Inhibition of toll-like receptor 4 breaks the inflammatory loop in autoimmune destructive arthritis

Authors

  • Shahla Abdollahi-Roodsaz,

    Corresponding author
    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    • Rheumatology Research and Advanced Therapeutics, Department of Rheumatology, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500HB Nijmegen, The Netherlands
    Search for more papers by this author
  • Leo A. B. Joosten,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
    • Drs. Joosten, Matera, van der Meer, Netea, and van den Berg hold a patent for Bartonella quintana lipopolysaccharide, a Toll-like receptor 4 antagonist.

  • Mieke F. Roelofs,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
  • Timothy R. D. J. Radstake,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
  • Giovanni Matera,

    1. University of Catanzaro, Catanzaro, Italy
    Search for more papers by this author
    • Drs. Joosten, Matera, van der Meer, Netea, and van den Berg hold a patent for Bartonella quintana lipopolysaccharide, a Toll-like receptor 4 antagonist.

  • Calin Popa,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
  • Jos W. M. van der Meer,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
    • Drs. Joosten, Matera, van der Meer, Netea, and van den Berg hold a patent for Bartonella quintana lipopolysaccharide, a Toll-like receptor 4 antagonist.

  • Mihai G. Netea,

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
    • Drs. Joosten, Matera, van der Meer, Netea, and van den Berg hold a patent for Bartonella quintana lipopolysaccharide, a Toll-like receptor 4 antagonist.

  • Wim B. van den Berg

    1. Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
    Search for more papers by this author
    • Drs. Joosten, Matera, van der Meer, Netea, and van den Berg hold a patent for Bartonella quintana lipopolysaccharide, a Toll-like receptor 4 antagonist.


Abstract

Objective

Degeneration of extracellular matrix of cartilage leads to the production of molecules capable of activating the immune system via Toll-like receptor 4 (TLR-4). The objective of this study was to investigate the involvement of TLR-4 activation in the development and progression of autoimmune destructive arthritis.

Methods

A naturally occurring TLR-4 antagonist, highly purified lipopolysaccharide (LPS) from Bartonella quintana, was first characterized using mouse macrophages and human dendritic cells (DCs). Mice with collagen-induced arthritis (CIA) and mice with spontaneous arthritis caused by interleukin-1 receptor antagonist (IL-1Ra) gene deficiency were treated with TLR-4 antagonist. The clinical score for joint inflammation, histologic characteristics of arthritis, and local expression of IL-1 in joints were evaluated after treatment.

Results

The TLR-4 antagonist inhibited DC maturation induced by Escherichia coli LPS and cytokine production induced by both exogenous and endogenous TLR-4 ligands, while having no effect on these parameters by itself. Treatment of CIA using TLR-4 antagonist substantially suppressed both clinical and histologic characteristics of arthritis without influencing the adaptive anti–type II collagen immunity crucial for this model. Treatment with TLR-4 antagonist strongly reduced IL-1β expression in articular chondrocytes and synovial tissue. Furthermore, such treatment inhibited IL-1–mediated autoimmune arthritis in IL-1Ra−/− mice and protected the mice against cartilage and bone pathology.

Conclusion

In the present study, we demonstrate for the first time that inhibition of TLR-4 suppresses the severity of experimental arthritis and results in lower IL-1 expression in arthritic joints. Our data suggest that TLR-4 might be a novel target in the treatment of rheumatoid arthritis.

Rheumatoid arthritis (RA) is an autoimmune disease of unknown etiology associated with chronic inflammation of peripheral joints. Today it is generally accepted that proinflammatory cytokines play an important role in the pathogenesis of RA (1); however, the mechanisms of initiation and perpetuation of the inflammatory cascade in RA are still unknown.

Toll-like receptors (TLRs) are a family of pattern recognition receptors that are involved in the recognition of conserved pathogen-associated molecular patterns (2). Ligand binding to TLRs initiates a signaling cascade that leads to the activation of the NF-κB and interferon regulatory factor 3 transcription factors and MAPKs, which in turn promote the production of inflammatory cytokines, chemokines, and tissue-destructive enzymes and the expression of costimulatory molecules on antigen-presenting cells (APCs). These costimulatory molecules provide a second signal to T cells to initiate the adaptive immune response (3, 4).

Considering the role of TLRs as a critical link between the innate and the adaptive immune responses, the idea has emerged that continuous activation or dysregulation of TLR signaling might contribute to the pathogenesis of autoimmune diseases such as RA. Indeed, TLR ligands of exogenous origin such as bacterial peptidoglycans and CpG-containing DNA, activating TLR-2 and TLR-9, respectively, have been found in the synovial fluid of patients with RA (5, 6). In experimental models of arthritis, TLR ligands have repeatedly been used to induce the disease in susceptible animals; for instance, intraarticular injection of streptococcal cell wall fragments, double-stranded RNA, or CpG-containing DNA, which mainly signal through TLR-2, TLR-3, and TLR-9, respectively, can induce arthritis (7–9). Furthermore, lipopolysaccharide (LPS) from the outer membrane of gram-negative bacteria, signaling through TLR-4, has been used extensively to aggravate or reactivate arthritis in distinct animal models (10–12). In addition, LPS has been demonstrated to circumvent the interleukin-1 (IL-1) dependence of the K/BxN serum transfer model of arthritis (13).

In spite of the arthritogenic properties of a variety of TLR ligands, it is still not clear whether TLR activation is present in RA. Alternatively, TLRs can also be activated by several endogenous ligands that are released from stressed or damaged host tissue. In this respect, TLR-4 can recognize extracellular matrix components such as heparan sulfate (HS) and extra domain A (ED-A) of fibronectin (14, 15), and both TLR-2 and TLR-4 can recognize the matrix component biglycan (16). TLR activation by these self antigens can potentially promote the development of autoimmune diseases. A critical role of TLRs in autoimmunity is supported by the finding that autoreactive B cells can be activated by RNA- as well as DNA-associated autoantigens via sequential engagement of the B cell receptor and TLR-7 or TLR-9, respectively (17, 18). This finding may have implications for our understanding of the pathogenesis of systemic lupus erythematosus.

Some of the endogenous TLR ligands can be found in arthritic joints. It was recently shown that RNA released from necrotic synovial fluid cells of RA patients can activate TLR-3 on RA synovial fibroblasts (19). The presence of endogenous TLR-4 ligands such as fibronectin fragments and heat-shock proteins has also been demonstrated in rheumatoid synovium (20–22), and it has been reported that rheumatoid synovial fibroblast–like cells synthesize ED-A–containing fibronectin (23). In addition, serum and synovial fluid from RA patients can activate a TLR-4–expressing Chinese hamster ovary cell line, suggesting the presence of TLR-4–activating substances in RA serum and joints (24).

In the present study, we investigated the involvement of TLR-4 activation in the development of chronic destructive arthritis, using the collagen-induced arthritis (CIA) model and the spontaneous arthritis model in IL-1 receptor antagonist–deficient (IL-1Ra−/−) mice. CIA is an autoimmune model of arthritis based on autoantibodies and T cell immunity against type II collagen (CII). Here, we demonstrate for the first time that inhibition of TLR-4 signaling in mice with non–LPS-accelerated CIA substantially suppresses both clinical and histologic characteristics of early-phase as well as established arthritis. The TLR-4 antagonist was the highly purified LPS from the gram-negative bacterium Bartonella quintana, which we previously reported to act as a specific TLR-4 antagonist (25). In the present study, we demonstrate that B quintana LPS can also inhibit cytokine production by endogenous TLR-4 ligands and dendritic cell (DC) maturation by Escherichia coli LPS. Suppression of arthritis by the TLR-4 antagonist was accompanied by a strong reduction of IL-1 expression in the joint. The protective effects of the TLR-4 antagonist in CIA were not mediated by a disruption of the adaptive immune response disturbing the development of autoimmunity in this model. In addition, treatment with the TLR-4 antagonist reduced the clinical and histopathologic features of arthritis in IL-1Ra−/− mice, in which an autoimmune T cell–mediated arthritis develops spontaneously due to unbalanced IL-1 signaling (26, 27).

MATERIALS AND METHODS

Animals.

Male DBA/1 mice were purchased from Janvier-Elevage (Le Genest St. Isle, France). IL-1Ra−/− mice on a BALB/c background were kindly provided by Dr. M. Nicklin (Sheffield, UK). The mice were housed in filter-top cages, and water and food were provided ad libitum. Age- and sex-matched animals were used in all experiments. Animal studies were approved by the Institutional Review Board and were performed according to the related codes of practice.

Preparation and purification of TLR-4 antagonist.

The TLR-4 antagonist was LPS derived from the cell membrane of the gram-negative bacterium B quintana. The B quintana Oklahoma strain was kindly provided by Prof. D. Raoult (Marseilles, France) and cultured on sheep blood agar at 37°C with 5% CO2. The 5-day cultures of B quintana were heat-killed at 56°C for 60 minutes, and LPS was extracted using a 2-step phenol–water extraction method to remove proteins and lipids as previously described (28).

Isolation, culture, and stimulation of murine resident peritoneal macrophages.

Murine macrophages were isolated from naive DBA/1 mice by lavage of the peritoneal cavity using 10 ml cold medium (Dulbecco's modified Eagle's medium plus 10% fetal calf serum). Adherent cells were harvested and cultured for 4 days before being used. For cytokine production, cells were incubated with purified E coli LPS (10 ng/ml), purified B quintana LPS (1 μg/ml), Pam3CysSerLys4 (Pam3Cys; 10 μg/ml), poly(I-C) (25 μg/ml), IL-1 (10 ng/ml), tumor necrosis factor α (TNFα; 10 ng/ml), ED-A of fibronectin (1 μM), and HS (10 μg/ml) for 24 hours. ED-A and HS were incubated with 10 μg/ml of polymyxin B for 30 minutes prior to use in order to disable possible LPS contamination. When used in combination with TLR-4 antagonist, cells were exposed to 10× higher concentrations of the antagonist for 30 minutes prior to stimulation. Both E coli LPS and B quintana LPS were double-purified by the phenol–water extraction method (28) before use to eliminate potential protein contaminations.

Measurement of cytokines.

Concentrations of cytokines (except transforming growth factor β [TGFβ]) in culture supernatants and mice sera were determined using the Bioplex (Luminex) cytokine assays from Bio-Rad (Hercules, CA). TGFβ concentrations were measured using enzyme-linked immunosorbent assay (ELISA; R&D Systems, Abingdon, UK), following the manufacturer's instructions.

Generation and maturation of monocyte-derived DCs.

Human immature DCs were generated from adherent monocytes as described by Roelofs et al (24). For DC maturation, 1 × 106 immature DCs were incubated with 2 μg/ml purified E coli LPS or 1 μg/ml TLR-7 ligand single-stranded polyuridine (ssPolyU) for 48 hours after a preincubation with 10× higher concentrations of TLR-4 antagonist. DC maturation was determined by measuring the up-regulation of class II major histocompatibility complex (MHC) molecules and “de novo” expression of CD83 using fluorescence-activated cell sorting analysis, as described previously (24).

Induction of CIA.

Arthritis was induced in 10–12-week-old DBA/1 mice. Bovine CII was dissolved in 0.05M acetic acid to a concentration of 2 mg/ml and emulsified in an equal volume of Freund's complete adjuvant (2 mg/ml of Mycobacterium tuberculosis strain H37Ra; Difco, Detroit, MI). Mice were immunized by intradermal injection of 100 μl of the emulsion at the base of the tail and were given an intraperitoneal (IP) booster injection of 100 μl of bovine CII dissolved in phosphate buffered saline without any adjuvant on day 21. Clinical onset and progression of arthritis were macroscopically evaluated by 2 observers in a blinded manner and scored on a scale of 0–2 for each paw, as described by Joosten et al (29).

Treatment of arthritis with TLR-4 antagonist.

To investigate the effects of TLR-4 inhibition on the development of arthritis, mice with CIA were treated using a total of 3 IP injections of TLR-4 antagonist (400 μg/kg body weight) once every 2 days starting on day 22 after immunization and before clinical onset of disease. For therapeutic treatment, mice received 4 daily injections of 2 mg/kg body weight TLR-4 antagonist after a macroscopic inflammation score of 0.5–1 (maximum possible score = 8) was reached (on day 24–25 after immunization). Development of arthritis was evaluated as described above. BALB/c IL-1Ra−/− mice received 400 μg/kg body weight TLR-4 antagonist (or saline as control) IP 3 times a week for 2 weeks, starting after the spontaneous onset of arthritis (mean starting score of 1 on a scale with a maximum possible score of 4).

Histology.

For histologic assessment of arthritis, total joints were isolated at the end point of the experiments and fixed for 4 days in 4% formaldehyde, then decalcified in 5% formic acid and embedded in paraffin. Tissue sections (7 μm) were stained with hematoxylin and eosin to study inflammatory cell influx and chondrocyte death, or with Safranin O to determine proteoglycan (PG) depletion and cartilage and bone destruction. Each parameter was scored on a 0–3-point scale by 2 observers in a blinded manner (29).

Immunohistochemistry.

Local expression of IL-1β was evaluated in paraffin sections of the knee joints. Sections were deparaffinized in xylol and rehydrated in serial dilutions of ethanol. Endogenous peroxidase was blocked using 1% H2O2 for 15 minutes. Tissue sections were incubated with 7.5 μg/ml rabbit anti-mouse IL-1β antibodies or rabbit normal IgG (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 hour, followed by incubation with biotinylated swine anti-rabbit antibodies and peroxidase-labeled streptavidin. Color was developed with diaminobenzidine, and tissues were counterstained with hematoxylin. IL-1β expression was scored on articular chondrocytes (scale 0–2) and synovial tissue around the patella, tibia, and femur (each scored between 0 and 2, then averaged to obtain overall expression in synovium).

Measurement of anticollagen antibodies.

Concentrations of anti-mouse CII IgG1 and IgG2a antibodies were determined using ELISA. Briefly, 96-well plates were coated with 0.1 μg of mouse CII (Chondrex, Redmond, WA). Nonspecific binding sites were blocked by a 5% solution of milk powder. Serial dilutions of mice sera were added, followed by incubation with isotype-specific goat anti-mouse antibodies (peroxidase labeled) and 5-aminosalicylic acid as substrate. Absorbance was measured at 450 nm.

Statistical analysis.

Group measures are expressed as the mean ± SEM. Statistical significance was assessed with a Student's unpaired 2-tailed t-test performed using GraphPad Prism 4.0 software (GraphPad Software, San Diego, CA). P values less than or equal to 0.05 were considered significant.

RESULTS

Characterization of the TLR-4 antagonist.

The TLR-4 antagonist used in this study was the double-extracted LPS derived from the cell membrane of the gram-negative bacterium B quintana. This highly purified LPS has been described to act as a potent TLR-4 antagonist, since it strongly blocks the induction of IL-1, TNF, and other proinflammatory cytokines by E coli LPS (25). To further characterize the TLR-4 antagonist, we determined the effects of B quintana LPS on the production of a wide range of pro- and antiinflammatory cytokines by mouse peritoneal macrophages. The TLR-4 antagonist did not by itself induce the production of proinflammatory cytokines including IL-1, TNFα, IL-6, and IL-12, nor did it induce the production of antiinflammatory cytokines such as IL-4, IL-10, and TGFβ. The only inflammatory mediator induced by B quintana LPS was the chemokine cytokine-induced neutrophil chemoattractant (KC) (343 pg/ml after stimulation with 1 μg/ml B quintana LPS). In comparison, stimulation of mouse peritoneal macrophages with only 10 ng/ml of purified E coli LPS resulted in the production of as much as 2,000 pg/ml KC (Figure 1A). Furthermore, the TLR-4 antagonist did not suppress cytokine production upon stimulation with IL-1, TNFα, and the synthetic TLR-2 and TLR-3 ligands Pam3Cys and poly(I-C), respectively (data not shown).

Figure 1.

Toll-like receptor 4 (TLR-4) antagonist (TLR-4 ant.) Bartonella quintana lipopolysaccharide (LPS) inhibits dendritic cell (DC) maturation upon stimulation with Escherichia coli LPS, and cytokine production upon stimulation with E coli LPS and endogenous TLR-4 ligands. A, Inability of B quintana LPS to induce cytokine production. Murine macrophages were stimulated with 10 ng/ml E coli LPS or 1 μg/ml B quintana LPS, and cytokines were measured by Luminex assay. B and C, Human immature DCs were incubated with 2 μg/ml E coli LPS or 1 μg/ml single-stranded polyuridine (ssPolyU) for 48 hours after preincubation with 10× higher concentrations of B quintana LPS. DC maturation was determined based on class II major histocompatibility complex (MHC) and CD83 expression (fluorescence-activated cell sorting analysis) (n = 2 healthy volunteers). D, Murine macrophages were incubated with 10 ng/ml E coli LPS, 10 μg/ml extra domain A (ED-A) of fibronectin, or 10 μg/ml heparan sulfate (HS), or left untreated after they were pretreated with 10× higher concentrations of B quintana LPS. ED-A and HS were mixed with 10 μg/ml polymyxin B (PMB) prior to use. Values are the mean or mean and SEM. IL-1β = interleukin-1β; TNFα = tumor necrosis factor α; G-CSF = granulocyte colony-stimulating factor; KC = cytokine-induced neutrophil chemoattractant; TGFβ = transforming growth factor β; Med. = medium; MFI = mean fluorescence intensity; ND = not detectable.

In addition, in contrast to E coli LPS, B quintana LPS did not induce the maturation of human monocyte–derived DCs in terms of induction of CD83 expression and class II MHC up-regulation. More important, B quintana LPS had no inhibitory effect on DC maturation that was induced by mechanisms other than TLR-4 activation (e.g., via stimulation of TLR-7) (Figures 1B and C).

Inhibition by TLR-4 antagonist of proinflammatory cytokine induction by endogenous TLR-4 ligands.

There is growing evidence that extracellular matrix components, which are generated by tissue damage during chronic inflammation, can activate TLR-4. Therefore, we examined the ability of the TLR-4 antagonist to block the inflammatory signal induced by some of these endogenous TLR-4 ligands (i.e., ED-A of fibronectin and HS). Mouse peritoneal macrophages were preincubated with the TLR-4 antagonist and then stimulated with LPS, ED-A, or HS. As we expected, stimulation of macrophages with ED-A or HS, which were premixed with excessive amounts of polymyxin B to inhibit the putative LPS contaminant, resulted in the production of IL-1β and TNFα. Preincubation of cells with the TLR-4 antagonist clearly inhibited TNF production upon stimulation with LPS and the endogenous TLR-4 ligands (Figure 1D), and similar results were obtained for IL-1β (data not shown).

Inhibition of disease progression by prophylaxis against CIA using TLR-4 antagonist.

To investigate the role of TLR-4 in the development of arthritis, we inhibited TLR-4 in an experimental model of arthritis. DBA/1 mice with CIA were treated with the TLR-4 antagonist before clinical manifestation of the disease. The disease incidence was not significantly reduced; however, macroscopic evaluation of the paws showed a significant reduction of disease severity in mice treated with the TLR-4 antagonist (Figure 2A).

Figure 2.

Suppression of clinical and histologic signs of collagen-induced arthritis by prophylactic treatment with Toll-like receptor 4 (TLR-4) antagonist. Mice received a total of 3 intraperitoneal injections of TLR-4 antagonist (400 μg/kg body weight) or 0.2 ml saline once every 2 days, starting before the onset of arthritis. A, Mean and SEM macroscopic arthritis score (scale of 0–2 for each paw). ∗ = P < 0.05, saline versus TLR-4 antagonist. B, Histologic analysis of the knee joints on day 6 of treatment, expressed as the mean ± SEM score (scale of 0–3) from 2 independent experiments. All parameters were scored by 2 observers in a blinded manner (≥10 mice per group per experiment). ∗ = P < 0.05 versus saline-treated mice. C, Representative histologic images. Tissue sections were stained with hematoxylin and eosin (top panels) to study inflammatory cell influx and chondrocyte death (arrow) or with Safranin O (bottom panels) to determine proteoglycan (PG) depletion and cartilage and bone destruction (arrows). PG depletion is apparent from loss of red staining. P = patella; C = cartilage; JS = joint space; F = femur; B = bone. (Original magnification × 100 in top panels; × 200 in bottom panels.)

Subsequently, we investigated the effect of prophylactic TLR-4 blocking on the inflammatory cell influx and various hallmarks of cartilage and bone damage. Analysis of the paraffin sections of the knee joints revealed that specific TLR-4 inhibition significantly suppressed PG depletion from the cartilage matrix, the earliest sign of cartilage damage in experimental arthritis. Furthermore, destruction of the cartilage matrix was markedly reduced in mice treated with the antagonist (P = 0.014 for both parameters).

The microscopic score of chondrocyte death was also decreased by the treatment, although nonsignificantly (P = 0.19). TLR-4 blocking had no effect on the inflammatory cell influx and the severity of bone erosion in this setting (Figure 2B). Representative images of histologic analysis demonstrating the effects of anti–TLR-4 treatment before the onset of CIA are shown in Figure 2C.

Strong suppression of joint pathology in ongoing disease by therapeutic treatment of CIA using TLR-4 antagonist.

To investigate whether TLR-4 blocking could ameliorate the ongoing disease in mice, mice with CIA with a macroscopic inflammation score of 0.5–1 were treated with the TLR-4 antagonist. As shown in Figure 3A, therapeutic treatment of CIA resulted in an ∼50% suppression of the clinical score for arthritis.

Figure 3.

Prevention of joint inflammation and cartilage and bone damage by therapeutic treatment of collagen-induced arthritis using Toll-like receptor 4 (TLR-4) antagonist. Mice received 4 daily intraperitoneal injections of TLR-4 antagonist (2 mg/kg body weight) or 0.2 ml saline after a macroscopic inflammation score of 0.5–1 (on a scale with a maximum possible score of 8 per mouse) was reached. A, Mean and SEM macroscopic arthritis score (scale of 0–2 for each paw). ∗ = P < 0.05, saline versus TLR-4 antagonist. B, Histologic analysis of the knee joints on day 4 of treatment, expressed as the mean ± SEM score (scale of 0–3). All parameters were scored by 2 observers in a blinded manner (n = 10 mice per group). ∗ = P < 0.05 versus saline-treated mice. C, Histologic images of the knee joints. Tissue sections were stained with hematoxylin and eosin (top panels) to study inflammatory cell influx and chondrocyte death (arrow) or with Safranin O (bottom panels) to determine proteoglycan (PG) depletion and cartilage and bone destruction (arrows). (Original magnification × 100 in top panels; × 200 in bottom panels.)

Histologic examination of the knee joints revealed that treatment with the TLR-4 antagonist strongly prevented PG depletion and destruction of the cartilage matrix. Chondrocyte death and infiltration of inflammatory cells into the joint space were also dramatically inhibited (P < 0.05 for all parameters). Furthermore, another hallmark of CIA, bone erosion, was suppressed by inhibition of TLR-4 signaling, although nonsignificantly (P = 0.067) (Figure 3B). Figure 3C shows representative images of the knee joints of mice with CIA treated with the TLR-4 antagonist compared with those treated with saline.

Serum cytokine and anti-CII antibody levels.

Treatment of mice with the TLR-4 antagonist before the onset of CIA did not result in any difference in serum concentrations of cytokines and chemokines. In contrast, inhibition of TLR-4 after the onset of arthritis led to a reduction in serum levels of IL-6 and KC, which corresponded well to a reduction of inflammatory cell influx into the joint (data not shown). IL-1β and TNFα concentrations were not different in the 2 groups. Importantly, serum levels of anti-mouse CII antibodies at the end point of the treatment with TLR-4 antagonist did not differ from those in the control groups, in both the prophylactic and therapeutic settings (Figures 4A and B). This finding confirms that treatment with the TLR-4 antagonist did not interact with the development of autoimmune responses driving the initiation and expression of CIA.

Figure 4.

Effects of Toll-like receptor 4 (TLR-4) antagonist on development of autoantibodies against murine type II collagen (CII) and on interleukin-1 (IL-1) expression in the joint. A and B, No effect on serum levels of anti-CII antibodies (as determined by enzyme-linked immunosorbent assay) by prophylactic (A) or therapeutic (B) treatment with the TLR-4 antagonist. OD = optical density. C and D, Lower expression of IL-1β protein in articular chondrocytes and synovial tissue resulting from prophylactic (C) and therapeutic (D) treatment with TLR-4 antagonist. See Materials and Methods for details of scoring of IL-1β expression. Values in AD are the mean ± SEM. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus saline-treated mice. E, Immunohistochemical staining of IL-1β in the knee joint. Arrows indicate high IL-1 expression in synovium at the site of bone erosion. C = cartilage; S = synovium; B = bone. (Original magnification × 200.)

Reduction of local production of IL-1β in the joint by treatment with TLR-4 antagonist.

IL-1 is considered the main mediator of cartilage PG depletion and destruction and bone erosion during CIA (30). Immunohistochemical staining of IL-1β showed that IL-1 was highly expressed in the synovium of control mice, especially at the sites of bone erosion. Inhibition of TLR-4 before as well as after the onset of CIA resulted in lower expression of IL-1β protein in the joints. IL-1β expression was reduced in chondrocytes of articular cartilage and also in synovial tissue surrounding patellar, tibial, and femoral surfaces of the knee joint (Figures 4C–E).

Blockage of progression of ongoing arthritis in IL-1Ra−/− mice by TLR-4 inhibition.

Spontaneous development of arthritis in IL-1Ra−/− mice reflects an IL-1–mediated autoimmune process that progresses with age (26). Arthritis in these mice is represented by an aggressive pannus-forming synovitis accompanied by cartilage and bone destruction. To confirm the relevance of TLR-4 activation in driving IL-1–mediated joint pathology during the chronic phase of arthritis, IL-1Ra−/− mice with ongoing disease were treated with a TLR-4 antagonist. Consistent with the findings in CIA, the severity of joint inflammation was clearly reduced in anti–TLR-4–treated mice compared with saline-treated control mice, in which arthritis became more aggravated in time (Figure 5A).

Figure 5.

Prevention of spontaneous joint inflammation and cartilage and bone damage in interleukin-1 receptor antagonist–deficient (IL-1Ra−/−) mice by therapeutic inhibition of Toll-like receptor 4 (TLR-4). A, Mean ± SEM macroscopic arthritis score (scale of 0–2 for each paw) in IL-1Ra−/− mice treated with TLR-4 antagonist (400 μg/kg body weight) 3 times a week after arthritis onset compared with that in saline-treated control mice. Only hind paws developed arthritis and were monitored. B, Histologic analysis of the ankle joints on day 14 of the treatment, expressed as the mean ± SEM score (scale of 0–3). ∗ = P < 0.05 versus saline-treated mice. PG = proteoglycan.

Histologic analysis of the ankle joints revealed that inhibition of TLR-4 significantly protected the cartilage against PG depletion and chondrocyte death (P = 0.033 and P = 0.041, respectively). Furthermore, treatment with the TLR-4 antagonist resulted in substantial suppression of cartilage destruction and bone erosion perceptible in several joints of the ankle (P = 0.028 and P = 0.032, respectively). Synovial inflammation (lining cell proliferation and invasion of inflammatory cells into the joint) was also reduced, although nonsignificantly (P = 0.09) (Figure 5B).

No induction of corticosteroids, antiinflammatory cytokines, or local chemotactic events in vivo by the TLR-4 antagonist.

To exclude the possibility that the inhibitory effect of the TLR-4 antagonist on progression of arthritis is mediated through the induction of antiinflammatory cytokines or corticosteroids, naive male DBA/1 mice were injected IP with 2 mg/kg body weight of the antagonist, 400 μg/kg body weight E coli LPS, or an equal volume of saline. Corticosterone and cytokine concentrations were measured in serum 90 minutes, 4 hours, and 24 hours after injection.

Table 1 shows that systemic injection of B quintana LPS, in contrast to E coli LPS, did not induce the production of TNFα, IL-1β, IL-6, and IL-10, indicating that the binding of B quintana–derived LPS to the TLR-4 receptor complex does not lead to the common NF-κB activation. Furthermore, no substantial levels of KC were found in serum after injection of B quintana LPS. These cytokines were not detectable in serum 24 hours after injection, and IL-4 was not detected at all. Injection of E coli LPS also led to the production of high levels of corticosterone within 90 minutes, whereas animals treated with B quintana LPS had corticosterone levels comparable with those of saline-treated animals. Furthermore, examination of the peritoneal cell population 24 hours after IP injection of B quintana LPS excluded a potential chemotactic activity of TLR-4 antagonist at the site of injection as a mechanism for the suppression of inflammation and tissue damage in arthritic joints, since both the total count and type of cells were not affected by the TLR-4 antagonist (data not shown).

Table 1. Cytokine and corticosterone levels after systemic injection of Bartonella quintana LPS or Escherichia coli LPS*
 90 minutes after injection4 hours after injection
SalineE coli LPSB quintana LPSSalineE coli LPSB quintana LPS
  • *

    Values are the mean ± SD. Naive DBA/1 mice were injected intraperitoneally with 400 μg/kg E coli lipopolysaccharide (LPS) or 2 mg/kg B quintana LPS, and corticosterone and cytokine concentrations were measured in serum 90 minutes and 4 hours after injection. TNFα = tumor necrosis factor α; IL-1β = interleukin-1β; ND = not detectable; KC = cytokine-induced neutrophil chemoattractant.

TNFα, pg/ml10.4 ± 3.71,965.2 ± 26.120.6 ± 8.85.6 ± 1.2269.5 ± 38.813.0 ± 9.4
IL-1β, pg/mlND15.2 ± 5.3NDND36.9 ± 13.22.0 ± 1.7
IL-6, pg/ml40.1 ± 22.38,941.7 ± 108.852.4 ± 31.340.5 ± 15.09,823.5 ± 246.950.7 ± 25.9
KC, pg/ml18.9 ± 5.25,302.0 ± 235.626.5 ± 4.815.3 ± 2.86,523.0 ± 108.117.4 ± 4.4
IL-10, pg/mlND232.2 ± 13.93.7 ± 0.1ND75.25 ± 23.6ND
Corticosterone, nmoles/liter295.3 ± 112.91,846.7 ± 523.5421.0 ± 135.8372.6 ± 112.92,010.3 ± 151.0348.3 ± 125.9

DISCUSSION

We report for the first time that inhibition of TLR-4 activation suppresses the clinical manifestations (i.e., swelling and redness) and histologic manifestations of arthritis in early-phase as well as established disease in mice. Inhibition of joint inflammation and cartilage damage was accompanied by reduced IL-1β expression. Our observations strongly suggest a proinflammatory role for TLR-4 in 2 chronic non–LPS-driven models of autoimmune arthritis.

TLRs were originally thought to have a function only in sensing pathogen-associated molecules. Activation of TLRs by these molecules has been proven to play a key role in the development and progression of various chronic infectious diseases depending on the expression of TLRs at sites of contact with bacteria. For instance, the Asp299Gly polymorphism of TLR-4, which is expressed on intestinal epithelial cells, was recently associated with Crohn's disease (31), and a synthetic TLR-4 antagonist was shown to inhibit the development of 2 experimental models of inflammatory bowel disease (32). Despite the concerns regarding possible LPS contamination, it is currently believed that some damage-associated components of extracellular matrix can activate TLR-4, and it was therefore hypothesized that TLR-4 activation may also be involved in several non–infectious disease conditions based on the “danger model” of autoimmunity (33). Consistent with this hypothesis, TLR-4–deficient mice have been shown to exhibit less myocardial and hepatic ischemia-reperfusion injury compared with wild-type animals (34, 35). Very recently, it was demonstrated that interaction of hyaluronan degradation products with TLR-2 and TLR-4 provides signals to initiate inflammation after noninfectious lung injury, whereas TLR-2 and TLR-4 serve to maintain epithelial cell integrity and tissue repair by sensing native high-molecular-mass hyaluronan (36).

In the present study, we explored the contribution of TLR-4 to the pathogenesis of RA, using the CIA and the spontaneous IL-1Ra−/− models of arthritis. TLR-4 is expressed by macrophages and fibroblasts in synovial lining and is up-regulated in moderately inflamed synovium of RA patients (37). CIA was chosen as a model of chronic joint inflammation that is accompanied by gradual cartilage and bone erosion and possibly leads to the production of endogenous TLR-4 agonists.

We blocked TLR-4 in this model using LPS from an intracellular gram-negative bacillus, B quintana. The antagonistic activity of B quintana LPS has been examined extensively in parallel studies, in which microarray analysis revealed that highly purified B quintana LPS does not affect gene expression in human peripheral blood mononuclear cells, and it completely blocks gene regulation by E coli LPS (25). Although B quintana LPS seems to induce KC production weakly in murine macrophages, it is still an effective TLR-4 antagonist in mice. The most likely explanation of the mechanism of TLR-4 inhibition by B quintana LPS is that it competes with TLR-4 agonists for a common binding site on TLR-4 or the TLR-4–myeloid differentiation 2 receptor complex; however, only ligand-binding assays can prove this hypothesis. The antagonistic activity of this LPS is probably caused by the cylindrical conformation of its lipid A due to the presence of a long-chain fatty acid, a characteristic that has been described for Bartonella henselae LPS and is known to reduce endotoxicity of LPS (38). Long-chain fatty acids are also present in other types of LPS with TLR-4–antagonistic properties, such as LPS of Helicobacter pylori and Rhodobacter capsulatus (39, 40). Further support for the TLR-4–blocking activities of B quintana LPS was provided by inhibition of cytokine release upon stimulation with endogenous TLR-4 ligands in the presence of excessive amounts of polymyxin B and by inhibition of DC maturation upon stimulation with E coli LPS.

Inhibition of the TLR-4 pathway using this TLR-4 antagonist substantially suppressed clinical and histologic manifestations of CIA. The dose of the antagonist in the prophylactic CIA study (400 μg/kg body weight) was based upon the reported dose of other TLR-4 antagonists with similar structure (lipid A analog E5564) (41). For therapeutic treatment, we enhanced the dosage and shortened the intervals, because we expected generation and release of more inflammation- and damage-associated TLR-4 agonists during existing disease. Inhibition of TLR-4 after onset of CIA proved to be more effective than such inhibition before onset. This might be explained by the presence of more endogenous TLR-4 agonists in ongoing disease.

We performed a broad range of control studies to exclude nonspecific mechanisms of inhibition of joint inflammation and damage such as induction of antiinflammatory cytokines and corticosteroids, as well as recruitment of polymorphonuclear cells to the peritoneal cavity after IP injection of the antagonist. None of these mechanisms was found to be responsible for the suppression of CIA in our experiments. Furthermore, the development of normal titers of autoantibodies against murine CII after treatment with TLR-4 antagonist showed that TLR-4 inhibition did not interfere with the humoral immune response against foreign CII and the cross-reactivity to self collagen. In addition, a recent study demonstrated that TLR-4–knockout mice exhibit normal delayed-type hypersensitivity and lymph node cell proliferation in response to a retinal antigen (42), indicating that the lack of TLR-4 activation does not disrupt the Th1-mediated immune response that is an essential participant in the induction of CIA.

To confirm the relevance of TLR-4 activation in arthritis, TLR-4 antagonist was administered to IL-1Ra−/− BALB/c mice, in which a chronic polyarthritis closely resembling RA develops spontaneously. Arthritis in these mice arises from disturbed immune homeostasis due to excessive IL-1 signaling, which results in the induction of costimulatory molecules such as OX40 and CD40 ligand on T cells, thereby enhancing APC–T cell interaction and mediating T cell autoimmunity (26, 27). For therapeutic treatment of IL-1Ra−/− mice, we chose a lower dose but a longer treatment course than that for existing CIA because of the milder joint destruction and the more chronic progress of arthritis in this model. Here we show that anti–TLR-4 treatment substantially improves both the clinical inflammation score and the histologic characteristics of arthritis in this model. While the inhibition of inflammatory cell influx into the joint was less pronounced, disruption of TLR-4 signaling had strong protective effects on cartilage and bone. Our recent studies, in which we crossed IL-1Ra−/− mice into TLR-4−/− animals on a BALB/c background, show that IL-1Ra−/− TLR-4−/− mice develop a clearly less progressive arthritis compared with IL-1Ra−/− TLR-4+/+ littermates at later stages of the disease (43). This underscores the involvement of TLR-4 activation in the chronic phase of the disease, when the existing inflammation allows the formation of endogenous TLR-4 agonists.

Synovial macrophages and fibroblasts may be the first responders to non–pathogen-associated endogenous TLR-4 agonists, thereby contributing to the development and progression of arthritis. TLR-4 activation of APCs stimulates these cells to produce neutrophil- and lymphocyte-attracting chemokines (44, 45) and to activate T cells by providing costimulatory signals (46, 47). The infiltrated cells also express TLR-4 and can be activated by products of extracellular matrix degradation, leading to persistent activation of the innate and adaptive immune systems. In an attempt to establish the presence of endogenous TLR-4 activation in RA, we cultured synovial biopsy specimens from patients with active RA with the TLR-4 antagonist B quintana LPS, and we found that inhibition of TLR-4 clearly reduced levels of IL-1 and TNF produced by cultured synovium (43).

An important mediator of cartilage and bone degradation is IL-1, which is produced upon TLR-4 activation (30). In cooperation with other cytokines, IL-1 promotes the production of nitric oxide and tissue-destructive enzymes, the activation of osteoclasts, and other catabolic events in the arthritic joint (1). Therefore, the protective effects of anti–TLR-4 treatment are, at least in part, attributed to reduced IL-1 production, as demonstrated by the lower expression of IL-1 in articular chondrocytes and synovial tissue of the knee joints after treatment with TLR-4 antagonist. This suggests that TLR-4 might function upstream of proinflammatory mediators such as IL-1 in certain stages of arthritis. Whether continuation of anti–TLR-4 treatment would have the same effects in later phases is an issue for further investigation. Another important question is whether short-term inhibition of TLR-4 would affect the further long-term progress of arthritis. Since endogenous TLR-4 agonists are expected to be more abundant in longstanding disease, and inhibition of IL-1 has been proved to be effective in late stages of CIA (29), we expect that anti–TLR-4 treatment at later phases will still suppress the disease in part.

To our knowledge, these data are the first to demonstrate the protective effects of anti–TLR-4 treatment in 2 autoimmune models of arthritis that are not driven by exogenous TLR-4–activating microbial adjuvants. The immunologic mechanisms of these protective effects involve reduced cytokine production and the inhibition of their deleterious effects in the joint. Further studies are warranted to investigate the exact source and nature of TLR-4–activating molecules and their arthritogenic capacity in RA joints, and especially the potential of TLR-4 blockade as a therapeutic strategy in chronic inflammatory diseases such as RA.

AUTHOR CONTRIBUTIONS

Dr. Abdollahi-Roodsaz had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Abdollahi-Roodsaz, Joosten, van der Meer, Netea, van den Berg.

Acquisition of data. Abdollahi-Roodsaz, Joosten, Roelofs, Popa, Netea, van den Berg.

Analysis and interpretation of data. Abdollahi-Roodsaz, Joosten, Roelofs, Radstake, Popa, van der Meer, Netea, van den Berg.

Manuscript preparation. Abdollahi-Roodsaz, Joosten, Matera, van der Meer, Netea, van den Berg.

Statistical analysis. Abdollahi-Roodsaz.

B quintanaLPS extraction. Matera.

Acknowledgements

We are grateful to Prof. D. Raoult (Marseilles, France) for providing B quintana and to Prof. A. Foca (Catanzaro, Italy) for supplying B quintana LPS. We thank Dr. M. Nicklin (Sheffield, UK) for providing IL-1Ra−/− mice. We would also like to thank M. Helsen, B. Oppers-Walgreen, and C. Arndtz for their support in in vivo studies and in tissue processing for histologic analysis and Dr. M. Koenders for valuable discussions.

Ancillary