Significance of valine/leucine247 polymorphism of β2-glycoprotein I in antiphospholipid syndrome: Increased reactivity of anti–β2-glycoprotein I autoantibodies to the valine247 β2-glycoprotein I variant
Hokkaido University Graduate School of Medicine, Sapporo, Japan
To clarify the consequences of the valine/leucine polymorphism at position 247 of the β2-glycoprotein I (β2GPI) gene in patients with antiphospholipid syndrome (APS), by investigating the correlation between genotypes and the presence of anti-β2GPI antibody. The reactivity of anti-β2GPI antibodies was characterized using recombinant Val247 and Leu247 β2GPI.
Sixty-five Japanese patients with APS and/or systemic lupus erythematosus who were positive for antiphospholipid antibodies and 61 controls were analyzed for the presence of the Val/Leu247 polymorphism of β2GPI. Polymorphism assignment was determined by polymerase chain reaction followed by restriction enzyme digestion. Recombinant Val247 and Leu247 β2GPI were established to compare the reactivity of anti-β2GPI antibodies to β2GPI between these variants. The variants were prepared on polyoxygenated plates or cardiolipin-coated plates, and the reactivity of a series of anti-β2GPI antibodies (immunized anti-human β2GPI monoclonal antibodies [Cof-19–21] and autoimmune anti-β2GPI monoclonal antibodies [EY1C8, EY2C9, and TM1G2]) and IgGs purified from patient sera was investigated.
A positive correlation between the Val247 allele and the presence of anti-β2GPI antibodies was observed in the patient group. Human monoclonal/polyclonal anti-β2GPI autoantibodies showed higher binding to recombinant Val247 β2GPI than to Leu247 β2GPI, although no difference in the reactivity of the immunized anti-β2GPI between these variants was observed. Conformational optimization showed that the replacement of Leu247 by Val247 led to a significant alteration in the tertiary structure of domain V and/or the domain IV–V interaction.
The Val247 β2GPI allele was associated with both a high frequency of anti-β2GPI antibodies and stronger reactivity with anti-β2GPI antibodies compared with the Leu247 β2GPI allele, suggesting that the Val247 β2GPI allele may be one of the genetic risk factors for development of APS.
The antiphospholipid syndrome (APS) is characterized by arterial/venous thrombosis and pregnancy morbidity in the presence of antiphospholipid antibodies (aPL) (1–3). Among the targets of aPL, β2-glycoprotein I (β2GPI), which bears epitopes for anticardiolipin antibodies (aCL), has been extensively studied (4–6). APS-related aCL do not recognize free β2GPI, but do recognize β2GPI when it is complexed with phospholipids or negatively charged surfaces, by exposure of cryptic epitopes (7) or increment of antigen density (8).
The significance of antigen polymorphism in the production of autoantibodies or the development of autoimmune diseases is now being widely discussed. It is speculated that amino acid substitution in antigens can lead to differences in antigenic epitopes of a given protein. In particular, β2GPI undergoes conformational alteration upon interaction with phospholipids (9). β2GPI polymorphism on or near the phospholipid binding site can affect the binding or production of aCL (anti-β2GPI autoantibodies), the result being altered development of APS. Polymorphism near the antigenic site, or which leads to alteration of the tertiary structure of the whole molecule, may affect the binding of autoantibodies. Five different gene polymorphisms of β2GPI attributable to a single-nucleotide mutation have been described: 4 are a single amino acid substitution at positions 88, 247, 306, and 316 (10), and the other is a frameshift mutation associated with β2GPI deficiency found in the Japanese population (11). In particular, the Val/Leu247 polymorphism locates in domain V of β2GPI, between the phospholipid binding site in domain V and the potential epitopes of anti-β2GPI antibodies in domain IV, as we reported previously (12). Although anti-β2GPI antibodies are reported to direct to domain I (13) or domain V (14) as well, it should be considered that a certain polymorphism alters the conformation of the molecule, affecting function or antibody binding at a distant site.
We previously reported that, in a group of British Caucasian subjects, the Val247 allele was significantly more frequent in primary APS patients with anti-β2GPI antibodies than in controls or in primary APS patients without anti-β2GPI antibodies (15), but the importance of the Val247 allele in patients with APS is still controversial. In this study, we analyzed the correlation between the β2GPI Val247 allele and anti-β2GPI antibodies in the Japanese population. We also investigated the reactivity of anti-β2GPI antibodies to recombinant Val247 β2GPI and Leu247 β2GPI, using a series of monoclonal anti-β2GPI antibodies and IgGs purified from sera of patients with APS. Finally, to investigate the difference in anti-β2GPI binding to those variants, we conformationally optimized to domain V and the domain IV–V complex of β2GPI variants at position 247, referring the crystal structure of β2GPI.
PATIENTS AND METHODS
Patients and controls.
The study group comprised 65 patients (median age 38 years [range 18–74 years]; 57 women and 8 men) who attended the Hokkaido University Hospital, all of whom were positive for aPL (IgG, IgA, or IgM class aCL, and/or lupus anticoagulant). Thirty-four patients had APS (16 had primary APS, and 18 had secondary APS), and 31 patients did not have APS (24 had systemic lupus erythematosus [SLE], and 7 had other rheumatic diseases). Among all subjects, 19 had a history of arterial thrombosis, and 6 had venous thrombosis. Of the 31 patients with a history of pregnancy, 8 experienced pregnancy complications (some patients had more than 1 manifestation of pregnancy morbidity). Anti-β2GPI antibodies were detected by enzyme-linked immunosorbent assay (ELISA) as β2GPI-dependent aCL (16). IgG, IgA, or IgM class β2GPI-dependent aCL were found in 30, 14, and 21 patients, respectively (some patients had >1 isotype), and 34 patients had at least 1 of those isotypes. Lupus anticoagulant, detected by 3 standard methods described previously (17), was found in 51 patients. The diagnoses of APS and SLE, respectively, were based on the preliminary classification criteria for definite APS (18) and the American College of Rheumatology criteria for the classification of SLE (19). Informed consent was obtained from each patient or control subject. The control group comprised 61 healthy individuals with no history of autoimmune, thrombotic, or notable infectious disease.
Determination of β2GPI gene polymorphism.
Genomic DNA was extracted from peripheral blood mononuclear cells (PBMCs) using a standard phenol–chloroform extraction procedure or the DnaQuick kit (Dainippon, Osaka, Japan). Polymorphism assignment was determined by polymerase chain reaction (PCR) followed by allele-specific restriction enzyme digestion (PCR–restriction fragment length polymorphism) using Rsa I (Promega, Southampton, UK) as described previously (15).
Purification of patient IgG.
Sera from 11 patients positive for IgG class β2GPI-dependent aCL were collected. The mean (±SD) titer of aCL IgG from these patients was 29.0 ± 21.5 IgG phospholipid (GPL) units (range 12.4 to >98 GPL units). IgG was purified from these sera using a protein G column and the MAbTrap GII IgG purification kit (Pharmacia Biotech, Freiburg, Germany), as recommended by the manufacturer.
Monoclonal anti-β2GPI antibodies.
Two types of anti-β2GPI monoclonal antibodies were used. Cof-19, Cof-20, and Cof-21 are mouse monoclonal anti-human β2GPI antibodies obtained from immunized BALB/c mice, directed to domains V, III, and IV of β2GPI, respectively. These monoclonal antibodies recognize the native structure of human β2GPI (12).
EY1C8, EY2C9, and TM1G2 are IgM class autoimmune monoclonal antibodies established from patients with APS (20). These antibodies bind to domain IV of β2GPI, but only after interaction with solid-phase phospholipids or with a polyoxygenated polystyrene surface. EY1C8 and EY2C9 were established from a patient whose genotype of β2GPI was heterozygous for Val/Leu247. The genotype of the patient with TM1G2 was not determined.
Preparation of recombinant β2GPI.
As previously reported, genes were expressed in Spodoptera frugiperda Sf9 insect cells infected with recombinant baculoviruses (12). A full-length complementary DNA of human β2GPI coding Val247 was originally obtained from Hep-G2 cells (21), and the valine residue was replaced by leucine, using the GeneEditor in vitro Site-Directed Mutagenesis System (Promega, Madison, WI). The sequence of the primers for a mutant Val247→Leu (GTA→TTA) is as follows: 5′-GCATCTTGTAAATTACCTGTGAAAAAAG-3′. A DNA sequence of the mutant was verified by analysis using ABI Prism model 310 (PE Applied Biosystems, Foster City, CA).
Binding assays of monoclonal anti-β2GPI antibodies and purified IgGs to the recombinant β2GPI (cardiolipin-coated plate).
The reactivity of a series of monoclonal anti-β2GPI antibodies and IgG fractions (purified from the sera of APS patients positive for IgG class anti-β2GPI) against 2 β2GPI variants was investigated using an ELISA. ELISAs were performed using a cardiolipin-coated plate as previously reported (16) but with a slight modification. Briefly, the wells of Sumilon Type S microtiter plates (Sumitomo Bakelite, Tokyo, Japan) were filled with 30 μl of 50 μg/ml cardiolipin (Sigma, St. Louis, MO) and dried overnight at 4°C. After blocking with 2% gelatin in phosphate buffered saline (PBS) for 2 hours and washing 3 times with 0.05% PBS–Tween, 50 μl of 10 μg/ml recombinant β2GPI and controls were distributed and incubated for 30 minutes at room temperature. Wells were filled with 50 μl of serial dilutions of monoclonal antibodies (Cof-19–21, EY1C8 and EY2C9, and TM1G2) or purified patient IgG (100 μg/ml), followed by incubation for 30 minutes at room temperature. After washing 3 times, 50 μl of alkaline phosphatase–conjugated anti-mouse IgG (1:3,000), anti-human IgM (1:1,000), or anti-human IgG (1:6,000) was distributed and incubated for 1 hour at room temperature. The plates were washed 4 times, and 100 μl of 1 mg/ml p-nitrophenyl phosphate disodium (Sigma) in 1M diethanolamine buffer (pH 9.8) was distributed. Optical density (OD) was read at 405 nm, with reference at 620 nm. One percent fatty acid–free bovine serum albumin (BSA) (A-6003; Sigma)–PBS was used as sample diluent and control.
Binding assays of monoclonal anti-β2GPI antibodies to recombinant β2GPI (polyoxygenated plate).
Anti-β2GPI antibody detection assay using polyoxygenated plates was performed as previously reported (22), with minor modifications. Briefly, the wells of polyoxygenated MaxiSorp microtiter plates (Nalge Nunc International, Roskilde, Denmark) were coated with 50 μl of 1 μg/ml recombinant β2GPI in PBS and incubated overnight at 4°C. After blocking with 3% gelatin–PBS at 37°C for 1 hour and washing 3 times with PBS–Tween, 50 μl of monoclonal antibodies, diluted with 1% BSA–PBS, were distributed and incubated for 1 hour at room temperature. The following steps were taken, in a similar manner.
Conformational optimization of domain V and the domain IV–V complex in human β2GPI variants at position 247.
A conformation of domain V in the valine variant at position 247 was first constructed from the crystal structure of the leucine variant (implemented in Protein Data Bank: 1C1Z) (23). Replacement of leucine by valine at position 247 was performed using the Quanta system (Molecular Simulations, San Diego, CA), and the model was optimized by 500 cycles of energy minimization by the CHARMm program (24), with hydrophilic hydrogen atoms and TIP3 water molecules (25). Molecular dynamics simulation (5 psec) of the model was then performed with 0.002 psec time steps. The cutoff distance for nonbonded interactions was set to 15Å, and the dielectric constant was 1.0. A nonbonded pair list was updated every 10 steps. The most stable structure of each domain in the dynamics iterations was then optimized by 500 cycles of energy minimization. The final structures of domain V consisted of 2,616 atoms, including 603 TIP3 water molecules, and had a total energy of −1.63 × 104 kcal/mole with a root-mean-square force of 0.869 kcal/mole.
Molecular models of a domain IV–V complex (leucine and valine variants at position 247) were further constructed by considering the location of the oligosaccharide attachment site in domain IV, the location of epitopic regions of the Cof-8 and Cof-20 monoclonal antibodies, the junction between domains IV and V, and molecular surface charges of both domains. These models were again optimized by molecular dynamics simulation and by energy minimization as described above. The final structures of the complex in the leucine and valine variants consisted of 3,773 and 3,778 atoms, respectively, including hydrophilic hydrogen atoms and 806 and 808 TIP3 water molecules, respectively, and had total energy of −2.07 × 104 and −2.03 × 104 kcal/mole with a root-mean-square force of 0.985 and 0.979 kcal/mole, respectively.
Correlations between the allele frequencies and clinical features such as the positiveness of β2GPI-dependent aCL were expressed as odds ratios (ORs) and 95% confidence intervals (95% CIs). P values were determined by chi-square test with Yates' correction. P values less than or equal to 0.05 were considered significant.
Val/Leu247 polymorphism of β2GPI and the presence of β2GPI-dependent aCL.
As shown in Table 1, the Leu247 allele was dominant in the population of healthy Japanese individuals, compared with Caucasians, which is consistent with a previous report (26). Japanese patients with anti-β2GPI had a significantly increased frequency of the Val247 allele, compared with Japanese patients without anti-β2GPI (P = 0.0107) or Japanese controls (P = 0.0209).
Table 1. Frequency of the Val247 allele of β2GPI in patients with APS*
Values are the number (%). β2GPI = β2-glycoprotein I; APS = antiphospholipid syndrome.
P = 0.0107 versus patients without anti-β2GPI (odds ratio [OR] 3.01, 95% confidence interval [95% CI] 1.26–7.16), and P = 0.0209 versus controls, by chi-square test (OR 2.15, 95% CI 1.09–4.23).
P = 0.204 versus patients without anti-β2GPI (OR 2.92, 95% CI 1.16–7.39), and P = 0.0396 versus controls, by chi-square test (OR 2.51, 95% CI 1.03–6.13).
The binding of autoimmune anti-β2GPI to recombinant Val247 and Leu247 β2GPI.
Representative binding curves using cardiolipin-coated plates and polyoxygenated plates are shown in Figure 1. Regardless of the type of plates, Cof-20 bound equally to valine and leucine variants of β2GPI (Figures 1a and c), in any concentration of Cof-20. The binding curves of Cof-19 and Cof-21 were similar to that of Cof-20 (results not shown). In contrast, EY2C9 showed stronger binding to Val247 β2GPI than to Leu247 β2GPI (Figures 1b and d). EY1C8 and TM1G2 also showed stronger binding to Val247 β2GPI. Figure 2a shows the binding of the monoclonal antibodies, on cardiolipin-coated plates, in the following concentrations: for Cof-19–21, 100 ng/ml; for EY1C8 and EY2C9, 1 μg/ml; and for TM1G2, 2.5μg/ml. In contrast with the close reactivity of Cof-19, Cof-20, and Cof-21 between Val247 β2GPI and Leu247 β2GPI, autoimmune monoclonal antibodies (EY1C8, EY2C9, and TM1G2) showed higher binding to Val247 β2GPI than to Leu247 β2GPI. The autoimmune monoclonal antibodies also showed a higher binding to Val247 β2GPI directly coated on polyoxygenated plates (Figure 2b). IgG in sera collected from 11 patients (100 μg/ml) also showed higher binding to Val247 β2GPI than to Leu247 β2GPI on cardiolipin-coated plates, regardless of the patients' genotypes (Figure 3).
Conformational alteration by leucine replacement by valine at position 247.
Each domain V conformation in 2 variants at position 247 is shown in Figure 4a. The root-mean-square deviations for matching backbone atoms and equivalent atoms in the leucine and valine variants were 0.76 and 1.11 Å, respectively. The largest shift was observed at Val303, one of the residues located on the backbone neighboring position 247. The shift seemed to be caused by weak flexibility of side chains consisting of Val247, Pro248, and Val249 and the electrostatic interactions between Lys250, Lys251, Glu307, and Lys308.
The molecular models of the IV–V complex in leucine and valine variants are shown in Figure 4b. The root-mean-square deviations for matching these backbone atoms and equivalent atoms were 1.72 and 2.03 Å, respectively. Electrostatic interactions and hydrogen bonds between Asp193 and Lys246/Lys250, Asp222 and Lys305, and Glu228 and Lys308 appeared in the IV–V complex, but the interaction between Glu228 and Lys308 was disrupted by the leucine replacement by valine, because direction of the Lys308 side chain was significantly changed in the complex. As a result, Trp235 of domain IV, located on the contact surface with domain V, was slightly shifted.
This study shows the positive correlation between the Val247 β2GPI allele and anti-β2GPI antibody production in a Japanese population, confirming the correlation observed in a British Caucasian population in our previous report (15). A positive correlation between the Val247 allele and the presence of anti-β2GPI antibodies was also reported in Asian American (26) and Mexican patients (27). However, this correlation was not observed in other American populations (26) or in patients with thrombosis or pregnancy complications in the UK (28). This discrepancy may be the result of the difference in the frequency of the Val247 allele among races, or the difference in the background of investigated patients. Another possibility is that the relationship between the Val247 allele and thrombosis in Caucasians may be controversial due to underpowered studies or to differences in the procedure used to detect anti-β2GPI antibodies. Methods for the detection of anti-β2GPI antibodies differ among laboratories. For example, cardiolipin-coated plates or oxygenated plates are used in some methods, whereas unoxygenated plates are used in others. In addition, bovine β2GPI is used instead of human β2GPI in some assays. The antibodies used for standardization also differ, although monoclonal antibodies such as EY2C9 and HCAL (29) have been proposed as international standards of calibration materials.
β2GPI is a major target antigen for aCL, and, according to our previous investigation, B cell epitopes reside in domain IV and are considered to be cryptic and to appear only when β2GPI interacts with negatively charged surfaces such as cardiolipin, phosphatidylserine, or polyoxygenated polystyrene surface (7), although other studies indicate that the B cell epitopes are located on domain I (13) or domain V (14). According to another interpretation for the specificity of aCL, increment of the local antigen density on the negatively charged surface also contributes to anti-β2GPI detection in ELISA (8, 30). Studies on the crystal structure of human β2GPI revealed that the lysine-rich site and an extended C-terminal loop region on domain V are crucial for phospholipid binding. Position 247 is located at the N-terminal side of domain V, and, around this position, Lys242, Ala243, and Ser244 were suggested to play a role in the interaction between domains IV and V (9, 23, 31).
Although the Val/Leu247 polymorphism may not be very critical for the autoantibody binding, the amino acid substitution at this point was revealed to affect the affinity of monoclonal aCL established from patients with APS and that of purified IgG from patients positive for β2GPI-dependent aCL. We conformationally optimized to domain V and the domain IV–V complex of β2GPI variants at position 247, referring the crystal structure of β2GPI. IgG aCL was screened using the standardized aCL ELISA, in which both the Leu247 and the Val247 allele of β2GPI are contained as antigen. Although biochemical characteristics and structure are similar between valine and leucine, the replacement of Leu247 by Val247 leads to a significant alteration in the tertiary structure of domain V and/or the domain IV–V interaction (Figure 4). It is likely that the structural alteration affects the affinity between anti-β2GPI autoantibodies and the epitope(s) present on its molecule. One explanation for this phenomenon is that this β2GPI polymorphism affects the electrostatic interaction between domain IV and domain V or the protein–protein interaction, resulting in differences in the accessibility of the recognition site by the autoantibodies, or the local density of β2GPI.
Another possible explanation of the correlation between the Val/Leu247 polymorphism of β2GPI and anti-β2GPI antibodies is T cell reactivity. Ito et al (32) investigated T cell epitopes of patients with anti-β2GPI autoantibodies by stimulating patients' PBMCs with a peptide library that covers the β2GPI sequence. Four of 7 established CD4+ T cell clones reacted to peptide fragments that include amino acid position 244–264, then position 247 is included among the candidate epitopes. Arai et al (33) found preferred recognition of peptide position 276–290 by T cell clones from patients with APS. They also found high reactivity to peptide 247–261 in one patient. We speculate that a small alteration in the conformation arising from the valine/leucine substitution at position 247 may affect the susceptibility to generate autoreactive T cell clones in patients with APS.
Our results in this study indicate that the Val/Leu247 polymorphism affects the antigenicity of β2GPI for anti-β2GPI autoantibodies, and that the Val247 allele can be a risk factor for having autoantibodies against this molecule. Therefore, the Val/Leu247 variation of β2GPI may be crucial for autoimmune reactivity against β2GPI. We further show the significance of the Val/Leu247 polymorphism of β2GPI in the strength of the binding between β2GPI and anti-β2GPI autoantibodies. The significance of antigen polymorphisms in the production of autoantibodies or in the development of autoimmune diseases is not well understood. To our knowledge, this report is the first to present a genetic polymorphism of autoantigen directly affecting its interaction with autoantibodies.