Toll-like receptor 3 activation promotes joint degeneration in osteoarthritis

Osteoarthritis (OA) is characterized by cartilage degradation that is induced by inflammation. Sterile inflammation can be caused by damage-associated molecular patterns that are released by chondrocytes and activate pattern recognition receptors. We evaluate the role of toll-like receptor-3-activating RNA in the pathogenesis of OA. Toll-like receptor 3 (TLR3) was detected by semiquantitative reverse transcriptase PCR, western blotting and microscopy. Rhodamine-labelled poly(I:C) was used to image uptake in chondrocytes and full-thickness cartilage. The production of IFNβ in chondrocytes after stimulation with poly(I:C) as well as in the synovial fluid of OA patients was measured using ELISA. Chondrocyte apoptosis was chemically induced using staurosporine. Immunohistochemistry was performed to examine TLR3 expression and apoptosis in human and murine OA cartilage. RNA in synovial fluid was quantified by RiboGreen assay. Destabilisation of the medial meniscus was performed in TLR3−/− and wildtype mice. OA was assessed after eight weeks using OARSI score. TLR3 expression was confirmed by western blot and RT-PCR. Poly(I:C) was internalised by chondrocytes as well as cartilage and caused an increase of IFNβ production in murine (11.46 ± 11.63 (wo) to 108.7 ± 25.53 pg/ml; N = 6) and human chondrocytes (1.88 ± 0.32 (wo) to 737.6 ± 130.5 pg/ml; N = 3; p < 0.001). OA cartilage showed significantly more TLR3-positive (KL0 = 0.22 ± 0.24; KL4 = 6.02 ± 6.75; N ≥ 15) and apoptotic chondrocytes (KL0 = 0.6 ± 1.02; KL4 = 9.78 ± 7.79; N ≥ 12) than healthy cartilage (p < 0.001). Staurosporine-induced chondrocyte apoptosis causes a dose-dependent RNA release (0 ng/ml = 1090 ± 39.1 ng/ml; 1000 ng/ml=2014 ± 160 ng/ml; N = 4; p < 0.001). Human OA synovial fluid contained increased concentrations of RNA (KL0-2 = 3408 ± 1129 ng/ml; KL4 = 4870 ± 1612ng/ml; N ≥ 7; p < 0.05) and IFNβ (KL0-2 = 41.95 ± 92.94 ng/ml; KL3 = 1181 ± 1865ng/ml; N ≥ 8; p < 0.05). TLR3−/− mice showed reduced cartilage degradation eight weeks after OA induction (OARSI WT = 5.5 ± 0.04; TLR3−/− = 3.75 ± 1.04; N ≥ 6) which was accompanied by gradually decreasing levels of TUNEL-positive cells (WT = 34.87 ± 24.10; TLR3−/ = 19.64 ± 7.89) resulting in decreased IFNβ expression (WT = 12.57 ± 5.43; TLR3−/− = 6.09 ± 2.07) in cartilage (p < 0.05). The release of RNA by apoptotic chondrocytes thus activating TLR3 signalling is one possible way of perpetuating inflammatory cartilage changes. The inhibition of TLR3 could be a possible therapeutic target for OA treatment.


INTRODUCTION
Osteoarthritis (OA) is a progressive joint disease, which is associated with severe pain and impairment of movement which results in a significant reduction of the quality of life. OA is a degenerative disease that is characterized by progressive structural changes in joint tissues. Especially the articular cartilage is associated with cartilage fibrillation and erosions accompanied by chondrocyte hypertrophic differentiation and changes in extracellular matrix composition. Different factors, including cytokines, growth factors and Wnts have been identified as being involved in OA disease progression. Innate inflammatory signals such as damage-associated molecular patterns (DAMPs) have been described to activate pattern-recognition receptors (PRRs), which induce the expression of matrix metalloproteinase (MMP) as well as disintegrin and metalloprotease with thrombospondin motif (ADAMTS), resulting in cartilage thinning with progressive loss of proteoglycans and collagen [1,2].
The expression of different TLRs (toll-like receptors) in cartilage has been described [3]. Current research, however, focuses mainly on the role of TLR2 and 4 in OA [4,5]. These TLRs have been studied in cartilage and chondrocytes extensively, however, not much is known about the role of nucleic-acid binding TLRs such as TLR3 and TLR7-9.
Toll-like receptor 3 (TLR3) is a PRR and type I transmembrane receptor with an extracellular domain located in endosomes [6,7]. Upon RNA or poly(I:C) binding and receptor dimerization TRIF is recruited inducing an activation pathway of IRF3, NFkB, INFβ and proinflammatory cytokine gene expression [8,9]. The group of TLRs plays an essential role in the pathogenesis of chronic inflammatory disease such as rheumatoid arthritis, interacting with endogenous ligands that originate from degraded extracellular matrix or dying cells [10,11]. Furthermore, epidemiologic data shows that polymorphisms in the promoter region of TLR3 are associated with primary osteoarthritis (OA) [12]. Gene expression studies identified a range of PRRs in OA cartilage including TLR3, retinoic acid-inducible gene 1 and nucleotide-binding oligomerization domain-like receptor X1, which are capable of binding endogenous nucleic acids [13]. Because MMP expression in chondrocytes is up-regulated after Poly(I: C) stimulation, the modulation of TLR signalling was initially suggested as a potential therapeutic strategy [14]. However, the simulation of a viral joint infection by intra-articular poly(I:C) injections induced signs of arthritis that were also seen in TLR3deficient (TLR3 −/ ) mice [15]. Although a recent study suggests that poly(I:C)-induced arthritis is regulated by the TLR3-p38 MAPK-NF-κB Il-33 pathway, which is modulated by the p65 and peroxisome proliferator-activated receptor-γ (PPARγ) complex, the role of TLR3 activation in the pathogenesis of osteoarthritis remains elusive [16].
Current research shows that nucleotides, being a potential ligand for nucleotide binding TLRs, are released during chondrocyte cell death after joint trauma [17]. Furthermore, in vivo chondrocyte depletion models suggest that the cartilage is protected from degenerative changes [18,19]. Thus, endogenous ligands might contribute to degenerative pathways in OA. The aim of this study is to investigate whether endogenous RNA could activate TLR3 in cartilage, thereby inducing a sterile inflammation and contributing to cartilage degeneration during OA.

MATERIALS AND METHODS OA cartilage samples
Human OA articular cartilage was obtained from patients undergoing joint replacement for knee OA after obtaining written consent (in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975, as revised in 2000). Healthy cartilage samples were harvested from body donors of the forensic department within 24 h after death (IRRB Magdeburg medical faculty: 23/16). Full thickness cartilage samples were dissected from the main loaded areas of the joint. Safranin-O staining was applied for OARSI scoring [40]. Two independent graders assessed the OARSI score in a blinded manner.
Quantitative Reverse-Transcriptase PCR  Table 1. GAPDH was used as housekeeping gene for normalisation. Relative quantification was performed using a standard curve.

Nucleic Acid Internalization
Human chondrocytes (P1) were plated at a density of 8 × 10 3 cells/cm 2 on a glass cover slip in a 24-well plate. Rhodamine-conjugated poly(I:C) (Invivogen, San Diego, USA) or FITC labelled CpG ODN (Invivogen, San Diego, USA) was added and cells were incubated for 24 h and mounted with ROTI ® Mount FluorCare DAPI (Carl Roth, Karlsruhe, Germany). Unconjugated rhodamine or FITC (Invivogen, San Diego, USA) at respective concentrations were used as negative control. Murine hip caps were obtained from 4 weeks old mice. Whole cartilage samples were incubated with rhodamine-conjugated poly(I:C), FITC labelled CpG ODN, unconjugated rhodamine or FITC for 36 h. Again, DAPI was used as counterstain. Samples were embedded in TissueTek (Sakura, Alphen aan den Rijn, Netherlands). Each experiment was performed at least 3 times and all samples were analysed using a Zeiss confocal laser scanning system LSM 510 meta. TLR3 immunostaining C-28/I2 human chondrocytes were seeded at a density of 8 × 10 3 cells/cm 2 on a glass cover slip in a 24-well plate (N ≥ 3). After 24 h cells were fixed with 4% PFA. Human cartilage sections (N ≥ 15) were pre-treated using sodium citrate buffer adjusted to pH6. Anti-TLR3 antibody (1:500, NBP2-24875, Novus, Centennial, USA) or murine IgG diluted 1:50 in Tris-buffered saline (TBS) were applied over night at 4°C. The Alexa Fluor® 488 anti mouse from donkey was used as secondary antibody. Fluorescence microscopy was performed using Axio Observer.Z1 (Zeiss, Oberkochen, Germany). TLR3 and DAPI positive cells were counted by the help of ImageJ.

TLR3 knockdown
Primary human chondrocytes (N = 3) were transformed using jetPRIME® (Illkirch-Graffenstaden, France) and either 50 µM siTLR3 (Silencer® Select, #s8862, AMBION GmbH, Kaufungen, Germany) or 50 µM siScrambled (Silencer® Negative Control #1, AMBION GmbH, Kaufungen, Germany) per well. To measure the knockdown efficacy a semiquantitative RT-PCR was performed.  [43]. Sham surgery was done on the contralateral limb by a medial skin incision and wound closure. Animal care was in accordance with institution guidelines. After eight weeks mice were euthanized and both knees were processed for histology. Two independent graders assessed the sections of the tibia and femur in a blinded manner using the OARSI scoring system [44].

TUNEL assay
Murine (N ≥ 6) and human (N > 8) cartilage sections were stained using the "in situ cell death detection kit" (Merck, Darmstadt, Germany) according to the manufacturer's instructions. Fluorescence microscopy was performed using Axio Observer.Z1 (Zeiss, Oberkochen, Germany). TUNEL and DAPI positive cells were manually marked and put into relation ((TUNEL positive cells/total number of cells)) *100) with the help of ImageJ.

Chemically induced apoptosis
Apoptosis in primary human chondrocytes was chemically induced by staurosporine (LKT laboratories, Minnesota, USA) using concentrations of 100 ng/ml, 200 ng/ml, 500 ng/ml and 1000 ng/ml for 24 h [45]. The amount of extracellular RNA in the supernatant was measured using RiboGreen ® assay (Thermo Scientific, Waltham, USA).

Statistical analysis
Data are presented as means + /-SEM (parametric) or median ± 95% CI (non-parametric). According to the data distribution, student's t-test, ordinary one way-ANOVA-analysis, Kruskal-Wallis test or Mann-Whitney-Utest were performed using GraphPad Prism Software, V.6.0 h (GraphPad Software Inc), with p < 0.05 determining the primary level of significance. All analyses were fully explorative without adjustment for multiple test problems. All results are interpreted accordingly.

Inhibition of TLR3-dependent signalling protects mice from OA-like cartilage changes
To evaluate if the knockout of TLR3 might protect mice from degradative cartilage changes, knee OA was induced using the DMM model in 10-week-old wild type and TLR3 −/− mice. After eight weeks there was a difference between TLR3 −/− and wildtype OA knees (

DISCUSSION
The most important finding of this study was that TLR3 contributes to degenerative changes in OA. In summary, the data shows that TLR3 is expressed in human and murine cartilage and that intra-articular release of ds nucleotides by apoptotic chondrocytes is one possible pathway causing TLR3 activation, subsequent IFNβ release and osteoarthritic changes within the joint.
TLR activation by DAMPs during OA has been demonstrated in recent studies [20][21][22][23]. A potential role of TLR3 in OA cartilage has been indicated by an up-regulation of gene expression and a description of functional polymorphisms in the promoter region that have been associated with a higher susceptibility for OA [12,14]. Li and colleagues observed that in supernatants of damaged cartilage, TLR3 activation was the main stimulus for IL-33, MMP-1 and MMP-3 expression in vitro [16]. Furthermore, increased IFNβ expression after TLR3 activation has already been shown in RA synovial tissue and fibroblasts [24]. However, while mounting evidence of TLR3 presence during the pathogenesis of OA has been presented, the underlying mechanisms of TLR3dependent signalling contributing to the disease progression remain elusive. To shed further light on the role of TLR3 in cartilage degradation during OA our study confirms the presence of TLR3 on the RNA and protein level in cartilage. Cartilage samples of patients with OA showed significantly more TLR3 positive cells than healthy controls, which proves the clinical relevance of TLR3 in cartilage.
Chondrocyte death has been identified in the early stages of OA [22,25]. Programmed or necrotic cell death causes the release of both dsDNA and dsRNA, making TLR3 an endogenous sensor of tissue necrosis [16,17,26]. Contrary to previous reports, recent literature suggests that chondrocytes are the key player in inflammatory changes in OA. Zhang et al. demonstrated that the induced death of superficial chondrocytes protects from degenerative cartilage changes [18]. Similarly, the removal of senescent chondrocytes in a post-traumatic OA mouse model attenuates disease progression [19]. However, the mechanisms driving these inflammatory changes are not yet clear. The data of this study shows a significant increase of dsRNA in the supernatant of human OA synovial fluid in vivo and in vitro after staurosporine stimulation in comparable concentrations. Chondrocytes are able to take up these potential TLR ligands even through the dense peri-cellular matrix to the endosomal location of TLR3 [27]. The confocal imaging in this study indicates that nucleic acid is taken up in cartilage tissue as well as in monolayer chondrocytes. Furthermore, biomechanical studies suggest that deterioration of the collagenproteoglycan network such as in OA increases cartilage permeability which again might ease the access of inflammatory ligands to the chondrocytes [17,[28][29][30]. Thus, ds nucleotides are a possible mediator between chondrocyte death and induction of inflammatory changes.
Our results show that TLR3 is up-regulated in OA, which induces a pro-inflammatory cell response [31]. TLR3 downstream signal transduction results in IFNβ production [16,32]. However, the effect of IFNβ on inflammation in OA continues to be a topic of basic and clinical research [33][34][35]. Furthermore, other PRRs such as the RIG-1, MDA-5 and TLR-7, -8, -9 also recognize nucleic acids and might also contribute to the described inflammatory changes [13]. This might be the reason why TLR3 knockdown chondrocytes did not show a complete suppression of IFNβ in this study. However, specific TLR3 inhibition by CU CPT 4a was highly effective. Future research is required to expand the knowledge about the role of other nucleotide-binding endogenous TLRs and PRRs in OA [36,37].
So far, there are two in vivo studies investigating the effect of TLR3 in mediating joint degeneration. Zare et al. used intraarticular poly(I:C) injections and observed no significant difference in synovial inflammation between TLR3 −/− and WT mice. However, they sacrificed the mice after three days, investigating only the initial inflammatory reaction [15]. Li et al. demonstrated that inflammatory changes caused by intra-articular Poly(I:C) injections were significantly dampened by an IL-33-neutralizing antibody [16]. Both models, however, did not investigate the longterm effects of joint destabilization in an OA mouse model. By destabilizing the medial meniscus we chose a well-established OA model that mimics post-traumatic OA rather than chemically induced OA [38]. We found that the knockout of TLR3 significantly reduced the OA-related cartilage degradation in the animals and slowed down OA progression by up to eight weeks. Future research needs to evaluate the systemic and synovial inflammatory effect of TLR3 on OA progression and long-term effects. Furthermore, the data of this study does not clarify to what extent TLR3 activation causes programmed chondrocyte death and whether joint injury causes cell death, then inducing TLR3-based inflammation [39]. While human OA cartilage samples show an increase of both TUNEL and TLR3 positive cells, TLR −/− mice showed significantly fewer TUNEL positive chondrocytes. This indicates that TLR3-induced programmed cell death plays a secondary role in the pathology of OA [36]. However, literature shows clear evidence of inflammation-based TLR3-mediated cartilage degeneration, which leaves this question unanswered [16].
Recent OA research has been experiencing a paradigm shift in that it shows that sterile inflammation caused by DAMPs contributes to the disease's progression [36]. Our work indicates that joint trauma induces chondrocyte death and the subsequent release of ds nucleotides. As suggested by other authors, chondrocytes seem to be key drivers of degenerative cartilage changes and are able to take up nucleotides [18,19]. Our in vivo experiments underline these in vitro findings and show that TLR3 knockout protects mice from OA-like cartilage degradation. Thus the inhibition of TLR3 signalling could be a possible therapeutic target for osteoarthritis.