Toll-like receptor 2 mediates the activation of human monocytes and endothelial cells by antiphospholipid antibodies

The antiphospholipid antibody syndrome (APS) is characterized by clinical manifestations such as arterial or venous thromboembolism and/or recurrent pregnancy complications, as well as the presence of antiphospholipid antibodies (aPLAs).¹  Endothelial cells, monocytes, and platelets are targeted by aPLAs, and inflammatory activation of these cells has been proposed as a pathogenic mechanism. aPLAs are composed of heterogenous auto-antibodies that recognize plasma proteins bound to phospholipid surfaces. One antigenic target of aPLAs is the plasma phospholipid–binding protein β2-glycoprotein 1 (β2GP1), which has been detected on the surface of endothelial cells and monocytes.^2,3  Patients with persistent aPLAs only occasionally present with thrombotic episodes, and sometimes bacterial or viral infections are associated with the clinical manifestations.⁴  These observations suggest that the presence of aPLAs alone is not sufficient to promote thrombosis. Most likely, a priming factor of infectious or inflammatory origin is needed. The implication of infection is particularly obvious in catastrophic APS, a rare but fatal subset of APS, which presents with characteristic features comparable with those occurring in septic shock.^5,6 

Stimulation of endothelial cells by aPLAs has been shown to be mediated by intracellular pathways dependent on NF-κB,^7,8  p38-MAPK,^9,10  myeloid differentiation factor 88, and TNF receptor-associated factor 6.¹¹  The latter mediates signaling by members of the TLR family. The TLRs are a family of integral membrane proteins that recognize conserved pathogen-associated molecular patterns. These receptors thereby function as the first line of defense against pathogens and are essential factors in the innate immune response.¹²  Among the 10 TLRs present in humans, TLR4/MD2 forms homodimers and recognizes lipopolysaccharide (LPS),¹³  whereas TLR2 heterodimerizes with TLR1 or TLR6 and recognizes bacterial triacylated or diacylated lipopeptides.¹⁴  TLRs work with accessory proteins, which help in ligand recognition and binding. One of these, CD14, functions as an accessory protein for both TLR4 and TLR2.^13,15,16  We demonstrated previously that TLR2 is required for the activation of mouse embryonic fibroblasts by aPLAs.¹⁷  In contrast, in a mouse thrombosis model, the absence of functional TLR4 reduced the prothrombotic effect of aPLAs.¹⁸  Monocytes constitutively express TLR2, TLR4, and CD14,^19,20  whereas endothelial cells express TLR4 and very low levels of TLR2 and CD14.^17,21,22  Expression of TLR2 by human endothelial cells is strongly increased after activation by inflammatory stimuli such as TNF, LPS, or IL-1β.^23,24 

The principal aim of the present work was to investigate the respective role of TLR2 and TLR4 in human cell activation by aPLAs. We used 3 cell types for our studies: monocytes, endothelial cells, and human embryonic kidney 293 (HEK293) cells expressing either TLR2 or TLR4. Cell activation by aPLAs was inhibited by antibodies to TLR2 or CD14, but not by antibodies to TLR4. Binding studies revealed an interaction of aPLAs with TLR2 but not with TLR4. Our results demonstrate that TLR2 is required for cell stimulation by some aPLAs and that CD14 enhances this response.

Human TNF was from BioGene, TLR4-grade LPS from Escherichia coli R515 was from Alexis, and lipoteichoic acid (LTA) from Staphylococcus aureus and monoclonal anti–human TLR2 blocking antibody (clone TL2.5) were from InvivoGen. Anti–human CD14-blocking antibody (clone M5E2), isotype-matched control IgG2a and IgG1, Alexa Fluor 647–conjugated mouse anti–human TLR2 (clone TL2.1), Alexa Fluor 647–conjugated isotype-matched control IgG2a, and anti–human CD32 antibody were from BioLegend. Monoclonal anti–human TLR4-blocking antibody (clone 7E3) was from Hycult Biotech. Alexa Fluor 647–conjugated mouse anti–human TLR4 (clone HTA125), FITC-conjugated mouse anti–human CD14, and FITC-conjugated isotype-matched control IgG1 were from AbD-Serotec (MorphoSys AG). Rabbit polyclonal anti-TLR4 (ab47839) and rabbit polyclonal anti-TLR2 (ab47840) antibodies were from Abcam. Rabbit control IgG (sc-2027) was from Santa Cruz Biotechnology.

Monocytes were isolated from blood buffy coats of healthy volunteers and provided by the Geneva Hospital blood transfusion center, as described previously.²⁵  Monocyte purity routinely consisted of > 90% CD14⁺ cells, < 1% CD3⁺ cells, and < 1% CD19⁺ cells as assessed by flow cytometry. Cells were cultured in RPMI medium (GIBCO-BRL/Life Technologies) containing 10% fetal bovine serum (FBS).

HEK293 cells stably transfected with human TLR4, MD2, and CD14 (HEK-Blue-4) or with human TLR2 and CD14 (HEK-Blue-2) and HEK293 cells stably transfected with human TLR2 (HEK-TLR2) or human TLR4 (HEK-TLR4) were obtained from InvivoGen and grown in Dulbecco modified Eagle medium containing 10% FBS (GIBCO-BRL/Life Technologies). HEK-Blue cells express the soluble alkaline phosphatase (sAP) reporter gene under the control of the NF-κB promotor, which enables the quantification of cell activation by measuring sAP activity in medium containing specific enzyme substrates.

Human umbilical cords were obtained with written consent from a parent and approval from the institutional ethics committee of the University Hospital of Geneva, in accordance with the Declaration of Helsinki. Human umbilical vein endothelial cells (HUVECs) were isolated from umbilical cord veins, as described previously.²⁴  Cells were cultured in EGM-2 medium and used at passage 1.

Thirty-two patients with aPLAs and clinical manifestations were selected at the hemostasis unit of the University Hospital of Geneva. Thirty-one of these patients had APS as defined by the revised Sapporo criteria,¹  3 had associated systemic lupus erythematosus, and 1 had systemic lupus erythematosus and aPLAs. A group of 19 healthy controls was included. Ten milliliters of blood was obtained from each patient or control with written consent and approval from the institutional ethics committee of the University Hospital of Geneva in accordance with the Declaration of Helsinki.

IgG fractions (aPLAs) were isolated from patient plasma on Protein-G CL-4B Sepharose (GE Healthcare). IgG was eluted with 0.1M glycine, pH 2.5, and immediately neutralized by addition of 1/4 volume of 1M Tris, pH 9.0. Protein levels were measured with the bicinchoninic acid protein assay (BCA Protein Assay Kit; Pierce). Negative control antibodies were isolated from plasma from 19 healthy donors (controls). β2GP1-immunopurified aPLAs were prepared by passing through an affinity column made by immobilizing 1 mg of β2GP1 per milliliter of Affigel-HZ (BioRad). The bound IgG was eluted with 0.1M glycine, pH 2.5, and immediately neutralized. Endotoxin levels were measured by the Limulus Amebocyte Lysate Endochrome Assay (Charles River Laboratories), and were found to be below the detection limit (0.25 EU/mL) for all IgG fractions at the concentration used in the assays. To exclude lipopeptide contamination of the IgG fractions, we depleted IgG from the aPLA fractions by one step of affinity adsorption to Protein G-Sepharose and tested the remaining supernatant for HEK-Blue-2 activation (a cell-based assay developed by InvivoGen). We found that IgG depletion, evaluated with the bicinchoninic acid protein assay, resulted in a decreased HEK-Blue-2 activation (supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). From these results, we concluded that the cell-activation potential of the aPLA preparations was IgG associated.

Different approaches were used to assess activation of the various cell types by the different agonists or antibody preparations.

For measuring cell activation by the alkaline phosphatase assay, HEK-Blue-4 or HEK-Blue-2 cells were plated at 4 × 10⁴ cells/well and incubated for 16 hours with LPS (100 ng/mL), LTA (0.5 μg/mL), or 500 μg/mL of aPLAs or control IgG in HEK-Blue Detection Medium (InvivoGen), which contains a specific sAP substrate. Cell activation was assessed by measuring the absorbance change of the detection medium at 650 nm.

For measuring cell activation by analysis of IL-8 secretion, HEK-Blue-2 and HEK-TLR2 cells were seeded in a 96-well plate at 4 × 10⁴ cells/well. After 48 hours, the medium was changed and the cells were incubated for 16 hours with various concentrations of LTA, aPLAs, or control IgG. The inflammatory chemokine IL-8 was quantified in cell culture supernatants using the human IL-8 ELISA kit from R&D Systems. The HEK-Blue-2 and HEK-TLR2 cell density was determined using the Cell Titer 96AQ_ueous One Solution Cell Proliferation Assay (Promega).

Cells (2.5 × 10⁴/well) were plated in 96-well plates. Half of the cells were maintained in EGM-2 medium (resting HUVECs), and the other half were prestimulated with 100 ng/mL of TNF for 24 hours, followed by an 8-hour period of TNF washout. Resting HUVECs and TNF-pretreated HUVECs were then incubated for 4 hours with 500 μg/mL of aPLAs or control IgG or 10 μg/mL of LTA in EGM-2 medium supplemented with 10% FBS. In some experiments, blocking anti-TLR2 or anti-TLR4 antibodies (10 μg/mL) were added to the TNF-pretreated HUVECs 30 minutes prior to incubation with the stimuli. Endothelial cell activation was assessed by measuring mRNA levels of the leukocyte adhesion molecule E-selectin.²⁴ 

Monocytes were seeded at 5 × 10⁵ cells/well in a 48-well plate. Blocking antibodies to TLR2, TLR4, or CD14 were added at a concentration of 10 μg/mL for 30 minutes before incubation with 1 μg/mL of LTA, 0.1 μg/mL of LPS, 500 μg/mL of aPLAs, or control IgG for 4 hours. The supernatant was collected for TNF quantification by ELISA (R&D Systems).

Tissue factor activity was quantified on monocytes (1 × 10⁵ cells/well in 96-well plates) after 6 hours of stimulation with aPLAs or control IgG (500 μg/mL) or β2GP1-immunopurified IgG (100 μg/mL) in the presence or absence of blocking antibodies to TLR2, TLR4, or CD14. Tissue factor activity was measured as described previously²⁶  using 150 nM FX (Diagnostica Stago), 5nM FVIIa (Novo Nordisk Pharma), 1mM CaCl₂, and 3.5mM of chromogenic substrate for FXa (Hyphen BioMed). Readings were performed at an optical density (OD) of 405 nm in the kinetic mode. The linear absorbance changes were converted to a concentration of generated FXa by reference to a standard curve made with a known amount of FXa (Hyphen BioMed).

Quantitative real-time RT-PCR (qRT-PCR) was done on a StepOne Real-Time PCR instrument (Applied Biosystems). For tissue factor mRNA quantification on monocytes, RNA was extracted using the RNeasy Mini kit (QIAGEN) and reverse transcribed using the ImPromII Reverse Transcription system (Promega). The tissue factor TaqMan probe (Hs00175225_m1) and TaqMan Master Mix Reagent were then used. Human GAPDH mRNA served as a control gene for the amount of cDNA present in each sample. Quantification of E-selectin mRNA in HUVECs was done using the TaqMan Gene expression Cell-to-CT kit (Ambion), and TaqMan probes for E-selectin gene expression (Hs00950407_m1) and human GAPDH gene expression were used with TaqMan Gene Expression Master Mix. Data were analyzed using the comparative ΔC_T method and Applied Biosystems StepOne software Version 2.0.1, according to the manufacturer's instructions.

The in situ proximity ligation assay²⁷  with the DuoLink kit (Olink Bioscience) was used to quantify the interaction of aPLAs or control IgG with TLR2 or TLR4 on the cell surface. HUVECs and HEK293 cells (HEK-TLR2 and HEK-TLR4) were grown on Lab-Teck II chamber slides (Nalgen Nunc) precoated with 0.1% gelatin (Bioconcept) or 15 μg/mL of human fibronectin (R&D Systems), respectively. For HUVECs, half of the wells were incubated with 100 ng/mL of TNF for 24 hours. Monocytes were seeded at 5 × 10⁴ cells per well and allowed to adhere on Lab-Tek II chamber slides. The cells were fixed with 4% paraformaldehyde solution for 5 minutes, rinsed with PBS, and incubated for 1 hour in the blocking buffer provided with the DuoLink kit. APLA (P1) or control IgG (C0) at 100 μg/mL or β2GP1-immunopurified IgG at 50 μg/mL were incubated for 1 hour, followed by 3 washes with Tris (10 mM)–NaCl (150 mM), pH 7.5, containing 0.05% Tween 20. For monocytes, FCγRII receptors were blocked by incubating cells with 0.5 μg/well of mouse anti–human CD32 antibodies for 10 minutes before the addition of aPLAs or control IgG. Rabbit anti-TLR2, anti-TLR4, or control rabbit IgG was incubated at 10 μg/mL for 1 hour, followed by 3 washes. The slides were then incubated for 1 hour with oligonucleotide-labeled anti–rabbit IgG and anti–human IgG, followed by hybridization, ligation, amplification, and detection using a fluorescent probe according to the manufacturer's instructions (Duolink Detection kit 563; Olink Bioscience). The slides were mounted using Vectashield mounting medium (Vector). The stained cells were analyzed with an LSM510 Meta Confocal microscope (Zeiss). A 40× objective (EC Plan Neofluar 40×/1.3 oil DIC; Zeiss) was used for all images. The images were collected using the LSM 510 Flex Scan package 2.5 and AxioVision LE v4.5 software (Zeiss). Quantification of the fluorescent surface per cell, identified by Hoechst 33342 blue nuclear staining, was performed using the MetaMorph v6.0 software (Visitron Systems). Results are expressed in arbitrary units: the experimental condition with the highest fluorescent surface per cell was arbitrarily placed at 100 AU.

Data are expressed as means ± SEM as analyzed by a 2-way ANOVA followed by a paired t test or by the Mann-Whitney test using Prism Version 5.0 software (GraphPad).

We investigated to what extent aPLAs were capable of activating human monocytes. In this study, isolated monocytes were used on the day of isolation and without attachment to plastic.²⁵  Freshly isolated monocytes were chosen because they express higher levels of TLR2, TLR4, and CD14 than monocytes obtained by adhesion to plastic (supplemental Figure 2A). Isolated monocytes showed specific responses to LTA (a TLR2 agonist) and LPS (a TLR4 agonist), as demonstrated by the inhibition of TNF secretion and the levels of tissue factor mRNA and activity when blocking antibodies to TLR2 or TLR4 were used (supplemental Figure 2B). Blocking antibody to CD14 weakly reduced monocyte responses to LTA and LPS at the agonist concentration used. We next investigated the activation of monocytes by aPLAs. First, a cutoff value for monocyte activation was established by measuring TNF concentrations in 4-hour conditioned medium and treating the cells with 19 control IgGs at 500μg/mL. The cutoff value was defined as the average value ± 3SD of the TNF protein results obtained with these control IgGs. Under our experimental conditions, the cutoff value was 0.44 ng/mL. We tested 32 aPLA preparations for their ability to activate monocytes, and observed that 19 induced a cell response above the cutoff value (Figure 1A). There was no significant association between cell-activation ability and patient characteristics or positivity in any aPLA assay (Table 1).

We investigated the role of TLR2 and TLR4 in monocyte activation by the 19 active aPLA preparations using specific blocking antibodies to TLR2 and TLR4. Monocyte activation was evaluated by quantification of TNF secretion and of tissue factor mRNA levels after 4 hours of stimulation, which corresponded to a maximal tissue factor mRNA response to LTA, LPS, and aPLAs (data not shown). Tissue factor activity was determined after 6 hours of stimulation. With the anti-TLR2 antibodies, we obtained a significant inhibition of monocyte TNF protein and tissue factor mRNA responses (P < .001 by 2-way ANOVA and paired t test) to the 19 tested aPLA preparations, whereas the anti-TLR4 antibodies (P = .7) had no effect (Figure 1B-C). Our observation that blocking antibodies to CD14 significantly (P < .001) reduced the responses to aPLAs implies a role for this protein in monocyte activation by these autoantibodies. Similar results were obtained for tissue factor activity (Figure 1D).

We also investigated what the effect of β2GP1-immunopurified IgG from 2 patients was on monocytes and whether the responses were reduced by preincubation with anti-TLR2 and anti-TLR4 antibodies. The monocyte response to β2GP1-immunopurified IgG was reduced 58% by anti-TLR2 antibodies, whereas anti-TLR4 antibodies or isotype-matched control antibodies (not shown) had no significant effect (Figure 1B-D). Blocking antibodies to CD14 reduced the monocyte responses to aPLAs by 57%.

In the experiments described in the previous section, we observed that antibodies to TLR2 had a much stronger inhibitory effect on aPLA-induced monocyte activation than antibodies to TLR4. Because these antibodies were developed to block responses to known TLR2 and TLR4 agonists, we cannot exclude the possibility that the blocking anti-TLR2 or anti-TLR4 antibodies are not optimal as inhibitors of cellular responses to aPLAs. To further clarify the role of each TLR, we took advantage of a well-characterized model system using HEK293 cells stably expressing TLR2 or TLR4 and MD2 (HEK-Blue-2 and HEK-Blue-4 cells) in combination with CD14.

We first made a detailed comparison of TLR2, TLR4, and CD14 expression by these cells, as well as their responses to LTA or LPS. Flow cytometric analysis revealed that HEK-Blue-2 cells expressed TLR2 but not TLR4, whereas HEK-Blue-4 cells expressed TLR4 but not TLR2 (supplemental Figure 3A). CD14 was highly expressed in both cell lines, with similar average fluorescence intensities (supplemental Figure 3A). Likewise, TLR2 mRNA was only detected in HEK-Blue-2 cells and TLR4 mRNA only in HEK-Blue-4 cells; expression of CD14 mRNA was identical in the 2 cell lines (supplemental Figure 3B). HEK-Blue-2 cells responded to LTA with an increased release of sAP, whereas LPS had no effect. In contrast, HEK-Blue-4 cells responded to LPS, but not to LTA (supplemental Figure 3C). These results indicate that the 2 cell lines contain all of the signaling intermediates required for inflammatory responses to TLR ligands and only differ in their expression of TLR2 or TLR4/MD2.

We tested the effect of 32 aPLA preparations on HEK-Blue-2 cells and HEK-Blue-4 cells, and observed that the 19 aPLA preparations that induced an inflammatory activation of monocytes also gave a positive response in HEK-Blue-2 cells, whereas the other aPLA preparations induced the same sAP activity as the 19 control IgG preparations. In contrast, none of the aPLA preparations induced a response in HEK-Blue-4 cells (Figure 2). The difference in response of HEK-Blue-2 and HEK-Blue-4 cells to the 19 monocyte-activating aPLAs was highly significant (P < .0001; 2-way ANOVA and paired t test).

Inhibition of CD14 reduced the response to aPLAs in monocytes (Figure 1B). To determine whether CD14 also contributes to HEK-Blue-2 activation, we compared responses to aPLAs of HEK-Blue-2 cells and HEK293 cells transfected with TLR2 alone (HEK-TLR2) by quantification of IL-8 release. The 19 cell-activating aPLA preparations induced a significantly (P < .001, 2-way ANOVA and paired t test) higher response in HEK-Blue-2 cells than in HEK-TLR2 cells; similar results were obtained with aPLA IgG immunopurified on immobilized β2GP1 (P < .01; Figure 3A). We used different concentrations of LTA and 3 different aPLA preparations. The results show that the presence of CD14 shifted the dose response to lower concentrations, but had no effect on maximal responses (Figure 3B-C).

Because endothelial cells are likely to contribute to the pathologic effects of aPLAs,^28,29  we determined the role of TLR2 in aPLA-induced activation of human endothelial cells. We used E-selectin mRNA as a very sensitive marker for inflammatory HUVEC activation. Treatment of these cells for 4 hours with the 19 cell-activating aPLAs induced a small but significant increase in E-selectin mRNA compared with control IgG (P < .001; Mann-Whitney test; Figure 4B). Incubation of the cells with immunopurified anti-β2GP1 antibodies from 2 different patients also produced a small increase in E-selectin mRNA. We have previously shown that 24-hour pretreatment of HUVECs with TNF, followed by an 8-hour TNF washout, resulted in high expression of TLR2 mRNA and TLR2 protein at the cell surface. The 8 hours of TNF washout was shown to be sufficient to bring down to almost basal level the inflammatory response of HUVECs²⁴  (Figure 4A media). TNF pretreatment and 8-hour TNF washout had no effect on subsequent responses of the HUVECs to TNF or LPS, which shows that the cells were not refractory to subsequent inflammatory stimuli. The higher response to LTA in the TNF-pretreated HUVECs shows the functional relevance of the up-regulation of TLR2 (Figure 4A). TNF pretreatment resulted in an 8- to 10-fold higher response of HUVECs to the 19 aPLAs and to the β2GP1-immunopurified antibodies (P < .0005 and P < .05, respectively, by 2-way ANOVA and paired t test; Figure 4B), whereas neither the control IgG nor the aPLA samples that did not activate monocytes (not shown) had an effect.

We used blocking antibodies to determine the respective roles of TLR2 or TLR4 in the aPLA responses in TNF-pretreated HUVECs. We observed that blocking anti-TLR2 antibodies reduced the responses to aPLAs between 80% and 95% (P < .01; 2-way ANOVA and paired t test), whereas anti-TLR4 antibodies were ineffective (Figure 4C).

We used the proximity ligation assay to analyze binding of aPLAs to TLR2 or TLR4.²⁷  On monocytes, binding of aPLAs to TLR2 was 10-fold higher than to TLR4. Binding of aPLAs to TLR2 on HEK-TLR2 cells was higher than binding of aPLAs to TLR4 on HEK-TLR4 cells. In HUVECs, aPLAs bound only to TLR2; the binding signal to TLR4 was comparable with the background control (between 2 and 5 AU). aPLA binding to TLR2 on HUVECs was increased by TNF pretreatment (Figure 5 and supplemental Figure 4). On all cell types, binding of control IgG was much lower than that of aPLAs and was comparable with the background control obtained by substituting the primary antibodies with control antibodies.

In addition, binding of one of the β2GP1-immunopurified IgG preparations to TLR2 on HEK-TLR2 cells and monocytes was 6.2 to 8 ± 1.8-fold (mean ± SEM; P < .01, Mann-Whitney test) higher, respectively, than that of the control IgG. Binding of β2GP1-immunopurified IgG to TLR4 on monocytes and on HEK-TLR4 cells was the same as that of control antibodies (P = 1.0).

Inflammatory activation of endothelial cells, monocytes, and platelets appears to be an important factor contributing to the complications of APS. Under our experimental conditions, a high proportion of APS patients, but none of the healthy controls, had cell-activating antibodies. The present study was undertaken to compare the contribution of TLR2 and TLR4 with inflammatory cell activation by aPLAs. We used 3 cell types to probe the role of these TLRs: monocytes, which express both TLR2 and TLR4; endothelial cells, which constitutively express TLR4 and in which TLR2 but not TLR4 is up-regulated by TNF²⁴ ; and a model system of HEK293 cells genetically modified to stably express either TLR2 or TLR4. In this study, we observed that the same aPLA preparations were able to activate monocytes, endothelial cells, and HEK293-expressing TLR2. aPLAs displayed their activating activity by interacting with TLR2 but not with TLR4. Our conclusions are based on 4 distinct results: (1) monocyte and HUVEC activation by aPLAs was inhibited by antibodies to TLR2 but not by antibodies to TLR4; (2) pretreatment of HUVECs with TNF increased TLR2 expression but not TLR4 expression and resulted in an increased subsequent inflammatory response to aPLAs; (3) HEK293 TLR2 cells are activated by aPLAs and LTA, whereas HEK293 cells stably expressing TLR4/MD2 do not respond; and (4) in the 3 different cell types, there was an interaction between aPLAs and TLR2, whereas there was little interaction between aPLAs and TLR4 and none between control IgG and TLR2 or TLR4.

The role for TLR2 in cell activation by aPLAs apparently contrasts with the pathogenic role for TLR4 described previously in a mouse model of thrombosis.¹⁸  Conceivably, under conditions of chronic or acute inflammation, TLR4-dependent induction of TLR2 expression by endothelial cells²⁴  might provide a mechanism by which an absence of functional TLR4 reduces the prothrombotic effect of aPLAs. This notion is in agreement with the 2-hit hypothesis for thrombotic complications of APS.⁵  The release of gut flora–derived endotoxin into the blood circulation³⁰  may result in a low level of chronic inflammation that is sufficient to increase TLR2 expression on endothelial cells in vivo in wild-type mice but not in TLR4-deficient mice. To clarify this issue, it will be important to determine in vivo TLR2 expression in large vessels in mice raised under sterile conditions, under normal conditions, and after challenge with inflammatory stimuli. In addition, the thrombogenic effects of aPLAs should be studied in parallel in wild-type mice, TLR2-deficient mice, and TLR4-deficient mice.

The requirement for TLR2 in aPLA-mediated activation of HEK293 cells, HUVECs, and monocytes raises the question of whether there is an interaction between aPLAs and TLR2. On all 3 cell types tested, we observed binding of aPLAs and β2GP1-immunopurified IgG to TLR2, and an increased binding to TLR2 on TNF-pretreated compared with nonstimulated HUVECs. We observed some binding of aPLAs to TLR4, but binding to TLR2 was between 6 and 10 times higher. Our results are in agreement with the binding of β2GP1 only to TLR2, as determined by mass spectrometry analysis after cross-linking of plasma membrane proteins of Eahy926 human endothelial-like cells.³¹  In contrast, one study demonstrated coimmunoprecipitation of β2GP1 and TLR4, but did not study coimmunoprecipitation with TLR2.³  The localization of TLR2 and TLR4 in the same lipid rafts³¹  may explain why the different binding studies found interactions between aPLAs or β2GP1 and TLR4.

In the present study, only approximately half of the aPLA samples were able to activate monocytes and endothelial cells. Previously, we observed that only a subset of aPLA samples was able to activate endothelial cells.^32,33  In other studies using freshly isolated monocytes, measurements of tissue factor mRNA revealed that only 66% of APS patients had activated circulating monocytes.³⁴  Because the aPLA concentrations used for cell activation (500 μg/mL) were 20-fold lower than IgG concentrations in vivo, we cannot exclude the possibility that some of the negative aPLA samples contained levels of activating antibodies that were below the amount necessary to activate monocytes or endothelial cells in in vitro tests. Our results do not rule out the alternative hypothesis that aPLA preparations that did not activate monocytes or endothelial cells might increase the risk for adverse clinical outcomes through other mechanisms, such as inhibition of the protein anticoagulant pathway or preactivation of platelets. The design of the study, as well as the limited number of aPLA samples used, does not permit conclusions as to an association between cell-activating antibodies and the clinical profile or pattern of positivity in the routine clinical aPLA assays. To resolve this very important clinical issue, it will be necessary to undertake specifically designed studies on a much larger scale using well-defined clinical criteria and standardized aPLA assays.

We observed that TLR2 blockade consistently reduced cell-activation responses to aPLAs by 50% on average, whereas responses to LTA were reduced by more than 80%. Several explanations may be proposed to explain this apparent discrepancy: (1) the site on TLR2 that interacts directly or indirectly with aPLAs may differ from the interaction site of LTA; (2) the concentration of the blocking antibodies was chosen to be optimal for the inhibition of LTA responses and may not be sufficient for inhibition of aPLAs, which are added in amounts exceeding those of the blocking antibodies; (3) the affinities of aPLAs toward their cell-activating targets may differ between patients; and (4) it cannot be ruled out that for some patients other receptors contribute. Our present results do not allow discrimination between these possibilities.

A clinically important target of aPLAs is β2GP1. In agreement with a pathogenic role for anti-β2GP1 antibodies, we observed activation of all 3 tested cell types with β2GP1-immunopurified antibodies, suggesting that β2GP1 can mediate cell activation. We did not observe any association between the cell-activation potential of aPLAs and their titer in routine anti-β2GP1 and anticardiolipin IgG (aCL) ELISAs or in lupus anticoagulant assays. A similar discrepancy between aPLA titers in routine assays and activation potential has been observed previously in a study using U937 monocyte-like cells.³⁵  In contrast, titers of aCL and anti-β2GP1 antibodies were correlated with the degree of expression of tissue factor by monocytes freshly isolated from APS patients.³⁴  It remains to be established to what extent these discrepancies are due to inadequacies of routine aPLA assays,³⁶  to the recognition by patient-derived aPLAs of different β2GP1 epitopes,³⁷  or to β2GP1 conformations,³⁸  of which only some may have cell-activating potential. In addition, we cannot exclude the possibility that other target proteins are capable of mediating cell activation. Specifically designed studies involving much greater numbers of patients than used thus far will be required to clarify these issues.

A hallmark of the TLR family of proteins is their cooperation with cofactor proteins to form a multivalent “sensing apparatus.”^15,16,39,40  One of these, CD14, is not essential for cell activation, but enhances cell activation via TLR2 and TLR4. Blocking antibodies to CD14 reduced the inflammatory activation of monocytes by aPLAs and LTA, and in HEK293 cells expressing both TLR2 and CD14, the reaction to aPLAs and LTA was stronger than in HEK293 cells expressing TLR2 alone. Our results imply a role for CD14 in increasing the sensitivity of monocytes for activation by aPLAs. In endothelial cells, CD14 is unlikely to play such a role, because CD14 is only weakly expressed on HUVECs and is down-regulated after inflammatory stimulation.²⁴  Furthermore, in addition to CD14, other TLR2 coreceptors such as CD36,⁴¹  CD11a (integrin α_L),⁴²  and CD61 (integrin β3),⁴³  are known to modulate TLR2 function. It may be of interest to evaluate a possible contribution of these coreceptors for cell activation by aPLAs.

Several receptors have been involved in cellular and platelet responses to aPLAs, including annexin 2,^44-46  ApoER2′,⁴⁷  and GPIbα,⁴⁸  and the results of the present study show that TLR2 can now be added to this list. It remains to be determined whether annexin 2, ApoER2, and GPIbα might form heterocomplexes with TLR2, or if subgroups of aPLAs deal preferentially with one or the other receptor. The fact that different receptors are involved in cell activation by aPLAs may reflect the heterogeneity of these antibodies.

Previous studies have shown that aPLAs activate both monocytes and endothelial cells,^33,49  and that leukocyte adhesion molecules on endothelial cells are important for the thrombogenic effects of aPLAs.^29,50  The increased expression of tissue factor in monocytes treated with aPLAs implies a role for these autoantibodies in tissue factor induction.^3,51,52  The aPLA-induced activation of resting endothelial cells suggests that the low amounts of TLR2 expressed by these cells are sufficient for generating a detectable inflammatory response. However, under conditions of increased TLR2 expression, such as that observed after TNF pretreatment, a much higher degree of endothelial cell activation is obtained. From these results, we may infer a model by which aPLAs increase the risk of thrombotic complications. In such a model, both monocytes and endothelial cells are weakly activated in APS patients; an exogenous inflammatory stimulus, such as LPS or TNF, leads to an increased TLR2-dependent activation of the endothelium and enhanced interaction between monocytes and the endothelium; and an additional thrombogenic stimulus, such as stasis, may then be sufficient to promote the formation of an intravascular thrombus.

In conclusion, our results provide evidence for a role of TLR2 in the activation of endothelial cells and monocytes by aPLAs from a large proportion of APS patients. Further progress in the understanding of APS will require the parallel analysis of all proposed pathogenic mechanisms, as well as the conduction of standardized routine aPLA assays, for a large number of patient samples to determine whether there is common underlying pathogenic mechanism for all aPLAs or whether the APS is composed of several distinct syndromes. Our observation that TLR2 expression by endothelial cells and the responses of endothelial cells to aPLAs were increased after an inflammatory stimulus may offer an explanation for the clinical observation that aPLAs normally circulate without thrombotic effects. It may also explain why complications of APS frequently occur in association with bacterial or viral infections.

We extend special thanks to Dr Bernadette Mermillod for her assistance in performing the biostatistical analysis.

This work was supported by a grant from the Swiss National Fonds (no. 310030-127639), by the Dr Henri Dubois-Ferrière-Dinu Lipatti Foundation, and by a grant from the ISTH2007 Presidential Fund.

Contribution: N.S. and E.K.O.K. designed the research; N.S. and C.F. performed the research; F.B., S.D.-G., and D.B. provided samples; N.S., E.K.O.K., G.R., and P.d.M. analyzed the data; and N.S., E.K.O.K., G.R., and P.d.M. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Philippe de Moerloose, Division of Angiology and Hemostasis, University Hospital of Geneva, 4, Rue Gabrielle Perret-Gentil, 1211 Geneva 14, Switzerland; e-mail: philippe.demoerloose@hcuge.ch.