Bacterial peptidoglycans but not CpG oligodeoxynucleotides activate synovial fibroblasts by toll-like receptor signaling

Authors


Abstract

Objective

To test the hypothesis that bacterial products acting as adjuvants, such as CpG oligodeoxynucleotides (ODNs) and peptidoglycans (PGs), are able to activate synoviocytes, and to determine the involvement of Toll-like receptors (TLRs) in this activation process.

Methods

Cultured synovial fibroblasts obtained from patients with rheumatoid arthritis (RA) or osteoarthritis (OA) were stimulated with CpG ODNs or PGs. The expression of various integrins was determined by fluorescence-activated cell sorting. TLR and matrix metalloproteinase (MMP) messenger RNA (mRNA) was measured by real-time polymerase chain reaction. Additionally, levels of interleukin-6 (IL-6) and IL-8 in the culture supernatants were assessed by enzyme-linked immunosorbent assay. Blocking experiments were performed by adding anti–TLR-2 and anti–TLR-4 monoclonal antibodies to cultures stimulated with bacterial PGs.

Results

Incubation of synovial fibroblasts with CpG ODNs resulted in neither up-regulation of the expression of integrins on the cell surface, up-regulation of MMP mRNA expression, nor IL-6 and IL-8 production. However, incubation of RA synovial fibroblasts as well as OA synovial fibroblasts with staphylococcal PGs led to an up-regulation of CD54 (ICAM-1) surface expression and to increased expression of MMP-1, MMP-3, and MMP-13 mRNA. Furthermore, production of the proinflammatory cytokines IL-6 and IL-8 was increased by treatment with PGs. We demonstrated that cultured synovial fibroblasts express low levels of TLR-2 and TLR-9 mRNA. TLR-2 was up-regulated after stimulation with PGs, whereas TLR-9 mRNA remained at baseline levels after stimulation with CpG ODNs. Anti–TLR-2 monoclonal antibodies significantly inhibited production of IL-6 and IL-8 induced by stimulation with PGs.

Conclusion

We demonstrate that bacterial PGs activate synovial fibroblasts, at least partially via TLR-2, to express integrins, MMPs, and proinflammatory cytokines. Inhibition of TLR signaling pathways might therefore have a beneficial effect on both joint inflammation and joint destruction.

Rheumatoid arthritis (RA) is thought to have an immunologic origin because of the abundance of immune complexes and rheumatoid factor found in patients, the close association of the disease with certain HLA–DRβ1 types, and the accumulation of activated lymphocytes and monocytes within the joints (1). Despite intensive investigations, no exogenous or endogenous antigens that drive rheumatoid synovitis have been identified so far. However, in spite of the inability to isolate a specific infectious organism from the joints of RA patients, transient exposure to one or more bacteria or viruses may possibly trigger disease, which could eventually perpetuate because of an abnormal host response. It is therefore of major interest to determine how foreign antigens can interact with a predisposed host to induce an abnormal synovial environment conducive to persistent inflammation.

Bacteria are not only a source of exogenous antigens, which potentially cross-react with those of the host, but can also exert adjuvant effects. It has become clear that conserved products of the microbial metabolism such as lipopolysaccharides, peptidoglycans (PGs), lipoteichoic acid, and other components of microbial cell walls, as well as bacterial DNA, activate the innate immune system via specialized pattern recognition receptors. These are termed Toll-like receptors (TLRs) after the gene Toll, first identified in the fruit fly Drosophila (2). To date, 10 members of the TLR family have been characterized (3). Recently, TLR-9 was identified as the receptor for unmethylated CpG DNA sequences, which are found in bacterial DNA and have been shown to be immunostimulatory (4). TLR-2 was previously demonstrated to bind bacterial PGs (5–7).

The role of innate immune recognition of bacterial products in the pathogenesis of arthritis is not known. However, the presence of bacterial PGs as well as bacterial DNA in the joints of patients with RA has been established (8). Furthermore, recent studies demonstrated transient arthritis in mice following intraarticular injection of CpG oligodeoxynucleotides (ODNs) as well as bacterial PGs (9, 10). These data suggest that bacterial products can give rise to an inflammatory reaction locally in the joint, in the absence of live bacteria.

To explore the functional consequences of deposition of bacterial products in joints, we assessed the effect of CpG ODNs and staphylococcal PGs on cultured synovial fibroblasts obtained from patients with RA and patients with osteoarthritis (OA). We show that CpG ODNs known to activate human leukocytes did not stimulate expression of integrins and other cell surface activation markers nor that of matrix metalloproteinases (MMPs) and interleukin-6 (IL-6), whereas bacterial PGs were efficient in inducing CD54 and MMP expression as well as the production of IL-6 and IL-8, in a TLR-2–dependent manner.

MATERIALS AND METHODS

Cell culture and media.

Synovial tissues were obtained from patients with RA and patients with OA who were undergoing synovectomy or joint replacement surgery. Immediately after surgery, the tissue was minced and digested with Dispase at 37°C for 60 minutes. After centrifugation at 400g for 10 minutes, the cells were grown in Dulbecco's modified Eagle's medium (Gibco Invitrogen, San Diego, CA) supplemented with 10% fetal calf serum (FCS), 50 IU/ml penicillin–streptomycin, 2 mML-glutamine, 10 mM HEPES, and 0.2% fungizone (all from Gibco). Cell cultures were maintained in a 5% CO2–humidified incubator at 37°C. For experiments, cultured synovial fibroblasts were used for passages 4 through 8. Blood samples were obtained from healthy volunteers. Peripheral blood mononuclear cells (PBMCs) were prepared by Ficoll-Paque density gradient centrifugation, according to the manufacturer's instructions (Amersham Biosciences, Freiburg, Germany).

ODNs and PG.

The following ODNs were used for the stimulation assays as described above and were synthesized by Microsynth (Balgach, Switzerland): GAC-30 and AGT-30 (11); PS 2006, PS 1982, PS 1968, PE 2078, and PE 2080 (12); PS 5007 (13); and D35, D122, and K3 (14). The sequences are shown in Table 1. If not indicated otherwise, phosphodiester as well as phosphorothioate-modified ODNs were used at 0.2, 1, and 3 μM, except for GAC-30 and AGT-30, which were used at 100 μg/ml. Staphylococcus aureus PG was purchased from Fluka (Buchs, Switzerland). Escherichia coli DNA was obtained from Sigma (St. Louis, MO). All ODNs, the PGs, and the E coli DNA were tested for endotoxin using the Limulus amebocyte cell lysate assay (BioWhittaker, Walkersville, MD). Endotoxin levels did not exceed 0.06 endotoxin units (EU)/ml in all tested samples.

Table 1. Sequences of the CpG ODNs and control ODNs*
ODNSequence 5′ to 3′
  • *

    ODNs = oligodeoxynucleotides. CpG ODNs are shown in boldface.

PE 2080TCG-TCG-TTC-CCC-CCC-CCC-CC
PE 2078TGC-TGC-TTC-CCC-CCC-CCC-CC
PS 2006TCG-TCG-TTT-TGT-CGT-TTT-GTC-GT
PS 1982TCC-AGG-ACT-TCT-CTC-AGG-TT
PS 1968TCG-TCG-CTG-TTG-TCG-TTT-CTT
PS 5007TCC-ATG-TCG-TTT-TTG-TCG-TT
PS D35GGT-GCA-TCG-ATG-CAG-GGG-GG
PS D122GGT-GCA-TTG-ATG-CAG-GGG-GG
PE K3ATC-GAC-TCT-CGA-GCG-TTC-TC
GAC-30ACC-GAT-GAC-GTC-GCC-GGT-GAC-GGC-ACC-ACG
AGT-30ACC-GAT-AGT-ACT-GCC-GGT-GAC-GGC-ACC-ACG

Stimulation assays.

Cultured synovial fibroblasts (80,000 cells/well) were grown in 12-well culture plates for 24 hours and were subsequently stimulated with ODNs, S aureus PGs (20 μg/ml), E coli DNA (30 μg/ml), or IL-1β (5 ng/ml; R&D Systems, Minneapolis, MN). Isolation of mRNA was performed after 24 hours; additional cultures were stimulated for 48 hours and analyzed by fluorescence-activated cell sorting (FACS). Culture supernatants were collected and stored at −80°C for subsequent enzyme-linked immunosorbant assay (ELISA) experiments.

For the blocking experiments, anti–TLR-2 and anti–TLR-4 antibodies (clones TL2.1 and HTA125; eBioscience, San Diego, CA) were added to the synovial fibroblast cultures at 20 μg/ml, 2 hours before stimulation with PGs. PBMCs (5 × 105 cells/well in 96-well plates) were stimulated with reagents for 48 hours and subsequently analyzed by FACS.

ELISA.

IL-6 and IL-8 proteins were detected in cell supernatants by ELISA, using the OptEIA human IL-6 and IL-8 sets according to the manufacturer's instructions (BD PharMingen, San Diego, CA). Briefly, IL-6 and IL-8 capture antibodies were coated overnight on 96-well ELISA plates (Nunc, Roskilde, Denmark). Plates were then blocked with phosphate buffered saline (PBS)/10% FCS. Diluted cell supernatants were added and incubated for 1 hour. IL-6 protein was detected with a complex of biotinylated anti–IL-6 and avidin–horseradish peroxidase conjugate. Tetramethylbenzidine hydrogen peroxide was used as substrate. Absorption was measured at 450 nm. Data were analyzed using Revelation software, version 4.22 (Dynex Technologies, Denkendorf, Germany).

Flow cytometry.

Synovial fibroblasts and PBMCs were harvested after incubation with CpG, PGs, IL-1β, or medium alone. Synovial fibroblasts were detached using 2.5 mM EDTA and washed once with PBS. Cells were then stained with antibodies. The following antibodies, which were either fluorescein isothiocyanate– or phycoerythrin-conjugated, were used to determine the expression of surface molecules: anti-CD20, anti-CD40, anti-CD40L, anti-CD35, anti-CD106, anti–HLA–DR/DP/DQ, anti-CD86, anti-CD54, anti-CD69, anti-CD45, anti-CD19, and anti-IgG1κ (all from BD Pharmingen); anti-CD44 and anti-CD29 (Immunotech, Marseilles, France); anti-Thy1 (clone ASO2; Dianova, Hamburg, Germany); and anti-CD49d (Endogen, Woburn, MA). Samples of 10,000 synovial fibroblasts or PBMCs were analyzed on a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA), and data were processed using CellQuest software (Becton Dickinson).

Real-time PCR.

For quantification of MMP and TLR mRNA, total RNA was isolated using the RNeasy Miniprep Kit (Qiagen, Basel, Switzerland), reverse transcribed with random hexamers, and amplified by single-reporter real-time PCR using the ABI PRISM 7200 Sequence Detection System (Applied Biosystems, Foster City, CA). The specific primers and probes for MMPs and TLRs are listed below. The endogenous control 18S complementary DNA (cDNA) was used for correcting the results with the comparative threshold cycle (CT) method for relative quantification, as described by the manufacturer. The differences of the CT values in sample and 18S cDNA were calculated (ΔCT). The following primers and probes were used: MMP-1 forward primer 5′-TGT-GGA-CCA-TGC-CAT-TGA-GA-3′, reverse primer 5′-TCT-GCT-TGA-CCC-TCA-GAG-ACC-3′, probe 5′-AGC-CTT-CCA-ACT-CTG-GAG-TAA-TGT-CAC-ACC-3′; MMP-3 forward primer 5′-GGG-CCA-TCA-GAG-GAA-ATG-AG-3′, reverse primer 5′-CAC-GGT-TGG-AGG-GAA-ACC-TA-3′, probe 5′-AGC-TGG-ATA-CCC-AAG-AGG-CAT-CCA-CAC-3′; MMP-9 forward primer 5′-GGC-CAC-TAC-TGT-GCC-TTT-GAG-3′, reverse primer 5′-GAT-GGC-GTC-GAA-GAT-GTT-CAC-3′, probe 5′-TTG-CAG-GCA-TCG-TCC-ACC-GG-3′; MMP-13 forward primer 5′-TCC-TAC-AAA-TCT-CGC-GGG-AAT-3′, reverse primer 5′-GCA-TTT-CTC-GGA-GCC-TCT-CA-3′, probe 5′-CAT-GGA-GCT-TGC-TGC-ATT-CTC-CTT-CAG-3′; MMP-14 forward primer 5′-TGG-AGG-AGA-CAC-CCA-CTT-TGA-3′, reverse primer 5′-GCC-ACC-AGG-AAG-ATG-TCA-TTT-C-3′, probe 5′-CCT-GAC-AGT-CCA-AGG-CTC-GGC-AGA-3′; TLR-2 forward primer 5′-GGC-CAG-CAA-ATT-ACC-TGT-GTG-3′, reverse primer 5′-AGG-CGG-ACA-TCC-TGA-ACC-T-3′, probe 5′-CTC-CAT-CCC-ATG-TGT-GCG-TGG-CC-3′; TLR-9 forward primer 5′-TGA-AGA-CTT-CAG-GCC-CAA-CTG-3′, reverse primer 5′-TGC-ACG-GTC-ACC-AGG-TTG-T-3′, probe 5′-AGC-ACC-CTC-AAC-TTC-ACC-TTG-GAT-CTG-TC-3′.

Statistical analysis.

Results are expressed as the mean ± SD or, where indicated, as the mean ± SEM. Expression of TLR and MMP mRNA as well as that of the indicated surface markers and interleukins was compared between stimulated and untreated control cultures using the 2-tailed Student's t-test. P values less than 0.05 were considered significant.

RESULTS

Induction of IL-6 and IL-8 by stimulation of synovial fibroblasts with PGs but not with CpG ODNs.

CpG ODNs have been demonstrated to activate human B lymphocytes and plasmacytoid dendritic cells (12, 15–17). Because it is not known whether synovial fibroblasts are stimulated by these sequences, we examined whether bacterial DNA might be contributing to the activated state of fibroblasts in the joints of patients with RA. Recently, increased IL-6 production by stimulation of gingival fibroblasts with CpG ODNs derived from Porphyromonas gingivalis DNA was reported (11). First, to assess the stimulatory activity of the various CpG ODNs, we cultured human PBMCs in the presence of the indicated ODNs for 48 hours and subsequently measured the expression of CD86 on the B lymphocytes. Use of all phosphorothioate and, to a lesser extent, the phosphodiester-modified ODNs led to a marked up-regulation of the activation marker CD86 (Figure 1).

Figure 1.

Stimulation of peripheral blood mononuclear cells (PBMCs) by CpG oligodeoxynucleotides (ODNs). PBMCs from healthy volunteers were prepared as described in Materials and Methods. Cells (5 × 105) were cultured in 96-well plates for 48 hours in the presence or absence of the indicated ODNs or genomic Escherichia coli DNA. Double bars indicate ODN treatment at 1 μM (left bar) and 3 μM (right bar). E coli was added at 30 μg/ml. Cells were then harvested and double stained with anti-CD20 and anti-CD86 antibodies and subsequently analyzed on a cytofluorometer as described in Materials and Methods. Values are the mean and SEM fluorescence intensity (MFI) of CD86 on B cells from 3 separate cultures.

We next used the CpG ODN sequences, E coli DNA, and PGs to stimulate synovial fibroblasts for 48 hours, as described in Materials and Methods, and tested the supernatants by ELISA for the presence of IL-6 and IL-8. As shown in Figure 2, CpG ODNs did not significantly up-regulate IL-6 or IL-8 expression, as compared with control ODNs. E coli DNA had a small but statistically significant stimulatory effect, whereas the addition of bacterial PGs led to a >100-fold increase in production of IL-6 (Figure 2A) and IL-8 (Figure 2B). This suggests that stimulation of synovial fibroblasts by bacterial PGs results in changes in their functional status, including the production of proinflammatory cytokines. Although CpG ODNs did not have a stimulatory effect in our experiments, the fact that E coli DNA led to increased IL-6 production leaves open the possibility that CpG ODNs (other than the ones tested in our experiments) with an optimal sequence might be stimulatory to RA synovial fibroblasts.

Figure 2.

Production of interleukin-6 (IL-6) and IL-8 by rheumatoid arthritis (RA) synovial fibroblasts following stimulation with peptidoglycan (PGN). Seven different RA synovial fibroblast cultures were stimulated with PGN or IL-1β or were left untreated. Three different RA synovial fibroblast cultures were stimulated with oligodeoxynucleotides (ODNs) or Escherichia coli DNA. IL-6 (A) and IL-8 (B) titers in the culture supernatants were determined after 48 hours. Values are the mean and SD. ∗ = P < 0.01, versus untreated cultures. Triple bars indicate ODN treatment at 0.2 μM (left bar), 1 μM (middle bar), and 3 μM (right bar). Concentrations of IL-8 below the detection level of 0.2 ng/ml (B) are shown as 0.2 ng/ml.

Expression of MMPs in RA and OA synovial fibroblasts cultured in the presence of CpG ODNs or staphylococcal PGs.

RA synovial fibroblasts produce MMPs, which play an important role in the joint destruction commonly observed in RA. To assess whether bacterial products have a stimulatory effect on the MMP expression of synovial fibroblasts, we incubated RA and OA synovial fibroblasts in the presence of CpG ODNs or control ODNs, or left them untreated. RNA was isolated from the harvested cells, and, after reverse transcription, expression of the indicated MMP genes was determined by real-time PCR (TaqMan). Expression of the MMPs was not significantly influenced by incubation with any of the CpG ODN sequences as compared with control ODNs (Table 2). Increased expression of MMP-3 and MMP-9 (∼4-fold) was seen in the presence of the phosphorothioate-modified CpG ODN (PS 2006); this increase was nonspecific, however, because the control ODN (PS 1982) had a comparable stimulatory effect. In contrast to CpG ODNs, in a separate set of experiments PG stimulation of synovial fibroblasts resulted in a marked increase in expression of MMP-1 and MMP-3 (∼40-fold and ∼250-fold, respectively). Expression of MMP-13 mRNA was up-regulated significantly in RA synovial fibroblasts only (Figure 3). Otherwise, no significant differences in the overall profiles of MMP expression, with and without stimulation, were detected between the RA synovial fibroblasts and the OA synovial fibroblasts that were tested.

Table 2. MMP mRNA expression of rheumatoid arthritis synovial fibroblasts cultured in the presence of CpG ODNs*
ODNMMP-1MMP-3MMP-9MMP-13MMP-14
  • *

    Values are the mean ± SD difference in the comparative threshold cycle, as obtained by real-time polymerase chain reaction. MMP = matrix metalloproteinase; ODNs = oligodeoxynucleotides; ND = not done.

Control12.4 ± 1.419.5 ± 3.817.9 ± 3.017.7 ± 1.18.5 ± 1.6
Interleukin-1β7.9 ± 2.18.7 ± 2.616.2 ± 2.213.6 ± 1.38.0 ± 1.8
AGT-3012.1 ± 1.321.9 ± 7.117.9 ± 2.217.3 ± 1.18.2 ± 0.9
GAC-3011.8 ± 0.618.8 ± 3.421.2 ± 8.016.8 ± 0.98.1 ± 1.5
PS 198210.9 ± 0.316.8 ± 4.715.9 ± 0.716.4 ± 1.67.8 ± 0.1
PS 200610.6 ± 0.916.4 ± 5.416.2 ± 0.316.5 ± 0.97.5 ± 0.5
PE 207813.57 ± 0.119.2 ± 0.318.9 ± 0.316.8 ± 0.3ND
PE 208012.5 ± 0.417.7 ± 0.517.7 ± 0.717.8 ± 0.6ND
D12217.9 ± 0.221.7 ± 0.721.3 ± 0.626.8 ± 6.9ND
D3517.7 ± 0.321.3 ± 0.522.2 ± 0.522.6 ± 1.1ND
Figure 3.

Matrix metalloproteinase (MMP) expression of synovial fibroblasts after stimulation with peptidoglycans (PGs). Rheumatoid arthritis synovial fibroblast (RA SF) and osteoarthritis synovial fibroblast (OA SF) cultures were stimulated for 24 hours with PG. Total RNA was extracted, and reverse transcriptase–polymerase chain reaction (PCR) was performed. Subsequently, expression levels of the indicated MMPs were determined by real-time PCR. Values are the mean up-regulation of MMP expression in cultures derived from 4–5 individual RA or OA patients. ∗ = P < 0.04, stimulated versus untreated cultures. The mean difference in the comparative threshold cycle values for untreated cultures was 11.1 for MMP-1, 16.9 for MMP-3, 17.4 for MMP-9, and 18.5 for MMP-13. Bars show the mean and SD.

Effects of CpG ODNs and PGs on activation markers of RA synovial fibroblast cultures.

To analyze the effect of CpG ODNs on cultured synovial fibroblasts obtained from patients with RA and patients with OA, various cultured synovial fibroblasts derived from patients with RA were incubated in the presence of the ODNs for a period of 48 hours. Subsequently, the cells were stained for surface expression of various activation markers (CD35, CD40, CD44, CD49d, CD54, CD86, CD106, CD154, and major histocompatibility complex class II). No differences in the expression of these activation markers on synovial fibroblasts were observed (Figure 4 and data not shown). However, when synovial fibroblasts obtained from patients with RA and patients with OA were stimulated with staphylococcal PGs, surface expression of CD54 (ICAM-1) was significantly increased. The other tested integrins and activation markers did not display any changes after stimulation. Representative results with CD106 and CD40 are shown in Figure 4. Similar results after PG stimulation were obtained with OA synovial fibroblast cultures (data not shown).

Figure 4.

Surface marker expression of rheumatoid arthritis (RA) synovial fibroblasts stimulated with CpG oligodeoxynucleotides (ODNs) or peptidoglycan (PGN). Cultured RA synovial fibroblasts were stimulated for 48 hours with CpG ODNs or control ODNs (PS 2006 and PS 1982 at 1μg/ml, AGT-30 and GAC-30 at 100 μg/ml), with PGN for 48 hours, or were left unstimulated. Cells were subsequently harvested and stained with fluorescein-conjugated anti-CD54, anti-CD106, and anti-CD40 antibodies as described in Materials and Methods. Values are the mean and SEM fluorescence intensity (MFI) of at least 3 different experiments (6 individual experiments for anti-CD54). ∗ = P < 0.02, versus untreated cells. IL-1β = interleukin-1β.

Inducible expression of TLR-2 but not TLR-9 on cultured synovial fibroblasts.

PGs and CpG ODNs exert their stimulatory effects through specific binding of distinct TLRs. To examine the role of TLRs in the development of synovial activation, we first assessed the expression of TLRs in cultured human synovial fibroblasts. Expression levels of TLR-2 and TLR-9 mRNA were measured by real-time PCR and were found to be very low at baseline (see Figure 5 for the respective ΔCT values). Stimulation with IL-1β or CpG ODNs did not increase the levels of TLR-9 mRNA (Figure 5B). However, expression of TLR-2 mRNA was up-regulated ∼5-fold after stimulation with IL-1β or PGs. The levels of expression at baseline were comparable in all of the tested individual cultures. However, their response to stimulation was variable, although all of them responded to PGs to some extent with an up-regulation of TLR-2.

Figure 5.

Expression of Toll-like receptor 2 (TLR-2) and TLR-9 in cultured rheumatoid arthritis synovial fibroblasts (RA SF) and osteoarthritis synovial fibroblasts (OA SF). RA SF and OA SF were cultured for 24 hours in the presence of interleukin-1β (IL-1β), the indicated CpG oligodeoxynucleotides (ODNs) and control ODNs (at 3 μM), peptidoglycan (PGN), or without stimulus (untreated). Expression of TLR-2 (A) and TLR-9 (B) was subsequently measured by real-time polymerase chain reaction, and the differences in the comparative threshold cycle (ΔCT) values were calculated as described in Materials and Methods. Expression values are shown as multiples of values obtained with untreated control cultures (arbitrarily set at 1). (Corresponding mean ΔCT values were as follows: for TLR-2, RA untreated 17.3, RA IL-1β 14.8, RA PGN 15.1, OA untreated 17.6, OA IL-1β 15.9, OA PGN 15.6; for TLR-9, untreated 18.9, IL-1β 18.5, D35 20.6, D122 20.1, 2080 18.7, 2078 19.1). Values are the mean and SD of 4 different OA SF cultures and 5 different RA SF cultures. ∗ = P < 0.02, PGN-stimulated versus control cultures.

There were no significant differences between RA and OA synovial fibroblasts with respect to expression of TLR-2 and TLR-9 at baseline as well as after stimulation with PGs (Figure 5A and data not shown). It must be noted that basal expression levels of TLR-2 and TLR-9 were much lower in synovial fibroblasts than in PBMCs (∼100-fold for TLR-2 and ∼10-fold for TLR-9). The functional relevance of the up-regulation observed for TLR-2 but not for TLR-9 is unclear. However, insufficient expression of the CpG receptor TLR-9 may be a possible explanation for our finding that PGs but not CpG ODNs efficiently stimulate RA synovial fibroblasts.

TR-2–dependent stimulation of synovial fibroblasts.

In studies using transfected fibroblast cell lines (6) and TLR-2–deficient mice (7), TLR-2 has been identified as the receptor of bacterial PGs. Therefore, we examined whether the stimulatory effects of PGs on synovial fibroblasts are dependent on signaling via TLR-2. Synovial fibroblasts were incubated in the presence of monoclonal anti–TLR-2 antibodies (clone TL2.1). Coincubation of synovial fibroblasts with anti–TLR-2 antibodies and PGs led to inhibition of CD54 up-regulation (∼30%), but statistical significance was not reached (data not shown). Determination by ELISA of IL-6 and IL-8 levels in supernatants of the corresponding cultures revealed significant inhibition (∼40%) in the presence of the anti–TLR-2 antibodies (Figure 6). Our results indicate that activation of synovial fibroblasts by PGs is at least partially mediated by TLR-2. Addition of anti–TLR-4 monoclonal antibodies did not lead to significant inhibition of IL-6 and IL-8 up-regulation, indicating that TLR-4 is not involved in the activation of synovial fibroblasts by PGs. This result also argues against the possibility that small amounts of lipopolysaccharides contaminating our PG preparation could stimulate synovial fibroblasts via TLR-4.

Figure 6.

Blocking of the stimulatory effect of peptidoglycans (PGN) by anti–Toll-like receptor 2 (anti–TLR-2) antibodies. Rheumatoid arthritis synovial fibroblasts were stimulated with PGN for 24 hours, as described in Materials and Methods, either in the presence or absence of anti–TLR-2, anti–TLR-4, or isotype control antibodies. Interleukin-6 (IL-6) and IL-8 titers in the culture supernatants were determined. Results are shown as percent expression compared with cultures stimulated with PGN as a reference (set at 100%). Values are the mean and SEM of at least 3 individual experiments. ∗ = P < 0.01, versus reference cultures.

DISCUSSION

Development of tissue injury in autoimmune diseases such as RA could be the end product of a cascade of events involving the induction of an inflammatory reaction. RA is characterized by synovial activation and chronic inflammation of the synovium. Microorganisms can induce and/or promote inflammatory events. Despite intensive investigations, no specific infectious agents have been isolated from the joints of patients with RA. However, not only intact live microorganisms but also bacterial products are able to stimulate the immune system by exerting adjuvant effects. Bacterial PGs as well as bacterial DNA have been found in the joint fluid of patients with RA (8). Research by Tokunaga et al demonstrated that bacterial DNA had immunostimulatory and antitumor effects (18). Unmethylated DNA containing CpG sequences were later identified as the mediators of this effect. Synthetic short CpG ODNs can mimic the immunostimulatory effects of bacterial DNA and were demonstrated to activate murine B lymphocytes as well as natural killer cells, macrophages, and dendritic cells (19–21). In humans, B lymphocytes and plasmacytoid dendritic cells have been identified as the main targets of CpG ODNs (12, 15–17).

Similarly to CpG ODNs, PGs have been previously hypothesized to play a role in the pathogenesis of RA (22–24). PGs were detected in the joint fluid of patients with RA, and in vitro analysis demonstrated that PGs derived from sterile human spleen induce production of proinflammatory cytokines in blood cells (25). Furthermore, bacterial PGs as well as CpG ODNs induce a transient arthritis in mice upon direct intraarticular injection (9, 10). Also, PGs isolated from the anaerobic bacterium Eubacterium aerofaciens, present in human feces, induce a chronic arthritis in rats (26).

Whereas the stimulatory effect of bacterial products on blood cells is well established, their effect on synovial fibroblasts is less clear. Hamilton et al previously suggested that streptococcal cell walls might stimulate synovial fibroblasts, indirectly via monocytes, to produce plasminogen activator (27). Stimulation of gingival fibroblasts by CpG ODNs resulting in enhanced IL-6 production was reported by Takeshita et al (11).

We examined the effect of the bacterial products CpG ODN and PG on synovial fibroblasts derived from patients with RA or OA, to assess the contribution of the pathways of innate immunity to the activated phenotype of RA synovial fibroblasts. First, we tested various CpG sequences reported to activate human cells (12), including the one reported by Takeshita et al. Stimulation of synovial fibroblasts by the ODN sequences used did not result in a significant up-regulation of activation marker expression on the cell surface, MMPs, or IL-6 and IL-8; however, a small but significant increase of IL-6 and IL-8 was observed when E coli DNA was used to stimulate RA synovial fibroblasts. Therefore, we cannot rule out the possibility that the sequences we used were not appropriate to stimulate fibroblasts in an optimal manner.

A recent study demonstrated that CpG ODNs stimulating natural killer cell lysis had very little effect on B lymphocyte proliferation and vice versa, suggesting that the CG-flanking sequences are important for the cell type specificity of CpG ODNs (26). However, the low magnitude of stimulation by the E coli DNA compared with stimulation by PGs argues against an important effect of CpG ODNs on RA synovial fibroblasts. TLR-9 has been identified as the receptor for bacterial DNA, recognizing CpG in a species-specific manner (4, 27). A recent analysis of human PBMC subsets indicated that TLR-9 expression is restricted to plasmacytoid dendritic cells and B lymphocytes; correspondingly, only these cells were responsive to CpG ODNs (28).

In our experiments, synovial fibroblast cultures displayed very low levels of TLR-9 mRNA expression, which did not change upon stimulation with CpG ODNs. The lack of a stimulatory effect of CpG ODNs in our experiments may be caused by insufficient expression of TLR-9 on synovial fibroblasts. Alternatively, the ODN sequences used in our studies may not have been optimal for stimulation of synovial fibroblasts. In contrast, synovial fibroblasts were stimulated by staphylococcal PGs to significantly increase expression of the surface marker CD54 (ICAM-1) as well as that of MMP-1, MMP-3, and MMP-13 mRNA. Additionally, secretion of IL-6 and the chemokine IL-8 was increased in culture supernatants after PG stimulation. Gram-positive cell wall components in general and PGs specifically activate macrophages via TLR-2, as has been shown using TLR-2–expressing cell lines in vitro and TLR-2–deficient mice (5–7). Whether synovial fibroblasts are activated by PGs via the same receptor is not known.

Asai et al recently reported that production of IL-8 by immortalized human gingival epithelial cells is dependent on TLR-2 (29). To assess the functional role of TLR-2 stimulation by PGs, we used anti–TLR-2 monoclonal antibodies (clone 2.1) to demonstrate significant inhibition of IL-6 and IL-8 production (∼40%) and insignificant inhibition of CD54 up-regulation by PGs. The magnitude of the inhibitory effect of the antibodies was comparable to that reported in another published study using TLR-2–transfected Chinese hamster ovary cells (5) but was somewhat less than the 60% inhibition reported for tumor necrosis factor α (TNFα) induction in human monocytes stimulated by heat-killed Listeria monocytogenes (30), which suggests that the antibody has reduced efficacy in blocking the effects of TLR-2 agonists. In the latter study, the additional use of anti-CD14 antibodies further inhibited production of TNFα. The incomplete inhibition observed after stimulation of synovial fibroblasts with PGs might therefore be attributable to involvement of other signaling pathways. CD14 is unlikely to be involved in activation of RA synovial fibroblasts by PGs, because it is not expressed by these cells (31, 32).

Contamination with lipopolysaccharides might activate synovial fibroblasts via TLR-4. We first addressed this possibility by measuring endotoxin levels in our PG preparation. Using a Limulus amebocyte cell lysate assay with a detection limit of 0.06 EU/ml, no endotoxin was found. Second, the addition of anti–TLR-4 antibodies to RA synovial fibroblast cultures before stimulation with PGs did not result in significant inhibition of cytokine production. This argues that activation of RA synovial fibroblasts by PGs is not attributable to lipopolysaccharide contamination. Whether signaling pathways other than TLR-2 may be activated by PGs in synovial fibroblasts remains to be clarified.

Our data support involvement of bacterial PGs in the development of the activated phenotype observed in synovial fibroblasts obtained from patients with RA. Besides bacterial components, endogenous ligands of various TLRs have been indentified, such as Hsp60 for TLR-4 (33). Necrotic cells were also shown to activate macrophages via TLR-2 (34). Other as yet unknown endogenous ligands of TLR-2 might be responsible for activation of RA synovial fibroblasts. Furthermore, reduced systemic levels of IgG against PGs from gut flora were demonstrated in RA patients, suggesting insufficient protection from spreading PGs, which was hypothesized to contribute to inflammatory processes (35).

In our experiments, PGs activated synovial fibroblasts derived from both patients with RA and those with OA. Because invasion into cartilage is a feature of synovial fibroblasts seen in RA but not in OA (36), additional mechanisms might be operating in a concerted manner to induce the aggressive phenotype. Expression of embryonic genes, such as Wnt-5A/frizzled 5 or the retrotransposon LINE 1, has been proposed in this context (37–39). Taken together, our results suggest that synovial fibroblasts, as part of the innate immune system, participate actively in the development of inflammatory arthritis, and highlight TLR signaling pathways as potential targets of both antiinflammatory treatments and inhibitors of joint destruction.

Ancillary