Our objectives were to obtain monoclonal anti-endothelial cell antibodies (AECA) from systemic lupus erythematosus (SLE) patients, to characterize their antigen specificity, and their capability to induce a pro-inflammatory and pro-adhesive endothelial phenotype, and to investigate the mechanism of endothelial cell (EC) activation in vitro. Monoclonal IgG AECA were generated by hybridoma formation with human SLE B cells. Antigen specificity was characterized by immunoblotting with enriched cell membrane fractions, by cytofluorimetry and by cell solid-phase ELISA. Endothelial activation was evaluated by measuring increases in U937 cell adhesiveness, adhesion molecule (E-selectin and ICAM-1) expression and IL-6 production. In addition, mechanisms of endothelial activation were investigated by assessment of NF-κB by measuring the loss of its inhibitor I-κB. mAb E-3 bound live EC and recognized a 42 kDa EC membrane protein, it enhanced U937 adhesiveness, E-selectin and ICAM-1 expression and IL-6 production, and caused the loss of I-κB. We conclude this is the first in vitro demonstration that a human monoclonal AECA from a SLE patient reacts with a constitutive endothelial membrane antigen and induces a pro-inflammatory endothelial phenotype through NF-κB activation.
Anti-endothelial cell antibodies (AECA) are a common finding in a number of systemic autoimmune diseases (reviewed in 1). Their occurrence has been correlated with both disease activity and vessel wall inflammation (reviewed in 2,3). The above-mentioned associations and the lack of AECA in conditions in which vessel damage was clearly induced by other effector mechanisms (such as circulating immune complexes in essential mixed cryoglobulinemia ) suggest a potential pathogenetic role for AECA.
In vitro studies have demonstrated the ability of AECA to (i) fix complement, at least in systemic lupus erythematosus (SLE) (5), (ii) mediate antibody-dependent cellular cytotoxicity by NK cells (4), and, particularly, (iii) activate endothelial cells (EC) by up-regulating adhesion molecule expression and the secretion of both pro-inflammatory cytokines and chemokines in primary autoimmune vasculitis (6,7) and in scleroderma (8). The demonstration that IgG AECA+ fractions from active SLE and thrombotic thrombocytopenic purpura (TTP) patients display comparable activity suggests that the AECA ability to activate EC might be a general characteristic of these auto-antibodies (9–11).
However, these studies present several problems. (i) The sera and polyclonal IgG fractions from patients with SLE, for example, might contain auto-antibodies directed against molecules that have been described to adhere to endothelial membrane surfaces such as DNA–histone complexes and phospholipid-binding proteins [such as β2-glycoprotein I (β2GPI)] (12–15). This makes it difficult to rule out a possible role for other auto-antibodies in activating EC. In addition, sera from patients suffering from active systemic autoimmune diseases may have very high levels of circulating cytokines and it is possible that free or IgG-bound cytokine contamination could explain these data, despite very careful preparation of the polyclonal antibodies (16,17). Finally, the anti-endothelial activity contained in these polyclonal preparations is apparently directed against a variety of constitutive antigens (18–21); thus, although the pro-inflammatory effects can be demonstrated, it remains impossible to attribute them to a specific auto-antigen.
The use of single specificity mAb isolated from patients with SLE would solve these problems. The aims of this study were to: (i) obtain mAb which display reactivity against human EC from the B cells of SLE patients, (ii) characterize the endothelial auto-antigen, (iii) investigate whether such a binding could also mediate endothelial activation comparable to that found with the polyclonal AECA preparations and (iv) investigate the signaling pathways responsible for EC activation.
Blood samples (30–50 ml) were taken from patients with active SLE with clinical evidence of skin or renal vasculitis and who fulfilled the American Rheumatology Association revised criteria for diagnosis (22). The patients sampled were all regularly attending the Dermatology, Rheumatology and Renal Clinics at Leeds General Infirmary.
Transformed human microvascular endothelial cells (HMEC-1) were a gift from the Centre for Disease Control (Atlanta, GA) (23). The mouse/human heteromyeloma cell line (CB-F7) was a gift from Dr Sigbert Jahn (24). HMEC-1 cells were grown in MCDB 131 medium (Clonetics, San Diego, CA) supplemented with 10 ng/ml epidermal growth factor (Sigma-Aldrich, Gillingham, UK), 1 μg/ml hydrocortisone and 10% FCS (Gibco/BRL Life Technologies, Paisley, UK). All adherent cells were propagated in 75 or 225 cm2 flasks (Corning, High Wycombe, UK). The cells for membrane preparation were mechanically detached to avoid antigen loss by trypsin digestion and processed as described below. All non-adherent cells were propagated in RPMI 1640 (Gibco/BRL Life Technologies).
Human umbilical cord vein endothelial cells (HUVEC) were isolated from normal term umbilical cord vein by collagenase perfusion and cultured as previously reported (25).
Production of heterohybridomas.
Heterohybridomas between human peripheral blood lymphocytes (PBL) and CB-F7 were produced according to the technique of Grunow et al. (24). PBL were isolated from SLE patients' blood by differential density centrifugation on lymphoprep (Nycomed, Birmingham, UK). Approximately 107 PBL and myeloma cells were rinsed twice in PBS and fused at 1:1 ratio using 1 ml 42% PEG-1500 (Merck, Lutterworth, UK) containing 15% DMSO (Sigma-Aldrich) in PBS. The fusion mixture was diluted with PBS and the cells centrifuged. The pellet was suspended in RPMI 1640 at a cell density of 106 PBL/ml and 100 μl/well was dispensed into microtiter tissue culture plates (Corning, High Wycombe, UK). After overnight incubation at 37°C, 100 μl double-strength selective medium containing 10–4 M hypoxanthine, 10–5 M aminopterin and 1.6×10–5 M thymidine (Gibco/BRL Life Technologies) was added to each well. After 2 weeks, the hybridomas were initially screened against endothelial cell membrane (ECM) lysates by ELISA. The positive hybridomas were expanded and cloned by limiting dilution. The monoclonality of a positive hybridoma was proven, at least twice, by limiting dilution. Supernatant was produced and concentrated 30- to 100-fold either by membrane filtration using filters (Amicon, Stonehouse, UK) with cut-off values of 100 kDa for IgM and 30 kDa for IgG hybridomas or by ammonium sulfate salt precipitation. IgG antibody preparations were then further purified by Protein G-affinity chromatography (Mab Trap-GII; Pharmacia-Biotech, Uppsala, Sweden) and dialyzed against PBS. Three IgG antibodies were selected for further study: E-3 and C11-2, which were isolated from patients with SLE, and E-1, which was isolated from a normal individual. Only the E-3 mAb displayed a binding activity against ECM lysates by ELISA.
Cell membrane preparations from HMEC-1 were isolated by differential centrifugation as described by McCrae et al. (26). In brief, cells were lyzed by freeze fracture followed by centrifugation at 10,000 g for 30 min in media containing the protease inhibitor mix (2 mM EDTA, 100 U/ml aprotinin, 100 μg/ml PMSF, 12.5 μg/ml leupeptin, 10 mM benzamidine, 10 μg/ml soybean trypsin inhibitor and 1 μg/ml pepstatin) as described (26). The pellet was then resuspended in inhibition medium and sonicated followed by centrifugation at 15,000 g for 30 min. Finally, the cytosolic fractions were removed by centrifugation at 4500 g. The protein concentration was determined by the Bradford reaction (27).
Characterization of hybridomas by ELISA
Anti-cell membrane ELISA. ELISA plates (Nunc; Gibco/BRL Life Technologies) were coated overnight at 4°C with 50 μl/well membrane preparations in 0.015 M sodium bicarbonate buffer (pH 9.6) at a protein concentration of 200 μg/ml. The plates were washed 3 times with PBS containing 0.05% Tween 20. After three washes with PBS/Tween, 50 μl/well hybridoma supernatant was added and incubated for 90 min at 37°C. The plates were then washed 3 times in PBS/Tween. Then 50 μl/well peroxidase conjugated rabbit anti-human Ig, specific for IgG or IgM or IgA, κ and λ chains (Dako, Ely, UK) diluted 1/1000 in PBS/Tween containing 5% fat-free milk powder (Marvel Premier Beverages, Knighton Adbaston, UK), was added and incubated for 1 h at 37°C. After three further washes, 50 μl/well o-phenylendiamine (0.5 mg/ml in sodium citrate buffer 10–1 M, pH 5, plus 0.01% H2O2) was added and incubated for 30 min at 37°C. The enzymatic activity was stopped with 5 M H2SO4 and the absorbance was measured in a multiscan plate reader (Biotec Instruments, Luton, UK).
Anti-cardiolipin (CL) and anti-β2GPI antibodies.
Anti-CL and anti-β2GPI were produced as previously reported (28).
A standard DNA ELISA technique was used to detect anti-DNA antibodies. In brief the plates were coated with 50 μg/ml/well protamine sulfate followed by 1 μg/ml/well calf thymus DNA. After overnight incubation the excess protamine sulfate binding sites were blocked with polyglutamate 10 μg/ml/well. The plates were then blocked with 3% BSA (Sigma-Aldrich) and 0.05% Tween 20 in PBS. The subsequent stages were the same as for the cell membrane ELISA.
Western blotting analysis
Western blotting after discontinuous SDS–PAGE (4% stacking gel and 12% resolving gel) was performed using membrane lysates of HMEC-1 preparations applying 8 μg per track. SDS–PAGE separated proteins were transferred to nitrocellulose and blocked with 5% fat-free milk PBS with 5% Tween. The nitrocellulose was then blotted with the appropriate mAb at a protein concentration 100 μg/ml for 2 h. After 3 washes with PBS 0.05% Tween, the nitrocellulose was incubated for a further 2 h with a peroxidase-conjugated anti-human IgG or IgM or IgA 1/1000 (Dako, Glostrup, Denmark). After a further 3 washes the blots were developed with chloronapthol.
Flow cytometry analysis
Cytofluorimetric analysis was performed with a FACSCalibur cytometer (Becton Dickinson Biosciences, San Jose, CA). The instrument was set up by using cells cultured in medium and incubated with FITC-conjugated rabbit anti-human IgG (Dako). This sample was employed to regulate the detectors for forward and side scatter. CellQuest software (Becton Dickinson Biosciences) was used to generate the plots.
HMEC-1 were used for cytofluorimetric studies. Cell monolayers were incubated in 1 ml PBS/2% FCS for 2 h at room temperature with human mAb at the final concentration of 100 μg/ml. After extensive washes with D-PBS, the cells were incubated with FITC-conjugated rabbit anti-human IgG for 90 min at room temperature, washed again and detached with 30 μM EDTA-PBS. After washing, recovered cells were resuspended in D-PBS.
AECA were detected by ELISA on confluent living HUVEC as previously described (30). Different final protein concentrations of the mAb preparations were used, ranging from 200 to 3 μg/ml.
Adhesion molecule expression on EC surface
Adhesion molecule cell-surface expression was evaluated by a cell ELISA as previously described (6,9,14,15). Briefly, the assay was performed on confluent HUVEC monolayers in 96-well microtiter plates. Cells were incubated in the presence of 10 U/ml human recombinant IL-1β (British BioTechnology, Oxford, UK) or medium alone at the final volume of 200 μl as positive or negative control respectively. A 4-h incubation was used for the evaluation of E-selectin expression and a 20-h incubation for ICAM-1 expression. Different final protein concentrations of the purified AECA mAb preparations were used, ranging from 100 to 0.15 μg/ml. After incubation, the cells were washed twice with RPMI 1640/2.5% FCS and incubated for 60 min at room temperature with 100 μl/well murine monoclonal IgG specific for E-selectin or ICAM-1 at a 1/1000 final dilution (Serotec, Kidlington, UK; cat. nos MCA883 and MCA675 respectively). After three more washes, the cells were fixed with 3% paraformaldehyde for 15 min at room temperature. Cells were then washed 3 times and incubated for another 60 min at room temperature with 100 μl peroxidase-conjugated goat anti-mouse IgG (Cappel, Cochranville, PA). After four washes with RPMI 1640/2.5% FCS and one with PBS alone, 100 μl o-phenylendiamine (0.5 mg/ml in Na citrate buffer 10–1 M, pH 5, plus H2O2 0.01%) was added. The optical density values were evaluated at 450 nm after 30 min incubation by a semiautomatic reader (Platereader; BioRad, Milan, Italy).
Functional adhesion assay
The assay was performed with [51Cr]Na (30 μCi/106 cells; Amersham International, Little Chalfont, UK)-labeled U937 cells as previously described (9,29). Briefly, the monocyte-macrophage cell line was labeled for 1 h at 37°C. Adhesion assays were performed on HUVEC monolayers 24 h pre-incubated with serial protein concentrations of the IgG AECA mAb and the irrelevant control (ranging from 100 to 0.15 μg/ml). Endothelial monolayers incubated with human recombinant IL-1β (British BioTechnology; 10 U/ml) or with medium alone served as positive or negative control respectively. The endothelial monolayers were extensively washed and radiolabeled U937 cells were added to each well. After 1 h adhesion at 37°C, the non-adherent cells were removed by washes and the adherent cells were lyzed with 0.1% SDS/0.025 M NaOH. Adhesion was quantified as c.p.m. and expressed as percentage of adhered cells referred to total radiolabeled U937 seeded on HUVEC monolayers (100%), as described (6,29). To rule out the potential involvement of U937 FcγR in the adhesion to EC monolayers sensitized by the IgG E3, control experiments were carried out in which radiolabeled U937 were pre-incubated with an anti-FcγRII blocking mAb (50 μg/ml; Medarex, Annandale, NJ) for 30 min at 4°C and then extensively washed. After the treatment, U937 cells were incubated with EC monolayers and the assay was performed as described above.
IL-6 production by EC
Confluent HUVEC monolayers in 96-well plates were incubated for 24 h with serial concentrations of purified AECA mAb at a final volume of 200 μl/well (in culture medium). Plates were then centrifuged at 800 r.p.m. at 4°C for 10 min and the cell-free supernatants used for IL-6 determination. Supernatants from HUVEC incubated with 10 U/ml human recombinant IL-1β (British Biotechnology) were used as positive controls. The level of IL-6 was measured using an ELISA kit (Amersham International) (and expressed as pg/ml) as previously described (6,9,14,15).
Activation of NF-κB
Cells were incubated in M199 containing the AECA mAb at a final concentration of 200 μg/ml at 37°C for appropriate times. The dose–response curve was constructed over the range 50–3.125 μg/ml and harvested after 30 min. Samples were prepared at 4°C by rinsing twice with ice-cold HPFEV buffer (50 mM HEPES, 10 mM Na4P2O7, 100 mM NaF, 4 mM EDTA and 2 mM Na3VO4) and then by addition of 300 μl Laemmli sample buffer at 65°C. Samples were then lyzed by passing several times through a 23 gauge needle, boiled (5 min) and analyzed by SDS–PAGE followed by Western blotting using a mAb against inhibitor κB (I-κB) (kind gift of Professor Ron Hay, St Andrews University, UK). Incubation of EC with 120 μg/ml tumor necrosis factor (TNF)-α (British BioTechnology) for 30 min was used as a positive control.
Measurement of JNK Activation
Cells were incubated in M199 containing the AECA mAb at a final concentration of 200 μg/ml at 37°C for appropriate times. Samples were prepared at 4°C by rinsing twice with ice-cold HPFEV buffer (50 mM HEPES, 10 mM Na4P2O7, 100 mM NaF, 4 mM EDTA and 2 mM Na3VO4) and then by addition of 300 μl Laemmli sample buffer at 65°C. Samples were then lyzed by passing several times through a 23 gauge needle, boiled (5 min) and analyzed by SDS–PAGE followed by Western blotting using a polyclonal antibody recognizing the phosphorylated forms of JNK (Promega, Southampton UK). Incubation of EC with 120 U/ml TNF-α (British BioTechnology) for 30 min was used as a positive control.
ELISA and cytofluorimetric studies.
In the present study we selected three different mAb: E-3 and C11-2 produced by B cell clones from SLE patients, and E-1 produced by B cell clone from a normal subject. Figure 1(A) shows the binding of the E-3, C11-2 and E-1 IgG mAb to plates coated with ECM at a concentration of 10 μg/ml antibody. Only E-3 mAb displayed a significant binding against ECM. In addition none of the mAb displayed any anti-DNA, anti-CL or anti-β2GPI activity (data not shown). Figure 1(B) shows the binding of the same mAb to unfixed HUVEC monolayers, E-3 mAb only displaying a clear dose-dependent binding. Figure 2 shows the cytofluorimetric analysis of the binding of E-3 and the irrelevant C11-2 mAb to live HMEC-1. Under the same experimental conditions the mAb E-1 did not display any cell staining (data not shown). Altogether these data show that E-3 recognizes EC and binds to antigens expressed on their cell surface.
Figure 3 shows the results of Western blotting analysis of the E-3 mAb against a membrane preparation of HMEC-1; it recognizes a single band of mol. wt 42 kDa at 100 μg/ml protein concentration. The control mAb shows no reactivity in Western blotting.
Leukocyte functional adherence and adhesion molecule expression
Endothelial monolayers incubated with E-3 showed increased adhesion of U937 cells in a dose-dependent manner; treatment of U937 with a blocking anti-FcγR mAb did not affect the cell adhesion, ruling out the possibility that adherence could be mediated by the binding of the Fc portion of the IgG AECA mAb to the monocytic cells (Fig. 4A). Incubation with either C11-2 or E-1 control mAb did not affect adhesion.
Figure 4(B and C) shows the effect of the E-3, C11-2 and E-1 mAb on the expression of E-selectin and ICAM-1 on HUVEC respectively. The induction of E-selectin can be seen at a concentration comparable with the lowest effective concentration for EC binding of U937; however, this is 30-fold lower than the lowest effective concentration for ICAM-1 induction.
Figure 6(A) shows the effect of the E-3 and C11-2 mAb on IL-6 production. It can be seen that the amount of cytokine produced in response to E-3 is comparable to that induced by incubation with human recombinant IL-1β and that it is produced in a dose-dependent manner (Fig. 6B). The concentration of antibody capable of inducing cytokine production is comparable to the binding curve against EC.
In addition, parallel experiments performed in the presence of Polymixin B (5 μg/ml) failed to inhibit the E-3-induced IL-6 secretion (data not shown).
NF-κB and JNK activation
Figure 7 shows the effect of E-3 mAb on the activation of NF-κB. The level of I-κB substantially decreased by incubation with the AECA E-3 mAb in a way inversely proportional to mAb concentration (Fig. 7A). The time course shows that I-κB is totally cleared by 15 min and starts to recover by 30 min (Fig. 7B). Figure 7(C) shows the relative controls represented by cells incubated with medium alone, with an irrelevant mAb or a positive control (TNF-α). HUVEC activation induced by E3 mAb is comparable to that obtained by TNF-α (120 μg/ml). Incubation of HUVEC with the C11-2 mAb did not affect I-κB expression. The JNK activation pathway was also studied but E-3 failed to activate this pathway (data not shown).
The E-3 mAb is an IgG anti-EC antibody that binds to the EC membrane of, at least, a proportion of EC. Westphal et al. (32) have recently suggested that anti-endothelial ELISA either using cell membrane preparations or whole cells overestimate AECA frequency when compared with cytofluorimetry. The human mAb E-3 not only reacts with the cell membrane and unfixed EC monolayers by ELISA, but also stains EC by cytofluorimetry. These data confirm that the antibody E-3 recognizes an epitope that is expressed on human EC surfaces.
Western blotting studies have identified the putative auto-antigen as having a mol. wt of 42 kDa. The exact nature of the epitope recognized by E-3 remains unclear and is under investigation in our laboratory.
The AECA E3 mAb induces a pro-adhesive and a pro-inflammatory endothelial phenotype in vitro similar to that previously reported with polyclonal IgG fractions from SLE sera (9,10). In this regard, E3 AECA mAb appears to be representative of the anti-endothelial antibodies spontaneously occurring in SLE and suggested to play a role in sustaining vessel wall inflammation (9,10).
These findings are comparable to those found with AECA from other autoimmune vasculitis, i.e. Wegener's granulomatosis, micropolyarteritis, Takayasu's arteritis, and from scleroderma and TTP (6–8,11). E-3 also behaves in a similar manner to a mouse anti-EC mAb directed against a 68 kDa auto-antigen on the surface of EC and derived from B cells of mice with an experimentally induced AECA-associated autoimmune vasculitis and to human mAb obtained from Takayasu's arteritis patients (7,29). Thus, it appears that the ability to modulate EC activation in vitro is a common characteristic of AECA present in different diseases that share in common an immune-mediated endothelial inflammation.
The endothelial stimulation is dose dependent and the effective concentrations are comparable to those previously reported for a murine mAb reacting with HUVEC (29) and for human AECA mAb from Takayasu's patients (7). The effective concentrations are comparable with those that demonstrate binding to whole live EC except in the case of E-selectin, which was up-regulated at concentrations lower than either ICAM-1 or IL-6. However, the dose that up-regulates E-selectin is comparable with that which up-regulates U937 adhesion. In this regard, it is useful to point out that we found that E-3-induced IL-1β secretion is able to further sustain endothelial activation through an autocrine loop, as previously shown for other polyclonal AECA (8,10,31). Thus, it is likely that low mAb amounts, even lower than those able to give the optimal antibody binding, might support E-selectin up-regulation either directly or through the autocrine effects of cytokines produced by the EC. An autocrine mechanism is supported by the studies of Carvalho et al. who showed that polyclonal SLE AECA IgG induce the release of at least two mediators, one as yet unidentified and the other IL-1, which, over different time courses, mediate EC activation (10). The possibility that E-3 mediates its effects by stimulating multiple autocrine factors may explain the different dose–response thresholds to E-3 of activation markers such ICAM-1 and E-Selectin.
Previous studies have shown activation of EC in response to cytokines such as TNF-α requires activation of both JNK–MAP kinases and NF-κB, although it is thought that these are on parallel pathways (33,34). This data suggests that activation of one pathways can occur in response to an appropriate stimulus without the recruitment of the other, and this activation is likely to lead to up-regulation of adhesion molecules and cytokine production.
Although the exact identity of the antigen recognized by E-3 remains unclear, the whole live cell binding data support the hypothesis that the antibody recognizes a cell-surface antigen. Auto-antibodies reactive with non-constitutive EC antigens have been described previously as pro-inflammatory. It has been demonstrated that antibodies reactive with β2GPI are pro-inflammatory, and stimulate both cytokine production and adhesion molecule expression (14,15,35,36). Chan et al. (13) have demonstrated that anti-DNA antibodies are capable of exerting pro-inflammatory effects on EC. Both of these antigens are planted antigens in that they are not normally expressed on EC surfaces. The lack of reactivity of E-3 against β2GPI and DNA and its preliminary characterization suggest it appears to react with a constitutive antigen of EC membranes as described for the murine mAb derived from an AECA-associated experimental model of vasculitis (29).
Furthermore, this is the first demonstration that an AECA mAb from a patient with lupus can strongly activate EC and suggests E-3 mediates its effects by transmembrane signaling, the usual route for activation of NF-κB. Although comparable data have been recently reported with human AECA mAb obtained from Takayasu's patients (7), further work is required with other pro-inflammatory AECA mAb to determine whether this is a common activation pathway.
Previous studies, which have demonstrated a pro- inflammatory effect of AECA, are subject to two major criticisms. First, that contamination by bacterial lipopolysaccharide could induce the same effect; however, treatment of the mAb with Polymixin B failed to alter the pattern of cytokine and adhesion molecule expression, thus eliminating this explanation for these data. The second criticism is that even purified IgG preparations may be contaminated by cytokines, particularly as they have been isolated from patients who have high levels of circulating cytokines (16,17). This criticism cannot be applied to mAb that have been in culture for as long as 2 years since the original isolation. Furthermore, other mAb from normal individuals and patients failed to have the same effect. Thus, the only interpretation of these data is that E-3 is a pro-inflammatory auto-antibody.
This in vitro study using a human mAb derived from a patient with SLE strongly supports the pro-inflammatory activity of AECA. In concert with the large body of evidence correlating their occurrence with disease symptoms and severity, and the in vitro pro-inflammatory effects of patients' sera, these data point in favor of a potential pathogenetic role for AECA in SLE.
This study was supported in part by Ministero Pubblica Istruzione, Progetti 40% and 60% 1997/1998, by Progetto a Concorso 1996/1997—IRCCS Policlinico, Milano, and by the Wellcome Trust and the Arthritis Research Campaign.