Chondrocytes were isolated from the femoral condyles of neonatal (1 day old) rats as previously described [10]. The cartilage canals in newborn rats do not form in the femoral condyles until 5 days postnatal and radiographic signs of the secondary ossification centre do not appear until about 10 days postnatal [16]. Furthermore, to avoid hypertrophic chondrocytes, the upper two-thirds of the cartilage was taken. Cells were plated onto tissue culture plastic (Falcon, Franklin Lakes, NJ, USA) at a density of ~2.5 × 10⁴ cells/cm². Under these conditions, the culture consists of an essentially pure chondrocyte population.

Monolayer chondrocyte cultures were grown in RPMI 1640 media (Invitrogen, Burlington, ON, Canada) supplemented with 5% foetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin and 1% HEPES buffer (Invitrogen) until approximately 90% confluence was reached (6 to 7 days). Prior to treatment, chondrocytes were incubated in serum-free media overnight. For inhibitor studies, chondrocytes were pretreated with the selective MEK1/2 inhibitor U0126 (10 μM; Promega, Thermo Fisher Scientific, Rockford, IL, USA) [17] for 30 minutes. As previously shown, U0126 has very low inhibitory activity towards other protein kinases [18]. Furthermore, previous studies in our laboratory have demonstrated that 24-hour treatment with 10 μM U0126 had no significant effect on the cell morphology or organization in culture [11]. As controls, cultures were treated in parallel with dimethyl sulfoxide (DMSO) (vehicle for inhibitors), U0124 (10 μM; Calbiochem, EMD Biosciences Inc., La Jolla, CA, USA) or the selective epidermal growth factor receptor inhibitor PD153035 (1 μM; Calbiochem, EMD Biosciences Inc.) [19]. Cultures were then treated with human recombinant TNFα (30 ng/ml; Endogen, Thermo Fisher Scientific) for 15 minutes to 24 hours.

Antibodies used in this study included anti-phospho-tyrosine-ERK1/2 (E4), anti-Egr-1 (588), anti-α-tubulin (E-19), and anti-NFκB p65 (C-20) antibodies (all from Santa Cruz Biotechnology, Santa Cruz, CA, USA). Horseradish peroxidase-conjugated goat-anti-rabbit or rabbit-anti goat secondary antibodies were obtained from Thermo Fisher Scientific.

Nuclear and cytoplasmic extracts were isolated using a modified method of Dignam and colleagues [20], as previously described [10]. Total cell extracts were isolated using RIPA buffer as previously described [21]. Protein concentration was determined using the Pierce BCA Protein assay kit (Pierce, Thermo Fisher Scientific), as per the manufacturers' instructions. For western blotting, 20 μg cytoplasmic protein was loaded into 10% polyacrylamide gels containing SDS and separated by electrophoresis. Proteins were transferred onto Protran™ nitrocellulose membranes (Whatman, Inc., Florham Park, NJ, USA) by electroblotting and were stained with Ponceau S to qualitatively determine equal loading of samples and efficient transfer of proteins. Membranes were blocked in 5% nonfat milk (Carnation, North York, ON, Canada) in 0.05% Tris-buffered saline containing 0.05% Tween-20 (TBST) for 1 hour followed by incubation with primary antibodies in blocking buffer overnight. Membranes were washed in TBST and incubated in 5% milk-TBST with appropriate secondary antibody for 45 minutes to 1.5 hours. Membranes were then washed with TBST and rinsed in Tris-buffered saline prior to incubation in Supersignal West Pico Chemiluminescent Substrate (Pierce, Thermo Fisher Scientific) and exposed to Amersham Hyperfilm ECL (GE Healthcare Bio-Sciences Inc., Baie d'Urfé, QC, Canada). Membranes were stripped using 1 M glycine, pH 2.5, and washed using TBST prior to reprobing.

Total RNA was isolated from cultures by Trizol (Invitrogen) followed by RNeasy clean-up (Qiagen, Mississauga, ON, Canada) as per the manufacturer's directions. Total RNA was quantified spectrophotometrically. High-quality RNA for use in the microarray analysis was confirmed by analysis in the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA).

Total mRNA (10 μg) from two biological replicates of cells treated with DMSO, U0126, TNFα or U0126 and TNFα, were amplified once and hybridized to RAT230_2.0 gene chips (Affymetrix, Santa Clara, CA, USA). Amplification, labelling, hybridization and detection were performed at the London Regional Genomics Centre (London, ON, Canada) according to the manufacturers' instructions.

The raw expression values were imported into Genespring GX 7.3 (Agilent Technologies). Raw expression values <0.01 were set to 0.01 and the normalization per chip was set to the 50th percentile. Relative gene expression of the 31,099 probe sets on the chip was determined by normalizing the raw expression values for each probe set to the DMSO control (= one-fold change for each probe set) from each independent experiment. To identify genes that were TNFα-regulated, probe sets that were altered ≥ 1.45 in DMSO/TNFα-treated cultures compared with DMSO-treated cultures were determined for each independent experiment. Probe sets identified as being TNFα regulated in both independent experiments were selected for further analysis. Genes whose transcript levels changed ≥ 1.45-fold were selected for study, as our microarray analysis revealed that aggrecan mRNA – a transcript previously shown to be TNFα sensitive [12] – was reduced approximately 1.45-fold and thus served as a positive control establishing the validity of our microarray data.

To identify probe sets whose changes were altered by TNFα in a MEK1/2-dependent fashion, we normalized the fold change in gene expression of U0126/TNFα-treated cultures to that of cultures treated with U0126 alone from both independent experiments. We determined probe sets that were altered <1.45-fold in response to DMSO/TNFα treatment, and hence were TNFα regulated in a U0126-sensitive fashion. The remainder of the genes on the lists of TNFα-regulated probe sets were determined to be TNFα regulated and MEK independent. Probe sets identified as being TNFα regulated and MEK/ERK dependent or MEK/ERK independent in both independent experiments were selected for further analysis.

Genes were also identified whose basal expression was sensitive to U0126 alone. Probe sets altered ≥ 1.45-fold in response to U0126 treatment relative to DMSO treatment were identified in both independent experiments. The limited number of genes that were altered with U0126 in both experiments (89/31,099) prevented the use of meaningful cluster analysis, but nonetheless served as a potent indication of the selectivity of the U0126 inhibitor. The generated list was then compared with the list of genes changing ≥ 1.45-fold with DMSO/TNFα to identify genes that were basal TNFα independent but MEK/ERK dependent and those genes that were both TNFα and basal MEK/ERK dependent.

The fold change in the transcript levels increased or decreased ≥ 1.45-fold in both independent experiments was averaged. The generated lists of genes determined as TNFα-activated MEK/ERK dependent and TNFα-activated MEK/ERK independent were analysed using the gene ontology browser in Genespring GX 7.3. Major cellular components and molecular functions subcategories of protein products from the list of genes were identified. The resulting list of cellular component ontologies was filtered such that a minimum of 10 genes must be in the initial group of annotated genes from the microarray and the resulting subcategory must be significantly represented (P < 0.05). Selected genes within the extracellular space ontology were then organized into subcategories that were significantly represented by the molecular function ontologies (P < 0.01).

Total RNA (25 ng) was amplified using the TaqMan One Step RT-PCR Master Mix (4309169; Applied Biosystems Inc., Streetsville, ON, Canada). Primer/probe sets to rat type II collagen (Col2a1, Rn00564954_m1), aggrecan 1 (Agc1, Rn00573424_m1), link protein (Hapln1, Rn00569884_m1), matrix metalloproteinase-9 (Mmp-9, Rn00579162_m1), matrix metalloproteinase-12 (Mmp-12, Rn00588640_m1), macrophage Csf-1 (Csf-1, Rn00576849_m1) and eukaryotic 18S rRNA (4352930E) were used to analyse relative transcript levels.

Reverse transcription and quantitative real-time PCR reactions were performed using the Prism 7900 HT Sequence Detector (Applied Biosystems Inc.). Samples were incubated at 48°C for 30 minutes to make cDNA templates. The resulting cDNA was amplified for 40 cycles. Cycles alternated between 95°C for 15 seconds and 60°C for 1 minute.

Results were analysed using SDS v2.1 software (Applied Biosystems Inc.). The ΔΔCt method was used to calculate gene expression levels relative to 18S and normalized to vehicle-treated cells. Data were log-transformed prior to analysis by one-way analysis of variance and Tukey's post-hoc test, paired t tests and Student's t tests, using Graphpad Software v. 4 (Graphpad Software, La Jolla, CA, USA).

Confluent cell cultures were detached using trypsin-ethylenediamine tetraacetic acid (Invitrogen), pelleted, and resuspended in serum-free culture medium. Cells were then plated into 48-well dishes (3.4 × 10⁴ cells/well) in 200 μl and were transfected with equal amounts of reporter plasmids. The reporter plasmids used in this study included the κB reporter (BD Biosciences, Mississauga, ON, Canada), comprising four tandem repeats of the κB response element upstream of the firefly luciferase reporter sequence and a type II collagen enhancer luciferase reporter (Sox9 reporter) containing four repeats of the 48-base-pair minimal enhancer of the type II collagen gene (pGL3 (4 × 48)) [22]. Each minimal enhancer sequence contains a binding site for Sox9. Multiple repeats of the minimal enhancer are required for optimal firefly luciferase expression [23]. Cells were transfected with 20 μl serum-free media containing the equivalent of 0.156 μg Sox9 reporter or NFκB reporter and 0.352 μl Fugene 6 transfection reagent (Roche Diagnostics Corporation, Indianapolis, IN, USA). In all experiments, chondrocytes were co-transfected with a 0.002 μg renilla luciferase plasmid (pRL-CMV; Thermo Fisher Scientific) to control for transfection efficiency. Cultures were transfected for 4 hours prior to addition of 200 μl foetal bovine serum containing media.

After overnight incubation, the media was aspirated off from the transfected cultures and replaced with serum-free media. Cultures were treated as indicated above and collected using Passive Lysis Buffer (Thermo Fisher Scientific) as directed by the manufacturer. Luciferase activity was measured using the Dual Luciferase Assay System (Thermo Fisher Scientific) in an L-max II microplate reader (Molecular Devices, Sunnyvale, CA, USA). Tanscription-factor-regulated firefly luciferase units were adjusted relative to constitutive cytomegalovirus-regulated renilla luciferase units obtained in control DMSO-treated, U0124-treated or U0126-treated cultures. Data were log-transformed prior to analysis by Student's t tests and one-way analysis of variance using Graphpad Software v. 4 (Graphpad Software).

Binding of nuclear protein complexes to the κB or Egr-1 cognate elements was determined as previously described [10,12]. The double-stranded oligodeoxynucleotides (ODNs) containing the κB cognate sequence (5'-AGTTGAGGGGACTTTCCCAGG-3'), the Egr cognate sequence (5'-GGATCCAGCGGGGGCGAGCGGGGGCGA-3') and the Egr mutant sequence (5'-GGATCCAGCTAGGGCGAGCTAGGGCGA-3') were purchased from Santa Cruz Biotechnology. Competition assays were performed by adding 100-fold molar excess of unlabelled probe to the nuclear extract-labelled probe mixture. Antibody interference assays were performed by adding 2 μg antibody against Egr-1 (specific) or NFκB (nonspecific) 1 hour prior to the addition of nuclear extract to the buffered radiolabelled DNA. Samples were loaded into 4% polyacrylamide gels and were electrophoresed for 3.5 hours. Following electrophoresis, gels were dried and exposed to Amersham Hyperfilm-MP (GE Healthcare Bio-Sciences Inc.) at -80°C.

Upstream regions proximal to the transcriptional start site of the rat Col2a1 and Agc1 genes have been described previously [24,25]. Upstream regions from the transcriptional start site (~5,000 base pairs) of the Rattus Norvegicus Col2a1 [GenBank:NM_012929.1] and Agc1 [GenBank:NM_022190] genes were obtained and analysed for putative transcription factor binding sites by TRANSFAC analysis [26].

Chondrocytes were plated at 1.2 × 10⁶ cells/well in six-well culture dishes. Single stranded, phosphorothiol-modified ODNs were annealed by heating complementary ODNs to 98°C for 20 minutes followed by cooling to room temperature for 3 to 4 hours. Chondrocytes were transfected with 2 μM double-stranded ODNs corresponding to the cognate EGR-1 binding sequence (5'-ggaTCCAGCGGGGGCGAGCGGGGgcgA-3') or the Egr mutant sequence (5'-ggaTCCAGCTAGGGCGAGCTAGGgcgA-3'; Sigma Genosys, Oakville, ON, CA) using 1% HiPerfect transfection reagent (Qiagen), as per the manufacturer's instructions. (Lowercase letters indicate phosphorothiol-modified bases.)

To optimize double-stranded ODN transfection conditions, chondrocytes were transfected cells with increasing concentrations of double-stranded, fluorescein-tagged and phosphorothiol-modified ODNs, and the cells were imaged by live-cell fluorescent microscopy (data not shown). Chondrocytes were allowed to grow for 24 hours in the presence of ODNs, after which cells were washed and cultured in serum-free RPMI media overnight. Chondrocytes were treated with TNFα for 24 hours, as described, and total RNA was collected for analysis by real-time PCR.

We have shown previously that TNFα induces ERK phosphorylation in primary articular chondrocytes 15 minutes post treatment [11]. To confirm and extend these results, we used western blot analysis to show that TNFα induced ERK1/2 phosphorylation 15 minutes post treatment (Figure 1), followed by a decrease in phosphorylation status (data not shown). ERK1/2 phosphorylation was again increased at 90 minutes post treatment (Figure 1). As anticipated, both the increases at 15 minutes and at 90 minutes could be inhibited by the MEK1/2 inhibitor U0126, but not its inactive isoform U0124 (Figure 1). Based on these data, we used U0126 as an inhibitor to assess the effect of blocking MEK1/2 on the mRNA expression pattern modulated by application of TNFα to chondrocytes.

To investigate the global impact of U0126 on TNFα-modulated gene expression in chondrocytes, we utilized microarrays to analyse changes in chondrocyte mRNA expression. Cells were serum-starved overnight and were treated with or without U0126 (10 μM, 30 min) prior to addition of TNFα for 24 hours. Cells were treated with TNFα for 24 hours as previous data showed that this length of TNFα treatment was necessary to generate a TNFα-mediated suppression of chondrocyte matrix genes, owing to the stability of chondrocyte matrix gene mRNAs [27-30].

Microarray analysis from two independent experiments determined that 629 genes were regulated by TNFα signalling in both sets of experiments by at least 1.45-fold, the majority of which were increased in response to TNFα (Figure 2). Of these genes, alterations of 138 (~22%) were attenuated with U0126. Furthermore, of the remaining genes that were not regulated by TNFα, 62 genes were regulated by U0126 alone, indicating that basal MEK/ERK activity may also play a role in chondrocyte gene regulation. Complete microarray data have been deposited in the Gene Expression Omnibus public repository [GEO:GSE14402].

We further analysed the lists of genes that were induced by TNFα using specific gene ontologies. Analysis of the list of TNFα-induced, MEK/ERK-dependent and MEK/ERK-independent probe sets indicated that there was significant representation (P < 0.05) of genes whose protein products localize to the extracellular space within both lists (Table 1). Further analysis of the list of TNFα-regulated, MEK/ERK-dependent genes – whose products are found in the extracellular space – indicated that some of these genes were significantly categorized by the molecular function of their protein products into categories that included hyaluronic acid binding activity (including Agc1 and Hapln1) and proteinase activity (including Mmp-9 and Mmp-12). Analysis of the TNFα-regulated, MEK/ERK-independent list of genes whose protein products were localized to the extracellular space determined that many of the protein products of these genes were involved in a variety of activities, including chemokine/cytokine activity – including macrophage Csf-1 – and various protease activities. The inflammatory genes, however, appeared to be primarily U0126 insensitive.

To validate the changes in gene expression in response to TNFα-induced MEK/ERK signalling determined by the microarray analysis, we identified the relative changes in transcript levels of the extracellular matrix components Agc1, Hapln1, and Col2a1, proteases Mmp-9 and Mmp-12, as well as the inflammatory cytokine macrophage Csf-1 (Figure 3). TNFα decreased Agc1 and Hapln1 (Figure 3a, b) and increased Mmp-9 and Mmp-12 (Figure 3e, f) in a MEK/ERK-dependent manner. In addition, Col2a1 – a gene not identified as MEK/ERK sensitive by microarray analysis – was also determined to be MEK/ERK sensitive (Figure 3c). Pretreatment with U0126, however, only partially attenuated the TNFα-induced reductions in Agc1, Hapln1 and Col2a1 transcript levels – to a level only moderately, but not significantly, lower than control treated cultures, suggesting the possible involvement of other pathways (Figure 3a to 3c). Conversely, TNFα-induced increases in macrophage Csf-1 were independent of MEK/ERK signalling (Figure 3d). As anticipated, the inactive U0126 analogue U0124 had no effect in any of the assays tested. Taken together, these results suggest that U0126 may attenuate the changes in chondrocyte gene expression towards a catabolic phenotype while allowing for inflammatory processes to be undisturbed.

We next wanted to determine the possible molecular basis for TNFα-modulated, U0126-sensitive gene expression. First, we investigated whether U0126 affected the ability of TNFα to regulate the activity of the transcription factors Sox9 and NFκB, which are known to be regulated by TNFα in chondrocytes [10,12]. As expected, TNFα significantly reduced the level of Sox9 activity and increased the level of NFκB activity in chondrocytes (P < 0.01; Figure 4a, b). There was no significant effect, however, on the level of inhibition or the induction of Sox9 and NFκB activity, respectively, by either U0124 or U0126 (P > 0.05; Figure 4a, b). Furthermore, we found that TNFα-induced DNA binding of NFκB was reduced by pretreatment with DMSO (vehicle for the inhibitors) and was not further reduced by pretreatment with U0124, U0126 or the selective epidermal growth factor receptor inhibitor, PD153035 (Figure 4c). These results indicate that transcription factors other than Sox9 and NFκB are targets of TNFα-induced MEK/ERK signalling.

To determine additional, candidate transcription factors that may regulated by MEK/ERK, we considered that Egr-1 is a known early target of MEK/ERK signalling and that IL-1 induction of Egr-1 inhibits the activity of the human type II collagen proximal promoter [31]. We therefore focused the remainder of our study on Egr-1 and its possible role in regulating U0126-sensitive TNFα-induced genes.

We identified multiple putative Egr-1 binding sites in the promoter regions of the rat Col2a1 and Agc1 genes that were proximal to the transcription initiation site and overlapped with putative Sp1 binding sites (Figure 5). TNFα treatment of chondrocytes over 24 hours did not alter the Egr-1 protein levels, and neither did treatment for 90 minutes alter the nuclear localization of Egr-1 (data not shown).

We then used electrophoretic mobility shift assays to investigate whether the binding of Egr-1 to DNA was dependent on TNFα-induced MEK/ERK signalling. Nuclear extracts from chondrocytes treated with TNFα for 90 minutes increased the DNA binding of two complexes containing Egr-1 to an Egr consensus DNA binding site (Figure 6, arrowheads). Both complexes were reduced when extracts were preincubated with a 100-fold molar excess of double-stranded cold Egr consensus ODNs, but not with cold mutant Egr ODNs or NFκB consensus ODNs (Figure 6, arrowheads). Compared with preincubation of extracts with the anti-NFκB p65 antibody, preincubation of extracts with the anti-Egr-1 antibody specifically reduced the DNA-protein complexes attributed by the Egr consensus ODN competition studies to be a result of Egr/DNA binding (Figure 6, arrowheads). Pretreatment of cells with U0126 attenuated the increase in complex formation of both identified complexes. The binding of the identified complexes to DNA was inhibited by pretreatment with U0126 but not with U0124, indicating DNA binding of Egr-1 is dependent on TNFα-activated MEK/ERK signalling.

To determine whether decreases in chondrocyte selective matrix gene expression in response to TNFα were dependent on the genomic DNA binding activity of Egr family members, we transfected cells with double-stranded ODNs containing phosphorothiolate modifications corresponding to the cognate and a mutated form of the Egr-DNA binding sequence (Figure 7). Transfection of cells with mutant double-stranded ODNs did not disrupt decreases induced by TNFα to Col2a1, Agc1 or Hapln1 transcript levels. Transfection using the cognate Egr double-stranded ODNs, however, attenuated the decreases in transcript levels of Col2a1, Agc1 and Hapln1 by TNFα. Egr-containing complexes, probably that include Egr-1, are therefore responsible for the reduced transcript levels of cartilage selective matrix genes in response to TNFα in chondrocytes.