A Novel Dimeric Inhibitor Targeting Beta2GPI in Beta2GPI/Antibody Complexes Implicated in Antiphospholipid Syndrome

Background β2GPI is a major antigen for autoantibodies associated with antiphospholipid syndrome (APS), an autoimmune disease characterized by thrombosis and recurrent pregnancy loss. Only the dimeric form of β2GPI generated by anti-β2GPI antibodies is pathologically important, in contrast to monomeric β2GPI which is abundant in plasma. Principal Findings We created a dimeric inhibitor, A1-A1, to selectively target β2GPI in β2GPI/antibody complexes. To make this inhibitor, we isolated the first ligand-binding module from ApoER2 (A1) and connected two A1 modules with a flexible linker. A1-A1 interferes with two pathologically important interactions in APS, the binding of β2GPI/antibody complexes with anionic phospholipids and ApoER2. We compared the efficiency of A1-A1 to monomeric A1 for inhibition of the binding of β2GPI/antibody complexes to anionic phospholipids. We tested the inhibition of β2GPI present in human serum, β2GPI purified from human plasma and the individual domain V of β2GPI. We demonstrated that when β2GPI/antibody complexes are formed, A1-A1 is much more effective than A1 in inhibition of the binding of β2GPI to cardiolipin, regardless of the source of β2GPI. Similarly, A1-A1 strongly inhibits the binding of dimerized domain V of β2GPI to cardiolipin compared to the monomeric A1 inhibitor. In the absence of anti-β2GPI antibodies, both A1-A1 and A1 only weakly inhibit the binding of pathologically inactive monomeric β2GPI to cardiolipin. Conclusions Our results suggest that the approach of using a dimeric inhibitor to block β2GPI in the pathological multivalent β2GPI/antibody complexes holds significant promise. The novel inhibitor A1-A1 may be a starting point in the development of an effective therapeutic for antiphospholipid syndrome.


Introduction
Beta2-glycoprotein I (b2GPI) is the major target for autoimmune antibodies associated with antiphospholipid syndrome (APS), an autoimmune disease characterized clinically by thrombosis and recurrent pregnancy loss [1,2,3,4]. Presently, APS patients with thrombotic complications who have high titers of antibodies are treated chronically with anticoagulants [5,6,7]. However, even continuous anticoagulation may not prevent recurrent thrombosis [5], emphasizing the need for a more effective antithrombotic therapy based on the thrombogenic mechanisms specific to APS.
b2GPI consists of five domains [8,9]. Flexible linkers between domains permit b2GPI to adopt different overall shapes such as a fishhook-like shape seen in the crystal structure [8,9], an S-shape observed by small angle x-ray scattering for b2GPI in solution [10] and a circular shape detected by electron microscopy [11]. The circular shape in which domain I is adjacent to domain V is the predominant conformation of b2GPI in normal human plasma [11]. Circular b2GPI can be converted to an extended form by altering pH and salt concentrations, binding to a high-affinity antibody directed to domain I or by the binding to cardiolipin [11]. b2GPI, which is abundant in plasma (about 170 mg/ml or 4 mM) [12], acquires its prothrombotic properties only in the presence of anti-b2GPI antibodies. Antibodies of the IgG isotype have the highest correlation with the clinical manifestations of APS compared to other identified antibodies [13,14]. Although anti-b2GPI antibodies in APS patients are highly heterogeneous in respect to their affinity for b2GPI and the location of their binding epitopes, autoantibodies against domain I are the most common and better correlate with thrombosis [15,16]. The presence of anti-b2GPI antibodies causes cellular activation both in vitro and in vivo [17,18,19]. Toll-like receptors, annexin A2, ApoE receptor (ApoER2), platelet receptor GPIb and anionic phospholipids exposed on cellular surfaces of activated cells are suggested to be pathologically important in APS [18,19,20,21,22,23,24,25,-26,27,28,29]. The binding sites for anionic phospholipids [30,31,32,33], lipoprotein receptors (including ApoER2) [34] and GPIb [21] are in domain V of b2GPI (b2GPI-DV).
In the present studies, we are suggesting a novel approach to interference with anti-b2GPI-dependent thrombosis in APS. To prevent the b2GPI/antibody complexes from the binding to receptors, we designed an inhibitor that a) targets b2GPI and b) binds tightly to b2GPI/antibody complexes expressing the dimeric b2GPI but binds weakly to b2GPI monomers. These requirements have the following rationale: First, complete b2GPI deficiency in humans, although rare, does not lead to apparent health problems [35,36,37], therefore the inhibitor that targets b2GPI will not disrupt normal biological processes. Second, b2GPI/anti-b2GPI antibody complexes expressing dimeric b2GPI but not monomeric b2GPI are pathologically important [28,38], therefore the inhibitor should bind preferentially to b2GPI/anti-b2GPI complex compared to b2GPI monomers. Anti-b2GPI antibodies constitute less than 3% of total IgG in patients with antiphospholipid syndrome and have weak affinity for b2GPI [39,40,41]. In contrast to b2GPI monomers which are abundant in plasma, b2GPI/anti-b2GPI complexes are present at low concentration.
In this study, we are focusing on the inhibition of the binding of b2GPI/anti-b2GPI antibody complexes to ApoER2 and to anionic phospholipids. ApoER2, like other members of the family of lipoprotein receptors, binds b2GPI via structurally homologous ligand-binding type A (LA) modules and the first LA module of ApoER2 is the most important for the binding [42,43,44,45]. Recently, we have shown that different LA modules bind to the same site on b2GPI and that b2GPI can not simultaneously bind an LA module and a cardiolipin-coated surface [46]. Therefore, an LA module bound to b2GPI has dual action: it inhibits both the binding of b2GPI to lipoprotein receptors and anionic phospholipids expressed on cells.
We created a dimeric inhibitor, A1-A1. To make this inhibitor, we isolated the first LA module from ApoER2 (A1) and connected two A1 modules with a flexible linker. In the present studies, we compared a monomeric A1 with A1-A1 on interfering with the binding of b2GPI/anti-b2GPI antibody complexes to anionic phospholipids. We tested the inhibition of b2GPI present in human serum, b2GPI purified from human plasma and domain V of b2GPI. b2GPI in serum is the circular form of b2GPI [11] and the individual domain V represents b2GPI in the extended conformation. We demonstrated that when b2GPI/antibody complexes are formed, A1-A1 is more effective than A1 in inhibition of b2GPI binding to cardiolipin, regardless of the source of b2GPI. Similarly, A1-A1 strongly inhibits the binding of dimerized domain V of b2GPI to cardiolipin compared to the monomeric A1 inhibitor. Moreover, A1-A1 preferentially binds b2GPI/anti-b2GPI antibody complexes and binds only weakly to monomeric b2GPI. The novel inhibitor A1-A1 may be a starting point in the development of an effective drug for prevention and treatment of b2GPI-dependent thrombosis in antiphospholipid syndrome.

Design of a dimeric inhibitor
Recently, we have shown that the first LA module from ApoER2 (A1) binds domain V of b2GPI with 1 mM affinity [46]. In order to target a multivalent b2GPI formed by anti-b2GPI antibodies, we made a dimeric inhibitor consisting of two A1 modules covalently connected by a linker. To allow largely unrestricted relative motion of two A1 modules in the A1-A1 molecule, we used a flexible linker Gly-Ser-Ser-Gly to connect A1 modules. In extended conformation, this four-residue linker is capable of separating the A1 modules in A1-A1 by up to 15 Å . We expressed and purified A1-A1 using the same procedure that we have previously used for the expression and purification of other LA modules including A1. Our previous analysis of different recombinantly expressed LA modules by solution NMR spectros-copy and crystallography demonstrated that the purified LA modules are properly folded, their structures are identical to the structures of these modules in full-length receptors and, in the presence of calcium, recombinant LA modules bind their ligands b2GPI, b2GPI-DV, RAP and ApoE [46,47,48,49]. Because calcium is essential for the function of LA modules and the formation of native disulfide bonds [50], we analyzed the folding of the dimeric molecule, A1-A1, in the presence and absence of calcium. We compared the oxidative refolding of A1-A1 to the refolding of A1, which yields a functional A1 module. The same quantities of the recombinant proteins were dialyzed in redox buffer containing either calcium or EDTA. After 36 hours (samples with A1) or 72 hours (samples with A1-A1) of refolding, the proteins were acidified with 0.1% TFA to stop the disulfide exchange and analyzed by a reversed-phase HPLC on an analytical C18 column. The A1 module contains six cysteine residues forming three disulfide bonds. In the presence of calcium, both A1 and A1-A1 converged to unique disulfide-bonded species out of many possibilities providing evidence that the presence of calcium guided formation of native disulfide bonds ( Figure 1A,B). For comparison, refolding of A1 and A1-A1 in the presence of EDTA yielded a distribution of multiple disulfide-bonded isomers.
Comparison of the dimeric inhibitor A1-A1 with monomeric A1. Inhibition of the binding of b2GPI to cardiolipin Anti-b2GPI antibodies create multivalent b2GPI/anti-b2GPI complexes that have pathological properties compared to pathologically inactive b2GPI monomers, which are normally present in plasma. The binding of b2GPI to anionic phospholipids in the presence of anti-b2GPI antibodies is one of the pathological mechanisms leading to thrombosis and pregnancy losses in antiphospholipid syndrome. We compared the dimeric inhibitor, A1-A1, to monomeric A1 on the inhibition of the binding of b2GPI/anti-b2GPI antibody complexes to cardiolipin coated on a plate. First, we used pooled normal human serum as a source of b2GPI. The majority of the b2GPI molecules in serum is in the circular form [11]. To select the appropriate concentration of b2GPI for the inhibition studies, we measured the binding curves ( Figure 2A). To create b2GPI/anti-b2GPI complexes, anti-b2GPI antibodies at constant concentration were added to b2GPI before the samples were applied to cardiolipin. The presence of anti-b2GPI antibodies significantly enhanced the binding of b2GPI to cardiolipin reaching the half-maximal binding at 0.02860.004% and 0.4060.05% of serum in the presence and in the absence of anti-b2GPI antibodies, respectively. We then compared the efficiency of A1-A1 and A1 for the inhibition of the binding of b2GPI in serum to cardiolipin in the presence of anti-b2GPI antibodies ( Figure 2B). 0.04% of human serum, which is in the linear region of the binding curve, was incubated on a cardiolipincoated surface in the presence of anti-b2GPI antibodies and the inhibitors. In the presence of anti-b2GPI antibodies, the dimeric molecule, A1-A1, inhibited the binding to cardiolipin of b2GPI in serum much stronger than monomeric A1. The half-maximal inhibition in the presence of anti-b2GPI antibodies was achieved at 1062 mM of A1-A1 and 218621 mM of A1. Also, we measured how A1-A1 and A1 inhibited the binding of b2GPI in serum in the absence of anti-b2GPI antibodies ( Figure 3). A 1% solution of human serum was titrated with A1-A1 or A1. b2GPI bound to cardiolipin was subsequently detected with anti-b2GPI antibodies. In the absence of anti-b2GPI antibodies, both A1-A1 and A1 were equally ineffective in inhibition of b2GPI. The concentration of the inhibitors at 50% inhibition of b2GPI was 189634 mM for A1-A1 and 176637 mM for A1. In sum, the efficiency of A1-A1 to inhibit the binding of b2GPI in serum to cardiolipin was significantly stronger in the presence of anti-b2GPI antibody than in the absence of antibodies. The inhibition efficiency of the monomeric A1 was practically the same and weak regardless of the presence or absence of anti-b2GPI antibodies. A1-A1 was more effective than A1 in inhibition of b2GPI in serum in the presence of anti-b2GPI antibodies.
Next, we analyzed how A1-A1 and A1 inhibit the binding of purified b2GPI to cardiolipin. Closed and extended conformations of b2GPI can be interconverted by altering of pH and concentrations of NaCl in the buffer [11], suggesting that the conformation of purified b2GPI may depend on the purification procedure. We used b2GPI purified from human plasma available from Haematologic Technologies, Inc. and analyzed the binding and inhibition of the binding of purified b2GPI by A1-A1 and A1 in the presence and in the absence of anti-b2GPI antibodies. The half-maximal binding was achieved at 2.460.4 nM and 4364 nM of the purified b2GPI in the presence and in the absence of anti-b2GPI antibodies, respectively ( Figure 4A). We incubated 10 nM of b2GPI with various concentrations of the dimeric, A1-A1, and monomeric, A1, inhibitors in the presence of anti-b2GPI antibodies. Similarly to what we observed for b2GPI in serum, A1-A1 was more effective in inhibition of the binding of b2GPI to cardiolipin in the presence of anti-b2GPI antibodies. The fit of the titration data to the one-site inhibition model resulted in 2663 mM of A1-A1 and 191634 mM of A1 at half-maximal inhibition of purified b2GPI in the presence of anti-b2GPI antibodies ( Figure 4B). In a separate experiment, we measured the inhibition of b2GPI in the presence of antibodies and compared the measured values to values predicted by the fit of the titration data ( Figure 4C). The measured values were close to those expected from the fit, additionally confirming that in the presence of anti-b2GPI antibodies a much lower concentration of A1-A1 was required to inhibit 50% of the binding of purified b2GPI to cardiolipin compared to A1. In the absence of anti-b2GPI antibodies, both A1-A1 and A1 only weakly inhibited the binding of the purified b2GPI to cardiolipin ( Figure 5). Similarly to what we observed for b2GPI in serum, the binding of purified b2GPI to cardiolipin in the presence of anti-b2GPI antibodies was inhibited more strongly by the dimeric inhibitor A1-A1 than by A1. Both, A1-A1 and A1, were ineffective in the inhibition of b2GPI in the absence of anti-b2GPI antibodies.
Crystal structure of the isolated domain V of b2GPI (b2GPI-DV) b2GPI binds anionic phospholipids and the A1 modules by its domain V [30,31,33,34,46]. In the crystal structures of a fulllength b2GPI in the extended conformation [8,9], domain V forms essentially no contacts with the adjacent domain. There are no glycosylation sites in b2GPI that could affect function of domain V indicating that the individual domain V (b2GPI-DV) dissected from the full-length b2GPI will function as domain V in the extended form of b2GPI. We solved the crystal structure of the isolated domain V to 1.9 Å resolution ( Table 1). As illustrated by Figure 6, the backbone conformation of b2GPI-DV is nearly identical to the structure of this domain in the full-length b2GPI. The largest difference between the structures is localized to a Cterminal loop. Experimental data strongly suggests that this loop is flexible in the native protein. For example, the residues from 311 to 317 comprising this loop are not defined in one of the crystal structures of the full-length b2GPI (PDB ID 1QUB) [8] and have large values of B-factors in the other (PDB ID 1C1Z) [9]. Also, the residues in the loop are either weak or missing from the NMR spectrum of b2GPI-DV in solution reflecting its internal flexibility [46]. The structural similarity between b2GPI-DV and domain V in the full-length b2GPI provides convincing evidence that the isolated recombinant b2GPI-DV mimics domain V in the extended form of b2GPI.
Comparison of the dimeric inhibitor A1-A1 with monomeric A1. Inhibition of the binding of the isolated domain V of b2GPI (b2GPI-DV) to cardiolipin in the presence of the dimerization antibodies To analyze how A1-A1 and A1 inhibit the binding of the extended form of b2GPI to cardiolipin, we used purified domain V of b2GPI (b2GPI-DV). We introduced a peptide tag at the Nterminus of b2GPI-DV and used an antibody directed to the tag to form dimeric b2GPI-DV/antibody complexes. We have previously demonstrated that the A1 module binds to the C-terminal part of b2GPI-DV [46] and, therefore, the N-terminal peptide tag on b2GPI-DV and the bound anti-tag antibody will not interfere with the binding of the A1 modules to b2GPI-DV.
As in the case of the full-length b2GPI, the presence of divalent b2GPI-DV/antibody complexes increased the attachment of b2GPI-DV to cardiolipin ( Figure 7A). The fit of the binding data to a one-site model resulted in 1961 nM of b2GPI-DV and 112621 nM of b2GPI-DV in the presence and in the absence of anti-tag antibodies. When 30 nM of b2GPI-DV in the presence of anti-tag antibody was incubated with the inhibitors, the halfmaximal inhibition was reached at 1262 mM of A1-A1 and 204633 mM of A1 ( Figure 7B). As we observed for the inhibition of the binding of b2GPI in human serum and purified b2GPI to cardiolipin-coated surfaces, the isolated domain V was inhibited much stronger by the dimeric inhibitor A1-A1 compared to monomeric A1 in the presence of dimeric b2GPI-DV/anti-tag antibody complexes.
Comparison of b2GPI in human serum with the isolated domain V of b2GPI (b2GPI-DV). Inhibition of the binding to cardiolipin by a monomeric A1 in the absence of antibodies We investigated if the binding of two forms of b2GPI, circular and extended, to cardiolipin is inhibited similarly by A1. We analyzed the inhibition of the monomeric molecules, b2GPI in serum and the isolated domain V, by monomeric A1. The majority of b2GPI in normal human serum is in a circular conformation [11]. The isolated domain V mimics this domain in the extended form of b2GPI. The same concentration of A1 was required to inhibit 50% of the binding of b2GPI in human serum and the individual domain V of b2GPI ( Figure 8). The concentration of A1 at half-maximal inhibition was 176637 mM for b2GPI in serum and 188644 mM for domain V. This observation demonstrates that A1 binds circular and extended b2GPI with the same affinity suggesting that the binding site for A1 is not obscured in the circular form of b2GPI.

Stability of the A1-A1 inhibitor in human serum at 37uC
To evaluate the susceptibility of the A1-A1 inhibitor to degradation by serum proteases, we incubated A1-A1 in serum at 37uC. Degradation of A1-A1 was monitored by the reversedphase HPLC by comparing the peak corresponding to the intact A1-A1 on chromatograms collected at different time intervals. The amount of A1-A1 that remained in serum was calculated from the area under the eluted peak. More than 35% of A1-A1 remained in serum after 15 days of incubation at 37uC, indicating that A1-A1 has a favorable stability in serum (Figure 9).

Discussion
The work reported here examines the effectiveness of a novel dimeric inhibitor A1-A1 to interfere with the binding of pathological b2GPI/anti-b2GPI antibody complexes to anionic phospholipids compared to monomeric A1. The dimeric inhibitor, A1-A1, consists of two ligand-binding A1 modules from ApoER2  connected by a flexible peptide linker. Biophysical characterization of A1-A1 by reverse-phased chromatography confirmed that it is correctly folded in a calcium-dependent manner. Recently, we determined that the bound A1 module prevents the association of b2GPI with anionic phospholipids [46]. Present studies confirmed our previous observations suggesting that the dimeric A1 inhibitor interferes with two pathologically important interactions: the binding of b2GPI/antibody complexes to anionic phospholipids expressed on activated cells and to ApoER2, a lipoprotein receptor on platelets [29,51].
Normally, b2GPI circulates in the blood plasma as a monomer. Anti-b2GPI antibodies in patients with antiphospholipid syndrome create multivalent b2GPI complexes that have much stronger   affinity for anionic phospholipids and lipoprotein receptors than the monomeric b2GPI [44,52]. Because b2GPI/antibody complexes expressing dimeric b2GPI have prothrombotic properties, in contrast to monomeric pathologically inactive b2GPI, we designed a dimeric inhibitor. We hypothesized that the dimeric molecule A1-A1 preferentially targets multivalent pathological b2GPI/anti-b2GPI antibody complexes leaving monomeric b2GPI, which is abundant in plasma, practically unaffected. To compare A1-A1 and A1 on the inhibition of the binding of b2GPI to anionic phospholipids, we used different preparations of b2GPI, such as b2GPI in normal human serum, b2GPI purified from human plasma and recombinant domain V of b2GPI. b2GPI is a flexible molecule that can adopt a circular [11] and extended conformation [8,9,10]. b2GPI in plasma is predominantly in a circular form [11] and the individual domain V closely resembles domain V in the extended conformation of b2GPI, as we demonstrated here by the X-ray crystallography.
We compared the binding curves for two preparations of b2GPI and for b2GPI-DV in the absence and in the presence of the dimerizing antibodies. In all three cases, we observed that the presence of antibodies significantly enhanced the binding of b2GPI and b2GPI-DV to cardiolipin, similarly to what was previously detected for chimeric dimers of b2GPI compared to b2GPI monomers [34]. The binding of b2GPI and b2GPI-DV to cardiolipin-coated surface in the presence of constant amounts of dimerizing antibodies increases with the addition of b2GPI or b2GPI-DV reaching saturation when all antibodies are engaged in complexes. For the inhibition studies, we used concentrations of b2GPI and b2GPI-DV in the linear region of the binding curves at about 50-60% of the maximal binding. Comparison of the binding curves in the presence and in the absence of antibodies suggests that the contribution of b2GPI or b2GPI-DV monomers to total binding in the presence of antibodies is negligible compared to the contribution of b2GPI/antibody or b2GPI-DV/antibody complexes and, therefore, the inhibition curves measured in the presence of antibodies describe the inhibition of a fraction of dimerized molecules.
We determined that, regardless of the source of b2GPI, 1) A1-A1 is much more efficient in inhibition of the binding of b2GPI/ antibody complexes to anionic phospholipids than A1 and 2) the  inhibition of the binding of monomeric b2GPI to anionic phospholipids by either A1-A1 or A1 is practically identical and weak. We also observed that the inhibition of both b2GPI in serum and the individual domain V by A1 is identical in the absence of dimerization antibodies, suggesting that A1 binds the circular and extended forms of b2GPI with the same affinity. Therefore, the binding site for A1 is not obscured on the circular form of b2GPI. Anti-b2GPI antibodies in patients with antiphospholipid syndrome are heterogeneous and their epitopes are scattered over domains I to IV of b2GPI [51,53]. Some antibodies in patients with antiphospholipid syndrome might bind circular b2GPI and some antibodies might need an extended b2GPI to have their epitopes exposed. Our results demonstrated that when b2GPI, whether circular or extended, is dimerized by anti-b2GPI antibodies, it is more strongly inhibited by A1-A1 than by monomeric A1 by forming stable b2GPI/anti-b2GPI/A1-A1 complexes ( Figure 10). We measured in vitro the serum stability of A1-A1. About 35% of A1-A1 remained intact after incubation in serum at 37uC for more than two weeks, indicating that A1-A1 might have favorable pharmacokinetic properties. Given that A1 modules are naturally expressed, the A1-A1 inhibitor is unlikely to be immunogenic.
In our previous work, we have shown that LA modules from different lipoprotein receptors bind to the same site on b2GPI-DV [46]. Therefore, A1-A1 inhibits the binding of b2GPI/antibody complexes not only to ApoER2, but to other lipoprotein receptors as well. Whether other lipoprotein receptors besides ApoER2 contribute to the pathology of antiphospholipid syndrome awaits further investigation. Our results suggest that A1-A1 may be a starting point in the development of the effective inhibitor that interferes with the binding of b2GPI/antibody complexes to anionic phospholipids and lipoprotein receptors. The binding affinity of A1-A1 for b2GPI/anti-b2GPI antibody complexes can be improved in two ways: by optimization of the linker between the two A1 modules and by improving the binding affinity of A1 for b2GPI-DV. Eventually, the A1-based inhibitor can be replaced with small molecule compounds in a dimerized form.
We believe the approach of using a dimeric inhibitor that blocks b2GPI in the pathological multivalent b2GPI/anti-b2GPI complexes holds significant promise. In these studies, we are inhibiting a well characterized binding site for lipoprotein receptors on b2GPI, instead of preventing the binding of antibodies to b2GPI, which are highly heterogeneous in APS patients. Our approach to target the dimerized b2GPI with a dimeric inhibitor could be applied to other pathologically important interactions of b2GPI/ antibody complexes. As soon as the binding sites on b2GPI for other APS-related receptors are mapped and characterized in detail, they can be targeted by dimeric inhibitors.
In conclusion, we developed and tested a novel dimeric inhibitor of the b2GPI/antibody complexes. This dimeric inhibitor preferentially targets b2GPI dimerized by anti-b2GPI antibodies compared to pathologically inactive monomeric b2GPI. It prevents the binding of b2GPI/antibody complexes to anionic phospholipids and ApoER2, and might eventually lead to a drug specific for antiphospholipid syndrome.

Protein expression and purification
Monomeric A1 is a fragment of mouse ApoER2 (residues 12-47 from the mature protein). The dimeric inhibitor, A1-A1, was constructed to contain two A1 fragments connected by a Gly-Ser-Ser-Gly linker. A1 and A1-A1 containing an extra N-terminal Ala and C-terminal Glu-Ala residues were expressed in E.coli as TrpLE fusion proteins and purified from inclusion bodies essentially as previously described [54]. Domain V of b2GPI (residues 244-326), was subcloned into a pET15b vector  (Novagen). The encoded protein has an N-terminal histidine tag followed by the sequence recognized by the Tobacco Etch Virus (TEV) protease so that the tag can be removed. To make the domain V of b2GPI recognized by antibodies directed to an HA peptide, the HA sequence, YPYDVPDYA, was introduced at the N-terminus of domain V right after the TEV cleavage site. Domain V with and without the peptide tag was expressed in E.coli, recovered from inclusion bodies, cleaved with TEV and refolded by dialysis at 4uC under conditions permitting disulfide exchange before final purification by reversed-phase HPLC on a C18 column. Protein concentrations were calculated from the measured absorbance of samples at 280 nm using extinction coefficients from the output of ExPASy Protparam tool (http:// expasy.org/tools/protparam.html). A full-length b2GPI was purchased from Haematologic Technologies, Inc. Concentrations of b2GPI were calculated using an extinction coefficient at 280 nm E 1% of 10 and molecular weight of 54200, as suggested by the supplier.
Crystallization, data collection and structure determination of b2GPI-DV Initial crystallization condition was determined in crystallization screen performed at the Hauptman-Woodward Medical Research Institute [55]. The best crystal of b2GPI-DV was obtained at room temperature in hanging drop by combining 1 mL of b2GPI-DV (7 mg/ml in 20 mM HEPES, pH 7.0) with 1 mL of reservoir solution containing 100 mM ammonium sulfate, 40% PEG 1500, 100 mM bis-Tris, pH 7.2. The reservoir solution supplemented with 20% glycerol was used as cryoprotectant. Data was collected from a single crystal at beamline X29A of Brookhaven National Laboratories (NSLS). The crystals belong to the space group P1 with two molecules of b2GPI-DV per asymmetric unit and a solvent content of 45%. Data was processed with MOSFLM [56]. A total of 5% of reflections were excluded and used for R free calculations. The structure was solved by molecular replacement with PHASER [57] using coordinates of domain V extracted from the crystal structure of b2GPI (PDB ID 1C1Z). The initial model determined by PHASER was adjusted with the program COOT [58] and refined using the program REFMAC5 [59]. The final refinement was performed with PHENIX software suit [60].
Assay for the binding and inhibition of the binding of b2GPI and b2GPI-DV to a cardiolipin-coated surface Cardiolipin-coated 96 well plates from the ImmunoWELL cardiolipin IgG test kit (GenBio) were blocked with 0.5% of skim milk and 2% BSA in 20 mM Tris, 100 mM NaCl, 2 mM CaCl2, pH 7.4. The assay buffer contained 20 mM Tris, 100 mM NaCl, 2 mM CaCl2, pH 7.4 with 2% BSA and the wash buffer was 20 mM Tris, 100 mM NaCl, 2 mM CaCl2, pH 7.4. When the purified b2GPI (Haematologic Technologies, Inc.) was used in experiments, 27 mM glycine was added to the assay buffer to account for glycine present in the stock solution of b2GPI. b2GPI bound to cardiolipin was detected with peroxidase-conjugated anti-b2GPI antibodies (Cedarlane, CL20021HP, 2 mg/ml) diluted 1:2500. To detect b2GPI-DV bound to cardiolipin, we used peroxidase-conjugated anti-HA antibody (Abcam, ab1265, 1 mg/ ml) directed to HA epitope tag at the N-terminus of b2GPI-DV diluted 1:2500. The peroxidase activity of the bound antibodies was detected using 2-29-azino-di-[3-ethylbenzthizzoline] sulfonate (ABTS) chromogenic reagent by measuring OD at 405 nm. All measurements were done in triplicates and corrected to blank before data fitting. The blank contained all components except for b2GPI, serum or b2GPI-DV. The binding and inhibition data was fitted to one-site models using the nonlinear least-squares Marquardt-Levenberg algorithm implemented in GNUPLOT program. The fits of the raw data and the titration data points were then normalized to the maximum binding determined from the fit to facilitate comparison.
For the binding studies, 50 ml of increasing concentrations of either b2GPI (Haematologic Technologies, Inc.), pooled normal human serum (Innovative Research) or the purified recombinant b2GPI-DV were applied to wells and incubated for 1 hour at room temperature. After washing, anti-b2GPI or anti-HA antibody was added to wells and incubated for 1 hour at room temperature before detection. In the second set of experiments, samples containing various concentrations of b2GPI, pooled normal human serum or b2GPI-DV were first incubated for 1 hour at room temperature with the anti-b2GPI or anti-HA antibodies. Then, the samples were applied to cardiolipin, incubated for 1 hour, washed, and bound b2GPI or b2GPI-DV was detected. Figure 10. Complex formation between b2GPI, inhibitor and anti-b2GPI antibody. The binding of A1-A1, but not monomeric A1, forms stable b2GPI/anti-b2GPI/A1-A1 complex regardless of localization of the epitope for anti-b2GPI antibody and whether circular or extended b2GPI is dimerized by antibody. b2GPI (blue), A1 or A1-A1 inhibitor (red) and anti-b2GPI antibody (green). doi:10.1371/journal.pone.0015345.g010 For the inhibition studies, increasing concentrations of A1 or A1-A1 were added to a constant amount of b2GPI (50 nM), normal human serum (1%) or b2GPI-DV (130 nM) and incubated for 1 hour at room temperature. Then, 50 ml of the mixtures were incubated on wells for the additional 1 hour. After washing, 50 ml of either anti-b2GPI or anti-HA antibody was added to wells and incubated for 1 hour before detection. In the second set of experiments, 50 ml of samples containing increasing concentrations of A1 or A1-A1 and the constant amounts of either b2GPI (10 nM), normal human serum (0.04%) or b2GPI-DV (30 nM) were first incubated for 1 hour at room temperature with the anti-b2GPI or anti-HA antibodies. Then, samples were incubated on wells for an additional 1 hour and, after washing, the bound b2GPI/anti-b2GPI or b2GPI-DV/anti-tag antibody complexes were detected.

Measurements of the stability of A1-A1 in serum
Lyophilized A1-A1 purified by reversed-phase chromatography was dissolved in water and its concentration measured by absorbance at 280 nm. The required amount of A1-A1 (180 mg) was then lyophilized and, subsequently, dissolved in 360 ml of pooled normal human serum (Innovative Research). Serum with A1-A1 was filtered through a 0.2 mm eppendorf centrifuge filter, divided into 40 ml samples and set for incubation at 37uC. At timed intervals, 900 ml of 10% acetonitrile with 0.1% TFA in water (buffer A) was added to a 40 ml sample of A1-A1 in serum. Filtered samples were analyzed by reversed-phase HPLC on a C18 column using a linear gradient of 0.1% per minute of buffer B (acetonitrile with 0.1% TFA) staring at 15 minutes from 26% of acetonitrile and monitored for 30 minutes.