Dr Pojen P. Chen Medicine/Rheumatology, 167022, UCLA, Room 32-59, 1000 Veteran Ave., Los Angeles, CA 90095-1670, U.S.A. e-mail: firstname.lastname@example.org.
Antiphospholipid antibodies (aPL), including antibodies detected in anti-cardiolipin (aCL) enzyme-linked immunosorbent assays and in lupus anticoagulant (LA) tests, are strongly associated with recurrent thrombosis and recurrent fetal loss, i.e. the antiphospholipid syndrome (APS). Although recent studies suggest that most APS-associated aCL are directed against the phospholipid (PL)-binding plasma protein β2-glycoprotein 1 (β2GP1), the precise nature of aCL binding specificities remains controversial. To address the issue of aCL specificity we generated five new monoclonal IgG aCL from two patients with APS. Characterization of these five aCL, as well as two previously published IgG aCL, revealed three patterns of reactivity: (1) four antibodies reacted strongly with human β2GP1-cardiolipin (CL) complexes and weakly with human β2GP1 alone; (2) two antibodies recognized bovine β2GP1, but not human β2GP1; (3) one antibody reacted with complexes of human β2GP1 and CL, but not with human β2GP1 alone. Only one monoclonal displayed weak LA activity. These patient-derived IgG monoclonal antibodies, and additional ones to be generated, may help define varying species of antibodies detected in aCL assays and identify the specific antibodies that may be pathogenic.
Patients with recurrent thrombosis and/or pregnancy loss often have serum antibodies detected in anticardiolipin antibody (aCL) enzyme-linked immunosorbent assay (ELISA) and/or lupus anticoagulant (LA) tests, as defined by their ability to prolong certain in vitro phospholipid (PL)-restricted blood clotting tests (Harris et al, 1983; Hughes, 1993; Khamashta & Hughes, 1994). LA and aCL are generally referred to as antiphospholipid antibodies (aPL); and the association of thrombosis and/or fetal loss with aCL and LA is recognized as the antiphospholipid syndrome (APS) or the Hughes syndrome (Khamashta & Asherson, 1995).
Although anti-β2GP1 antibodies are likely to represent a major species of autoantibodies in APS, the conventional aCL and LA assays may detect additional autoantibodies which may be pathogenic. Such antibodies may include: autoantibodies against prothrombin, protein C or protein S, and/or complexes of those proteins with PL (Oosting et al, 1993; Rao et al, 1996). Thus, to better define antibodies detected in the conventional aCL assay, we continued our efforts to generate and characterize IgG monoclonal aCL from APS patients. Patient-derived monoclonal antibodies will be useful in cataloguing all characteristic antibodies in patients with APS and identifying the pathogenic antibodies, and eventually in the elucidation of pathogenic mechanisms.
METHODS AND MATERIALS
Patient 1 has been described previously (Olee et al, 1996). Briefly, this patient is a 19-year-old woman with primary APS. She experienced a deep venous thrombosis at age 16. For the past 2 years she has continuously had high titres of IgG aCL, but displays fluctuating LA activities. There has been no clinical or laboratory evidence of systemic lupus erythematosus (SLE).
Patient 2 is a 17-year-old Caucasian male. He was in good health until age 12 when he developed a right common femoral vein thrombosis and bilateral pulmonary emboli. Coagulation studies demonstrated prolongation of the prothrombin time of 15.1 s (mean normal 10.3 ± 2 s), and a partial thromboplastin time of 74.9 s (mean normal (26.7 ± 5 s) which failed to correct a 1:1 mix with normal plasma (40.9 s). The patient's dilute Russell's viper venom time (dRVVT; LA-Screen, CenterChem, Ct.) was 83.6 s versus a control of 36.4 s, with a ratio of 2.30. A dRVVT-confirmatory test (LA-Confirm, Gradipore, North Ryde, Australia) was performed in which the ratio of the dRVVT/(dRVVT confirm) was equal to 2.16. The patient had normal protein C, protein S, antithrombin III levels, a normal platelet count and no factor V Leiden mutation. He had a heterozygous factor VII deficiency. He had additional findings suggestive of SLE (fulfilling three of the 1982 revised ACR criteria for SLE) (Tan et al, 1982); a positive ANA (>1/640), double-stranded DNA binding of 10% (normal 10%), mildly depressed C3 and C4, and the nephrotic syndrome secondary to biopsy-proven membranous nephropathy. Treatment was begun with intravenous heparin followed by coumadin. His renal disease markedly improved on prednisone (initially 60 mg/d, now 10 mg every other day). Although he has had no thrombotic episodes since presentation, his LA and aCL have remained strongly positive, 8–13 standard deviations above the mean of normal for aCL, and he has continued on coumadin to the present time. His diagnosis is secondary APS. The blood sample for the present study was obtained while the patient was on full-dose anticoagulation and 10 mg prednisone every other day.
Generation of monoclonal IgG aCL
This was done as described previously (Olee et al, 1996). Briefly, peripheral blood mononuclear cells (PBMC) were isolated from patients, transformed with Epstein-Barr virus (EBV), and plated out in 96-well plates. The supernatants were screened for IgG aCL by ELISA as described in detail previously (Harris et al, 1994). Briefly, plates were precoated with cardiolipin (CL; 50 μg/ml in ethanol; Avanti), air-dried, and then blocked with 10% bovine serum (BS) in phosphate-buffered saline (PBS). Supernatants were distributed to wells, and bound human IgG was detected with affinity-purified horseradish peroxidase-labelled goat anti-human IgG (γ-chain specific) and the substrate TMB/H2O2. Results were read at a wavelength of 450 nm against a background of 650 nm on the Thermomax plate reader (Molecular Devices Inc., Sunnyvale, Calif.).
Cells from each positive well were subcloned twice at one cell per well to yield monoclonal cell lines. Thereafter, each monoclonal EBV transformed cell line was fused with human–mouse heterohybridoma K6H6/B5 as described (Lu et al, 1993). Again, cells with aCL activity were subcloned twice at one cell/well. Occasionally, when a particular EBV-transformed aCL cell line could not be subcloned at one cell/well after repeated trials, cells were fused directly, and the resultant hybridomas were then subcloned twice at one cell/well. To further ensure that each resultant aCL was monoclonal, the light chain isotypes and IgG subclasses of each aCL were determined with ELISA using isotype and subclass-specific reagents.
Affinity purification of antibodies
Initially, each monoclonal antibody was purified from culture medium containing 10% fetal calf serum by two sequential affinity columns, HiTrap Protein G column (Pharmacia, Piscataway, N.J.) and goat anti-human IgG. Purified IgG was dialysed against PBS or Tris buffer (50 mm Tris/150 mm NaCl, pH 7.5) or Hepes buffer (20 mm Hepes, 150 mm NaCl, pH 7.4).
Subsequently, to avoid the need for the second affinity column, aCL hybridomas were switched to a serum-free culture medium before isolation. The concentrations of purified IgG were determined in ELISA using human IgG as standard.
Immunological characterization of monoclonal IgG aCL
Purified aCL were first studied for their serum dependence. Briefly, plates were precoated with CL in ethanol (or with ethanol alone as background control) and blocked with 0.25% gelatin in PBS. Test samples were diluted 1:1 with PBS containing either 10% BS or 1% bovine serum albumin (BSA; Sigma Chemical Co., St Louis, Mo.), and distributed to wells in duplicate. Binding to the ethanol-treated wells was subtracted from binding to CL-coated wells (Harris et al, 1994).
Second, monoclonal aCL were analysed by ELISA for their reactivities with the following antigens: chicken ovalbumin (OVA), type VI collagen, bovine thymus single-stranded DNA (ssDNA), and keyhole limpet haemocyanin (KLH); all were from Sigma. Briefly, plates were coated with the indicated antigens at 2 μg/ml in PBS and blocked with 0.25% gelatin/PBS; and monoclonal aCL were used at the indicated concentrations in PBS containing 5% BS.
Third, monoclonal aCL were studied for their reactivities with CL in the presence of human serum (HS) or human β2GP1, or with human β2GP1 alone. Human β2GP1 was purified according to the published protocol (Roubey et al, 1995). For binding to CL in the presence of HS or human β2GP1, plates were coated with CL, and then blocked with 0.25% gelatin. Thereafter, test IgG in PBS containing either 5% HS (which had been depleted of human IgG) or 20 μg/ml of β2GP1 were added to wells. For binding to β2GP1 alone, plates were coated with human β2GP1 at 20 μg/ml in PBS and then blocked with 0.25% gelatin.
Fourth, for BS-dependent aCL, they were analysed for reactivity with bovine β2GP1. Bovine β2GP1 was kindly provided by Dr Hisao Kato (Osaka, Japan) (Kato & Enjyoji, 1991), and was used at 5 μg/ml in borate-buffered saline (BBS; 0.2 m borate, 0.075 m NaCl, pH 8.4) to coat ELISA plates. After an overnight incubation at 4°C, plates were blocked with BBS containing 0.5% BSA and 0.4% Tween-80 (Sigma), and antibody samples diluted in blocking buffer were applied to wells. Bound antibody was quantitated using alkaline phosphatase-conjugated goat anti-human IgG (Zymed Laboratories, South San Francisco, Calif.) and p-nitrophenyl phosphate. Diluted antibodies were tested in triplicate with coefficients of variation <10%.
The lupus anticoagulant test
LA activity was determined by a modified dRVVT. 10 μl of test IgG was incubated with 40 μl of normal pooled plasma for 3 min at 37°C. Then 50 μl of the prewarmed dRVVT reagent (LA-Screen) was added to initiate clotting. Test samples were performed in duplicate with a ST4 coagulizer (Diagnostica Stago, Parsippany, N.J.). If the ratio of the clotting time of a test IgG to that of a normal IgG exceeded 1.2 (Exner et al, 1978), its LA activity was confirmed by a confirmatory test in which the assay was repeated exactly as above except that the dRVVT-confirm reagent (LA-Confirm) contained excess PL to neutralize the LA effect. An LA activity was confirmed positive if the clotting time ratio of dRVVT/dRVVT-confirm was also >1.2.
Generation of five monoclonal IgG aCL from two APS patients
From patient 1, two monoclonals were obtained, designated IS3 and IS4. Combined with the two previously reported ones from this patient (i.e. IS1 and IS2) (Olee et al, 1996), four IgG aCL were derived from this patient. Subsequently we obtained three monoclonal IgG aCL from patient 2; they were designated CL1, 15 and 24.
It should be noted that all five new human IgG monoclonal aCL had each been subcloned at least twice at one cell/well. Each aCL had only one light chain isotype and one IgG subclass. Specifically, IS3, CL15 and CL24 had κ light chains, while IS4 and CL1 had λ light chains. For heavy chains, all five aCL were of the γ3 subclass; this differed from previous IS1 and IS2, both of which were γ1 subclass together with κ light chains.
Immunological characterization of APS-derived monoclonal IgG aCL
A major characteristic of aCL from APS patients is their dependence on serum factors in their binding to CL, as compared with aCL from individuals with infections which bind to CL directly (Sammaritano et al, 1992). Accordingly, monoclonal aCL were first examined for serum-dependence in their binding to CL. The previously reported aCL IS1 and IS2 were included for comparison (Olee et al, 1996). Fig 1 shows that, in the presence of BS, all seven aCL bound to CL (filled bars), whereas polyclonal human IgG and a randomly selected human monoclonal IgG control (i.e. H2) did not. All antibodies were used at 1 μg/ml, except for CL15 and CL24 which were used at 2 μg/ml. These antibody concentrations had been found to be the optimal concentrations, and were used through most of the subsequent studies, except when specific concentrations were indicated. In contrast, in the absence of BS, the binding to CL was either lost completely (such as IS1, IS2, IS4 and CL24, open bars in Fig 1) or reduced substantially (such as IS3, CL1 and CL15). These data demonstrated that all seven APS-derived aCL displayed serum-dependence in their binding to CL, and therefore were likely to be specific for APS.
Second, monoclonal aCL were analysed for their binding specificities by comparing their binding to CL and four unrelated antigens, including OVA, collagen, ssDNA and KLH. As can be seen in Fig 2, IS1, IS2 and CL15 bound only to CL, whereas each of the four remaining monoclonal aCL displayed weak reactivities toward various irrelevant tested antigens. Overall, the data indicated that all seven APS-derived IgG aCL were more specific for CL plus bovine serum.
Third, monoclonal aCL were analysed for their binding to CL in the presence of BS, HS or human β2GP1. Fig 3 shows that IS1 and IS2 bind to CL only in the presence of BS, but not HS or human β2GP1. In contrast, the five remaining aCL bound to CL in the presence of BS, or HS or human β2GP1. To detect possible weak binding by IS1 and IS2 to the CL–human β2GP1 complexes, both were analysed at a series of two-fold higher concentrations (from 1 to 64 μg/ml). Even at 64 μg/ml both IS1 and IS2 did not react with the CL–human β2GP1 complexes (data not shown). Since none of these seven monoclonal aCL reacted with only CL-HS, it was unlikely that any of these aCL recognized complexes of CL with non-β2GP1 serum factors (such as prothrombin, protein C or protein S).
Thereafter, monoclonal aCL were examined for direct binding to human β2GP1. As can be seen in Fig 4, IS3, IS4, CL1 and CL24 bound to β2GP1, whereas the other three aCL did not. However, the binding of these four positives to β2GP1 was much weaker than their binding to CL in the presence of BS. Since CL15 at 2 μg/ml bound strongly to the CL–human β2GP1 complexes, to detect possible weak reactivity with human β2GP1, CL15 was analysed at 32 and 64 μg/ml for direct binding to human β2GP1. The result showed that, even at 64 μg/ml, CL15 did not react with human β2GP1 (data not shown). On the other hand, because IS1 and IS2 bound to CL only in the presence of BS, but not HS, it was possible that both aCL might be specific for bovine β2GP1. As is evident in Fig 5, both BS-dependent IS1 and IS2 bound directly to bovine β2GP1.
Lupus anticoagulant activities of monoclonal aCL
Monoclonal aCL were analysed for their LA activity by dRVVT. Considering that the human serum IgG concentration is about 10 g/l and that a particular IgG antibody species may account for maximally 1% of total serum IgG, monoclonal aCL were used in the 0.1 g/l range. As shown in Table I, experiments 1 and 2 show that IS3 and CL24 displayed borderline LA activities, prolonging the clotting time in the in vitro PL-restricted coagulation test by 21% and 17%, respectively. When these two aCL were subjected to the dRVVT confirmatory test, both were negative. Since the negative results might be due to insufficient IgG concentrations in both aCL preparations, they were concentrated by 7–10-fold (resulting in 1 g/l for IS3 and 0.5 g/l for CL24) and tested again. Experiment 3 shows that IS3 remained negative, whereas CL24 was weak positive. Taken together with the previous LA data for IS1 and IS2 (Olee et al, 1996), 6/7 monoclonal aCL from two APS patients did not possess any LA activities.
Table 1. Table I. LA activity of monoclonal aCL.
* Each figure represents an average of duplicate tests with very good consistency.† Each figure represents the ratio of the clotting time of a test IgG to that of a normal IgG.‡ N/A, not applicable, as only the samples positive for the dRVVT test (i.e., having a ratio 1.2) require dRVVT confirmatory test. Each figure represents the clotting time ratio of dRVVT/dRVVT-confirm for the same test sample. LA activity of a test IgG was confirmed positive if this ratio was also >1.2.
To better define aCL in APS, we generated five new human monoclonal IgG aCL. These current five aCL and the two previously reported IgG aCL were characterized for their binding properties and anticoagulant activities. The results are summarized in 2Table II. Based on the species origins of β2GP1 and the presence or absence of CL, these seven IgG aCL can be classified into three groups. The first group includes IS1 and IS2. These aCL bind to CL in the presence of BS, but not HS; and they bind to bovine β2GP1, but not to human β2GP1. Thus, these two aCL are species-specific, similar to those anti-β2GP1 antibodies observed by Arvieux et al (1996). The second group includes IS3, IS4, CL1 and CL24. These aCL bind strongly to the CL–β2GP1 complexes, but weakly to human β2GP1 alone. The third group includes only CL15. It reacts with CL in the presence of human β2GP1, but does not bind to human β2GP1 alone. Combined, these data from two APS patients clearly show that there are at least three distinctive species of aCL: one is specific for bovine β2GP1, one for CL–β2GP1 complexes and human β2GP1 alone, and one for CL–β2GP1 complexes. Most of these three kinds of aCL displayed no LA activity, except for the human β2GP1-reactive CL24 which had weak LA activity.
Table 2. Table II. Summary of the immunological and functional properties of seven monoclonal IgG aCL from two patients*.
* For each aCL, its reactivity with CL/BS is assigned arbitrarily as ++, and its binding reactivities with other antigens are expressed in relative term in comparison to that with CL/BS: +++, stronger; ++, equal; +, weaker; ±, borderline reactivity; −, no reactivity.† nd, not done.
Including the five present IgG aCL/anti-β2GP1, 15 monoclonal aCL from APS patients have been studied to date (Hasegawa et al, 1994; Ichikawa et al, 1994; Harmer et al, 1995; Olee et al, 1996). Of these, eight were of the IgM isotype and seven IgG. When five of these seven IgM aCL and all seven IgG aCL were examined for binding to human β2GP1, five IgM and four IgG could bind to human β2GP1 alone (Ichikawa et al, 1994). On the other hand, when these 15 monoclonal aCL were examined for LA activities, two IgM and one IgG aCL displayed LA activity (Hasegawa et al, 1994; Ichikawa et al, 1994; Harmer et al, 1995; Olee et al, 1996). Interestingly, when LA of some monoclonal aCL was done in the presence of 2 μmβ2GP1, three of the four LA-negative IgM aCL (i.e. EY1C8, EY2C9 and GR1D5) were found to inhibit the RVV-induced thrombin generation in diluted plasma (Takeya et al, 1997). Taken together with the present data, a majority of aCL (9/12 tested monoclonals) could react directly with β2GP1 and therefore are anti-β2GP1 antibodies; and most aCL (12/15 tested monoclonals) do not possess LA activity.
Recently, a murine model was established which demonstrated that aCL can promote and/or sustain thrombosis (Pierangeli et al, 1994). Specifically, the thrombus was larger and persisted longer in mice treated with affinity-purified aCL from an APS patient with thrombosis than in mice treated with normal human IgG (Pierangeli et al, 1995). Importantly, using this murine model, aCL/anti-β2GP1 IS2 was found to be thrombogenic (Olee et al, 1996). Therefore all six remaining IgG monoclonal aCL/anti-β2GP1 will need to be studied with this murine model to determine their thrombogenic potentials.
Various mechanisms have been proposed to explain the thrombogenic properties of aPL in APS patients. First, polyclonal and monoclonal aPL (i.e. the TM1G2 and EY1C8 IgM aCL/anti-β2GP1) were found to bind to endothelial cells (EC) and to induce expression of adhesion molecules (including E-selectin, ICAM-1 and VCAM-1) and monocyte adhesion (Simantov et al, 1995; Del Papa et al, 1997). These antibody activities were dependent on β2GP1. Along this line, it was reported that incubation of monocytes with aCL led to surface expression of tissue factor (TF); and that TF expression on monocytes was increased in APS patients, particularly those positive for IgG aCL (Kornberg et al, 1994; Cuadrado et al, 1997). Moreover, it was shown very recently that incubation of PBMC from normal individuals with either IgG anti-β2GP1 or three β2GP1-reactive monoclonal IgM aCL (i.e. EY1C8, EY2C9 and GR1D5) induced TF expression and procoagulant activity (Amengual et al, 1998). Combined, these data suggest that aPL may interact with EC and induce a TF-dependent procoagulant state. Second, some β2GP1-reactive aPL were shown to enhance the binding of β2GP1 to EC and trophoblasts, and thus caused the loss of surface annexin V, which was hypothesized to be a physiological anticoagulant by covering up the procoagulant PL surface (Rand et al, 1997). Third, some aPL were shown to interfere with the regulation of coagulation by inhibiting activation of protein C and/or function of the activated protein C (Cariou et al, 1988; Marciniak & Romond, 1989; Malia et al, 1990; Borrell et al, 1992; Smirnov et al, 1995; Galli et al, 1998). In the future it will be important to study all seven IgG aCL in the three aforementioned systems. Hopefully, the results from these studies, together with those from the in vivo murine thrombosis model, may help us define the phatogenic aPL and delineate the underlying mechanisms of the aPL-related thrombosis in APS.
The authors thank Dr Hisao Kato (Osaka, Japan) for kindly providing us with purified bovine β2GP1, and gratefully acknowledge the secretarial assistance of Ms Leslie Mehana in the preparation of the manuscript. This work was supported by grants AR42506 and AI32243 from the National Institutes of Health.