Binding of β₂–glycoprotein I to anionic phospholipids facilitates processing and presentation of a cryptic epitope that activates pathogenic autoreactive T cells

Antiphospholipid syndrome (APS) is characterized by arterial and venous thrombosis as well as recurrent intrauterine fetal loss in association with antiphospholipid antibodies.¹  β₂–glycoprotein I (β₂GPI), a plasma protein that binds negatively charged substances including phospholipids (PLs), is the most common target for the antiphospholipid antibody associated with the clinical features of APS.²  Pathogenicity of the anti-β₂GPI antibody has been demonstrated in animal models, including normal mice immunized with human β₂GPI³  and severe combined immunodeficiency mice into which peripheral blood lymphocytes from patients with APS were transferred.⁴  We recently identified autoreactive CD4⁺ T cells to β₂GPI that promote anti-β₂GPI antibody production in patients with APS.^5-7  β₂GPI–specific CD4⁺ T cells recognize amino acid residues 276-290 (p276-290), which define the immunodominant β₂GPI peptide, in the context of DRB4*0103 (DR53). This epitope is located on domain V and contains the major PL-binding site at amino acids 280-288.⁸  The p276-290–reactive T-cell clones did not respond to functional antigen-presenting cells (APCs) bearing native β₂GPI but did to those bearing chemically reduced β₂GPI or recombinant β₂GPI fragments produced in bacteria.⁶  Given that β₂GPI-reactive T cells are also detected in some healthy individuals,⁵  the p276-290 epitope defined by β₂GPI-specific T cells is “cryptic,” since it is generated at a subthreshold level by the processing of native β₂GPI under normal circumstances.⁹ 

There is increasing interest in the possibility that crypticity is an important characteristic of epitopes recognized by the autoreactive T cells and thus is relevant to autoimmune pathogenesis.¹⁰  T cells recognizing self-determinants generated in sufficient amounts in APCs undergo deletion in the thymus or anergy in the periphery. On the other hand, T cells specific for cryptic self-determinants are a component of the normal T-cell repertoire but normally do not encounter antigenic peptides in the periphery. These T cells might become activated and autoaggressive if the previously cryptic self-determinants were presented at a higher concentration. This concept represents the major hypothesis for the pathogenesis of autoimmune diseases, but the fundamental question is how epitopes that are normally cryptic become visible to the immune system and elicit a sustained pathogenic response. In this study, p276-290–specific T-cell lines generated from patients with APS were used to investigate the mechanisms that induce the efficient processing and presentation of cryptic p276-290 as a consequence of antigen processing.

Peripheral blood T cells from 5 Japanese patients with APS were analyzed in this study. All patients fulfilled the preliminary classification criteria for APS proposed by the International Workshop.¹¹  Primary APS was diagnosed in 3 of the patients, whereas the remaining 2 had secondary APS accompanied by systemic lupus erythematosus. At the time of blood examination, all the patients were on low-dose corticosteroids (< 10 mg/d) and aspirin. Samples from 6 healthy individuals possessing DRB4*0103 confirmed by polymerase chain reaction–based genotyping¹²  were used in experiments on priming the p276-290–specific T-cell response. All samples were obtained after the patients and control subjects gave their written informed consent in accordance with the Declaration of Helsinki, as approved by the Keio University Institutional Review Board (Tokyo, Japan).

Native β₂GPI was purified from normal pooled human plasma as described elsewhere.¹³  Nicked and reduced β₂GPI was prepared by treating β₂GPI with plasmin¹⁴  and dithiothreitol,⁵  respectively. Fusion recombinant maltose-binding proteins (MalBPs) expressed in Escherichia coli included GP-F and GP3, which encoded the entire amino acid sequence (amino acids 1-326) and domains IV and V (amino acids 182-326), respectively, of human β₂GPI.⁵  MalBP, a fusion partner, was also prepared as a control antigen. GP-F and GP3 lacking MalBP, GP-F/MalBP(–), and GP3/MalBP(–) were prepared by incubating MalBP fusion proteins with factor Xa followed by the removal of MalBP and factor Xa by passing the mixture through amylose resin and benzamidine-Sepharose columns, respectively. A recombinant polypeptide encoding domain V of β₂GPI (amino acids 242-326; rDomain V) was expressed by a Pichia pastoris expression system.¹⁵  Peptides encompassing amino acids 276-290 and 306-320 of β₂GPI (p276-290 and p306-320) were synthesized using a solid-phase multiple synthesizer (Advanced ChemTech, Louisville, KY) and purified by high-performance liquid chromatography. These peptides had potential DR53-binding anchor residues,¹⁶  but the binding capacity to the DR53 molecule was not examined. Capacity of individual antigen preparations to bind anionic PL was evaluated by competitive inhibition of an interaction between native β₂GPI and immobilized cardiolipin.¹³  Briefly, cardiolipin-coated plates were incubated with native β₂GPI in the presence of an excess amount of individual antigen preparations. The β₂GPI-cardiolipin complex was detected by incubation with a monospecific APS serum positive for a high titer of anti-β₂GPI antibodies.

Dipalmitoylphosphatidylserine (DPPS), phosphatidylserine from bovine brain (BBPS), and cardiolipin from bovine heart were purchased from Sigma Chemical (St Louis, MO); dilauroylphosphatidylserine (DLPS), dimyristoylphosphatidylserine (DMPS), dioleoylphosphatidylserine (DOPS), dioleoylphosphatidylcholine (DOPC), and monooleoylphosphatidylserine (MOPS) were from Avanti Polar Lipids (Alabaster, AL); and a lyso form of BBPS (lyso-BBPS) was from Doosan Serdary Research Laboratories (Englewood Cliffs, NJ). All chemicals were of reagent-grade quality. The fatty acid chains of bovine tissue–derived BBPS and cardiolipin were not well characterized, but all other PLs were chemically synthesized. Liposomes were prepared principally as described previously,¹⁷  with a lipid composition of DOPC at a molar ratio of 7:3 with the following PLs: DOPS, DPPS, DLPS, DMPS, MOPS, BBPS, lyso-BBPS, and cardiolipin. A mixture of the desired lipids in chloroform-methanol (1:1) was placed in a pear-shaped flask and solvent was removed in a rotary evaporator under reduced pressure. The dried lipids were dispersed with a vortex mixture in sterilized 0.3 M glucose solution. The liposome solutions were then sonicated for 1 minute at 70°C with a bath-type sonicator and adjusted to 1 μmol lipid/mL. These PL liposomes were preincubated with or without native β₂GPI (100 μg/mL) for 30 minutes at room temperature before addition to the cultures. The capacity of PLs to bind β₂GPI was evaluated by an assay as described¹³  with some modifications. Briefly, individual PLs were coated on microtiter plates and subsequently incubated with β₂GPI. The PL-β₂GPI complex was detected by anti-β₂GPI monoclonal antibody (mAb) Cof-23.¹⁸ 

All cultures were incubated in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum, 2 mM l-glutamine, 10 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid, 50 U/mL penicillin, and 50 μg/mL streptomycin in a humidified atmosphere of 5% CO₂ at 37°C. Before use in culture, the fetal bovine serum was depleted of bovine β₂GPI using heparin-Sepharose as described previously⁵  to avoid its potential influence on the generation of the epitope peptide.

T-cell lines reactive with p276-290 were generated from peripheral blood T cells by the repeated stimulation with GP-F followed by limiting dilution as described previously.⁶  A total of 7 CD4⁺ T-cell lines established from patients with APS were used in this study. Four of them (KS3, OM2, OM7, and EY3) were confirmed to be clones based on the single functional T-cell receptor β-chain and were reported in detail previously.^6,7  The clonality of the remaining 3 lines (OM-b, KY-a, KM-b) was not determined. All the T-cell lines recognized p276-290 in the context of HLA-DRB4*0103, had the T-helper 0 (Th0)/Th1 phenotype expressing interferon γ (IFN-γ), and had the capacity to induce anti-β₂GPI antibody production from autologous B cells. T-cell lines were maintained by repetitive stimulation with GP-F, recombinant human interleukin 2 (IL-2; 100 U/mL), and irradiated autologous APCs at 7- to 10-day intervals.

Epstein-Barr virus–transformed lymphoblastoid B-cell line cells (LBLs) were generated from all patients with APS. Circulating monocytes and B cells were isolated from peripheral blood mononuclear cells using anti-CD14 or anti-CD19 mAb-coupled magnetic beads (Miltenyi Biotech, Bergisch Gladbach, Germany) followed by magnetic-activated cell separation (MACS) column separation according to the manufacturer's protocol. Flow cytometric analysis revealed that purity of monocyte and B-cell fractions was greater than 98%. Macrophages and immature monocyte-derived dendritic cells (DCs) were obtained from plastic adherent peripheral blood mononuclear cells in the presence of macrophage-colony stimulating factor (R&D Systems, Minneapolis, MN) or granulocyte-macrophage-colony stimulating factor and IL-4 (PeproTech, Rocky Hill, NJ), respectively, according to previously published methods.¹⁹  Macrophages and immature DCs were pulsed with antigen and subsequently incubated with 50 ng/mL tumor necrosis factor α (TNF-α; PeproTech) for 24 hours to induce activation/maturation before being used in the assays. Flow cytometric analysis revealed that the activated macrophage fraction contained greater than 98% CD14⁺CD80⁺ cells, and the mature DC fraction contained greater than 95% CD83⁺HLA-DR⁺ cells. Allogeneic splenocytes were obtained from DR53-carrying patients with gastric cancer who required splenectomy as part of the dissection of tumor tissues.²⁰  In some experiments, endosomal processing inhibitor chloroquine (0.1 μM; Sigma Chemical) or brefeldin A (1 μg/mL; Sigma Chemical) was added to the APC cultures 1 hour before the addition of antigen.

Antigen-specific proliferation of T-cell lines was determined principally as described previously.⁶  T-cell lines (2 × 10⁴/well) were cultured with various combinations of irradiated APCs (2 × 10⁴/well) and antigen. The APCs were autologous LBLs, monocytes, B cells, monocyte-derived DCs, macrophages, and allogeneic CD53-carrying splenocytes. Native β₂GPI, nicked β₂GPI, reduced β₂GPI, GP-F, GP3, GP-F/MalBP(–), GP3/MalBP(–), rDomain V (10 μg/mL), and p276-290 (5 μg/mL) were used as antigens containing amino acids 276-290 of human β₂GPI. Control antigens were MalBP (10 μg/mL) and p306-320 (5 μg/mL). PL liposomes that were preincubated with native β₂GPI, GP3/MalBP(–), or rDomain V were used at a final concentration of 0.1 μmol lipid and 10 μg protein per 1 mL. A combination of immobilized anti-CD3 mAb (30 ng/mL) and phytohemagglutinin (1 μg/mL) was also used to exclude nonspecific unresponsiveness. LBLs were irradiated at 100 Gy and all other APCs at 40 Gy before being mixed with T-cell lines. After 60 hours of incubation with antigen, 0.5 μCi/well (0.0185 MBq) [³H]-thymidine was added to the cultures for 16 hours. The cells were then harvested and [³H]-thymidine incorporation was determined in a Top-Count microplate scintillation counter (Packard, Meriden, CT). The antigen-induced T-cell response was also evaluated from the production of IFN-γ as described previously.²¹  In some experiments, anti–HLA-DR (L243; immunoglobulin G2a [IgG2a]), anti–HLA-DQ (1a3; IgG2a), or an isotype-matched control mAb (1 μg/mL; Leinco Technologies, Ballwin, MO) were added at the initiation of the cultures. All experiments were carried out in duplicate or triplicate, and the values are the mean of multiple determinations.

Peripheral blood T cells (2 × 10⁶) isolated from peripheral blood mononuclear cells using anti-CD3 mAb-coupled magnetic beads (Miltenyi Biotech) were cultured with autologous mature DCs or TNF-α–stimulated macrophages that were previously pulsed with a mixture of BBPS liposomes and β₂GPI, BBPS liposomes, or β₂GPI alone. On day 3, IL-2 (30 units/mL) was added to the cultures. On day 10, viable T cells were harvested and we examined the capacity to produce IFN-γ in response to antigenic stimulation with autologous LBLs pulsed with MalBP, GP-F, native β₂GPI, p276-290, or p306-320. All culture experiments were carried out in duplicate, and all values represent the mean of duplicate determinations. Results were expressed after the background IFN-γ production was deducted.

Various combinations of antigens and APCs were tested for their ability to stimulate p276-290–reactive CD4⁺ T-cell lines generated from patients with APS. In this in vitro assay system, a response of p276-290–reactive T-cell line can be used as an indicator for the efficient presentation of p276-290 by APCs as a consequence of antigen processing. First, various types of DR53-carrying APCs pulsed with native β₂GPI were examined for their capacity to stimulate p276-290–reactive T-cell lines. Autologous LBLs, monocytes, B cells, mature DCs, activated macrophages, or allogeneic splenocytes bearing native β₂GPI failed to induce a proliferative response of the T-cell lines OM7 and KS3 (Figure 1A). Identical results were obtained using all 7 T-cell lines. Next, autologous LBLs were pulsed with various forms of β₂GPI and cultured with p276-290–reactive T-cell lines OM7 and KS3 (Figure 1B). A significant response was detected in the cultures with reduced β₂GPI and recombinant fusion proteins expressed in a bacterial expression system (GP-F and GP3), as shown in our previous study.⁶  Interestingly, the capacity of GP-F and GP3 to stimulate p276-290–reactive T-cell lines was largely reduced when the fusion partner MalBP was removed from these recombinant fusion proteins. The addition of MalBP to these cultures did not reverse the response (data not shown), indicating that MalBP expressed as a fusion protein played a role in the T-cell response. In contrast, LBLs pulsed with native and nicked forms of β₂GPI or with rDomain V expressed in the eukaryotic expression system did not induce the T-cell response. Analogous findings were obtained from all 7 T-cell lines examined.

None of the individual β₂GPI preparations that stimulated p276-290–reactive T-cell lines were able to bind anionic PL (Figure 1B). Since β₂GPI binds anionic PLs mainly through the major PL-binding site located on a surface-exposed turn,²²  we hypothesized that the loss of the binding capacity in the β₂GPI preparations was due to internalization of the major PL-binding site by a structural modification and that this feature is important for the subsequent presentation of p276-290 that occurs as a result of antigen processing. If this were the case, the β₂GPI-PL complex, in which the major PL-binding site is covered by anionic PL, should stimulate p276-290–reactive T-cell lines. To test this possibility, immature DCs and macrophages were pulsed with DOPS liposomes that were preincubated with or without native β₂GPI, induced to mature/activate by TNF-α treatment, and used to stimulate the p276-290–reactive T-cell lines KS3 and EY3 (Figure 2). DCs bearing DOPS-bound β₂GPI induced a T-cell response, as was observed in those pulsed with GP-F or p276-290, but those bearing native β₂GPI or DOPS liposomes alone did not. In contrast, macrophages pulsed with DOPS-bound β₂GPI were less efficient in stimulating p276-290–reactive T-cell lines. Analogous findings were obtained from all 7 T-cell lines used in this study.

We tested various PL liposomes for their ability to stimulate the p276-290–reactive T-cell lines KS3, EY3, OM-b, and KM-b in the presence of β₂GPI using DCs as APCs. All p276-290–reactive T-cell lines proliferated upon recognition of DCs preincubated with β₂GPI and liposomes containing DOPS or BBPS (Figure 3A). There was a borderline response to a mixture of β₂GPI and cardiolipin-containing liposomes, but all other PLs failed to induce a proliferation irrespective of the presence or absence of β₂GPI. When T-cell response was evaluated by IFN-γ release in response to antigenic stimulation, the response of T-cell lines was specifically induced by liposomes containing DOPS or BBPS and to a lesser extent by liposomes containing cardiolipin (Figure 3B). All 4 T-cell lines represented similar findings, and the IFN-γ release assay appeared to be more sensitive than the proliferation assay. The assay to evaluate capacity of individual PLs to bind β₂GPI revealed that DOPS, BBPS, and cardiolipin were able to bind β₂GPI but the others were not (Figure 3C). These findings indicate that the capacity of individual PLs to induce a T-cell response was correlated with their β₂GPI-binding capacity. We further tested whether β₂GPI preparations with capacity to bind anionic PL (GP3/MalBP(–) and rDomain V) that were preincubated with PL liposomes induced a response of p276-290–reactive T-cell lines KS3 and EY3. DCs bearing DOPS liposomes preincubated with GP3/MalBP(–) or rDomain V induced a T-cell response, although these 2 β₂GPI preparations alone failed to induce a response.

In addition, the T-cell response induced by BBPS-bound β₂GPI was completely abolished by the pretreatment of antigen-captured DCs with chloroquine or brefeldin A, which impair the antigen-processing pathway (Figure 3D). Moreover, the T-cell response induced by BBPS-bound β₂GPI was completely blocked by an anti–HLA-DR antibody during the DC–T-cell interaction. These findings indicate that the response of p276-290–reactive T-cell lines induced by β₂GPI-PL complex–pulsed DCs is endosomal antigen-processing dependent and HLA class II dependent.

We further tested whether β₂GPI bound to anionic PLs primes a T-cell response to p276-290 in healthy individuals in vitro. In this experiment, BBPS was used since liposomes containing this PL constantly induced a strong response of p276-290–reactive T-cell lines in the presence of β₂GPI. Peripheral blood T cells from 6 healthy individuals carrying DR53 were stimulated once with autologous DCs or macrophages pulsed with BBPS-bound β₂GPI, BBPS liposomes, or β₂GPI alone, and the antigen-specific T-cell response was evaluated by IFN-γ production in response to various antigens, including p276-290. Figure 4A shows representative results obtained using DCs as APCs. T cells stimulated with BBPS-bound β₂GPI showed profound responses to GP-F and p276-290, whereas those stimulated with BBPS liposomes or β₂GPI alone did not. Successful priming of p276-290–reactive T cells was obtained in 5 of 6 individuals when DCs were used as APCs but detected in only one when macrophages were used instead (Figure 4B).

The present study demonstrates that the binding of β₂GPI to anionic PL surfaces renders the molecule highly immunogenic by an enhanced generation of the cryptic T-cell determinant as a direct consequence of antigen processing in functional APCs. To date, little information has been obtained about the mechanisms that induce the expression of cryptic self-peptides and elicit autoimmunity in human autoimmune diseases. Potential mechanisms that reveal cryptic self-determinants in APCs include modulation of antigen processing and/or increased antigen delivery to the processing compartment.¹⁰  One possible explanation is that the PL binding physically shields the p276-290 determinant from proteolytic attack in endocytic compartments. In this regard, Simitsek et al²³  reported that antibody binding to the antigen suppresses the generation of some epitopes and boosts that of others. Remarkably, both suppressed and boosted epitopes were present within a protein domain that was “fingerprinted” by the antibody, whereas epitopes that lay outside this domain were not affected. Based on these findings, they speculated that the antibody, by binding and stabilizing a protein domain, might influence the accessibility of the site to proteases during antigen processing. In addition, processing of the thyroglobulin-autoantibody complex has been described to promote generation of cryptic pathogenic peptides in APCs in murine models for autoimmune thyroid disease.²⁴  A similar mechanism can be proposed for the generation of p276-290 from the processing of PL-bound β₂GPI. The major PL-binding site located on the surface of a β₂GPI molecule²²  should be easily accessed by proteases in endocytic compartments during antigen processing, and therefore, the peptides containing the intact major PL-binding site would not be generated from native β₂GPI. In contrast, the binding of β₂GPI to anionic surfaces may protect the major PL-binding site from protease attack by masking the site, resulting in the appearance of the previously cryptic peptide containing the entire major PL-binding site. Anionic PLs on apoptotic bodies and platelet microparticles were potentially present in our cultures, but p276-290–reactive T-cell lines did not respond to native β₂GPI in the absence of PL liposomes. The precise reason for this phenomenon is unclear, but quantity and quality of the lipid vesicles may be important in inducing the presentation of the previously cryptic p276-290 in APCs.

The capacity of individual PLs to induce a response of the p276-290–reactive T-cell line was principally correlated with their β₂GPI-binding capacity. However, β₂GPI had a strong binding affinity to solid-phase cardiolipin, but the cardiolipin-β₂GPI complex demonstrated only a weak T-cell stimulatory capacity. β₂GPI interacts with lipid vesicles containing anionic PLs via the PL-binding patch (ie, a cluster of basic amino acid residues and a hydrophobic flexible loop in the domain V).²²  In contrast to the β₂GPI interaction with solid-phase cardiolipin, binding capacity of β₂GPI to lipid vesicles depends not only on the strength of the negative charge but also on fluidity of the lipid vesicles. In general, the membrane fluidity is influenced by length of fatty acid chains, number of unsaturated double bonds, and/or phase transition temperature of each lipid composed in liposomes. Therefore, it is likely that such structural differences also contribute to the efficiency to reveal the cryptic epitope containing the intact major PL-binding site in APCs.

The in vitro priming experiment strongly suggested that an event inducing the presentation of p276-290 by DCs would activate β₂GPI-reactive CD4⁺ T cells in the normal T-cell repertoire in genetically susceptible individuals. Therefore, exposure of p276-290 to the immune system might be a critical step for inducing APS by triggering the activation of disease-relevant β₂GPI-reactive CD4⁺ T cells. Activated β₂GPI-reactive T cells would subsequently stimulate B cells to produce pathogenic anti-β₂GPI antibody through the expression of CD40 ligand and IL-6, as reported previously.⁶  Since β₂GPI is a plasma protein abundant in the circulation (∼ 200 μg/mL), excessive exposure to anionic surfaces, such as microorganisms and apoptotic cells, may induce the formation of a large quantity of β₂GPI bound to anionic surfaces in vivo. In this regard, associations between various types of infections and the production of antiphospholipid antibodies with or without APS manifestations have been reported,²⁵  and infection is one of the major precipitating factors contributing to the development of catastrophic APS.²⁶  Furthermore, an enhanced yield of cryptic determinants and T-cell stimulatory capacity can be achieved by highly potent DCs that have specialized mechanisms for antigen capture and increased expression of HLA class II and costimulatory and adhesion molecules. These 2 mechanisms may act synergistically to elicit the β₂GPI-specific T-cell response, but additional factors, such as impaired regulatory function and nonspecific inflammation mediated by cytokines and toll-like receptor ligands, are apparently required to initiate the pathogenic autoimmune response. Once the T-cell response to p276-290 is primed, the specific T-cell response could be sustained and amplified by professional APCs that have taken up the β₂GPI complexed with anionic surfaces that are normally present in a small quantity in vivo, such as apoptotic cells, platelet microparticles, and oxidized low-density lipoprotein.²⁷  In addition, this response can be further boosted by the formation of immune complexes consisting of anti-β₂GPI antibodies and β₂GPI bound to anionic surfaces.

In summary, our finding is the first demonstration of a mechanism that elicits pathogenic autoreactive T-cell responses in APS. Further studies examining anionic surfaces that bind to β₂GPI and induce the presentation of the cryptic peptide of β₂GPI in vivo would be useful in clarifying the pathogenesis of APS. In addition, it is likely that modulation of antigen processing is an inevitable consequence of the high-affinity binding and influence processing of autoantigens that are bound by high-affinity ligands. This theory encourages further research examining the possibility that the unveiling of cryptic self-determinants by the altered processing of autoantigens complexed with certain ligands is a major mechanism of the initiation of the autoimmune spiral in other autoimmune diseases.

We thank Drs Takahide Arai and Kazue Yoshida for helpful discussions.