Domain V of β2-Glycoprotein I Binds Factor XI/XIa and Is Cleaved at Lys317-Thr318*

The fifth domain (DV) of β2-glycoprotein I (β2GPI) is important for binding a number of ligands including phospholipids and factor XI (FXI). β2GPI is proteolytically cleaved in DV by plasmin but not by thrombin, VIIa, tissue plasminogen activator, or uPA. Following proteolytic cleavage of DV by plasmin, β2GPI retains binding to FXI but not to phospholipids. Native β2GPI, but not cleaved β2GPI, inhibits activation of FXI by thrombin and factor XIIa, attenuating a positive feedback mechanism for additional thrombin generation. In this report, we have defined the FXI/FXIa binding site on β2GPI using site-directed mutagenesis. We show that the positively charged residues Lys284, Lys286, and Lys287 in DV are essential for the interaction of β2GPI with FXI/FXIa. We also demonstrate that FXIa proteolytically cleaves β2GPI at Lys317-Thr318 in DV. Thus, FXIa cleavage of β2GPI in vivo during thrombus formation may accelerate FXI activation by decreasing the inhibitory effect of β2GPI.

␤ 2 -Glycoprotein I (␤ 2 GPI) 1 (also known as apolipoprotein H, apoH) is a constituent of human plasma that circulates in free and bound forms associated with lipoproteins. ␤ 2 GPI is a single chain glycoprotein containing 326 amino acids that comprise five complement control protein module-type repeats, also known as short consensus repeats or Sushi domains. Short consensus repeats are characterized by disulfide bridges joining the 1 st -3 rd and 2 nd -4 th cysteine residues (1)(2)(3)(4). The first four domains of ␤ 2 GPI structurally resemble each other, whereas DV has three internal disulfide bonds and an extra C-terminal loop encompassing residues Cys 306 -Cys 326 . The Cterminal loop is surface-exposed and susceptible to proteolytic cleavage (5,6). The region defined by Cys 281 -Cys 288 is critical for phospholipid and heparin binding and is highly conserved (5,(7)(8)(9)(10)(11)(12). We have previously reported that ␤ 2 GPI is proteolytically clipped between Lys 317 and Thr 318 in DV, abolishing its binding to anionic phospholipids but not heparin (5,11). Cleavage at Lys 317 -Thr 318 is generated in vitro by plasmin (13). Heparin greatly enhances the plasmin-mediated cleavage of the Lys 317 -Thr 318 site in ␤ 2 GPI at concentrations that can be achieved in vivo during anticoagulation therapy (11). The cleavage at Lys 317 -Thr 318 is generated in vitro by plasmin and at extremely low efficiency by factor Xa (13) and in vivo in pathological states of increased fibrinolysis (13)(14)(15).
␤ 2 GPI is the principle antigenic target for antiphospholipid antibodies in patients with the antiphospholipid syndrome (APS) (16,17). APS is a condition associated with recurrent arterial or venous thrombosis, complications of pregnancy including fetal loss and preeclampsia, and the presence of antiphospholipid antibodies (8,18,19). The binding of autoantibodies to ␤ 2 GPI is now generally accepted as an important feature of APS, and a number of studies have shown that there is a significant correlation between thrombotic manifestations and the presence of anti-␤ 2 GPI antibodies (8,20). The pathophysiological mechanisms by which autoantibodies exert their adverse effects remain unknown. However, ␤ 2 GPI autoantibodies have been implicated in the aberrant activation of endothelial cells (21)(22)(23) by binding to ␤ 2 GPI complexed with annexin II (24) or members of the toll-like receptor family (25). Additionally, autoantibody-bound ␤ 2 GPI may activate platelets via interaction with the apolipoprotein E receptor 2Ј (26). We have recently reported that ␤ 2 GPI binds factor XI (FXI) and inhibits its activation to factor XIa (FXIa) by thrombin and factor XIIa (27). ␤ 2 GPI cleaved by plasmin at Lys 317 -Thr 318 bound FXI as well as intact ␤ 2 GPI but did not inhibit its activation by thrombin (27). This may be an important mechanism for regulation of the amplification pathway of the coagulation cascade, and autoantibodies in patients with APS may interfere with this inhibition, leading to thrombosis.
FXI is a unique zymogen that circulates in plasma as a disulfide-bound homodimer (28). FXIa catalyzes the activation of factor IX to its active form factor IXa (29). FXI/FXIa binds to activated human platelets in the presence of high molecular weight kininogen (HK) and zinc or prothrombin and calcium ions in a specific and reversible manner (30 -32). Activated platelets promote FXI activation by thrombin in the presence of HK or prothrombin, thereby initiating the intrinsic coagulation pathway independent of contact proteins and amplifying the generation of thrombin. ␤ 2 GPI proteolytically clipped at Lys 317 -Thr 318 has been thought to be physiologically inactive as it loses its phospholipid binding properties. However, Koike and co-workers (33) have recently demonstrated that ␤ 2 GPI proteolytically cleaved by plasmin further suppresses plasmin generation by binding to Glu-plasminogen via DV. This interaction suggests a negative feedback loop controlling extrinsic fibrinolysis (33). Increased levels of clipped ␤ 2 GPI (c␤ 2 GPI) have been reported in the plasma of patients with lupus anticoagulant and in patients with disseminated intravascular coagulation (13). Moreover, increased fibrinolysis and thrombin generation occurs in patients with disseminated intravascular coagulation and antiphospholipid antibodies (13)(14)(15).
In this study, we identified specific residues on ␤ 2 GPI involved in binding FXI/FXIa and report that DV of ␤ 2 GPI is the major FXI/FXIa binding site. Moreover, FXIa cleaved ␤ 2 GPI at Lys 317 -Thr 318 in DV. This cleavage of ␤ 2 GPI may represent a unique mechanism in the control of FXI activation and fibrinolysis.
Recombinant Human ␤ 2 GPI Preparations-Eight ␤ 2 GPI mutants were utilized in this study to define the binding site on ␤ 2 GPI for FXI. Recombinant full-length human ␤ 2 GPI (rh␤ 2 GPI) and domain deletion mutants of human ␤ 2 GPI were generated as described previously (34 -36). The cDNAs encoding human ␤ 2 GPI were inserted into the baculovirus vector pBacPak6, and the nucleotide sequences of the cDNAs were confirmed using standard sequencing technology. Sf9 cells grown in monolayers were infected with the resultant constructs and were cultured 3-5 days at 25°C in serum-free medium. Affinity chromatography with a nickel column or a rabbit polyclonal anti-␤ 2 GPI antibody was used to purify the resulting recombinant proteins in the conditioned medium. The purified proteins were analyzed by SDS-PAGE/ immunoblot and subjected to automated Edman sequencing using Applied Biosystems (Foster City, CA). An oligonucleotide-directed in vitro mutagenesis kit was used to convert Lys 286 and Lys 287 to Glu (mutant 2K) and Lys 284 , Lys 286 , and Lys 287 to Glu (mutant 3K) in DV of ␤ 2 GPI as previously described (12). The nucleotide sequence of each mutated cDNA was confirmed by standard sequencing technology. Each cDNA was subcloned into the pBacPak6 vector, and the recombinant proteins were expressed and purified as outlined above.
Radiolabeling of Factor XI with 125 I-FXI was radiolabeled with 125 I using the Iodogene method (IODO-BEAD, Pierce, Rockford, IL) as described previously (27).
Protein Determination-Concentrations of 125 I-FXI were determined by Micro BCA protein assay kit (Pierce) according to the instructions provided by the manufacturer. Specific radioactivity was determined as described by Baird and Walsh (37). The specific activity of 125 I-FXI was 2.12 ϫ 10 18 cpm/mol. SDS-PAGE analysis of the iodinated proteins revealed only one radioactive band, indicating that there was no contamination of FXI with FXIa.
Binding of 125 I-FXI to Mutants of ␤ 2 GPI-The binding of 125 I-FXI to immobilized rh␤ 2 GPI was performed using Lockwell TM microtiter plates (Nunc, Roskilde, Denmark) as described (27,38). Microtiter wells were coated with 100 l of n␤ 2 GPI, rh␤ 2 GPI, DI, DI-II, DI-III, DI-IV, DII-V, DV, mutant 2K, or mutant 3K mutant of ␤ 2 GPI or human serum albumin (3.12-200 nM) by incubation overnight at 4°C. The plate was washed five times with PBS, 0.1% Tween 20 (PBST) using an automated microplate washer (Beckman Coulter Inc., Fullerton, CA). The wells were blocked with 2% BSA/PBST for 2 h at 25°C. The plate was washed five times with PBST and then five times with PBS. 100 l of 125 I-FXI (0.28 nM) in 0.5% BSA/PBS was added to individual wells and incubated for 4 -5 h at 25°C. The wells were then washed five times with 0.5% BSA/PBS, air-dried, and counted in a ␥-counter. The number of counts/min were measured and converted to femtomoles of FXI bound.
Competitive Inhibition Assays-The effect of native recombinant fulllength and domain deletion mutants of ␤ 2 GPI on the binding of 125 I-FXI to rh␤ 2 GPI was studied using Lockwell microtiter plates, the wells of which were coated with 100 l of rh␤ 2 GPI (50 nM) by incubation overnight at 4°C. The plate was treated as described above before 50 l of 125 I-FXI (0.56 nM), and 50 l of various ␤ 2 GPI preparations or BSA (1.3 nM-4 M) in 0.5% BSA/PBS was added to the wells and incubated for 4 -5 h at 25°C. The wells were washed five times with 0.5% BSA/PBS, air-dried, and counted in a ␥-counter. The number of counts/min bound were measured and converted to the percentage of total binding by dividing by the counts/min bound in the absence of competitors. The IC 50 was calculated by non-linear regression (one-site binding model, GraphPad Prism 3.03).
Saturation Binding of FXIa to ␤ 2 GPI-Saturation binding of FXIa to immobilized rh␤ 2 GPI or HK was performed using Lockwell microtiter plates, the wells of which were coated with 100 l of HK, rh␤ 2 GPI, or BSA (100 nM) by incubation overnight at 4°C. Plates were washed five times with PBST using a microplate washer. The wells were blocked with 2% BSA/PBST for 2 h at 25°C and then washed five times with PBST and five times with PBS. 100 l of various concentrations of FXIa (0.09 -50 nM) in PBS then was added to individual wells and incubated for 3 h at 37°C. The wells were washed five times with 0.5% BSA/PBS. 200 l of FXIa substrate (S2366, 600 M) was then added to individual wells and incubated for 1 h at 25°C, and then the optical density was measured at 405 nm using a microplate-scanning spectrophotometer (Power Wave TM , BIO-TEK Instruments Inc., Winooski, VT). The amount of FXIa bound was derived from a standard curve constructed with the known concentrations of FXIa. There was a linear correlation between the reciprocal of the A 405 value and the reciprocal of the amount of FXIa within 0.008 and 0.125 pmol (r 2 ϭ 0.9945). K d was calculated by non-linear regression (GraphPad Prism 3.03).
Binding of FXIa to Native, Recombinant, and Domain Deletion Mutants of ␤ 2 GPI-The binding of FXIa to various preparations of immobilized ␤ 2 GPI was performed using Lockwell microtiter plates. The wells were coated with 100 l of HK (single chain), n␤ 2 GPI, rh␤ 2 GPI, DI-IV, DII-V, DV of ␤ 2 GPI, or BSA (6.25-400 nM) by incubation overnight at 4°C. The plate was treated as described above and 100 l of FXIa (10 nM) in 0.5% BSA/PBS was then added to individual wells and incubated for 3 h at 37°C. The wells were washed five times with 0.5% BSA/PBS. FXIa substrate (S2366, 600 M) was then added to individual wells, and the amount of FXIa bound was derived as described above.
Time Course of FXIa Cleavage of ␤ 2 GPI-To investigate the time course of ␤ 2 GPI cleavage by FXIa, n␤ 2 GPI (30 M) was incubated with FXIa (0.1 M) in Tris-buffered saline buffer (pH 7.6) at 37°C. At defined time points, 130 l of aliquots were diluted 20-fold in heparin affinity column binding buffer (0.01 M Tris-HCl, 0.03 M NaCl (pH 8.0)) and then applied to heparin affinity columns (1 ml). cn␤ 2 GPI and intact n␤ 2 GPI were separated with a linear gradient (0.10 -0.35 M NaCl, 0.01 M Tris-HCl (pH 8.0)) as described previously (11). The eluted peak area of cn␤ 2 GPI and intact n␤ 2 GPI was calculated using Primeview Evaluation software (Amersham Biosciences), and the amount of cn␤ 2 GPI and intact n␤ 2 GPI was expressed as a percentage of total n␤ 2 GPI binding to the heparin affinity column. The purification of all of the proteins was completed at 4°C using the ⌬KTA prime purification system (Amersham Biosciences).
N-terminal Sequencing of n␤ 2 GPI Incubated with and without FXIa-n␤ 2 GPI was incubated with and without FXIa in Tris-buffered saline buffer (pH 7.6) at 37°C for 0, 6, and 24 h. Aliquots of each reaction were subjected to automated Edman sequencing carried out by the Australian proteome analysis facility (Sydney, New South Wales, Australia) using an Applied Biosystems 494 Procise protein sequencing system (Foster City, CA).
Statistical Analysis-Data are expressed as the mean Ϯ S.E. Differences between the groups were evaluated using Student's t test.

FXI/FXIa Bind ␤ 2 GPI via Lysine Residues in DV-
To investigate the FXI binding site on ␤ 2 GPI, we measured 125 I-FXI binding to ␤ 2 GPI and various mutants coated on microplate wells. 125 I-FXI bound to n␤ 2 GPI, rh␤ 2 GPI (27), domain deletion mutants DV, and DII-V but not to DI, DI-II, DI-III, or DI-IV (Fig. 1). These findings indicate that 125 I-FXI binds to the C-terminal domain of ␤ 2 GPI. To further characterize the residues in DV that are critical for binding FXI, we tested ␤ 2 GPIcontaining point mutations in DV. The binding of FXI was negligible with Lys 286 and Lys 287 mutated to Glu (Fig. 1, 2K  Mutant) and with Lys 284 , Lys 286 , and Lys 287 mutated to Glu (Fig. 1, 3K Mutant). Thus the positively charged residues in DV of ␤ 2 GPI are critical for the interaction of FXI with ␤ 2 GPI.
To confirm the interaction of DV with FXI, we used the panel of ␤ 2 GPI mutants in fluid phase to inhibit binding to immobilized rh␤ 2 GPI. The binding of 125 I-FXI to rh␤ 2 GPI was competitively inhibited in a dose-dependent manner by n␤ 2 GPI, rh␤ 2 GPI, DII-V, or DV with IC 50 values of 0.115, 0.167, 0.075, and 0.079 M, respectively (Fig. 2). There was negligible inhibition with domain deletion mutants DI, DI-II, DI-III, and DI-IV at final concentrations of up to 2 M (Fig. 2). Thus, the results confirm that DV contains the major FXI binding site on ␤ 2 GPI. To further define DV residues critical for binding, mutants 2K and 3K were tested in similar inhibition experiments. Neither mutant inhibited the binding of 125 I-FXI to rh␤ 2 GPI at concentrations of up to 2 M (Fig. 2).
Interaction of FXIa with ␤ 2 GPI was assessed in direct binding experiments using FXIa enzymatic activity on the chromogenic substrate S2366 as described under "Experimental Procedures." The K d of FXIa binding to ␤ 2 GPI (3.86 Ϯ 0.91 nM) was similar to HK (3.65 Ϯ 0.20 nM) (Fig. 3). The B max of FXIa binding to HK (0.14 Ϯ 0.0042 pmol) was 3.18-fold greater than the B max of FXIa binding to ␤ 2 GPI (0.045 Ϯ 0.005 pmol). The K d of FXIa binding to ␤ 2 GPI was much lower than that (K d ϭ 15.43 Ϯ 1.00 nM) of FXI binding to ␤ 2 GPI (27), indicating a higher affinity interaction with FXIa. Direct binding experiments demonstrated that FXIa bound to n␤ 2 GPI, rh␤ 2 GPI, DII-V, and DV but not to DI-IV or BSA (Fig. 4), confirming that FXIa also binds to DV of ␤ 2 GPI. Previous studies have shown that the peptide loop defined by Cys 281 -Cys 288 is critical for binding to phospholipids and heparin (9,11,27). Taken together, these results support the hypothesis that the peptide loop defined by Cys 281 -Cys 288 and containing Lys 284 , Lys 286 , and Lys 287 may mediate multiple binding functions in vivo.
FXIa Enzymatic Activity Is Not Influenced by ␤ 2 GPI-As previously reported (27), ␤ 2 GPI did not influence the enzymatic activity of FXIa in the amidolytic assay with its chromogenic substrate (data not shown).
FXIa cleaves ␤ 2 GPI at Lys 317 -Thr 318 -N␤ 2 GPI was incubated with FXIa at a molar ratio of 300:1 for up to 24 h. When applied to a heparin affinity column, the ␤ 2 GPI consisted of two elution peaks (Fig. 5A). This chromatogram has a profile similar to that seen when n␤ 2 GPI is subjected to plasmin cleavage (11,13,39). The amount of cn␤ 2 GPI generated from n␤ 2 GPI with FXIa increased over time, and greater than 50% n␤ 2 GPI had been cleaved by FXIa at 5 h of incubation (Fig. 5B). The amount of cn␤ 2 GPI reached a plateau (80% cleavage of n␤ 2 GPI) at 24 h of incubation with FXIa (Fig. 5B).
N-terminal sequencing showed that there were two N termini in the n␤ 2 GPI preparations incubated with FXIa for 6 and 24 h (Table I). The affinity-purified cn␤ 2 GPI preparation also had two N termini (Table I). The n␤ 2 GPI preparations incubated without FXIa for 0 and 24 h at 37°C and with FXIa at 0 time point revealed a single sequence with one N terminus ( Table I). The results indicate that FXIa cleaved n␤ 2 GPI at Binding of 125 I-FXI to rh␤ 2 GPI was inhibited in a concentration-dependent manner by n␤ 2 GPI, rh␤ 2 GPI, and domain deletion mutants containing DV. The binding of 125 I-FXI to rh␤ 2 GPI was not inhibited by deletion mutants lacking DV, 2K mutant, 3K mutant, or BSA. 125 I-FXI was incubated with serial dilutions of the indicated proteins at 25°C for 4 -5 h in microtiter plates coated with rh␤ 2 GPI to determine capacity of the proteins to inhibit binding. The amount of 125 I-FXI binding to immobilized ␤ 2 GPI was expressed as a percentage of that bound with no inhibitor. The data shown are an average of results from three separate experiments.
FIG. 3. Saturation binding of FXIa to ␤ 2 GPI. The K d of FXIa binding to ␤ 2 GPI was similar to HK, whereas the B max of FXIa binding to HK is ϳ3-fold greater than FXIa binding to ␤ 2 GPI. FXIa was incubated in microtiter plate coated with that of rh␤ 2 GPI, HK, or BSA. Open squares, HK; solid squares, rh␤ 2 GPI.
Lys 317 -Thr 318 (Table I), which is the same site on ␤ 2 GPI cleaved by plasmin (11-13, 27, 39) DISCUSSION We recently reported that ␤ 2 GPI binds FXI and inhibits its activation by thrombin and factor XIIa in the presence of dextran sulfate at concentrations lower than those normally found in human plasma (27). In this study, we established that the interaction of ␤ 2 GPI with FXI/FXIa occurs in DV of ␤ 2 GPI. There was no binding of FXI/FXIa to domain deletion mutants of ␤ 2 GPI that lacked DV. Furthermore, the peptide region spanning residues Cys 281 -Cys 288 in DV is the FXI binding site. Positive-charged residues in this region at Lys 284 , Lys 286 , and Lys 287 were found to be critical in the interaction of ␤ 2 GPI with FXI. The same lysine residues mediate binding to phospholipids and heparin (10,11). In addition, FXIa cleaves ␤ 2 GPI at Lys 317 -Thr 318 within DV. The cleavage site on ␤ 2 GPI for FXIa is the same as that for plasmin (13)(14)(15). Functional studies demonstrate the significance of maintaining the integrity of the C-terminal loop in DV, because cleavage of DV at Lys 317 -Thr 318 retained the binding of ␤ 2 GPI with FXI but attenuated or abolished the ability of ␤ 2 GPI to inhibit activation of FXI by thrombin (27).
␤ 2 GPI consists of five complement control protein domains, the first four being more typical of the complement control protein family (1). DV terminates in a disulfide bridge and has a lysine surface patch that is predicted to be very mobile between Leu 313 and Lys 317 , thus being poorly resolved on crystallographic studies (6,40). The hydrophobic segment corresponding to residues Leu 313 -Trp 316 is critical in binding with anionic phospholipids but not with heparin (11,41).
The C-terminal loop and the peptide region Cys 281 -Cys 288 in DV have been predicted to be involved in the interaction with lipid membrane surfaces (8,40). This C-terminal loop is identical in human, bovine, and canine sequences and only differs by one residue in the rat and two residues in the mouse (8). The loop region is surface-exposed and is susceptible to cleavage at Lys 317 -Thr 318 by plasmin and reported to be cleaved extremely weakly (13) or not at all (15) by factor Xa. However, the proteases thrombin, tissue-type plasminogen activator, urokinase, and tissue factor/factor VIIa (13-15) are unable to cleave ␤ 2 GPI.
Cleavage at Lys 317 -Thr 318 produces two short polypeptide segments that remain linked to DV by disulfide bonds (5). This cleavage alters the spatial array of the three critical Lys residues and abolishes the ability of ␤ 2 GPI to bind anionic phospholipid surfaces (5). It has been reported that cleavage of ␤ 2 GPI at Lys 317 -Thr 318 disturbs the nearby electrostatic environment (42). It is believed that the integrity of the 317/318 peptide bond is important in tethering the cluster of positively charged and hydrophobic residues in this region. Although the hydrophobic C-terminal loop is important in phospholipid binding, we have recently reported that a preparation of ␤ 2 GPI cleaved at Lys 317 -Thr 318 by plasmin bound to FXI as well as rh␤ 2 GPI but did not inhibit thrombin activation of FXI (27).
Proteolytically cleaved ␤ 2 GPI at Lys 317 -Thr 318 is found in the plasma of patients with lupus anticoagulant, disseminated intravascular coagulation, ischemic stroke, and in healthy individuals with lacunar infarct(s) (14,15,17,33). It has been proposed that FXI plays an important role in the down-regulation of fibrinolysis (Fig. 6) (43). This is thought to occur as a FIG. 4. Direct binding of FXIa to various preparations of ␤ 2 GPI. FXIa bound to n␤ 2 GPI, rh␤ 2 GPI, DII-V, and DV but not DI-IV or BSA, confirming that FXIa also binds to DV of ␤ 2 GPI. The amount of FXIa binding to immobilized proteins was determined after incubation in microtiter plate coated with serial dilutions of the indicated proteins.
FIG. 5. Time course of FXIa cleavage of ␤ 2 GPI. n␤ 2 GPI was cleaved by incubation with FXIa in Tris-buffered saline buffer at 37°C. The area of clipped n␤ 2 GPI and intact n␤ 2 GPI was calculated using Primeview Evaluation software, and the amount of clipped n␤ 2 GPI was expressed as a percentage of total n␤ 2 GPI binding to heparin column. A, heparin affinity column chromatogram of ␤ 2 GPI purification following treatment with FXIa. B, time course of FXIa cleavage of n␤ 2 GPI.
consequence of a FXI-dependent burst in thrombin generation, which results in the activation of thrombin-activatable fibrinolysis inhibitor, which inhibits the activation of plasminogen (43)(44)(45). The conversion of plasminogen to plasmin by twochain tissue plasminogen activator is a key event in the fibrinolytic system, because plasmin is a critical protease in this system. Regulation of plasmin generation is important in in vivo fibrinolytic homeostasis. Proteolytically cleaved ␤ 2 GPI at Lys 317 -Thr 318 has recently been demonstrated to inhibit plasmin generation by tissue plasminogen activator (33). The clipped form of ␤ 2 GPI correlated with in vitro markers of fibrinolytic activity. The C-terminal hydrophobic loop of c␤ 2 GPI has been confirmed by heteronuclear magnetic resonance to be tightly fixed by electrostatic interaction with the lysine cluster at the phospholipid binding site while at the same time enhancing stability and neutralizing the positive charge in this region (42). Thus, the C-terminal loop of clipped ␤ 2 GPI is more mobile than that of the intact molecule, possibly allowing the interaction of ␤ 2 GPI with FXI/thrombin complexes, such that it binds FXI but does not inhibit its activation by thrombin (27).
In vivo, it would appear that activated platelets provide an appropriate procoagulant surface for assembly of surface bound protease substrate complexes. FXI/FXIa bind activated platelets in a saturable and reversible manner with K d ϭ 10 and ϳ 0.8 nM, respectively (46,47). HK and zinc ions or prothrombin and calcium ions have been shown to promote this binding (30 -32). ␤ 2 GPI binds FXI and could possibly substitute for HK or prothrombin in vivo in the interaction of FXI with activated platelets. Thus, the clipping of ␤ 2 GPI by FXIa bound in a complex with FXIa on activated platelets could abolish its phospholipid binding but not its FXI/FXIa binding, a negative feedback that counteracts its inhibition of FXI activation by thrombin, because in the presence of activated platelets, thrombin is the preferred activator of FXI (48).
The interaction of ␤ 2 GPI with FXI/FXIa on the surface of activated platelets would facilitate the inhibition of FXI activation and the subsequent generation of c␤ 2 GPI by FXIa. The FXIa cleavage of ␤ 2 GPI in this complex would regulate the activation of FXI by thrombin, providing a negative feedback loop in FXI activation. Furthermore, c␤ 2 GPI binds Glu-plasminogen-suppressing plasmin generation, thus providing a negative feedback loop controlling fibrinolysis (33). A schematic representation of the effect of c␤ 2 GPI and intact ␤ 2 GPI on this pathway is shown in Fig. 6.
Plasmin and FXIa cleave ␤ 2 GPI, but in vivo, it is not clear which protease is responsible for the elevated levels of c␤ 2 GPI demonstrated in certain patient populations. Even though 50% of the plasma ␤ 2 GPI was cleaved at Lys 317 -Thr 318 in some of the patients with lupus anticoagulant, there was a minimal increase in plasmin inhibitor complex and D-dimer levels, indicating that plasmin is not the major protease in vivo responsible for this cleavage of ␤ 2 GPI. Because FXIa cleaves ␤ 2 GPI in vitro as efficiently as plasmin and factor Xa is the only other plasma protease reported to cleave ␤ 2 GPI with very poor efficiency by one group (13) but not confirmed by Horbach et al. (15), it is likely that FXIa is the major in vivo protease responsible for this cleavage in patients with lupus anticoagulant.
In conclusion, we first have demonstrated that ␤ 2 GPI binds FXI/FXIa by its C-terminal domain and that lysine residues in the region Cys 281 -Cys 288 are critical in this interaction and in the inhibition of FXI activation by thrombin (27). Second, we have demonstrated that FXIa cleaves ␤ 2 GPI at Lys 317 -Thr 318 at its C terminus, which abolishes its inhibition of FXI activation by thrombin (27). We propose that ␤ 2 GPI in vivo is proteolytically cleaved by FXIa and that c␤ 2 GPI represents a unique mechanism in the control of FXI activation and fibrinolysis.