ω-Carboxyl variants of 7-ketocholesteryl esters are ligands for β2-glycoprotein I and mediate antibody-dependent uptake of oxidized LDL by macrophages

β2-Glycoprotein I (β2-GPI) is a major antigen for anticardiolipin antibodies (aCL, Abs) present in patients with antiphospholipid syndrome. We recently reported that β2-GPI specifically binds to oxidized LDL (oxLDL) and that the β2-GPI's major ligand, oxLig-1 is 7-ketocholesteryl-9-carboxynonanoate (Kobayashi, K., E. Matsuura, Q. P. Liu, J. Furukawa, K. Kaihara, J. Inagaki, T. Atsumi, N. Sakairi, T. Yasuda, D. R. Voelker, and T. Koike. 2001. A specific ligand for β2-glycoprotein I mediates autoantibody-dependent uptake of oxidized low density lipoprotein by macrophages. J. Lipid Res. 42: 697–709). In the present study, we demonstrate that ω-carboxylated 7-ketocholesteryl esters are critical for β2-GPI binding. A positive ion mass spectrum of a novel ligand, designated oxLig-2, showed fragmented ions at m/z 383 and 441 in the presence of acetone, which share features of oxLig-1 and 7-ketocholesterol. In the negative ion mode, ions at m/z 627, 625, and 243 were observed. oxLig-2 was most likely 7-ketocholesteryl-12-carboxy (keto) dodecanoate. These ligands were recognized by β2-GPI. Liposome binding to macrophages was significantly increased depending on the ligand's concentration, in the presence of β2-GPI and an anti-β2-GPI Ab. Synthesized variant, 7-ketocholesteryl-13-carboxytridecanoate (13-COOH-7KC), also showed a significant interaction with β2-GPI and a similar binding profile with macrophages. Methylation of the carboxyl function diminished all of the specific ligand interactions with β2-GPI. Thus, ω-carboxyl variants of 7-ketocholesteryl esters can mediate anti-β2-GPI Ab-dependent uptake of oxLDL by macrophages, and autoimmune atherogenesis linked to β2-GPI interaction with oxLDL.

The autoimmune disorder, antiphospholipid syndrome (APS), is characterized by the presence of a group of heterogeneous antiphospholipid antibodies (aPL Abs), such as anticardiolipin Abs (aCL) and lupus anticoagulants (LA), in blood, and by occurrence of thromboembolic complications in the arterial and/or venous vasculatures of the patients (1,2). In 1990, three groups of investigators independently reported that a plasma/serum cofactor complexed with negatively charged phospholipids (PLs), such as cardiolipin, is an antigenic target for aCL (3)(4)(5). It is now widely accepted that ␤ 2 -glycoprotein I ( ␤ 2 -GPI) is the major antigen for aCL. However, the mechanisms for interaction between ␤ 2 -GPI and anti-␤ 2 -GPI Abs are still uncertain. Two currently proposed mechanisms are: i ) Binding of ␤ 2 -GPI to PL induces a conformational change in the ␤ 2 -GPI molecule, thus exposing a cryptic epitope on the protein for the auto-Ab binding, and/or, ii ) ␤ 2 -GPI binding to anionic PL increases the local concentration of ␤ 2 -GPI, thus promoting an increase in intrinsic affinity and Ab binding to the protein (6)(7)(8)(9)(10)(11)(12)(13)(14).
␤ 2 -GPI is a 50 kDa protein present in plasma at approximately 200 g/ml. It binds to negatively charged molecules, including PLs (15), heparin (16), and plasma membranes of activated platelets, and apoptotic cells on which phosphatidylserine (PS) is exposed (17,18). ␤ 2 -GPI is a member of the short consensus repeats of the complement control protein superfamily, and its fifth domain has a PL binding region. The X-ray crystal analyses (19) showed that the PL binding is provided by a patch consisting of 14 residues of positively charged amino acids and by a flexible loop of S 311 -K 317 in domain V. Recent analysis with domain V's mutant proteins demonstrated interactions of the flexible loop with hydrophobic ligands (20,21). However, the full details of the structure of in vivo lipid ligands participating in the ␤ 2 -GPI binding remains unclear.
In the present report, we now demonstrate that oxLDL recognition by ␤ 2 -GPI and an anti-␤ 2 -GPI Ab is provided by an -carboxyl function introduced by oxidation of an unsaturated acyl chain of cholesteryl esters.

Preparation of human ␤ 2 -GPI
␤ 2 -GPI was purified from normal human plasma as described (36) with slight modification. Pooled plasma from healthy subjects was subsequently chromatographed on a heparin-Sepharose column, on a DEAE-cellulose column, and on an anti-␤ 2 -GPI affinity column. To remove any contamination by IgGs, the ␤ 2 -GPI-rich fraction was further passed through a protein A-Sepharose column. The final ␤ 2 -GPI fraction was delipidated by extensive washing with n -butanol.

Isolation and oxidation of LDL
LDL (d ϭ 1.019-1.063 g/ml) was isolated by preparative ultracentrifugation from fresh normal human plasma, as described (39). The LDL was adjusted to 100 g/ml and oxidized with 5 M CuSO4 in PBS for 8 h at 37 Њ C. To terminate the oxidation, 1 mM EDTA was added and dialyzed extensively against PBS containing 1 mM EDTA. The degree of oxidation was estimated as thiobarbituric acid reactive substance (TBARS) value (40) and as migration in agarose electrophoresis.

Ligand blot analysis on a TLC plate
For analytical TLC ligand blot, lipids were spotted on a Polygram silica gel G plate (Machery-Nagel, Duren, Germany) and developed in solvent A or in chloroform-methanol (8:1, v/v, solvent B). Ligand blot analysis was performed, as described previously (35). Briefly, after drying and blocking with PBS containing 1% BSA, the plate was simultaneously incubated with ␤ 2 -GPI and an anti-␤ 2 -GPI Ab (WB-CAL-1 or EY2C9) for 1 h. In case of Cof-22 Ab, ␤ 2 -GPI and the Ab were subsequently incubated for 1 h each. Horseradish peroxidase (HRP)-labeled antimouse IgG or HRP-labeled anti-human IgM was then incubated for 1 h. In each step, a plate was extensively washed with PBS. The color was developed with H 2 O 2 and 4-methoxy-1-naphtol (Aldrich, Milwaukee, WI). On control TLC plates, ligands were separated and stained with I 2 vapor or with a spray of molybdenum blue.

HPLC
The ␤ 2 -GPI-specific ligand, oxLig-2, was purified from the ligand-enriched fraction by a reversed-phase HPLC on a Sephasil Peptide C18 column (4.6 mm ϫ 250 mm; Amersham-Pharmacia Biotech). The scraped band, Band-2, was eluted using a linear gradient of 50-100% solvent C (acetonitrile-isopropanol, 30:70, v/v) against solvent D (water containing 0.2% acetic acid), over 15 min, then 100% solvent C for the following 15 min, at a flow rate of 0.5 ml/min, and absorbance was monitored at 210 nm or 234 nm. The eluate was fractionated every 2 min (1ml/tube). Each fraction was spotted on a TLC plate and subjected to ligand blot analysis with ␤ 2 -GPI and EY2C9 Ab.

Fig. 1. TLC and ligand blot of lipids extracted from LDLs.
Lipids extracted from LDLs were spotted on a silica gel plate and developed in solvent A (A and B) and solvent B (C), respectively. The plate was stained with I 2 vapor and molybdenum blue (A). Ligand blot of lipids extracted from LDLs and incubated with ␤ 2 -GPI and an anti-␤ 2 -GPI Ab (WB-CAL-1 or EY2C9) are shown in B. TLC of the isolated Band-1 and Band-2, in solvent B was followed by ligand blot with ␤ 2 -GPI and EY2C9 Ab (C).

Fig. 2.
Elution profiles of Band-2 in reversed-phase HPLC. Isolated Band-2 was applied to a Sephasil-Peptide column with a linear gradient of 50-100% solvent C (acetonitrile-isopropanol 30/70, v/v) against solvent D (water containing 0.2% acetic acid). Absorbance was detected at 210 nm (A) and 234 nm (B) using a flow rate of 0.5 ml/min. Fractions (1 ml/tube) were collected and subjected to ligand blot with ␤ 2 -GPI and EY2C9 Ab (B, insert). Fraction # 14 from the first HPLC was re-chromatographed on the same HPLC to confirm its purity (C and D).

Methylation of lipid ligands
1-Methyl-3-nitro-1-nitrosomethylguanidine (0.20 g) was added to a mixture of 2 M sodium hydroxide (10 ml) and diethyl ether (10 ml) in an ice bath. The mixture was shaken several minutes and the pale yellow ethereal solution separated was used for methylation. The diazomethane solution (2 ml) was added dropwisely to a solution of lipid ligand (1.0 mg) in diethyl ether (1 ml) at 0ЊC. Each mixture was stored in refrigerator overnight. TLC of the mixture showed complete disappearance of the starting materials. The solvent was removed by blowing air to give the methyl ester as a white amorphous powder.

ELISA for anti-␤ 2 -GPI Ab binding
Anti-␤ 2 -GPI Ab binding was performed as described (35). Briefly, the lipid ligand (50 g/ ml, 50 l/well) was adsorbed by evaporation on a plain polystyrene plate (Immulon 1B, Dynex Technologies Inc., Chantilly, VA) and the plate was then blocked with 1% BSA. A monoclonal anti-␤ 2 -GPI Ab (WB-CAL-1, or EY2C9, 1.0 g/ml, 100 l/well) was incubated in PBS containing 0.3% BSA with ␤ 2 -GPI (15 g/ml) for 1 h. In case of Cof-22 Ab, ␤ 2 -GPI and the Ab were subsequently incubated for each 1 h. Ab binding was probed using HRP-labeled anti-mouse IgG or anti-human IgM. The color was developed with H 2 O 2 and o-phenylenediamine and absorbance was measured at 490 nm. Between each step, extensive washing was performed using PBS containing 0.05% Tween 20.

Preparation of liposomes
Liposomes were prepared as described (42), with the following lipid compositions. Lipid molar ratios of 0, 10, 25, 30, and 50% ligand-containing liposomes were made with DOPC-ligand-   www.jlr.org Downloaded from 8 ϫ 10 5 cells/ml and incubated for 24 h at 37ЊC. The medium was replaced with Celgrosser-P medium (Sumitomo Pharmaceutical Co., Tokyo, Japan). After 2 h preincubation at 37ЊC, 50 l of liposomes (50 nmol lipid/well) with/or without ␤ 2 -GPI (200 g/ml) and WB-CAL-1 (100 g/ml) were added to each well, and the cells were then incubated at 4ЊC for 2 h. The wells were washed with chilled PBS, and the cells were lysed with 1ml of 0.1 N NaOH. An aliquot was taken to determine cellular proteins and radioactivity associated with the cells. Protein concentration was determined using the BCA protein assay reagent (Pierce Chemical Co., Rockford, IL).

Detection of ␤ 2 -GPI-specific ligands
We first detected extracted lipids from native and Cu 2ϩoxdized LDL preparations, by different staining procedures applied to TLC plates, developed in solvent A (Fig.  1A, B). With the I 2 vapor and the molybdenum blue spray, the major change observed due to the Cu 2ϩ -oxidation was a small increase in lysophosphatidylcholine (lysoPC). To define the ligands targeted by ␤ 2 -GPI and an anti-␤ 2 -GPI auto-Ab (i.e., WB-CAL-1 or EY2C9), ligand blot analysis was performed on the TLC plate. Two major bands and a diffuse lipid band were stained with ␤ 2 -GPI and either anti-␤ 2 -GPI auto-Abs. The reactive lipids migrated at similar R f positions to those of cardiolipin and glycolipids, such as galactosylceramide (Gal-Cer) and glucosylceramide (Glu-Cer). The bands detected by ligand blot were not stained with molybdenum blue spray, indicating that they are not phospholipid (Fig. 1A, B). ␤ 2 -GPI ligand-enriched lipids (i.e., Band-1 and -2, indicated by arrows in Fig. 1B) were scraped from the TLC plate (in solvent A) and were subjected to another TLC development in solvent B and subsequent ligand blot with ␤ 2 -GPI and EY2C9 (Fig. 1C). The ligand corresponding to the upper band (Band-1) has already been reported to contain oxLig-1 (35). The lower band (Band-2) was further purified by reversed-phase HPLC.

Purification and characterization of a novel ligand, oxLig-2
The HPLC yielded a novel ligand, we named oxLig-2, from the scraped Band-2 ( Fig. 2A, B). A peak that revealed binding specific for ␤ 2 -GPI and EY2C9 was eluted at approximately 26.7 min (equivalent to 13.4 ml of elution volume). To confirm the purity of oxLig-2 (fraction #14), the fraction was re-chromatographed under the same HPLC conditions (Fig. 2C, D) and subjected to the analysis by LC/MS.

Liposome binding to macrophages
Direct binding of liposomes containing oxLig-1, oxLig-2, or 13-COOH-7KC to mouse macrophages, i.e., J774A.1 cells, was compared with that of liposomes containing DPPS. DPPS-containing liposomes showed binding dependent upon the concentration of DPPS. In contrast, the liposomes containing oxLig-1, oxLig-2, or 13-COOH-7KC displayed relatively weak or negligible binding to the cells (Fig. 6A). Further, we have done inhibition experiments to see whether scavenger receptor(s) is involved in the binding of liposomes containing ␤ 2 -GPI ligands. As shown in Fig. 6B and Table 2, binding of oxLig-1-liposomes to macrophages was inhibited by the addition of poly(I) or fucoidan but not by poly(C). Similar results were obtained with DPPS-liposomes. In contrast, the binding of oxLig-2 or 13-COOH-liposomes was not affected even by the addition of poly(I) or poly(C). These results indicate that the scavenger receptor(s) may primarily be involved in binding of liposomes containing DPPS to macrophages but may only be weakly involved in those with ␤ 2 -GPI ligandcontaining liposomes. Conversely, the uptake of oxLig-1, oxLig-2, and 13-COOH-7KC-containing liposomes by J774A.1 cells was significantly enhanced in the presence of both ␤ 2 -GPI and an anti-␤ 2 -GPI Ab (WB-CAL-1), as compared with control binding of cholesteryl linoleate-liposomes (Fig. 7A-D). In contrast, binding of liposomes was almost completely eradicated by methylation of oxLig-1, oxLig-2, or 13-COOH-7KC (Fig. 7C, D). The ␤ 2 -GPI and anti-␤ 2 -GPI Ab-mediated binding of ligand-containing liposomes was not affected either by poly(I) or poly(C).

DISCUSSION
We previously reported that the major lipid ligand, oxLig-1, specific for ␤ 2 -GPI and anti-␤ 2 -GPI auto-Abs de-   rived from the lipid oxidation of LDL (35), is oxLig-1 (Fig.  4). In the present study, we isolated another ligand (oxLig-2), which we characterized to be keto-dodecanoate. Derivatization of such ligands now demonstrate that an -carboxyl function introduced by Cu 2ϩ -oxidation is critical for an interaction with ␤ 2 -GPI and its ligands.
Foam cell formation is regarded as the hallmark of early atherogenesis, and LDL is the major source of the lipid deposited in foam cells (26). Native LDL, under normal physiological conditions, cannot induce foam cell formation. The binding of modified LDL to scavenger receptors and possibly other cell surface sites on macrophages leads to unregulated cholesterol accumulation and the formation of foam cells with development of atherosclerotic lesions (44,45).
Although the nature of the agents responsible for LDL oxidation in vivo is unknown, several candidates have been proposed (46)(47)(48)(49). LDL oxidized with Cu 2ϩ ion in vitro exhibits the physicochemical and immunological properties of oxLDL extracted from atherosclerotic lesions (28).
The Cu 2ϩ -dependent oxidative products include cholesterol /or oxysterols esterified with 9-or 13-hydroperoxy (or hydroxy)-octadecadienoate, with 9-oxononanoate, or with 9-caboxynonanoate, some of which have also been shown to be present in atherosclerotic plaques (43,50,51). In the present study, we provide evidence that oxLig-2 is keto-dodecanoate (Fig. 4). Methylation of oxLig-2 indicated the presence of a carboxyl function on its acyl chain, but the exact location of the ketone-group cannot be assigned by mass spectrometry. Nevertheless, cholesteryl linoleate, which constitutes one of the major cholesteryl esters of LDL, has four carbons with double bonds susceptible to oxygenation at positions C9, C10, C12, and C13.
It is now well established that anti-␤ 2 -GPI Abs, found in APS patients, bind a complex of ␤ 2 -GPI and negatively charged PLs, such as cardiolipin, PS, and phosphatidic acid (36). In our recent (35) and present studies, however, negatively charged PLs are very minor components of oxLDL.
The flexible loop in the C-terminus and a particular cluster consisting of 14 residues of positively charged amino acid residues in domain V of ␤ 2 -GPI has a critical role for interaction with amphiphilic compounds such as cardiolipin, PS, phosphatidic acid, and phosphatidylglycerol (19)(20)(21)36). Although ␤ 2 -GPI did not bind to cholesterol, 7-ketocholesterol, or cholesteryl linoleate, significant binding was observed to oxLig-1, oxLig-2, and 13-COOH-7KC. Thus, these oxysterol esters having an -carboxylated acyl chain, constitute a new class of an amphiphilic ligand suitable for ␤ 2 -GPI. Further, the observation that methylation of these ligands diminished the ␤ 2 -GPI interaction indicates that a free carboxyl residue is required for the recognition.
As previously described, in vivo uptake of oxLDL via scavenger receptor(s) of macrophages may play a central role in atherogenesis. The term, scavenger receptor(s), refers to a number of different cell-surface proteins that can bind chemically or biologically modified lipoproteins. Various scavenger receptors that bind oxLDL have been found on macrophages, including class A scavenger receptors (56), CD36 (57), human homolog CD68 (58), a lectin-like oxLDL receptor-1 (LOX-1) (59), and Fc␥ receptor (60).
In the present study, we demonstrate that oxidized cholesteryl esters, especially those with 7-ketocholesterol and an -carboxyl function in the acyl chain are ligands for ␤ 2 -GPI and anti-␤ 2 -GPI auto-Abs. Such auto-Abs are found in APS patients and in an animal model, the WB F1 mouse. Furthermore, one major class of biochemically oxidized compounds derived from plasma LDL consists of -carboxylated oxysterols such as oxLig-1 and oxLig-2. Although 13-COOH-7KC is an artificially synthesized compound, it also showed significant binding to ␤ 2 -GPI as well as oxLig-1 and -2.
Most recently, we observed that high levels of circulating immune complexes containing oxLDL, ␤ 2 -GPI, and anti-␤ 2 -GPI auto-Abs in the blood stream, were associated with development of arterial thrombosis in APS patients (unpublished observations). Thus, -carboxylation of oxysterol esters to form the autoantigenic complex of ␤ 2 -GPI bound to oxLDL may have patho-physiologically (etiologically) important roles, especially in development of APS and atherosclerosis.
This work was supported in part by a grant for Scientific Research from the Ministry of Education, Science, Sports, and Culture of Japan and by grants from the Ministry of Health and Welfare of Japan.