Induction of anti-β2-glycoprotein I autoantibodies in mice by protein H of Streptococcus pyogenes


Philip G. de Groot, Department of Clinical Chemistry and Hematology, University Medical Center, 3548 CX, Utrecht, the Netherlands.
Tel.: +31 88 7557769; fax: +31 88 7555418.


Summary. Background: The antiphospholipid syndrome (APS) is characterized by the persistent presence of anti-β2-glycoprotein I (β2-GPI) autoantibodies. β2-GPI can exist in two conformations. In plasma it is a circular protein, whereas it adopts a fish-hook conformation after binding to phospholipids. Only the latter conformation is recognized by patient antibodies. β2-GPI has been shown to interact with Streptococcus pyogenes. Objective: To evaluate the potential of S. pyogenes-derived proteins to induce anti-β2-GPI autoantibodies. Methods and results: Four S. pyogenes surface proteins (M1 protein, protein H, streptococcal collagen-like protein A [SclA], and streptococcal collagen-like protein B [SclB]) were found to interact with β2-GPI. Only binding to protein H induces a conformational change in β2-GPI, thereby exposing a cryptic epitope for APS-related autoantibodies. Mice were injected with the four proteins. Only mice injected with protein H developed antibodies against the patient antibody-related epitope in domain I of β2-GPI. Patients with pharyngotonsillitis caused by S. pyogenes who developed anti-protein H antibodies also generated anti-β2-GPI antibodies. Conclusions: Our study has demonstrated that a bacterial protein can induce a conformational change in β2-GPI, resulting in the formation of antiβ2-GPI autoantibodies. This constitutes a novel mechanism for the formation of anti-β2-GPI autoantibodies.


The antiphospholipid syndrome (APS) is characterized by the persistent presence of antiphospholipid antibodies in plasma samples of patients with thrombotic events or obstetric complications [1]. The presence of these autoantibodies can be measured with a prolongation of clotting assay, known as lupus anticoagulant, and with an ELISA set-up with either cardiolipin or β2-glycoprotein I (β2-GPI) as antigen. Although the names of the assays suggest otherwise, β2-GPI is the major antigen for the autoantibodies detected with all three assays [2,3]. β2-GPI is a 43-kDa protein consisting of five complement control protein domains. The first domain contains an epitope recognized by the subpopulation of autoantibodies that correlate best with the clinical manifestations, whereas domain V contains a patch of positively charged amino acids with a hydrophobic insertion loop harboring the phospholipid-binding site [4,5]. The epitope within domain I of β2-GPI recognized by the autoantibodies includes amino acids Arg39–Arg43 [6–9]. The autoantibodies do not recognize this epitope when β2-GPI circulates in blood. However, β2-GPI undergoes a major conformational change when it binds to anionic phospholipids. β2-GPI circulates in blood in a circular conformation but, when it binds to negatively charged phospholipids with its positively charged domain V, the interaction of domain V with domain I of β2-GPI is disturbed [10]. The closed conformation of β2GPI opens up, and the site within domain I containing amino acids Arg39–Arg43 becomes exposed on the outside of the molecule [4,10]. This epitope can now be recognized by the autoantibodies that characterize the syndrome [11].

Several publications have linked infections to the cause of APS, but the etiology of the autoantibodies is not well understood. So far, no evidence is available that links the presence of anti-β2-GPI antibodies to infections, although this idea is generally accepted. In a review, Sene et al. [12] suggested that autoantibodies found during an infection are directed against cardiolipin, independently of β2-GPI, but many publications have shown otherwise [14–20]. A theory to explain the formation of autoantibodies directed against β2-GPI is molecular mimicry, in which the immune system develops antibodies directed against viral or bacterial antigens that mimic peptide sequences present in self-proteins [21]. However, only limited evidence has been published supporting the idea that molecular mimicry can induce the anti-β2-GPI autoantibodies and lupus anticoagulant activity that are characteristic for the serology of APS [22,23]. It has been suggested that children with varicella infection and a coinfection with Streptococcus are prone to developing lupus anticoagulant [24]. Streptococcus pyogenes is an important bacterial pathogen of humans, causing a variety of diseases, ranging from a mild phenotype to life-threatening infections [25]. In this article, we demonstrate that S. pyogenes surface protein H can interact with β2-GPI and induce a conformational switch within this protein. This conformational switch is sufficient to induce anti-β2-GPI autoantibodies in vivo.

Materials and methods

Proteins and purification

Human β2-GPI was purified as previously described [26] Plasma-derived β2-GPI had a closed conformation as shown by electron microscopy (JEOL, Sollentuna, Sweden). Closed β2-GPI was converted into the fish-hook conformation by dialyzing it against 20 mmol L−1 Hepes containing 1.15 mol L−1 NaCl (pH 11.5) for 48 h at 4 °C, and then against 20 mmol L−1 Hepes and 150 mmol L−1 NaCl (pH 7.4) [10]. cDNA of murine β2-GPI was commercially obtained (Genescript, Piscataway, NJ, USA). The cDNA was subcloned into the expression vector HisN-Tev (Promega, Madison, WI, USA) and expressed in HEK293E cells. Murine β2-GPI was purified via its His-tag with nickel–Sepharose beads, and eluted with 25 mmol L−1 Tris, 500 mmol L−1 NaCl, and 500 mmol L−1 imidazole (pH 8.2). Human β2-GPI cDNA was used for the construction of the individual domains I, II, IV and V of β2-GPI as previously described [6]. M1 protein, protein H, streptococcal collagen-like protein A (SclA) and streptococcal collagen-like protein B (SclB) were purified as described previously [27–29]. Protein concentrations were determined with the bicinchoninic acid protein assay (Thermo Fisher Scientific LSR, Rockford, IL, USA).

Surface plasmon resonance

Surface plasmon resonance was performed with a BIAcore 2000 (GE Healthcare, Piscataway, NJ, USA). Purified human-derived β2-GPI or recombinant domain I, domain II, domain IV and domain V were immobilized on an activated C-1 sensor chip according to the manufacturer’s instructions. Binding to the proteins was corrected for non-specific binding to an unmodified control channel. M1 protein, protein H, SclA or SclB at various concentrations in a buffer containing 20 mmol L−1 Hepes, 150 mmol L−1 NaCl, 15 μmol L−1 ZnCl2 and 0.005% Tween-20 (pH 7.4) (flow buffer) were injected for 3 min at a flow rate of 30 μL min−1. The dissociation was followed for a period of 10 min. Regeneration of the sensor chip was achieved with a 30-s wash of 1/6 ionic buffer (92 mmol L−1 KSCN, 0.366 mol L−1 MgCl2, 0.184 mol L−1 urea, 0.366 mol L−1 guanidine) and subsequent equilibration with flow buffer.

Purification of IgG of APS patients

Antibodies against β2-GPI from the sera of three individual APS patients were purified by applying sera, diluted 1 : 4 in phosphate-buffered saline (PBS), to a HiTrap Protein G column (GE Healthcare). Subsequently, the column was washed with 25 mL of PBS and eluted with 25 mL of 0.5 mol L−1 acetic acid (pH 2.8). Eluted samples were dialyzed against PBS and stored at − 20 °C for analysis. The patient plasmas were positive for both lupus anticoagulant and anti-β2-GPI antibodies. The presence of lupus anticoagulant and anti-β2-GPI IgG and IgM was determined as previously described [30]. Patient samples were collected with approval of the local ethics committee of the University Medical Center Utrecht. Informed consent was obtained in accordance with the Declaration of Helsinki.

Negative staining transmission electron microscopy

β2-GPI in Tris-buffered saline (TBS) (pH 7.4) was analyzed by negative staining electron microscopy as described previously [31]. Solutions of β2-GPI (5–10 nmol L−1) with or without preincubation with M1 protein, protein H, SclA or SclB were placed on a carbon-coated copper grid and negatively stained with uranyl formate.

Immunosorbent assay with patient antibodies

NUNC MaxiSorp High Protein-Binding Capacity ELISA plates (Nalge Nunc International, Roskild, Denmark) were coated with 5 μg mL−1 purified IgG isolated from plasmas of three APS patient in 50 mmol L−1 carbonate buffer (pH 9.6), 100 μL in each well, overnight at 4 °C. After being washed with TBS-T (50 mmol L−1 Tris, 150 mmol L−1 NaCl, and 0.1% Tween-20, pH 7.4, wash buffer), the plates were blocked with 250 μL of 2% bovine serum albumin in TBS-T (block buffer) for 1 h at 37 °C. After washing, 100 μL of 1 μg mL−1β2-GPI was incubated per well alone or in combination with 1 μg mL−1 protein H, M1 protein, SclA, or SclB. β2-GPI was detected with an in-house polyclonal rabbit anti-β2-GPI antibody and a peroxidase-conjugated anti-rabbit antibody (Dako, Ely, UK). Peroxidase activity was measured by the addition of 100 μL per well of TMB substrate (Tebu-bio Laboratories, Le-Perray-en-Yvelines, France), and color development was stopped by the addition of 50 μL of 1 mol L−1 sulfuric acid to each well. The optical density was measured at 450 nm with a spectrophotometer (Molecular Devices, Wokingham, UK).

Immunization protocol

Forty-eight BALB/c cAnNCrl mice (Charles River Laboratories, L'Arbresle, France) were injected intraperitoneally every 4 weeks with 200 μL of PBS containing 25 μg of human serum albumin, M1 protein, protein H, SclA, SclB or buffer in the absence of any adjuvant. Two weeks before the first protein boost and every 2 weeks after the boosts, 200 μL of blood was drawn in 3.2% citrate via the submandibular veins. Each mouse was boosted six times. Two weeks after the last protein boost, the mice were killed and blood was collected in 3.2% citrate via a heart puncture. All experimental protocols were approved by the institutional Animal Care and Use Committee of the University Medical Center Utrecht.

Characterization of mouse antibodies

Hydrophobic Costar 2595 plates (Costar, Cambridge, MA, USA) were coated with 1 μg mL−1 protein H, M1 protein, SclA or SclB diluted in TBS (20 mmol L−1 Tris, 150 mmol L−1 NaCl, pH 7.4). Hydrophilic Costar 9102 plates were coated with 5 μg mL−1 recombinant mouse or plasma-purified human β2-GPI. After being washed with wash buffer, the plates were blocked with 200 μL of block buffer for 1 h at room temperature. After washing, 100 μL of 1 : 100 diluted mouse plasma in TBS high salt (50 mmol L−1 Tris, 500 mmol L−1 NaCl, pH 7.4) was applied. After washing, 100 μL of 1 : 5000 anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) in block buffer was applied to each well. After removal of unbound antibodies by washing with wash buffer, the peroxidase activity of the bound antibody was measured as described above. Human and mouse β2-GPI were also coated on a Costar 9102 plate and tested with the same protocol as described for the detection of anti-β2-GPI antibodies. The mouse plasma was also tested for the presence of anti-β2-GPI IgM and IgA (both from Sigma Aldrich, St Louis, MO, USA).

Hydrophobic Costar 2595 plates were coated with recombinant human domain I, domain II, domain IV, domain V or domain III–V in 50 mmol L−1 carbonate buffer (pH 9.6), 100 μL in each well, overnight at 4 °C, and tested with the same protocol as described for the detection of anti-β2-GPI antibodies.

Mouse IgG isolation

Plasmas of mice injected six times with protein H were pooled, diluted 1 : 10 in TBS, and applied to a Protein G column (GE Healthcare). Subsequently, the column was washed with 15 mL of TBS and eluted with 0.1 mol L−1 glycine (pH 2.4). Fractions containing IgG were pooled and dialyzed against TBS. This was also done for non-immunized pooled mouse plasma. IgG (5 μg mL−1, 100 μL) was coated in 100 mmol L−1 NaHCO3 in NUNC MaxiSorp High Protein-Binding Capacity ELISA plates (Nalge Nunc International) overnight at 4 °C. After being washed with wash buffer, the plates were blocked with block buffer for 1 h at 37 °C. After washing, 100 μL of 1 μg mL−1 of either fish-hook or circular β2-GPI was incubated for 2 h, and β2-GPI was detected with a polyclonal anti-β2-GPI horseradish peroxidase-conjugated antibody (Cedarlane Laboratories, Burlington, Ontario, Canada) [10].

Coagulation assays

Both an activated partial thromboplastin time (APTT) assay and a diluted Russell’s viper venom time (dRVVT) clotting assay were used to analyze the anticoagulant activity of mouse IgG. The APTT was measured with PTT-LA (Diagnostica Stago, Gennevilliers, France) and Actin FS (Siemens Healthcare Diagnostics, Marburg, Germany). The dRVVT was measured with LA-1 and LA-2 (both from Siemens Healthcare Diagnostics). All coagulation measurements were carried out in a coagulometer (KC 10; Amelung, Lemgo, Germany). First, 50 μL of normal pooled plasma (pool of 200 healthy volunteers) was mixed with 10 μg mL−1 mouse IgG. This was incubated at 37 °C for 2 min; 50 μL of PTT-LA or Actin FS was then added, and incubation was continued for 2 min. Coagulation was initiated by the addition of 50 μL of CaCl2 (25 mmol L−1), and clotting time was recorded. For the dRVVT, 50 μL of normal pooled plasma was mixed with 10 μg mL−1 mouse IgG. This was incubated at 37 °C for 2 min; the dRVVT mixture (at 37 °C) was then added, and clotting time was recorded.

Patient samples

Serum samples were collected from patients treated at the Clinic for Infectious Diseases, Lund University Hospital, Lund, Sweden. Thirteen patients had S. pyogenes bacteremia, and four of these presented with streptococcal toxic shock syndrome including circulatory failure. Six patients with pharyngotonsillitis were included in the study. Acute-phase serum (days 1–3 after onset of symptoms) was collected from all patients. Nineteen patients treated for erysipelas were also sampled. They had typical signs of a bacterial skin infection, with fever and rapid spread of a painful erythema on a lower limb or arm. From these patients, acute-phase sera were collected between days 0 and 5 after onset of symptoms. The study was approved by the Research Ethics Committee of Lund University. For these patient samples, the presence of anti-protein H or anti-β2-GPI IgGs was determined according to the same protocol as for the mice, except for the use of alkaline phosphatase-conjugated anti-human IgG (Sigma Aldrich). The plasmas from the pharyngotonsillitis patients were also tested for β2-GPI domain specificity. Patients were considered to be positive for antibodies when the antibody level exceeded the mean + 3 standard deviations of a plasma pool of 40 healthy individuals.


Interaction of β2-GPI with S. pyogenes surface proteins

It has been shown that β2-GPI interacts with S. pyogenes [32]. We isolated different surface proteins from S. pyogenes and studied their interaction with plasma-purified β2-GPI. Surface plasmon resonance studies revealed that β2-GPI bound to all four tested bacterial proteins: M1 protein (Fig. 1A), protein H (Fig. 1B), SclA (Fig. 1C), and SclB (Fig. 1D). Biacore experiments with the individual domains of β2-GPI revealed that interaction sites were located within domains I and V of β2-GPI for all four bacterial proteins. Competition experiments showed that the binding of protein H to β2-GPI could be completely abolished by the addition of 10 μg mL−1 domain I of β2-GPI but not by the same concentration of domain V, suggesting that the interaction between protein H and β2-GPI predominantly takes place via domain I of β2-GPI.

Figure 1.

 Binding of β2-glycoprotein I (β2-GPI) to bacterial protein was investigated with surface plasmon resonance. β2-GPI, domain I, domain II, domain IV and domain V were immobilized on C-1 sensor chips, and binding of 50 nmol L−1 (A) M1 protein, (B) protein H, (C) streptococcal collagen-like protein A (SclA) and (D) streptococcal collagen-like protein B (SclB) was investigated by surface plasmon resonance. After adjustment for binding to a blank signal, the response of the bacterial proteins at equilibrium was determined, and the amount of bound bacterial protein per fmol of immobilized β2-GPI or domain of β2-GPI was calculated.

Protein H influences the conformation of β2-GPI

Binding to a negatively charged surface induces a conformational change within β2-GPI, resulting in the exposure of the epitope that is recognized by the autoantibodies. To establish whether interaction of β2-GPI with the four bacterial proteins coincides with a conformational change, electronmicrographs were taken of β2-GPI (Fig. 2A) and β2-GPI after incubation with M1 protein (Fig. 2B), protein H (Fig. 2C), SclA (Fig. 2D), or SclB (Fig. 2E). M1 protein and protein H are linear proteins, whereas SclA and SclB consist of a linear segment and an additional globular domain [29]. β2-GPI remained in a circular conformation when bound to M1 protein, SclA, or SclB, but after interaction with protein H, the conformation of β2-GPI changed from a circular to a fish-hook shape (Fig. 2C). The electronmicrographs of protein H and β2-GPI clearly showed that protein H bound to the first domain of β2-GPI, because the stretched end of the fish-hook interacted with protein H.

Figure 2.

 The conformation of β2-glycoprotein I (β2-GPI) in the presence of bacterial proteins. (A) Plasma β2-GPI, in a circular conformation. Bar: 100 nm. (B–E) Electronmicrographs; arrows point at β2-GPI. (B) First panel: M1 protein. Other panels: β2-GPI incubated with M1 protein. β2-GPI remains in a closed conformation. (C) First panel: protein H. Remaining panels: β2-GPI incubated with protein H. β2-GPI is in the fish-hook conformation. (D) First panel: streptococcal collagen-like protein A (SclA). Remaining panels: β2-GPI incubated with SclA. β2-GPI remains in a closed conformation. (E) First panel: streptococcal collagen-like protein B (SclB). Remaining panels: β2-GPI incubated with SclB. β2-GPI remains in a closed conformation. Bar in (E): 25 nm. (F–H) ELISA plates were coated with purified IgG from three different antiphospholipid syndrome (APS patients), three different systemic lupus erythematosus (SLE) patients, and normal pooled plasma (NPP), and a control plate was coated with anti-β2-GPI mAb (3B7, anti-domain I). The plates were incubated with 1 μg mL−1 plasma β2-GPI alone, in a 1 : 1 molar ratio with M1 protein, protein H, SclA, or SclB, or with the individual bacterial proteins. Binding of β2-GPI was measured with a polyclonal anti-β2-GPI antibody. OD, optical density.

To further establish whether protein H induces a conformational change within β2-GPI, we determined whether autoantibodies isolated from patients could recognize β2-GPI bound to the bacterial proteins in solution. Total IgG was isolated from three different patients suffering from APS, coated on a microtiter plate, and incubated with plasma-purified β2-GPI. No binding of plasma β2-GPI could be observed. Additionally, no binding of β2-GPI to the patient antibodies could be detected when the antibodies were incubated with β2-GPI in combination with M1 protein, SclA, or SclB. However, the patient antibodies recognized plasma-derived β2-GPI in the presence of protein H (Fig. 2F–H), indicating that, after interaction with protein H, β2-GPI changed its conformation from a circular to a fish-hook shape, thereby exposing the cryptic epitope for the antibodies. Total IgG from patient plasmas negative for anti-β2-GPI but positive for anticardiolipin antibodies did not bind to β2-GPI (Fig. 2I–K).

Injection of mice with protein H induces anti-β2-GPI antibodies

To determine whether a conformational change of β2-GPI is sufficient to induce the development of anti-β2-GPI autoantibodies, groups of eight mice were injected at six successive time points 4 weeks apart with M1 protein, protein H, SclA, or SclB. The proteins were injected without any adjuvant. After the first boost, all mice started to develop antibodies directed against the injected respective bacterial protein (data not shown). After two protein boosts, mice challenged with protein H developed anti-murine β2-GPI IgM (Fig. 3B). After four protein boosts, mice injected with protein H developed anti-murine β2-GPI IgGs. Anti-β2-GPI IgG or IgM did not develop in mice injected with the other three proteins, even after six boosts (Fig. 3A). None of the mice developed anti-murine β2-GPI IgA. The antibodies induced in the mice injected with protein H recognized not only murine β2-GPI, but also human β2-GPI. This is not surprising, because the overall identity between the proteins of both species is 76%, and the cryptic epitope in β2-GPI for the autoantibodies in domain I is completely conserved between human and mouse. This high level of identity enabled the analysis of the domain specificity of these antibodies with recombinant domains of human β2-GPI. The anti-β2-GPI autoantibodies that developed in mice challenged with protein H were mainly directed against domain I (Fig. 3C).

Figure 3.

 Mice were injected six times, 4 weeks apart, with bacterial proteins. Blood was drawn before injection and 2 weeks after each injection. (A) The level of anti-β2-glycoprotein I (β2-GPI) IgG after six boosts with a bacterial protein. (B) Development in time of the levels of anti-β2-GPI antibodies after injection with protein H. (C) Domain specificity of the anti-β2-GPI antibodies in mice injected with protein H after six protein boosts. (D) Activated partial thromboplastin time (APTT) and (E) diluted Russell’s viper venom time (dRVVT) after addition of 10 μg mL−1 mouse IgG purified from mice injected with protein H or normal pooled mouse IgG to human plasma. Data are presented as mean ± standard deviation. D, domain; NPP, normal pooled plasma; OD, optical density; PL, phospholipids; SclA, streptococcal collagen-like protein A; SclB, streptococcal collagen-like protein B.

Injection of mice with protein H induces lupus anticoagulant activity

The plasmas of the eight mice collected after the sixth boost with protein H were pooled, and total IgG was isolated. Figure 3D shows that IgG derived from these mice prolonged the APTT when added to human plasma. No prolongation was observed with IgG from control mice. The prolongation induced by the added IgG disappeared when the APTT assay was performed in the presence of high phospholipid concentrations, the classic confirmation of the presence of lupus anticoagulant activity. In contrast, no lupus anticoagulant was observed when a dRVVT-based assay was used (Fig. 3E). Total IgG was also isolated from pooled plasma of mice injected six times with M1 protein. These IgGs did not show lupus anticoagulant activity (data not shown).

Anti-β2-GPI antibodies recognize a cryptic epitope in β2-GPI

The total IgG isolated from the pooled plasma of mice injected with protein H was coated on ELISA wells. As a control, a murine mAb against human β2-GPI was used that recognized both fish-hook and circular β2-GPI. The IgG from mice injected with protein H did not recognize plasma-derived β2-GPI, whereas an interaction was found when the IgGs were incubated with β2-GPI that was converted into a fish-hook structure (Fig. 4). The control antibody recognized both plasma β2-GPI and fish-hook β2-GPI at similar levels. These data indicate that the antibodies present in mice boosted with protein H recognize a cryptic epitope in plasma β2-GPI.

Figure 4.

 Control mAb anti-β2-glycoprotein I (β2-GPI) or IgG from mice that were immunized with protein H was coated on an ELISA plate and incubated with either circular or fish-hook β2-GPI. Binding of β2-GPI was determined with a polyclonal antibody directed against β2-GPI. As a control, an anti-domain I anti-β2-GPI antibody mAB (3B7) was used. This antibody recognizes both circular and fish-hook β2-GPI. OD, optical density.

S. pyogenes infection in humans

Mice developed autoantibodies against their own β2-GPI upon challenge with protein H. To determine whether there is a similar mechanism in humans, anti-protein H and anti-β2-GPI IgG were measured in patients suffering from S. pyogenes infections. Samples were collected from three different patient groups: patients with sepsis, with erysipelas, and with pharyngotonsillitis. Three of 13 patients with sepsis were positive for anti-β2-GPI antibodies and anti-protein H antibodies (Table 1). Of the 19 erysipelas patients included in this study, six were positive for both anti-β2-GPI IgG and anti-protein H antibodies. Of the six pharyngotonsillitis patients, four were positive for both anti-β2-GPI antibodies and anti-protein H antibodies. The anti-β2-GPI IgGs in these four patients were mainly directed against the first domain of β2-GPI (Fig. 5).

Table 1.   The presence of anti-protein H and anti-β2-glycoprotein I (β2-GPI) IgG and IgM was determined in three patient groups with a Streptococcus pyogenes infection: sepsis, erysipelas, and pharyngotonsillitis
Negative anti-protein H IgG/IgM and negative anti-β2-GPI IgG/IgM162
Positive anti-protein H IgG/IgM and negative anti-β2-GPI IgG/IgM970
Negative anti-protein H IgG/IgM and positive anti-β2-GPI IgG/IgM000
Positive anti-protein H IgG/M and positive anti-β2-GPI IgG364
Figure 5.

 The presence of anti-β2-glycoprotein I (β2-GPI) antibodies in humans with Streptococcus pyogenes infection. Domain specificity of anti-β2-GPI IgG of pharyngotonsillitis patients. D, domain; OD, optical density.


A role for infections in the development of antiphospholipid antibodies has been an important topic of investigation over the years [12–20]. In this article we describe, for the first time, the in vivo development of autoantibodies against a cryptic epitope within native β2-GPI. The interaction of protein H from S. pyogenes with plasma-derived β2-GPI resulted in a conformational change in β2-GPI. The conformational change resulted in the exposure of an epitope in domain I of β2-GPI that is normally shielded from the circulation. Domain I is of particular interest, because this epitope is recognized by autoantibodies identified in patients with APS with the highest correlation with thrombosis [9,33]. Exposure of this cryptic epitope in β2-GPI caused by binding of protein H to β2-GPI resulted in the development of anti-β2-GPI autoantibodies in all eight mice injected with protein H. The observations made in this mouse model were supported by those made in individuals with S. pyogenes infections, where the presence of anti-β2-GPI antibodies coincides with anti-protein H antibodies. Thirteen of 32 patients with S. pyogenes infection developed both anti-β2-GPI antibodies and anti-protein H antibodies. No individuals were found with anti-β2-GPI antibodies but without anti-protein H antibodies. The percentage of individuals positive for anti-β2-GPI antibodies was highest in the group suffering from pharyngotonsillitis. Because of intracellular survival in the throat, S. pyogenes could establish a reservoir of bacteria causing recurrent pharyngotonsillitis infections [34]. We hypothesize that recurrent or long-lasting infections are necessary for the induction of anti-β2-GPI antibodies. It remains to be determined whether these infections result in transient or persistent anti-β2-GPI antibodies. Although it is generally accepted that transient anticardiolipin antibodies do not contitute a risk factor for thrombosis [35], no information is available on whether there is a difference in risk between the transient and permanent presence of anti-β2-GPI autoantibodies, except for the time period during which they circulate. It has been shown that the anti-β2-GPI antibodies that are present in patients with cytomegalovirus infections correlate with thrombosis [36,37].

The anti-β2-GPI antibodies found in mice after injection of protein H possessed lupus anticoagulant activity when measured with an APTT-based assay but not when measured with a dRVVT assay (Fig. 3D,E). There is ample evidence that many APS patients are only positive in coagulation tests representing the intrinsic coagulation system, probably because the APTT assay used is very sensitive for lupus anticoagulant [38]. Alternatively, lupus anticoagulant activity may be caused by a heterogeneous population of antibodies, and we may have induced a specific APTT-dependent lupus anticoagulant activity by injecting mice with protein H.

Many studies have shown that the presence of anti-β2-GPI autoantibodies is associated with infections in mice. The prevailing theory to explain this correlation is molecular mimicry. Sequence similarities between foreign proteins and self-proteins are sufficient to induce a loss of immune tolerance, resulting in the formation of autoantibodies. The group of Shoenfeld has shown homology between the peptide TLRVYK in domain III of β2-GPI and various microbial agents, and the presence of these antibodies in mice resulted in fetal resorption [14]. The importance of these observations for the human situation is questionable, as there are no indications that anti-domain III antibodies correlate with increased thrombotic manifestations, and their presence only weakly correlates with recurrent spontaneous abortions in patients with APS [39]. In another study, Gharavi et al. [40] injected mice with a peptide derived from cytomegalovirus with homology to an amino acid sequence present in domain V of β2-GPI. They found induction of IgMs directed against β2-GPI with functional properties similar to those found in APS. However, these antibodies were not observed in patients with acute cytomegalovirus infections [41]. In a third study, Krause et al. [23] identified cross-reactivity between antibodies against the cell wall of Saccharomyces cerevisiae and β2-GPI in patients with APS. However, the presence of these antibodies was not associated with any specific manifestations of APS. Sequence analysis showed a complete lack of homology between protein H and β2-GPI. Moreover, the epitope within domain I is not linear, but a three-dimensional conformational epitope created by the constraints of two disulfide bridges within this domain. It is unlikely that the domains of protein H have adopted the short consensus repeat-like conformation of domain I of β2-GPI, because protein H lacks disulfide bridges. Moreover, oxidized recombinant domain I of β2-GPI, but not reduced recombinant domain I, was recognized by APS patient antibodies, indicating the importance of intact disulfide bridges in the recognition of domain I of β2-GPI by patient antibodies [42]. Altogether, it seems highly unlikely that the antibody development after injection with protein H was attributable to molecular mimicry.

Rose [43] suggested that infectious agents can serve as adjuvants. An important role of an adjuvant in the induction of antibodies is the unfolding of the injected protein, resulting in the exposure of antigenic epitopes that are normally shielded from the circulation [44]. Here, we show that this mechanism can also induce the development of autoantibodies against the self-protein β2-GPI. A problem in the understanding of the link between infection and autoimmunity is the observation that many different infections can induce the same autoimmune condition. Indeed, at least 14 distinct microorganisms have been associated with the etiology of antiphospholipid antibodies [12]. We recently reported that binding to anionic phospholipids resulted in a conformational change in β2-GPI [10]. Recently, we have found that binding of lipopolysaccharide (LPS) also induces a conformational change in β2-GPI [45]. It has been shown previously that injection of LPS leads to the development of anti-β2-GPI autoantibodies in rabbits [46]. We have now identified protein H as a third inducer of a conformational change in β2-GPI. We hypothesize that other (bacterial, viral or parasite protens, or self-proteins) or anionic phospholipids exposed on apoptotic cells or microparticles could also induce this conformational change, and we speculate that proteins present on other microorganisms may also bind to β2-GPI and induce a conformational change. We propose that the conformational change in β2-GPI is the common denominator in the development of anti-β2-GPI autoantibodies.

Recently, we found that M1 protein and protein H bind full-length β2-GPI, and thereby prevent the processing of β2-GPI by proteases from neutrophils into antibacterial peptides [32]. Here, we provide evidence that the body has developed an alternative strategy to fight S. pyogenes infection. The induction of autoantibodies could help the plasma proteins in their defense mechanism. Anti-β2-GPI antibodies could support the immune system by stimulating the clearance of apoptotic cells. The presence of anti-β2-GPI antibodies increased the rate of clearance of phosphatidylserine-exposing bodies more than two-fold [47,48]. This theory is supported by the frequently detected prolonged APTT in children with recent infections [49] and the presence of transient anti-β2-GPI antibodies and anticardiolipin antibodies in infectious diseases [50]. Greinacher recently suggested that positively charged plasma proteins with a poorly understood function, such as β2-GPI and platelet factor 4, may be representatives of a previously unrecognized charge-related system in host defense [51].

Figure 6 gives a schematic representation of our current view on the development of anti-β2-GPI antibodies. β2-GPI is present in plasma in a circular conformation. A lasting infection with S. pyogenes can lead to an interaction between protein H and β2-GPI. This interaction leads to a conformational change in β2-GPI to the fish-hook shape. Fish-hook β2-GPI reveals a cryptic epitope that is normally shielded from the circulation. This (repetitive) exposure of the cryptic epitope induces the development of anti-β2-GPI antibodies.

Figure 6.

 Schematic representation of the etiology of anti-β2-glycoprotein I (β2-GPI) antibodies. β2-GPI exists in plasma in the circular conformation. When β2-GPI comes into contact with bacteria, the surface protein H from Streptococcus pyogenes can alter the conformation of β2-GPI to a fish-hook shape. In this fish-hook conformation, a cryptic epitope becomes exposed. During a long-lasting infection and probably repetitive interactions between β2-GPI and protein H, the fish-hook conformation of β2-GPI triggers the immune system, resulting in the development of antibodies against this cryptic epitope in domain I of β2-GPI. SclA, streptococcal collagen-like protein A; SclB, streptococcal collagen-like protein B.


J. C. M. Meijers, H. Herwald, and P. G. de Groot: designed the study; G. M. A. van Os, J. A. Marquart, and M. Morgelin: performed the experiments; G. M. A. van Os, P. G. de Groot, J. C. M. Meijers, C. Agar, M. V. Seron, R. T. Urbanus, H. Herwald, and M. Morgelin: interpreted and critically discussed the data; H. Herwald, P. Akesson, and R. H. W. M. Derksen: provided essential materials; G. M. A. van Os, J. C. M. Meijers, and P. G. de Groot: wrote the manuscript. The other authors critically reviewed the manuscript.


Domains III–V of β2-GPI were a generous gift from G. M. Iverson (La Jolla Pharmaceuticals, San Diego, CA, USA), and human β2-GPI cDNA was kindly provided by T. Kristensen (University of Aarhus, Aarhus, Denmark). We acknowledge A. D. Barendrecht for the artistic impression of the etiology of anti-β2-GPI autoantibodies (Fig. 6). We gratefully acknowledge the skillful work of M. Baumgarten, and thank R. Wallén (Cell and Organism Biology) for help with electron microscopy. This work was supported by a grant of the Netherlands Organization for Scientific Research (ZonMW 91207002 to J. C. M. Meijers and P. G. de Groot), the Swedish Research Council (grant 7480), and an AMC stimulation grant (to J. C. M. Meijers).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.