Tumor necrosis factor alpha-dependent aggrecan cleavage and release of glycosaminoglycans in the meniscus is mediated by nitrous oxide-independent aggrecanase activity in vitro

Introduction Little is known about factors that induce meniscus damage. Since joint inflammation appears to be a causative factor for meniscal destruction, we investigated the influence of tumor necrosis factor (TNFα) on glycosaminoglycan (GAG) release and aggrecan cleavage in an in vitro model. Methods Meniscal explant disks (3 mm diameter × 1 mm thickness) were isolated from 2-year-old cattle. After 3 days of TNFα-treatment GAG release (DMMB assay), biosynthetic activity (sulfate incorporation), nitric oxide (NO) production (Griess assay), gene expression of matrix-degrading enzymes (quantitative RT-PCR, zymography), and immunostaining of the aggrecan fragment NITEGE were determined. Results TNFα induced release of GAG as well as production of NO in a dose-dependent manner, while sulfate incorporation was decreased. TNFα increased matrix metalloproteinase (MMP)-3 and a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS)-4 mRNA expression, whereas collagen type I was decreased, and aggrecan, collagen type II as well as MMP-1, -2, -13 and ADAMTS-5 were variably affected. Zymography also showed a TNFα-dependent increase in MMP-3 expression, but pre-dominantly in the pro-form. TNFα-dependent formation of the aggrecanase-specific aggrecan neoepitope NITEGE was induced. Tissue inhibitor of metalloproteinases (TIMP)-3, but not TIMP-1 or -2 inhibited TNFα-dependent GAG release and NITEGE production, whereas inhibition of TNFα-dependent NO generation with the NO-synthetase inhibitor L-NMMA failed to inhibit GAG release and NITEGE production. Conclusions Our study shows that aggrecanase activity (a) is responsible for early TNFα-dependent aggrecan cleavage and GAG release in the meniscus and (b) might be involved in meniscal degeneration. Additionally, the meniscus is a TNFα-dependent source for MMP-3. However, the TNFα-dependent NO production seems not to be involved in release of proteoglycans under the given circumstances.


Introduction
Meniscal function and integrity are crucial for a healthy knee joint, because damage to the tissue subsequently leads to articular cartilage destruction and further degenerative diseases such as osteoarthritis (OA) [1][2][3]. In order to restore the meniscal function it is important to understand the pathomechanisms of meniscal destruction.
Increased levels of nitric oxide (NO) and pro-inflammatory cytokines, such as TNFα and IL-1, have been found in the synovial fluid and tissues of inflamed joints [4,5]. It is also well established that cytokines can be involved in cartilage tissue or proteoglycan degradation [6]. It has recently been shown in a serum-containing porcine in vitro model that these cytokines are able to inhibit the intrinsic meniscal repair response [7,8], and part of this effect has been found to be mediated by the ADAMTS activation of matrix metalloproteinases (MMPs) [9,10]. The patterns of enzyme expression during experimental OA suggest that there are similarities in the involvement of MMPs and aggrecanases in the degradation of menisci and articular cartilage [11]. It is therefore suggested that members of the MMPs as well as the a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS) family, such as ADAMTS-4 (aggrecanase-1) and ADAMTS-5 (aggrecanase-2), must also be involved in cytokine-dependent degradation of proteoglycans in the meniscus. Meniscal expression and biomechanical regulation of all these enzymes has recently been shown in a porcine tissue explant model [12]. Aggrecanases are known to be responsible for aggrecan degradation in articular cartilage in diseases such as OA and rheumatoid arthritis (RA) [13], and cleave the aggrecan core protein at several specific sites; one is between Glu 373 and Ala 374 which generates the G1-NITEGE fragment [14,15].
In order to investigate the influence of TNFα on the meniscus we present a bovine in vitro model that allows the isolation of meniscal tissue explants of defined geometry and anatomical location. Using this model we study the effect of TNFα on glycosaminoglycan (GAG) release, biosynthetic activity, NO production, aggrecan fragmentation (because aggrecan has been described as one of the major proteoglycans in the meniscus [24]), and gene expression of matrix molecules, MMPs and aggrecanases in the meniscus. We demonstrate that within three days of incubation there is a TNFα-dependent up-regulation of MMP-3 and ADAMTS-4 expression, as well as aggrecanase activity. The latter induces GAG release, cleaves aggrecan at the NITEGE site and is independent of the TNFαinduced NO production.

Isolation and culturing of meniscal explant disks
Meniscal explant disks were isolated from bovine menisci (from 16 to 24 month old cattle), procured from a local abattoir with authorization from the relevant meat inspectors. This study does not involve human subjects, human tissue or experimentation of animals. Up to four full thickness tissue cylinders (10 mm in diameter) per meniscus were punched perpendicular to the meniscus bottom surface. Tissue disks 1 mm in thickness were sliced including the original meniscal surface using a sterile scalpel blade, and four to five smaller explant disks (3 mm in diameter × 1 mm thick) were isolated using a biopsy punch (HEBUmedical, Tuttlingen, Germany) and cultured in DMEM (supplemented with 100 U/ml penicillin G, 100 μg/ml streptomycin, and 0.25 μg/ml amphotericin B; Sigma-Aldrich, St. Louis, MO, USA) in a 37°C, 5% CO 2 environment after measurement of wet weight. The total of up to 60 explants per animal (2 knee joints including medial/lateral menisci) were randomised among the different experimental groups matched by their anatomical location for every single experiment and cultured in the absence or presence of varying concentrations of recombinant human TNFα (R & D Systems, Minneapolis, MN, USA). In most of the experiments a concentration of 100 ng TNFα/ml was used. Three explant disks per well of a 24well plate were cultured in 1 ml medium. After three days of culture the medium and explants were used for measurements. For inhibitory studies different tissue inhibitor of metalloproteinases (TIMPs; R & D Systems, Minneapolis, MN, USA) and the NO synthetase inhibitor L-NMMA were used. For these investigations only one meniscal explant per well was cultured for three days in 200 μl medium in 96-well plates.

Immunohistochemistry
The meniscal explants were fixed overnight in 4% paraformaldehyde and embedded in paraffin. Serial sections (7 μm) were cut sagittally through the entire thickness of the explant disks, immobilised on glass slides, and deparaffinised. After incubation for 2.5 minutes in a digester at 100°C (in 0.01 M citric acid, pH 6.0), they were incubated overnight at 4°C with the primary antibody (anti-NITEGE; 1:50 dilution in 1% BSA; ABR Affinity BioReagents, Golden, CO, USA), rinsed in Tris-NaCl three times for five minutes and incubated with the secondary antibody AlexaFluor 488 goat anti-rabbit IgG (1:500; Invitrogen, Carlsbad, CA, USA) for one hour at room temperature. After further washing, the sections were labeled for nuclear staining with bisbenzimide (Sigma, St. Louis, MO, USA), mounted with fluorescence mounting medium (Dako, Glostrup, Denmark), and visualised using the Apotome (ZEISS, Jena, Germany) fluorescence microscope. For measurement of GAG release or content the media were collected after cytokine treatment or the papain-digested explants were used (see above), and GAG content was determined by DMMB dye assay photometrically at a wavelength of 520 nm (Photometer Ultraspec II, Biochrom, Cambridge, UK) using shark chondroitin-sulfate as standard. Values were presented as μg GAG per mg wet weight of the explants.

Measurement of biosynthetic activity, glycosaminoglycans and nitric oxide production
Generation of NO was determined by measuring nitrite accumulation in culture supernatants using Griess reagent (1% sulfanilamide and 0.1% N-(1-naphtyl)-ethylene diamine-dihydrochloride in 5% H 3 PO 4 , Sigma-Aldrich, St. Louis, MO, USA). A 100 μl aliquot of each sample and 100 μl Griess reagent were mixed and incubated for five minutes, and the absorption was determined in an automated plate reader (SLT Reader 340 ATTC, SLT-Labinstruments, Achterwehr, Germany) at 540 nm. Sodium nitrite (NaNO 2 , Merck, Darmstadt, Germany) was used to generate a standard curve for quantification.

Quantitative RT-PCR
After three days of incubation, quantitative real-time RT-PCR was performed using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as reference gene to determine gene expression levels. Meniscal explants (approximately 100 mg from each group) were frozen immediately in liquid nitrogen. Total RNA was extracted after pulverisation of the tissue using the TRIZOL reagent (1 ml/100 mg wet weight tissue; Invitrogen, Carlsbad, CA, USA) followed by extraction with chloroform and isopropanol precipitation. The concentration of extracted RNA was quantified spectro-photometrically at OD 260 /OD 280 nm. Before real-time RT-PCR was performed using the Qiagen QuantiTect SYBR ® Green RT-PCR Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions the extracted RNA was digested with DNase (65°C for 10 minutes; Promega, Madison, WI, USA) to remove any traces of DNA. Bovine primers were designed using Primer3 Software [25] and used at a concentration of 0.5 μM (Table 1). Conditions for real-time RT-PCR were as specified Table 1 List of primers used for real time RT-PCR

Target
Sequence (

Statistics
Quantitative data are presented as mean ± standard error of the mean, n represents the number of independent experiments. Statistical analysis of data was made using a one-way analysis of variance (ANOVA) indicating significant differences, and comparisons among the various experimental groups were made using the two-tailed Student's t-test. Differences were considered significant if P ≤ 0.05.

TNFα-dependent GAG release
We have established an in vitro model for the investigation of bovine meniscal tissue destruction where tissue explant disks (3 mm in diameter and 1 mm thick) were isolated from the meniscal bottom surface (facing the tibial articular cartilage). Mean GAG content of freshly isolated explants was 14.2 ± 0.8 μg/mg wet weight (n = 8). After three days of culture, 4.8 ± 0.3 μg/mg of GAG was released into the media in control explants (normalised to the mean GAG content of fresh explants about one-third of explant GAG is being released during culture). Stimulation with TNFα induced a dose-dependent increase in GAG release: using a concentration of 1 ng/ml caused an additional but non-significant increase in GAG release of approximately 8.8 ± 3.7% compared with control release. With 10 ng TNFα/ml, GAG release increased significantly by 30 ± 12% (n = 11), and 100 ng TNF α/ml (chosen for all subsequent experiments; Figure 1a) increased GAG release significantly by 24 ± 10% (n = 11). In order to distinguish between the release of existing GAG or newly synthesised GAG, radiolabeled sulfate was incorporated after cytokine treatment. TNFα induced a significant reduction in sulfate uptake (controls: 100 ± 12% vs TNFα: 55 ± 11%; n = 4), suggesting that the TNFα-dependent increase in media GAG content must be predominantly the result of an increased matrix degradation, rather than an increased biosynthetic activity.

TNFα-dependent NO production
TNFα induced a dose-dependent (not shown) and significantly increased production of NO in meniscal explants which increased about four-fold in comparison to the un-stimulated control (Figure 1b). The NO-synthetase inhibitor L-NMMA reduced the basal NO production of the tissue significantly and prevented the TNFα-mediated increase in NO completely.

Influence of NO synthetase inhibition and TIMPs on TNFα-dependent GAG release
It has been described that proteoglycan degradation in cartilage tissues can be mediated by both the production of NO and the involvement of matrix-degrading enzymes. We therefore studied the influence of the NO-synthetase inhibitor L-NMMA on meniscal tissue. L-NMMA had no significant influence on the basal GAG release and did not reduce the TNFαinduced effect (Figure 1a). There was a slight, but not significant, increase of GAG release instead. In order to support the hypothesis that aggrecanases are involved in TNFα-dependent GAG release, we studied the influence of TIMP-1, -2 and -3. TIMPs are known as specific inhibitors of MMPs, but it has been reported that TIMP-3 has the additional ability to inhibit the aggrecanases ADAMTS-4 and -5 [26,27]. TIMPs did not affect the GAG release in control cultures (not shown). However, the TNFα-induced GAG release was significantly reduced by TIMP-3 by approximately 52% (Figure 1c), whereas TIMP-1 and TIMP-2 showed a trend to increase the TNFα-induced GAG release, although this effect was not significant.

Expression of matrix molecules and matrix degrading enzymes
To further determine the mechanisms of TNFα-dependent GAG release, the mRNA of meniscal explants was analyzed after a three-day incubation by quantitative RT-PCR. GAPDH had been used as a reference gene, and it had been tested that there is no significant alteration in the C T values of GAPDH expression under the influence of TNFα (control: 27.1 ± 1.7 versus TNFα: 27.3 ± 0.9; n = 4 independent experiments). Additionally, GAPDH expression had been tested in relation to another housekeeping gene, 18sRNA: the ratio of (page number not for citation purposes) GAPDH expression remained unaffected under the influence of TNFα (1.03).
The mRNA levels of most of the genes tested were quite variable under the influence of TNFα except for the matrix-degrading enzymes MMP-3 and ADAMTS-4 (see below). Collagen type I mRNA was decreased in all cases (0.75 ± 0.15), while aggrecan and collagen type II as well as MMP-1 and MMP-13 showed both increases and decreases depending on the experiment. ADAMTS-5 was not detectable in some cases or not increased by TNFα. MMP-3 and ADAMTS-4 showed a mean TNFα-dependent 6.9 ± 2.1 and 3.7 ± 0.8-fold increase of mRNA expression (Figure 1d). Comparing delta-C T -values (C TGAPDH -C Tgene of interest ) of controls and TNFα-stimulated meniscal explants allows a statistical analysis and showed a significant mean change of about 2.5 ± 0.58 for MMP-3 and 1.86 ± 0.16 for ADAMTS-4, indicating a clear up-regulation of these enzymes in all four independent experiments. The TNFα-dependent MMP-3 expression was also detectable in the supernatants of the cultures by casein zymography (Figure 2). There was only one band detectable in the gels, which was missing or expressed at lower levels in controls, but strong in TNFα-stimulated cultures. This band was not visible in gelatin zymograms (not shown), and had a molecular size of about 57 kDa (typical size for MMP-3, [28]). TIMP-3 as well as L-NMMA had no influence on the band intensity. However, the enzyme activator substance APMA altered the size of the band, indicating that most of the enzyme was expressed as a pro-form [28].

Aggrecan degradation
Immunostaining of the aggrecanase activity-specific aggrecan neoepitope NITEGE showed very low signals in control tissue with a clear TNFα-dependent increase in staining in all meniscal tissue areas that could be characterised as fibrous cartilage (Figures 3a and 3d). Co-incubation with the NO- synthetase inhibitor L-NMMA failed to influence the TNFαdependent NITEGE formation (Figures 3c and 3f), whereas TIMP-3 clearly inhibited this effect (Figures 3b and 3e).

Discussion
Cartilage catabolism is initiated by proteoglycan degradation followed by that of collagen fibers. Therefore, our study focused on the TNFα-dependent depletion of proteoglycans in a three-day bovine in vitro meniscal model [29]. TNFα induced a dose-dependent increase in GAG release supporting data from other investigations on pro-inflammatory cytokines in which IL-1 promoted GAG release in lapine and porcine meniscal tissue [19,21]. TNFα, therefore, appears to be another key factor in meniscal diseases.
To study the mechanisms of TNFα-dependent proteoglycan degradation we investigated the transcription of different matrix-degrading enzymes. One limitation in our study is that aggrecanases had been detected on the mRNA level only; there is no measurement of enzyme proteins, which could help to specify the degradative potencies of enzymes involved in TNFα-dependent proteoglycan degradation. A reason for the missing protein detection is that enzyme levels in the tissue are quite low compared with the large amounts of matrix proteins. We performed immunostainings in tissue sections (not shown), but differences in ADAMTS-4 expression were hard to differentiate, probably due to the fact that immunohistochemistry is not useful for the differentiation of slightly variable expression levels. We therefore mainly focus on the effect of inhibitors such as TIMPs or NO synthetase inhibitor (L-NMMA), and the cleavage products of aggrecan (NITEGE), which both suggest that aggrecanases must be involved in the early TNFα-dependent aggrecan degradation and GAG release in the meniscus (see below).
Increased concentrations of MMPs have been found in animal models of OA, in osteoarthritic human articular cartilage and in the synovial fluid of RA and OA patients [11,[30][31][32][33], but only little is known about the extent to which the meniscus might be involved in the production of these enzymes. We demonstrate that the meniscus can be an additional source for MMP-3 production, especially under the influence of TNFα. Wilson and colleagues [34] emphasise the importance of MMP activity in meniscal proteoglycan degradation after a 12-day incubation of bovine meniscal tissue from one to two-weekold calves with 20 ng/ml IL-1 and different enzyme inhibitors, but the authors do not specify the kind of MMPs. Additionally, Wilusz and colleagues [9] found MMPs to be responsible for some of the repair inhibition by pro-inflammatory cytokines in a serum-containing porcine model. However, in our study most of the MMP-3 in the culture supernatant was in the pro-form, and it remains unclear to what extent this enzyme might have been involved in the present GAG release. But it is reasonable to believe that MMP-3 will be involved in the subsequent TNFαdependent matrix degradation, as indicated by Wilson and colleagues [34]. TIMP-3, but not the other TIMPs, were able to inhibit the TNFα-induced GAG release and NITEGE production. This suggests that in the early three-day phase of meniscal proteoglycan degradation, aggrecanases must be involved. TIMPs are able to inhibit the active forms of almost all MMPs by binding to the C-terminal site of these enzymes [35]. However, TIMP-3 additionally inhibits ADAMTS-4 and -5 activity, whereas TIMP-1 and TIMP-2 have no effect on or even increase the activity of aggrecanases at concentrations of 1 μM or less [27,[36][37][38][39][40][41][42][43]. According to our mRNA study,

ADAMTS-4 might be one of the aggrecanases involved in
TNFα-dependent proteoglycan degradation in bovine meniscal tissue, even though final evidence is still missing. This is supported by the fact that TIMP-3 inhibited, whereas TIMP-1 and -2 increased, the TNFα-dependent GAG release (in contrast to TIMP-3, TIMP-1 and -2 are known to stimulate the activity of ADAMTS-4 under certain conditions [43]).
ADAMTS-4 mRNA has also been found in degenerated human menisci [44]. Therefore, it is likely that there might be similar effects in the human meniscus. Other studies showed that ADAMTS-5 mRNA was expressed next to ADAMTS-4 in osteoarthritic rabbit menisci [11]. Therefore, it is possible that both aggrecanases may play a role in the degradation of meniscal tissue. However, in the present investigation there was a basal meniscal mRNA expression of ADAMTS-4 in the bovine meniscus which increased with TNFα-treatment, whereas ADAMTS-5 mRNA expression was low or not detectable.
We were able to localize the NITEGE fragment in meniscal tissue by immunostaining in TNFα-treated explants, while it was almost non-detectable in control tissue. This is another strong indicator for aggrecanase involvement, according to many articular cartilage studies [14,15,45,46]. Additionally, TNFαdependent NITEGE-formation could be blocked by TIMP-3, while TIMP-1 and -2 had no inhibitory effects (not shown). TIMP-3 is not a specific aggrecanase inhibitor. It has to be mentioned that it also regulates the activity of members of the membrane-bound ADAM-family, sheddases (a disintegrin and metalloproteinase: ADAM-10, -12 and -17; TACE [47][48][49]).
The importance of these enzymes should therefore also be investigated in future studies.
We found a significant TNFα-dependent increase in meniscal NO production, which could be blocked completely by the common NO synthetase inhibitor L-NMMA. Although NO has been described as a meniscal product in several joint diseases and as an important mediator of meniscal tissue degradation in several studies [4,[16][17][18][19][20][21][22][23], we did not see a stimulating influence of NO on the TNFα-induced GAG release or aggrecan cleavage. Our study suggests that NO is not involved in the early degradation of aggrecan in the meniscus. The slight but not significant increase in TNFα-induced GAG release after incubation with L-NMMA might reflect a protective function of endogenous NO in this context, as it has been shown previously by others [21].

Conclusions
TNFα-treatment of meniscal tissue causes a reduced biosynthetic activity, release of GAG, degradation of aggrecan, and up-regulation of MMP-3 expression and aggrecanase activity. To our knowledge, this is the first report, showing that aggrecanase activity might be involved in the early TNFα-mediated aggrecanolysis in the meniscus. Inhibition of aggrecanase activity or TNFα-activity might therefore help to prevent meniscal destruction. TNFα also induces NO production, but it remains unknown what role NO might play in meniscal proteoglycan destruction because there is no evidence for a definite influence of endogenous NO on GAG release or aggrecan cleavage at the NITEGE site in this study.