Oxygen tension modulates the effects of TNFα in compressed chondrocytes

Full depth slices of articular cartilage were isolated from freshly slaughtered cattle aged <18 months (Dawn Cardington, Bedfordshire, UK), as previously described [35]. Cartilage tissue was pooled from two joints, diced and incubated on rollers for 1 h at 37 °C in Dulbecco’s modified Eagle’s medium supplemented with 20 % (v/v) foetal calf serum (DMEM + 20 % FCS), 2 µM l-glutamine, 5 µg ml penicillin, 5 μg/ml streptomycin, 20 mM Hepes buffer, and 0.85 μM l-ascorbic acid + 700 unit/ml pronase, and for a further 16 h at 37 °C in DMEM + 20 % FCS, supplemented with 100 unit/ml collagenase type XI (Sigma Chemical Co., Poole, UK). The cell suspension was washed and viable chondrocytes counted using a haemocytometer and Trypan blue. Cells were resuspended in medium at a cell concentration of 8 × 10⁶ cells/ml. The cell suspension was added to an equal volume of molten 6 % (w/v) agarose type VII (Sigma Chemical Co., Poole, UK) in Earle’s Balanced Salt Solutions (EBSS, Sigma Chemical Co., Poole, UK) to yield a final cell concentration of 4 × 10⁶ cells/ml in 3 % (w/v) agarose. The chondrocyte/agarose suspension was transferred into a sterile stainless steel mould, containing holes 5 mm in diameter and 5 mm in height and allowed to gel at 4 °C for 10 min to yield cylindrical constructs. All constructs were equilibrated in culture in 1 ml DMEM + 20 % FCS at 5 or 21 % oxygen tension for 72 h.

The effect of 5 and 21 % oxygen tension were examined in constructs cultured under free-swelling conditions in a glove-box style workspace integrated within a Biospherix incubator to ensure that the experimental conditions during setup and experimentation were uninterrupted, as previously described [34]. Constructs were cultured with TNFα (Peprotech EC Ltd, London, UK) at concentrations ranging from 0.1, 1, 10 and 100 ng/ml in the presence and absence of 1 mM l-N-(1-iminoethyl)-ornithine (L-NIO) for up to 48 h. L-NIO inhibits all isoforms of the nitric oxide synthase enzymes (Merck Chemicals, Nottingham, UK).

In separate experiments, a novel ex vivo bioreactor (Bose ElectroForce, Gillingham, UK) was used to apply dynamic compression to constructs cultured at 5 or 21 % oxygen tension, as previously described [34]. Constructs were transferred into individual wells of a 24-well culture plate (Costar, High Wycombe, UK) and mounted within the bioreactor system that is integrated within the Biospherix incubator. The medium was supplemented with either 0 or 10 ng/ml TNFα in the presence and absence of 1 mM L-NIO and the experimental conditions during setup and culture were uninterrupted. Constructs were subjected to intermittent compression under unconfined conditions, with a profile of 10 min compression followed by a 5 h 50 min unstrained period for both the 6 and 48 h culture periods, as previously described [34]. The ex vivo conditions were found to be optimal when measuring gene expression and protein synthesis at these time points [34]. The compression regime was applied in a dynamic manner with a strain amplitude of 0–15 % using a sinusoidal waveform at a frequency of 1 Hz and resulted in duty cycles which ranged from 600 to 4800 cycles for the 6 and 48 h culture periods, respectively. Control constructs were maintained in an unstrained state in the bioreactor system and cultured for the same time period. At the end of the experiment, the constructs and corresponding media were stored at −70 °C prior to analysis.

NO and PGE₂ production were determined in supernatant by Griess and EIA (GE Healthcare, Buckinghamshire, UK) using well established methods [34, 35]. Total MMP activity was measured in media samples using a fluorogenic substrate assay. 25 µl sample was incubated with 2.5 µM amino-phenyl mercuric acetate (APMA) for 1 h at room temperature to activate latent MMPs. Each sample was subsequently mixed with an equal volume of 10 µM Dnp-PChaGCHAK(Nma) fluorogenic MMP substrate, 50 µl buffer (500 mM HEPES, 100 mM CaCl₂, 0.5 % Brij-35, pH 7.0) in a 96-well plate (Enzo Life Sciences, Exeter, UK) and reactions measured at excitation and emission values of 340 and 440 nm, respectively. The change in fluorescence was calculated in the linear region of the kinetic assay for each sample between a period of 5 and 60 min at 37 °C. GAG synthesis was measured in constructs and media samples by DMMB assay and normalized to DNA values measured using the Hoescht dye 33258 in agarase/papain digests, as described [34, 35].

RNA was isolated from chondrocytes cultured in agarose as previously described (Qiagen, West Sussex, UK) [34, 36]. RNA was quantified on the Nanodrop ND-1000 spectrophotometer (LabTech, East Sussex, UK) and reverse transcription was performed using the Enhanced Avian RT First Strand cDNA synthesis kit, oligo(dT)₂₃ primer, and a total of 200 ng of RNA (Sigma Genosys, Cambridge, UK) according to the manufacturer’s protocols. Real-time quantitative PCR were performed in 10 µl reaction mixtures containing 2.5 µl cDNA, 5 µl Kapa SYBR^® FAST Universal qPCR Master Mix (2X) (Kapa Biosystems, Wilmington, Massachusetts, USA), primer pairs at final concentrations of 0.5 μM and nuclease-free PCR-grade water to 10 µl (Sigma Genosys, Cambridge, UK). The following specific primer sequences were used: ADAMTS-5 (NM_001166515) Forward: 5′-GCCCTGCCCAGCTAACGGTA-3′, Reverse: 5′-CCCCCGGACACACACGGAA-3′, MMP-13 (NM_174389.2) Forward: 5′-CCCTTGATGCCATAACCAGT-3′, Reverse 5′-GCCCAAAATTTTCTGCCTCT-3′ and 18S (NR_036642.1) Forward: 5′-GCAATTATTCCCCATGAACG-3′, Reverse: 5′-GGCCTCACTAAACCATCCAA-3′. Each sample was run in duplicate on the 96-well thermal system of the Mx3000P quantitative PCR instrument (Stratagene, Amsterdam, The Netherlands). Thermocycling conditions comprised an initial polymerase activation step at 95 °C for 3 min, followed by denaturation of 35 cycles at 95 °C for 30 s, annealing at 55 °C for 1 min, and extension at 72 °C for 1 min. Fluorescence data were collected during the annealing stage of amplification, and data were analysed on the MxPro qPCR software (version 3, Stratagene). Baselines and thresholds were automatically set by the RG-3000 qPCR software and used after manual inspection. The cycle threshold (Ct) value for each duplicate reaction was expressed as the mean value, and the results were exported into Microsoft Excel for further analysis. The data obtained by PCR assay for 18S were validated as a reference gene by displaying the Ct values as box-and-whisker plots, and the distribution examined under the treatment conditions (data not shown). The Ct values for 18S remained stable, with no changes detected under all culture conditions, suggesting its suitability as a reference gene. Relative quantifications of ADAMTS-5 and MMP-13 signals were estimated by normalizing each target to the reference gene, 18S, and to the calibrator sample by a comparative Ct approach. For each sample, the ratio of target ∆Ct and reference ∆Ct was calculated, as previously described [36, 37]. Ratios were expressed on a logarithmic scale (arbitrary units).

For free-swelling studies, data represent the mean and SEM values for 6–18 replicates from 3 to 4 separate experiments. For the ex vivo bioreactor experiments, data represent the mean and SEM values of 8–12 replicates from two separate experiments. Statistical analysis was performed by a two-way analysis of variance (ANOVA) and the multiple post hoc Bonferroni-corrected t tests to compare differences between the various treatment groups as indicated in the figure legend. In all cases, a level of 5 % was considered statistically significant (p < 0.05).

At 5 and 21 % oxygen tensions, the levels of NO were enhanced by TNFα, with a significant increase at 1, 10 and 100 ng/ml, when compared to untreated controls (all p < 0.001; Fig. 1a). The NOS inhibitor abolished cytokine-induced NO release with levels returning to basal values at both 5 and 21 % oxygen. At 10 and 100 ng/ml TNFα, the levels of NO release were greater at 5 % oxygen tension when compared to 21 % (both p < 0.001). TNFα dose-dependently increased PGE₂ release with values greater at 5 % than 21 % oxygen tension (all p < 0.001; Fig. 1b). These effects were partially reversed with the NOS inhibitor (all p < 0.001). TNFα concentrations greater than 1 ng/ml increased MMP activity in a dose-dependent manner at both oxygen tension when compared to untreated controls (Fig. 1c). Co-incubation with the NOS inhibitor partially reversed this effect in TNFα-treated constructs cultured at either 5 or 21 % oxygen. At 10 and 100 ng/ml, MMP activity was significantly greater at 5 % oxygen when compared to 21 % (all p < 0.001).

In the absence of the cytokine, GAG synthesis was greater at 21 % oxygen when compared to 5 % (p < 0.001; Fig. 2a). At low cytokine concentration (0.1 ng/ml), TNFα did not significantly influence GAG synthesis when compared to untreated controls cultured at either 5 or 21 % oxygen (Fig. 2a). At TNFα concentrations of 1–100 ng/ml, the cytokine downregulated GAG synthesis (p < 0.05) but the response was not influenced by the NOS inhibitor, except at 100 ng/ml of TNFα. At the highest cytokine concentration (100 ng/ml), the inhibitory effect on GAG synthesis was greater at 5 % oxygen when compared to 21 % (p < 0.01). In the absence of the cytokine, GAG loss was inhibited with L-NIO at both 5 and 21 % oxygen (p < 0.01 and p < 0.001, respectively; Fig. 2b). At TNFα concentrations of 10–100 ng/ml, the cytokine increased GAG loss (p < 0.05), and this response was reversed with the NOS inhibitor. At 100 ng/ml TNFα, the increase in GAG loss was greater at 5 % oxygen when compared to 21 % (p < 0.01).

Figure 3 reveals that in the absence of the cytokine, dynamic compression did not significantly influence NO release at either 5 or 21 % oxygen tension. In unstrained constructs, TNFα enhanced NO production with a greater effect at 5 % oxygen (41.6 µM) when compared to 21 % oxygen (28.5 µM), and the response was reduced with dynamic compression (all p < 0.001; Fig. 3a, b). The magnitude of inhibition was greater at 5 than 21 % oxygen, as indicated in Table 1, and the inhibitory effect abolished with L-NIO (p < 0.001).

In the absence of the cytokine, dynamic compression did not significantly influence PGE₂ release at either oxygen tension (Fig. 3c, d). In unstrained constructs, TNFα increased PGE₂ release when compared to untreated controls (p < 0.001), with values greater at 5 % (1517.5 pg/ml) than 21 % (795.1 pg/ml) oxygen tension. Stimulation with either dynamic compression or the NOS inhibitor reduced cytokine-induced PGE₂ release (all p < 0.001) with the magnitude of inhibition greater at 5 % than 21 % oxygen tension (Table 1). Co-stimulation with both compression and the NOS inhibitor induced a further reduction in PGE₂ release (all p < 0.001) with the magnitude of inhibition greater at 5 % than 21 % (Table 1).

In the absence of the cytokine, dynamic compression did not significantly influence MMP activity at either oxygen tension (Fig. 3e, f). In unstrained constructs, the cytokine increased MMP activity when compared to untreated controls (p < 0.001), with values marginally greater at 5 % than 21 % oxygen (p < 0.001). Stimulation with dynamic compression or the NOS inhibitor reduced cytokine-induced MMP activity (all p < 0.001) with the magnitude of inhibition greater at 5 % than 21 % oxygen (Table 1). In the presence of TNFα, co-stimulation with compression and L-NIO reduced MMP activity (all p < 0.001) with the magnitude of inhibition broadly similar at both oxygen tensions (Table 1).

GAG synthesis was enhanced by dynamic compression when compared to unstrained controls (all p < 0.001; Fig. 4) with a greater magnitude in stimulation for constructs cultured at 21 % oxygen than 5 % (Table 1). At both oxygen tensions, TNFα reduced GAG synthesis (all p < 0.01; Fig. 4a, b). Dynamic compression increased GAG synthesis in cytokine-treated constructs (both p < 0.001), with the magnitude of stimulation greater at 21 % oxygen than 5 %. Co-stimulation with dynamic compression and L-NIO increased GAG synthesis (p < 0.001, Fig. 4a, b) with the magnitude of stimulation by dynamic compression greater at 5 % oxygen than 21 % (90.9 vs 64.5 %, respectively). GAG loss was not influenced by compression at either oxygen tension (Fig. 4a, b, inset).

We examined whether oxygen tension influenced gene expression of MMP-13 and ADAMTS-5 in chondrocytes cultured with TNFα and subjected to dynamic compression. At 5 and 21 % oxygen tensions, the presence of TNFα increased gene expression of MMP-13 and ADAMTS-5 when compared to untreated controls (both p < 0.001; Fig. 5). In unstrained constructs, we observed greater levels of MMP-13 (p < 0.05) and ADAMTS-5 (p < 0.01) gene expression at 5 % oxygen tension than 21 % oxygen tension. The induction of MMP-13 and ADAMTS-5 gene expression by TNFα at 5 and 21 % oxygen tension were reduced with the NOS inhibitor or abolished by stimulation with dynamic compression.