% Tween 20, EDTA (2 mM), and benzamidine (5 mM) and a final change of buffer without EDTA. After dialysis, each sample was estimated for protein concentration and used for FIX binding as described above. FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9

RESULTS
Effect of the Isolated Heavy Chain (FXIa-HC) of FXIa on the activation of FIX by FXIa-LC-A plausible explanation for the observation that the isolated catalytic domain of FXIa is a poor activator of FIX compared with wild-type FXIa, despite its intact enzymatic activity against small peptide substrates, is that FIX, bound to the exosite in the heavy chain of FXIa, undergoes a conformational change that facilitates optimal cleavage rates of the two FIX peptide bonds, leading to the formation of FIXa. If this hypothesis is correct, then activation of FIX by FXIa-LC should be potentiated by preincubation of FIX with FXIa-HC. This hypothesis was tested by examining the rate of activation of FIX preincubated with buffer or FXIa-HC. As shown in Fig. 1A, preincubation of FIX with the isolated heavy chain of FXIa does not potentiate the rate of FIX activation by FXIa-LC. Cleavage patterns of FIX activation by FXIa-LC in the presence and absence of FXIa-HC were also examined by Western blot analysis and are shown in Fig. 1B. We also examined activation of FIX using a higher concentration of FXIa-LC (50 nM) after preincubating FIX (1 M) with varying concentrations of FXIa-HC (1-5 M) for 15 min at room temperature. The results of these experiments (Fig. 1C) demonstrate no discernible difference in the cleavage rate or cleavage pattern of FIX activation by the catalytic domain of FXIa in the presence or absence of the FXIa heavy chain.
Effects of FXIa/S557A and FXIa-HC on the Rate of FIX Cleavages by FXIa-If the affinity of the substrate-enzyme interaction is solely determined through interaction of FIX with the heavy chain of FXIa, then a saturating concentration of isolated FXIa-HC should deplete the substrate, making it unavailable for binding to the enzyme, and thereby inhibiting the rate of FIX activation. We therefore compared the cleavage of FIX (1 M) by FXIa (1.0 nM) in the presence or absence of a 200-fold molar excess of the FXIa-HC (200 nM) (Fig. 2B). The result of this experiment demonstrated no discernible effect of the isolated FXIa heavy chain on the rate or cleavage pattern of FIX activation. On the contrary; the intact molecule with the active site Ser mutated to Ala (FXIa/S557A) and, at a similar concentration (200 nM), was able to inhibit the cleavage rate of FIX activation by FXIa (Fig. 2A). The effects of higher concentrations of FXIa-HC as well as of FXIa/S557A were also examined, as shown in Fig. 2, D and C, respectively. Although FXIa/S557A inhibited cleavage of the substrate almost completely at 500 nM, no demonstrable inhibition could be observed, even at concentrations of FXIa-HC as high as 4 M.
Factor IX Binding to FXIai, Isolated FXIa-HC, FXIa-LCi, and Zymogen FXI-Because FXIa-HC was ineffective in potentiating FIX activation by FXIa-LC or inhibiting the rate of FIXa generation by full-length FXIa, we examined binding of the substrate, FIX to FXIa and to the isolated heavy and light chains of FXIa in an ELISA assay. Because FXIa or FXIa-LC would cleave FIX during long incubation on the plate, both enzymes were inactivated by the irreversible serine protease inhibitor 4-amidinophenylmethanesulfonyl fluoride, followed by dialysis before being used in the ELISA assay. In this assay, FIX was coated on the wells of a microtiter plate, followed by incubation with FXIai, FXIa-HC, FXIa-LCi, or zymogen FXI. Binding of these proteins to immobilized FIX were then detected as described under "Experimental Procedures." The results of these assays are shown in Fig. 3A. Although FXIai was found to bind to FIX with high affinity (K d ϳ0.25-0.5 nM), no detectable binding to FIX of the zymogen FXI or the isolated heavy or light chain could be demonstrated. In a separate experiment, binding to FIX was examined using high concentrations of FXI, FXIa-HC, and FXIa-LCi. As shown in Fig. 3B, binding of these proteins was extremely poor, even at a concentration of ϳ2000fold higher than the apparent K d value of FXIai binding to FIX.
Comparison of Binding to FIX of FXIa/S557A and FXIa/S557A/C362S/C482S-To rule out the possibility that the isolation procedure may have been responsible for the loss of the capacity of FIX to recognize FXIa-HC and FXIa-LC, we investigated the binding to FIX of FXIa/S557A and FXIa/    FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9 linked at Cys 321 and the other consisting of the isolated light chain (Fig. 4B, lane 7). The results of binding to FIX of FXIa/ S557A and FXIa/S557A/C362S/C482S are shown in Fig. 4C. Although FXIa/S557A, like FXIai, binds to FIX with high affinity (K i ϳ0.25-0.5 nM), the FXIa/S557A/C362S/C482S mutant demonstrated no detectable binding.

Cys 362 -Cys 482 Linkage in FXIa Required for FIX Recognition
It is possible that FXIa/S557A/C362S/C482S does indeed bind to FIX in solution but not to FIX coated to the surface of a microtiter plate. To examine this possibility we designed the binding assay as follows. A fixed concentration of FIX (5 nM) was incubated with varying concentrations of either FXIa/ S557A or FXIa/S557A/C362S/C482S in solution for 15 min at room temperature, followed by addition of 100 l of these preincubated samples to the wells of a microtiter plate coated with FXIa/S557A. FIX bound to the wells was detected using HRPconjugated polyclonal antibody to FIX. Because FIX was preincubated with FXIa/S557A or FXIa/S557A/C362S/C482S before adding to the well coated with FXIa/S557A, we would expect to observe a concentration-dependent inhibition of FIX binding to the coated FXIa/S557A in the wells of the plate. The results are shown in Fig. 4D  Comparison of Plasma FXI, FXI/S557A, and FXI/S557A/ C362S/C482S Binding to Monospecific Polyclonal Antibody Raised against FXI-To examine whether there is any major structural alteration of the FXI mutants compared with plasma FXI, we compared binding of FXI antibody to these proteins. As shown in supplemental Fig. 1C, there was no detectable difference between plasma FXI and either of the two mutants in their zymogen forms. The same three proteins in their activated forms also demonstrated indistinguishable binding (data not shown).
Binding of FIX to Partially Reduced and Blocked FXIa/S557A (FXIa/S557A-R/B) and to FXIa/S557A after Partial Reduction followed by Oxidation (FXIa/S557A-R/O)-The experimental results obtained above support the conclusion that, even though the isolated or separated heavy and light chains, both of which contain exosites required for efficient FIX activation, are structurally and functionally intact, they fail to bind FIX and facilitate efficient FIX activation. This suggests the possibility that disulfide linkage between Cys 362 and Cys 482 within the heavy and light chains of FXIa is also required. To examine this possibility, we utilized mild reducing conditions (15) to selectively cleave the interchain disulfide bonds between Cys 362 and Cys 482 (as well as those between the two Cys 321 residues in each FXI subunit) minimally affecting other intrachain disulfides. The resulting protein was then either treated with iodoacetamide to block the two thiols and prevent reestablishment of the disulfide bond or subjected to oxidizing conditions to facilitate partial reformation of the interchain disulfide bond between Cys 362 and Cys 482. As shown in supplemental Fig. 2A, partial reduction of FXIa/S557A with 150 M DTT produced three products, a monomeric FXI/FXIa molecule formed by reduction of the disulfide bond between the two Cys 321 residues in each subunit and the heavy and the light chains formed by reduction of both the intersubunit disulfide bond as well as the disulfide bond between the two cysteines (Cys 362 and Cys 482 ) joining the heavy and the light chains (see supplemental Fig. 2A, non-reduced section). The reduced portion of the gel shows the presence of only the heavy and the light chains because the disulfide bond joining the heavy and the light chains in the monomer is reduced (supplemental Fig. 2A, reduced section). One-half of the reduced protein was then alkylated using a 2-fold molar excess of iodoacetamide to block the reduced cysteine residues. Both of the blocked and unblocked samples were then dialyzed as described under "Experimental Procedures." It is clear that although the reduced and blocked protein FXIa/ S557A-R/B (supplemental Fig. 2B, lane 3) was unable to form any dimeric FXIa, the unblocked protein FXIa/S557A-R/O (supplemental Fig. 2B, lane 2) produced a significant amount of dimeric FXIa, some of which was formed by oxidation of the intersubunit free Cys 321 bonds of the monomeric form and some by disulfide bond formation between the free Cys 362 in the heavy chain and the free Cys 482 in the light chain. In addition to dimeric FXIa, two new bands appeared below the dimeric protein (supplemental Fig. 2B, lane 2). Because the reduced gel (supplemental Fig. 2B, lane 5) showed only two bands comprising the heavy and light chains of FXIa, the band at ϳ90 -100 kDa most likely represents the dimeric heavy chain, and the band at ϳ125-130 kDa represents one monomeric FXIa (80 kDa) joined through Cys 321 to one heavy chain (50 kDa). Total protein concentrations of the dialyzed FXIa/ S557A-R/B and FXIa/S557A-R/O samples were then determined and examined in our FIX binding assay. The results (supplemental Fig. 2C) clearly demonstrate that at all three concentrations of the proteins tested, binding to FIX of FXIa/ S557A-R/O was significantly greater than that of FXIa/S557A-R/B. Thus, formation of the disulfide bond between Cys 362 and Cys 482 within the heavy and the light chains promotes enhanced FIX binding, whereas prevention of this disulfide bond formation inhibits FIX binding. These results support the conclusion that the binding of FXIa to FIX and the efficient activation of FIX by FXIa requires that the heavy and light chains of FXIa be joined together by the disulfide bond between Cys 362 and Cys 482 .

DISCUSSION
We showed previously that the macromolecular substrate (FIX) binds to the enzyme (FXIa) at an exosite remote from its active site (14,15). This suggests that FXIa is unique among coagulation enzymes by possessing an endogenous substratebinding site within the non-catalytic domain that serves a function similar to that of various cofactor molecules (e.g. FVIIIa, FVa, and TF) that potentiate substrate activation by the enzyme. Subsequently, numerous studies have demonstrated that macromolecular substrates in the coagulation cascade interact with various enzymes to determine the affinity and specificity of the Michaelis complex. These extended interactive sites are distinct from active site residues and have been appropriately termed "exosites" (26 -28). During FIX activation either by the FVIIa⅐TF complex or by FXIa, the zymogen is cleaved at Arg 145 -Ala 146 and Arg 180 -Val 181 bonds forming FIXa, a two-chain molecule joined by a disulfide bond, and releasing an activation peptide with M r ϳ11 kDa (5-7, 29, 30). This study was undertaken to examine the interaction of FIX with exosites in FXIa in the generation of FIXa.
It has been documented (14 -16, 18 -20) that, compared with FXIa, the isolated light chain is a poor activator of FIX, although hydrolysis rates of the small peptidyl substrate S-2366 by the two enzymes are similar. The weak enzymatic activity of the isolated light chain against FIX is not due to allosteric modification of FIX upon binding to the FXIa-HC because our data ( Fig. 1) clearly demonstrate that preincubation of FIX with saturating concentrations of FXIa-HC did not potentiate the cleavage rate of FIX by FXIa-LC.
Does the interaction of FIX with the isolated FXIa-HC alone determine the affinity of the enzyme-substrate interaction? If this were the case, then the rate of activation of FIX by FXIa should be inhibited in the presence of FXIa-HC as a result of substrate depletion because of complex formation of FIX with the isolated FXIa-HC. However, although the rate of FIX cleavage by FXIa was inhibited in the presence of FXIa/S557A (Fig. 2,  A and C), it was unaltered in the presence of FXIa-HC (Fig. 2, B  and D).
Although kinetic studies examining the activation of FIX by FXIa strongly suggest the presence in the FXIa-HC of an enzyme-substrate interactive site remote from the active site, the kinetic data do not provide information on the nature and architecture of the enzyme-substrate interactive site(s). Therefore, we examined the direct binding to FIX in an ELISA assay of FXIai, compared with the isolated heavy chain or light chain. Although FXIai bound to FIX with high affinity (K D ϳ0.25-0.5 nM), no binding could be demonstrated for FXIa-HC or FXIa-LCi (Fig. 3A). When FIX binding was examined at high concentrations (15 nM-1 M, i.e. 30 -2000-fold in excess of the K d ), only FXIa-HC and FXI demonstrated very low level binding (Fig. 3B).
To examine whether the isolation procedure contributes to the loss of the FIX binding properties of the isolated heavy and light chains of FXIa, ELISA assays were carried out to examine the binding to FIX of FXIa/S557A and of the triple mutant, FXIa/S557A/C362A/C482A, without isolation of the heavy or the light chain. The results (Fig. 4C) demonstrate that although FXIa/S557A binds to FIX with high affinity (K i ϳ0.25-0.5 nM), the activated triple mutant fails to bind FIX. The impaired binding of the triple mutant is not an artifact resulting from the solid phase assay because FIX, after preincubation in solution with varying concentrations of FXIa/S557A, inhibited FIX binding dramatically, whereas FXIa/S557A/C362S/C482S demonstrated only minor (Ͻ 15%) inhibition at very high concentrations (1 M) (Fig. 4D). Therefore, once the light chain and the heavy chain of FXIa are separated from one another by the absence of the covalent bond between the heavy and light chains, the recognition site for FIX is lost.
Mild reduction and alkylation of FXIa/S557A produced a single chain subunit as well as free heavy and light chains (supplemental Fig. 2B, lane 3). Each subunit of FXIa interacts with FIX as an independent enzyme, accounting for the fact that binding to FIX of the reduced and alkylated FXIa/S557A is diminished but not abolished (supplemental Fig. 2C). In contrast, the reduced and oxidized sample (supplemental Fig. 2B, lane 2) demonstrated binding to FIX that was consistently higher than that of the reduced and alkylated sample (supplemental Fig. 2C). Thus, although blocking disulfide bond formation results in loss of FIX binding, its reformation restores the interaction between FIX and FXIa.
Why do the heavy and the light chains of FXIa need to be disulfide-linked for interaction with FIX? FXIa, unlike other coagulation enzymes, is a homodimer, and it has been suggested that the relative paucity of accumulation of FIX␣ (FIX cleaved only at Arg 145 ) during FXIa-catalyzed FIX activation might be explained by the simultaneous cleavage of the two scissile bonds of a single FIX molecule by the two active sites of dimeric FXIa (31,32). This hypothesis is rendered extremely unlikely because FXIa/G326C (22) and several FXIa/C321A mutants (33) that exist predominantly as monomers can cleave FIX, generating FIXa at rates identical to those catalyzed by dimeric wtFXIa. Recently, FXIa with only one subunit in the activated form has been shown to activate FIX as efficiently as FXIa with two active sites (34). Therefore, it is reasonable to conclude that each subunit of FXIa acts as an independent enzyme and that (a) FIX interactive site(s) is/are present in each subunit of FXIa, provided the heavy chain and light chain are joined by a disulfide bond.
Thus, the following facts have been established concerning the activation of FIX by FXIa. 1) Each subunit of FXIa acts as an independent enzyme, cleaving FIX at Arg 145 -Ala 146 and Arg 180 -Val 181 . 2) FXIa and not FXI binds to FIX and, therefore, the FIX recognition site is formed only when the Arg 369 -Ile 370 bond of each subunit is cleaved. 3) (A) FIX binding exosite(s) is/are present both on the heavy and the light chain of each subunit of FXIa. 4) The FIX interactive site within FXIa is formed only when the heavy and the light chains are disulfide-linked.
The most logical hypothesis that is consistent with the above facts is that the covalent attachment of the heavy and light chains of FXIa is required to bring the two exosites, one in the heavy chain and the other in the light chain, into close proximity to form a single substrate binding exosite for efficient FIX activation. Detailed information on the structure of the enzyme, FXIa, would be required to confirm or refute this hypothesis. The crystal structures of FXI (35) and catalytic domain of FXIa (36,37) have been published, but no such data are available for the full-length enzyme. However, small angle x-ray scattering and ultrastructural (electron microscopic) data on FXIa suggested that FXI activation is accompanied by a major conformational change from an elongated to a more compact arrangement of the domains (32). These data and our present studies suggest that upon activation of FXI to FXIa, a cryptic or a newly formed FIX binding exosite, possibly consisting of residues from both the heavy and light chains, is made available for interaction with the substrate. Alternatively, such a substrate-binding exosite could consist of residues involving or in close proximity to the disulfide bond (Cys 362 -Cys 482 ) connecting the heavy and light chains of FXIa. In conclusion, the endogenous cofactor function of the heavy chain requires its covalent linkage to the active-site containing FXIa light chain.