Twenty-eight immunoglobulin G (IgG) aCL-positive outpatients were enrolled after clinical referral to the Division of Rheumatology of the University of Rome 'La Sapienza.' All patients were positive for medium-to-high levels of IgG aCL according to our standard enzyme-linked immunosorbent assay (ELISA) (β₂-GPI-dependent) 89 GPL, range 40 to 120). Eighteen patients had APS according to the Sapporo criteria 3 primary (n = 6) or secondary (n = 12) to systemic lupus erythematosus (SLE), and 10 patients had SLE fulfilling American College of Rheumatology criteria 26. The clinical and serological features of the patients are summarised in Table 1. We enrolled 24 healthy subjects as controls (13 female and 11 male; mean age 34 years, range 22 to 52 years). Finally, to establish whether aPL reactivity from patients with infections was different than that from patients with APS and SLE, we selected three sera from 37 hepatitis C virus (HCV) patients previously deemed positive for IgG aCL by ELISA and thin-layer chromatography (TLC) immunostaining 27. None of the healthy subjects or the selected HCV patients experienced arterial or venous thrombosis or recurrent foetal loss. After informed consent was obtained, each subject underwent peripheral blood sample collection and the sera were then stored at -20°C until assayed.

CL (bovine heart) was obtained from Sigma-Aldrich (St. Louis, MO, USA). MCL, dilysocardiolipin (DCL), and hydrogenated ('reduced') CL (HCL) were obtained from Avanti Polar Lipids, Inc. (Alabaster, AL, USA). TLC was performed as previously described 28 to assess the purity of the phospholipid preparations (data not shown). Human β₂-GPI was obtained from Chemicon International (Temecula, CA, USA).

In addition to human sera, the following antibodies were used: rabbit polyclonal anti-human β₂-GPI (Chemicon International), mouse monoclonal anti-transferrin receptor (CD71; BD Pharmingen, San Diego, CA, USA), goat polyclonal anti-voltage-dependent anion channel-1 (VDAC-1)/porin (N18; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), and mouse monoclonal anti-subunit IV of cytochrome c oxidase (COX-IV) (Molecular Probes, now part of Invitrogen Corporation, Carlsbad, CA, USA); alkaline phosphatase-conjugated secondary antibodies (goat anti-human and mouse anti-rabbit IgG) were purchased from Sigma-Aldrich.

Human IgG fractions were first isolated with 33% ammonium sulphate fractionation from plasma of patients with APS and from healthy donors as previously described 2122. Protein concentration was measured with the method of Lowry and colleagues 29, and the purity of the IgG preparations was confirmed by SDS-PAGE.

Briefly, pure phospholipids (50 μg/ml) in ethanol were used to coat microtitre plates by incubation at 4°C overnight. In this way, oxidation of phospholipids occurs during the coating process as previously reported 19. After four washes with phosphate-buffered saline (PBS), plates were blocked for 1 hour at room temperature with PBS containing 10% foetal calf serum (PBS-F). After four washes with PBS-F, plates were incubated for 90 minutes at room temperature with sera samples diluted at 1:50 or human IgG (100 μl of concentrated solutions of 4.8 mg/ml) in PBS-F. Titration of aCL/aMCL-positive sera was performed by serial dilution (1:25 to 1:1,000) in PBS-F by using measurements in triplicate. Moreover, a rabbit polyclonal anti-β₂-GPI was used to detect the levels of β₂-GPI bound to lipids. After four washes with PBS-F, the plates were incubated for 90 minutes at room temperature with secondary anti-human IgG and anti-rabbit IgG (Sigma-Aldrich) diluted to 1:1,000 in PBS-F; after multiple washes, immunoreactivity was developed using the alkaline phosphatase substrate (paranitrophenyl phosphate in ethanolamine). The enzyme reaction was evaluated from the absorbance at 405 nm in a plate reader. To account for the different molecular weights of CL derivatives, we performed ELISA with lipids coated to the plate at the same molarity. Finally, to perform β₂-GPI-independent ELISA, 1% bovine serum albumin (BSA) or 0.25% gelatine was used in the blocking and washing steps. All assays were performed at least in duplicate, and the non-specific binding was evaluated by subtracting the absorbance of non-specific binding of each serum in wells without antigens.

To investigate the specificity of the assay, competitive inhibition tests were performed according to technique described previously 3031. Briefly, serum samples (1:50 in PBS containing 1% BSA) were pre-incubated for 60 minutes at 37°C with increasing amounts of CL or its derivatives dried onto the surface of glass tubes. Subsequently, the tubes were centrifuged (10,000 g for 30 minutes) and the supernatants were tested for reactivity toward CL derivatives.

The immunostaining of TLC plates (Merck, Darmstadt, Germany) was performed as described previously 28, using 2.5 μg of CL, MCL, DCL, and HCL. IgG fractions (2.4 mg/ml) from both APS sera (which had been deemed aCL-positive by standard ELISA screening) and normal human sera were diluted 1:100 or 1:1,000 in PBS containing 0.5% (wt/vol) gelatine. Parallel blots were processed without primary antibody or without antigen as control for non-specific reactivity.

BSA, gelatine, and human β₂-GPI were run on 12% SDS-PAGE and electro-transferred to nitrocellulose membrane (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Membranes were blocked with PBS containing 5% defatted dry milk and then probed with rabbit polyclonal anti-β₂-GPI antibodies. Antibody binding was detected using alkaline phosphatase-conjugated secondary antibodies and visualised with a kit containing chromogenic alkaline phosphatase substrate (Bio-Rad Laboratories, Inc.) according to manufacturer's instructions. The reaction was stopped by washing in distilled water.

Cellular indirect immunofluorescence was carried out to analyse CL expression on the PM of HUVECs, which were isolated by collagenase perfusion from normal-term umbilical cord veins as described previously 32. Cells were cultured in M-199 medium (Sigma-Aldrich) supplemented with 20% foetal calf serum (FCS), 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin at 37°C in a humidified atmosphere of 5% CO₂. HUVECs were grown to 60% to 70% confluence, transferred at 5 × 10⁶ cells per well on glass coverslips, either untreated or treated with 20 ng/ml TNF-α and 10 μg/ml cycloheximide for 16 hours 3334, and fixed in PBS containing 4% formaldehyde for 1 hour at 4°C. Apoptosis was evaluated by morphologic analysis and by propidium iodide staining according to Nicoletti and colleagues 35. After three washes with PBS, cells were incubated for 1 hour at 4°C with purified human IgG from APS patients (aCL- and aMCL-positive) in PBS containing 1% BSA. Alternatively, after previous absorption with CL or MCL as reported above, cells were incubated for 1 hour at 4°C with purified human IgG from APS patients. Fluorescein isothiocyanate-conjugated anti-human IgG (γ-chain specific; Sigma-Aldrich) were then added and incubated at 4°C for 30 minutes. After a washing with PBS, fluorescence was analysed with an Olympus U RFL microscope (Olympus, Tokyo, Japan). In parallel experiments, cells were directly stained before formaldehyde fixation. Alternatively, cells were processed for a second formaldehyde fixation immediately after the incubation with purified IgG and before the addition of the secondary antibody. Neither fixation procedure affected the staining on the cell surface 21.

Subcellular fractions were isolated from HUVECs as previously reported 22. Briefly, control untreated cells or cells treated with 20 ng/ml TNF-α and 10 μg/ml cycloheximide for 16 hours were rinsed with cold isolation buffer (0.25 M mannitol, 1 mM EDTA, 10 mM K-HEPES, 0.2% BSA, pH 7.4) containing a cocktail of protease inhibitors (Sigma-Aldrich), resuspended in 1 ml of the same buffer, and homogenised vigorously. After a brief centrifugation at 600 g in the cold isolation buffer, pellet and supernatant were combined, rehomogenised, and centrifuged at 800 g for 10 minutes at 4°C. The pellet was discarded and the supernatant was further centrifuged at 10,000 g for 10 minutes at 4°C. The pellets (P10) were washed three times with assay buffer (0.12 M mannitol, 0.08 M KCl, 1 mM EDTA, 20 mM K-HEPES, pH 7.4, containing a cocktail of protease inhibitors) and then resuspended in the same buffer. The supernatant was further centrifuged at 100,000 g for 1 hour at 4°C to obtain the cytosolic supernatant (S100) and the light membrane pellet (P100). The latter was dissolved in assay buffer, normalised for protein content by Bio-Rad protein assay (Bio-Rad Laboratories, Inc.), split into two aliquots, and analysed for phospholipid composition by TLC and for the presence of mitochondrial and endosomal contaminants by Western blotting. Phospholipids were extracted according to the technique described by Folch and colleagues 36 and separated by TLC by using high-performance TLC (HPTLC) silica gel 60 (10 × 10) plates (Merck). Chromatography was performed in chloroform/methanol/acetic acid/water (100:75:7:4) (vol/vol/vol/vol). After exposure to iodide vapours, the bands comigrating with the CL and MCL standard were scraped, eluted from silica with chloroform/methanol (2:1) (vol/vol), and dried under nitrogen. Phospholipids were run using methanol/chloroform/NH₃ (7:13:1) and stained with copper acetate. Alternatively, light membranes were diluted with assay buffer containing protease inhibitors and adjusted to a final protein concentration of 0.5 to 1 mg/ml with concentrated SDS sample buffer. Protein samples were separated by SDS-PAGE and blotted in PBS containing 0.05% Tween-20 with the following antibodies to exclude the presence of mitochondrial contaminants: mouse monoclonal anti-transferrin receptor (CD71), goat polyclonal anti-VDAC-1/porin, and mouse monoclonal anti-COX-IV. Blots were visualised by chemiluminescence reaction by using the ECL Western detection system (Amersham Biosciences, now part of GE Healthcare, Little Chalfont, Buckinghamshire, UK). Protein loading was evaluated by india ink staining 37.

Statistical analysis of data was carried out using Mann-Whitney's U test for comparison of means between different groups of subjects. Wilcoxon paired test was used to compare differences between aCL derivative profiles in each group of patients and controls. Correlation analysis was carried out by the Spearman test. P less than 0.05 was considered statistically significant.

The levels of serum antibodies reacting with CL, MCL, DCL, and HCL in patients and controls were quantified by ELISA, and the overall data are shown in Figure 1. Similar results were obtained when purified human IgG were used instead of sera (data not shown). The average immunoreactivity to CL and MCL was significantly higher than that for DCL and HCL, both in APS (P < 0.0001) and SLE patients (P < 0.001) (Figure 1). In addition, the same reactivity profile was observed when CL derivatives were coated onto plates at the same molarity, even if under these conditions the amount (weight) of coated lipids may have been slightly different. All three patients with HCV positive for aCL IgG were also positive for aMCL. Thus, our results indicated that, at least in solid-phase immunoassays, a key derivative of CL such as MCL represents a lipid easily recognised by sera autoantibodies. To study this novel observation in more detail, we performed titrations with representative sera showing strong reactivity for both CL and MCL (Figure 2a). The concentration dependence of the antibody titre indicated that binding to MCL saturated as efficiently as that to CL, whereas reactivity toward both DCL and HCL remained very low and did not saturate at all (Figure 2a). To further characterise the IgG specificity of binding, three sera positive for both aCL and aMCL were tested in ELISA after absorption with increasing amounts of CL and its derivatives. As shown in Figure 2b, pre-absorption with both CL and MCL significantly decreased the binding of aCL to CL. Similar results were observed when binding of aMCL to MCL was analysed after absorption with CL and MCL (Figure 2c). In contrast, when sera were pre-absorbed with either DCL or HCL, no significant competition in the binding of both aCL and aMCL was observed (Figure 2b,c). These findings suggest that CL and MCL present identical epitopes, or at least overlapping epitopes.

All the aCL/aMCL-positive sera of patients with APS and SLE were found to react also with β₂-GPI as detected by standard ELISA (data not shown). Conversely, none of the controls (healthy subjects and patients with HCV) was positive for anti-β₂-GPI.

We observed that the isolated β₂-GPI protein could bind to MCL other than CL (Figure 3a). Intriguingly, the protein also showed a significant binding to DCL, with an affinity that appeared to be comparable with that for CL. So, although β₂-GPI attached itself to both MCL and CL, it equally reacted with DCL in vitro (Figure 3a), whereas APS sera displayed only low reactivity toward DCL (Figures 1 and 2). In this respect, it is interesting to note that the reactivity of SLE and APS sera toward other negatively charged lipids such as lyso(bis)phosphatidic acid (LBPA) 2738 could be mediated by the relatively non-specific interaction that serum proteins like β₂-GPI may have with chemically different phospholipids sharing a negative charge.

To determine whether aCL/aMCL binds directly to lipids, we performed ELISA in the absence of serum proteins (for example, FCS) or in the presence of other proteins like BSA or gelatine. Clearly, patients' autoantibodies retained high reactivity toward both CL and MCL in the absence of FCS and irrespective of the exogenous proteins used for blocking the plate matrix (Figure 3b). To exclude the possibility that BSA or gelatine was contaminated with the serum protein β₂-GPI that avidly binds CL and its derivatives (Figure 3a), we performed Western blot analysis with a specific anti-β₂-GPI antibody. The results did not show detectable levels of β₂-GPI in the samples of BSA or gelatine used in the experiments (data not shown).

In another approach, we tested the immunoreactivity of aCL/aMCL-positive sera toward pure lipids separated by TLC in the complete absence of exogenous proteins that might interfere with the autoantibody reactivity with phospholipids. TLC immunostaining of five representative aCL/aMCL-positive sera showed strong reactivity to a combination of CL and MCL (Figure 3c). No sera showed reactivity against DCL and HCL. Next, we performed experiments to assess the possible contribution of β₂-GPI toward aCL and aMCL reactivity in TLC immunostaining. Considering that β₂-GPI is normally present in concentrations of approximately 200 μg/ml in human sera, we performed immunostaining on TLC with sera that had been diluted to 1:1,000. With such a dilution, β₂-GPI levels would fall below the minimum requirement (>0.5 μg/ml) for allowing binding of aCL to anionic phospholipids 39. As previously demonstrated 4041, TLC immunostaining showed a reactivity with CL and MCL up to a dilution of 1:1,000 (Figure 3c, lane 4), thereby excluding a major contribution of serum proteins to the observed immunoreactions to CL derivatives.

To evaluate whether apoptosis may represent a possible trigger for autoantibody production toward CL and MCL, human IgG fractions from APS patients were used to analyse the distribution pattern of CL (and MCL also) on the surface of apoptotic HUVECs. Reactivity showed specificity for CL and MCL, given that no significant reactions were detected by TLC immunostaining between IgG fractions and other phospholipids such as phosphatidylserine, phosphatidylinositol, and LBPA. No antinuclear reactivity of IgG fractions was observed by standard indirect immunofluorescence on Hep2 cells. To promote apoptosis, we used treatment with TNF-α plus cycloheximide under conditions that were previously shown to induce high levels of endothelial cell death 3334. Fluorescence microscopy analysis revealed that human IgG fractions displayed a stronger binding to the surface of HUVECs undergoing apoptosis as compared with untreated cells (Figure 4a,b). The enhanced staining onto the PM appeared uneven, indicating that the reactivity was concentrated in localised areas on or near the cell surface. Most likely, these areas coincided with PM patches where active membrane traffic and re-modelling were most intense (for example, blebbing precursors, which subsequently give rise to apoptotic bodies) (Figure 4b). A virtual absence of immunolabelling was observed across internal membranes, suggesting that intracellular CL was somehow shielded from autoantibody reactivity. Interestingly, a very low staining was observed after previous absorption with CL or MCL in both untreated cells (Figure 4c and 4d, respectively) or cells treated with TNF-α plus cycloheximide (not shown). As a negative control, human IgG from healthy subjects showed negligible surface staining of HUVECs, either untreated or after TNF-α and cycloheximide treatment (not shown).

To further investigate the above observations, we isolated light membrane fractions containing PMs of apoptotic and untreated HUVECs. Phospholipids were extracted and the presence of CL and MCL was verified by HPTLC (Figure 4e). Strong bands corresponding to CL and MCL were identified only in the membrane fractions of apoptotic cells. The PM enrichment of the light membrane fractions was confirmed by Western blot analysis, which showed strong reactivity against the membrane protein marker transferrin receptor, with the virtual absence of mitochondrial (VDAC-1/porin and COX-IV) contaminants (Figure 4f). Hence, the results obtained with HUVECs undergoing apoptosis were similar to those reported previously with U937 cells 2122 and indicated that CL and MCL become surface-exposed and much more accessible to autoantibody recognition after induction of apoptosis.