NFAT3 and TGF-β/SMAD3 regulate the expression of miR-140 in osteoarthritis

Human cartilage was obtained from femoral condyles and tibial plateaus. Normal human cartilage was obtained from individuals within 12 hours of death (n = 8, 60 ± 19 years (mean ± SD)) and OA cartilage from patients undergoing total knee arthroplasty (n = 48, 66 ± 9 years). Normal individuals had no history of joint disease and died of causes unrelated to arthritic diseases. The cartilage was examined macroscopically and microscopically to ensure that only normal tissue was used. All OA patients had been evaluated by a certified rheumatologist and diagnosed as having OA according to the American College of Rheumatology (ACR) criteria [18]. These specimens represented moderate to severe OA. The Institutional Ethics Committee Board of the University of Montreal Hospital Research Centre (CRCHUM) approved the use of the human articular tissues. Patients signed informed consent and post-mortem tissue was obtained with the consent of a family member or authorized individual.

Chondrocytes were seeded directly from the digested cartilage (primary chondrocytes) as described [5] and the SW1353 chondrosarcoma cell line was purchased from American Type Culture Collection, Manassas, VA, USA (#HTB-94). Briefly, the cells were seeded at high density (10⁵/cm²) and cultured in (Dulbecco’s) modified Eagle’s medium ((D)MEM; Wisent, St-Bruno, QC, Canada) supplemented with 10% heat-inactivated fetal calf serum (FCS; PAA Laboratories Inc, Etobicoke, ON, Canada) and an antibiotic mixture (100 units/ml penicillin base and 100 μg/ml streptomycin base; Wisent) at 37°C in a humidified atmosphere. Primary chondrocytes were used when comparing expression levels in normal and OA chondrocytes to avoid loss of the chondrocyte phenotype and first-passage chondrocytes for all other experiments involving cultured chondrocytes. In the experiments, the culture medium was replaced with (D)MEM containing 0.5% FCS 24 hours before the treatment. The ionophore ionomycin (1 μM; Sigma, Oakville, ON, Canada), NaCl (100 mM), and TGF-β (10 ng/ml, Feldan Bio Inc., Montreal, QC, Canada) were added for 18 hours and the specific inhibitor of SMAD3 phosphorylation (SIS3, 8 μM; Calbiochem, La Jolla, CA, USA) [19] for 24 hours.

Total RNA was extracted and quantified as described [5] except that 10 μg glycogen was added to the precipitation step to enrich for miRNAs. mRNA levels were quantified by real-time PCR with the SYBR Green PCR Master Mix (Qiagen, Valencia, CA, USA). When quantifying expression between normal and OA chondrocytes, internal standards were added at known concentrations in the PCR reactions and amplified by the same primers as the specific target mRNAs as to give absolute numbers. The values of each sample were calculated as the ratio of the number of molecules of the target gene/number of molecules of the housekeeping gene. When evaluating the effect of a treatment on a given cell culture, the expression level of each control was assigned an arbitrary value of 1, and the treated cells were evaluated as fold change over control and calculated as 2^-Δ(ΔCt). Basal expression values of the control specimens (as determined from the number of molecules of the target gene/housekeeping gene) are shown in Additional file 1: Figure S1. Primer efficiencies for the genes under study were the same as those for the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The sequences of the human specific primers used are listed in Table 1. miRNAs were quantified with the TaqMan MicroRNA Reverse Transcription kit and TaqMan MicroRNA Assays specific for each mature miRNA (Applied Biosystems, Burlington, ON, Canada) as described [5]. Normalisation of the miRNA expression data was done using the GAPDH gene. The expression level of each control was assigned an arbitrary value of 1 and the treated cells were evaluated as fold change over control.

siRNA pools specific for NMP4, MAZ, NFAT1 (NFATc2), NFAT2 (NFATc1), NFAT3 (NFATc4), NFAT4 (NFATc3), NFAT5 (TonEBP) and SMAD3 were purchased from Dharmacon (Lafayette, CO, USA) and transfected with the HiPerfect Transfection Reagent (Qiagen) at a concentration of 100 nM. Total RNA was extracted after 48 hours and mRNA levels measured by quantitative PCR and normalised to the control gene. Cells transfected with non-targeting (random) siRNAs served as controls. Silencing gene expression was carried out by transfecting OA chondrocytes for 48 hours with small interfering RNAs (siRNAs), resulting in a silencing efficiency of more than 80% as described [20] and as illustrated in Additional file 2: Figure S2.

1.883 kb of the 5’-flanking region directly upstream of pre-miR-140 (regulatory sequence miR-140 (rsmiR-140)) was cloned into the luciferase reporter vector pGL3-basic (Promega, Madison, WI, USA). The fragment was amplified by PCR using genomic DNA from human chondrocytes and the primers 5’-TGAGCTAGC GGTGCTTATGACCGCAGTTTTC (sense) and 5’-CAGAAGCTT ACCAAGCAGAGCCTGGAGAGGAG (antisense). NheI and HindIII restriction sites (bold) were added to the sense and antisense primers, respectively, to facilitate the cloning into the NheI and HindIII sites of the pGL3-basic vector. A smaller plasmid was constructed similarly by using the sense oligonucleotide 5’-TGAGCTAGC GTGCCCCGGAAGGCTGCCCTGTAC, resulting in a cloned fragment of 1.153 kb. Both plasmids were sequenced to confirm the integrity of the cloned DNA.

Wild-type and mutated oligonucleotides were chemically synthesised and used in site-directed mutagenesis assays with the QuickChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA, USA) as described [21]. Bases were mutated to change the consensus NFAT binding site GGAAA to GGTAC, and the SMAD3 consensus binding sites from CAGA to TACC and from TTGGTGTTGG to TTGCATATGG. Each plasmid was verified by DNA sequencing.

The SW1353 cells were transfected with the TransFectin transfection reagent (Bio-Rad, Mississauga, ON, Canada), incubated for 48 hours in (D)MEM/10% FCS and lysed in Reporter Lysis buffer (Promega). The transfected cells were then incubated for an additional 18 hours with the factor under study. Luciferase activity was measured with the Lumat LB 9507 luminometer (EG&G Berthold, Bad Wildbad, Germany). To monitor transfection efficiency, a plasmid coding for a green fluorescent protein was used. Total protein was quantified by the bicinchoninic acid method (Pierce, Rockford, IL, USA). Plasmid activity was calculated as relative luciferase units per μg of protein. The pGL3-basic control was assigned an arbitrary value of 1 for each experiment and the activity evaluated as fold change over control. Basal expression values of the control specimens as determined by luciferase units/μg of protein are shown in Additional file 1: Figure S1.

OA chondrocytes were treated with ionomycin (1 μM, 60 minutes), NaCl (100 mM, 60 minutes) or TGF-β (10 ng/ml, 30 minutes). Nuclear proteins from the control and treated cells were extracted with the Nuclear Extraction kit (Cayman Chemical Company, Ann Arbor, MI, USA) and processed (5 μg) for Western blotting as previously described [21]. The primary antibodies were a rabbit anti-human SMAD3 (Cell Signaling Technology, Danvers, MA, USA; dilution 1/2,000), a mouse anti-human NFAT3 (Santa Cruz Biotechnology, Santa Cruz, CA, USA; dilution 1/2,000), and a rabbit anti-NFAT5 (Novus Biologicals, Littleton, CO, USA; dilution 1/5,000). The secondary antibodies were an anti-mouse or anti-rabbit immunoglobulin G (IgG; Pierce, Rockford, IL, USA; dilution 1/10,000). The nuclear protein nucleolin (Active Motif, Carlsbad, CA, USA; dilution 1/5000) was used as a housekeeping (control) protein.

OA chondrocytes were treated with TGF-β (10 ng/ml, 30 minutes) or ionomycin (1 μM, 60 minutes) and processed with the EZ-Magna ChIP A/G Chromatin Immunoprecipitation Assay kit (Millipore, Temecula, CA, USA) as recommended by the manufacturer and as described [22]. The antibodies used in the chromatin immunoprecipitation (ChIP) reactions were an anti-human SMAD3 (Cell Signaling Technology) and an anti-human NFAT3 (Santa Cruz Biotechnology). Pre-immune IgGs were used as negative control. The PCR primers were 5’-GCATTGTCTTGCCTTCACCC (sense, located 52 bp upstream of the NFAT3 binding site) and 5’-TGGAGAGGAGGTCACCACGG (antisense, located 200 bp downstream from the NFAT3 site). These primers were also used for the SMAD3 ChIP assay, as the NFAT3 and SMAD3 sites are separated by less than 40 bp. The primer sequences of the MAP1A gene (microtubule-associated protein 1A, used as an unrelated control gene) were 5’-GACCTTATGCCAGGACAGGA (sense) and 5’-AGACAAGAGCTCCCTCCACA (antisense). The pre-immune IgG ChIP results were used as negative control and the data from the NFAT3 and SMAD3 experiments were normalized to this control. The genomic DNA (1% of the starting lysate input) was used as positive control. The amplified PCR products were analysed on agarose gels and quantitated by qPCR. The results were compared to those of the pre-immune assays, and the effect of the treatment measured as fold change over the control which was assigned an arbitrary value of 1. Basal expression values of the control specimens (as determined by PCR control values per input control DNA) are shown in Additional file 1: Figure S1.

Nuclear translocation of SMAD3 and NFAT3 was assessed in OA chondrocytes by immunocytochemistry and its effect by qPCR.

We previously reported that miR-140-5p expression was significantly reduced in human OA chondrocytes [5]. Here, we followed this by comparing its expression to that of miR-140-3p, and their host gene, WWP2, in normal and OA human chondrocytes. Data showed (Figure 1A, B) that the expression levels of miR-140-5p and -3p were both markedly and significantly reduced in OA chondrocytes. Interestingly, WWP2 expression (Figure 1C) was only slightly reduced (about 21%) in OA and this did not reach statistical significance. Of note, these results represent the global expression of the mRNAs and miRNAs at a given point in time.

The WWP2 gene codes for three variants or isoforms. Variant 1 (FL-isoform) is the longest variant with 25 exons; variant 2 (C-isoform) has its ATG in exon 14 of variant 1 and the same TAA stop codon as variant 1; variant 3 (N-isoform) has the same ATG as variant 1 but with a stop codon in exon 10. Preliminary RT-PCR experiments using primers specific for variants 2 and 3 [23] have shown that these variants were not as strongly expressed as variant 1 in human chondrocytes. Variant 3 was expressed in both normal and OA chondrocytes and the expression pattern was similar to that observed with variant 1 (data not shown); variant 2 was either not expressed or very weakly expressed in chondrocytes (data not shown). We have routinely used primers located in exons 4 and 5 (variants 1 and 3, Figure 1C). Experiments performed with primers located in exons 14 and 16 (variants 1 and 2) (Additional file 3: Figure S3 A) showed similar data to those with the variants 1 and 3 and yielded no significant differences between the two sets of primers. Subsequent experiments were done using primers located in exons 4 and 5.

To determine whether the differential expression level between WWP2 and miR-140 was due to a different miRNA processing in OA cells, we determined the levels of two unrelated intronic miRNAs: miR-33a (Figure 1D) present in one intron of the sterol regulatory element binding factor-2 (SREBF2) gene and miR-151 (Figure 1F), present in one intron of the protein tyrosine kinase or focal adhesion kinase (PTK2/FAK) gene. The expression levels of miR-33a (Figure 1D) and its host gene SREBF2 (Figure 1E) were similar in normal and OA chondrocytes; however, the expression level of miR-151 (Figure 1F) was significantly decreased compared to that of its host gene PTK2/FAK (Figure 1G). These findings indicate that the reduced expression of miR-140 in OA chondrocytes is not due to a general processing and likely results from an additional level of regulation specifically directed at this miRNA.

Intronic miRNAs can be regulated independently of their host gene by sequences located directly upstream of their precursor sequence. To determine if such was the case for miR-140 and WWP2, we cloned 1.883 kb and 1.153 kb (Figure 2A) of the sequence located directly upstream of the pre-miR-140 [GenBank:NC_000016]. Both plasmids promoted similar transcriptional activity (Figure 2B), indicating the presence of regulatory elements in the sequence upstream of pre-miR-140. Further experiments were conducted with the 1.153 kb cloned DNA, designated in the text rsmiR-140.

The rsmiR-140 sequence has several potential binding sites for transcription factors, such as NMP4 (CAAAAA), MAZ (GAGAGA), NFAT (GGAAA) and SMAD3 (CAGA, TTGGTGTTGG) (Figure 2A). To examine whether these factors could be responsible for the differential regulation of miR-140 and WWP2, their expression was silenced in OA chondrocytes and the miR-140 and WWP2 levels were determined (Figure 3). As both miR-140 s are similarly decreased in OA chondrocytes, further experiments were carried out with miR-140-5p.

Silencing NFAT3 significantly decreased (P ≤0.01) miR-140 expression without affecting WWP2, and silencing SMAD3 significantly increased (P ≤0.05) miR-140 without significantly affecting WWP2. Silencing NFAT5 significantly decreased (P ≤0.003) miR-140 and, to a lesser extent, WWP2 expression (P ≤0.05). Silencing NMP4 significantly increased WWP2 (P ≤0.05) but not miR-140, and silencing NFAT1, NFAT2, NFAT4 or MAZ did not significantly affect either miR-140 or WWP2 levels. To further investigate SMAD3, OA chondrocytes were treated with the specific inhibitor of SMAD3 phosphorylation SIS3 (Figure 3C, D). Data showed a pattern similar to that of the silenced SMAD3; WWP2 expression was not affected (Figure 3C) and miR-140 expression was significantly increased (P ≤0.003) (Figure 3D).

The effects of SMAD3, NFAT3 and NFAT5 were investigated in OA chondrocytes by activating SMAD3 by TGF-β, NFAT3 by increasing calcium flux with ionomycin and NFAT5 by hypertonic stress via increasing NaCl concentration. The concentrations used and durations of exposure resulted in an accumulation of the molecules in the nucleus as shown by Western blot (Figure 4). As illustrated in Figure 5A and B, ionomycin significantly increased (P ≤0.05) miR-140 but not WWP2 expression, while NaCl increased both WWP2 and miR-140 expression levels but significance was reached only for WWP2 (P ≤0.03). TGF-β significantly decreased (P ≤0.05) miR-140 but had no true effect on WWP2 levels. Complementary experiments evaluating the expression of the WWP2 variants 1 and 2 also showed no effect of ionomycin or TGF-β, but a significant increase with NaCl treatment (Additional file 3: Figure S3 B). Additional experiments were performed to determine the expression level of a known miR-140 direct target, IGFBP5. Data demonstrated (Figure 5C) that the increased expression of miR-140 following stimulation by ionomycin and NaCl resulted in decreased expression of IGFBP5, and the TGF-β-induced decrease in miR-140 led to an increased expression of IGFBP5, indicating that a change in miR-140 expression was reflected on its target gene.

All together, these findings confirm that NFAT3 and SMAD3 affect miR-140 expression independently of WWP2. It is possible that the increased expression of miR-140 caused by NFAT5 under hypertonic conditions results in part from increased expression of its host gene WWP2.

We further examined whether the regulation by NFAT3, NFAT5 and SMAD3 of miR-140 occurred at the level of rsmiR-140. SW1353 cells were transfected with the rsmiR-140 plasmid and treated with ionomycin, NaCl and TGF-β (Figure 5D). rsmiR-140 was significantly stimulated by ionomycin (P <0.0003), decreased by TGF-β (P <0.01) and not affected by NaCl. To verify that SMAD3, but not SMAD1, was involved in miR-140 regulation, the cells were treated with BMP2 (10 ng/ml); rsmiR-140 activity was not affected by this factor (data not shown). This result agrees with our previous finding that BMP2, as opposed to TGF-β, does not significantly affect miR-140 expression [5].

We next investigated whether NFAT3 and NFAT5 could act through TGF-β, as the TGF-β promoter contains potential NFAT binding sites [GenBank:J04431.1]. NFAT3 and NFAT5 expression was silenced in OA chondrocytes and the TGF-β levels determined (Figure 5E). Interestingly, NFAT3 did not affect TGF-β expression, but NFAT5 significantly decreased its levels (P = 0.054).

Together, these results indicate that the TGF-β-induced miR-140 down-regulation is the result of SMAD3 activation and that NFAT3 regulates miR-140 directly, likely at the rsmiR-140 level. NFAT5 could indirectly contribute to the down-regulation of miR-140 by up-regulating the expression of TGF-β, which in turn inhibits miR-140 expression.

rsmiR-140 has consensus binding sites for NFAT (GGAAA at position -234 bp) and SMAD3 (CAGA at position -209 bp and -195 bp, and TTGGTGTTGG at -120 bp) (Figure 2A). To determine if NFAT3 and SMAD3 directly acted through those sites, SW1353 cells were transfected with rsmiR-140 without or with the mutated sites and treated with ionomycin and TGF-β. Mutation of the NFAT site (Figure 5F) significantly decreased basal (P ≤0.0001) as well as the ionomycin-induced (P ≤0.0001) expression, indicating the involvement of this site in the positive regulation of miR-140 by NFAT3. The -195 bp CAGA mutation (Figure 5G) resulted in a significant increase in basal (P ≤0.003) and TGF-β-induced (P ≤0.005) expression, suggesting its involvement in the negative regulation of miR-140 by TGF-β through the inhibitory action of SMAD3. Mutating the -120 bp TTGGTGTTGG and -209 bp CAGA sites did not affect either basal or TGF-β-induced expression (data not shown). It was also observed that TGF-β treatment of the -195 bp mutated construct resulted in a slightly decreased expression. TGF-β has a number of roles and acts through many intermediates in addition to SMAD3; it is thus possible that TGF-β could act indirectly on rsmiR-140 through other (unknown) binding sites.

ChIP assays were done to determine if NFAT3 and SMAD3 physically associated with the identified sites. OA chondrocytes were treated with ionomycin and TGF-β and processed for ChIP assays (Figure 6). The results (Figure 6A, B, D, E) showed that treatment with ionomycin and TGF-β significantly enriched the DNA sequences containing the binding sites of NFAT3 and SMAD3 (P ≤0.05, P ≤0.01, respectively). Similar experiments done with primer pairs located 800 bp upstream of the binding sites and on the unrelated negative control gene MAP1A revealed no significant binding (Figure 6C, F), suggesting that the increased binding seen with treatment with TGF-β and ionomycin was specific for rsmiR-140.

As TGF-β production is significantly increased in OA [24, 25], we examined whether TGF-β could interfere with the translocation of NFAT3 and prevent its action. As expected (Figure 7A, B), ionomycin significantly triggered the translocation of NFAT3 (47%, P ≤0.001), and TGF-β triggered that of SMAD3 (52%, P ≤0.0001) in OA chondrocytes. Treatment with TGF-β alone had no significant effect on NFAT3 translocation and treatment with ionomycin alone had no effect on SMAD3 translocation. However, when TGF-β was added for the last 30 minutes of the 90-minute ionomycin treatment, there was a significant decrease in the number of NFAT3-positive nuclei (P ≤0.05) when compared to ionomycin alone (Figure 7A). When ionomycin was added for the last 60 minutes of the 90-minute TGF-β treatment, there was a slight but non-significant decrease in SMAD3-positive nuclei (Figure 7B). Similar experiments done with human normal chondrocytes revealed the same pattern, that is, TGF-β interfered with the ionomycin-induced translocation of NFAT3 into the nucleus (data not shown).

Further experiments were done to verify whether such interference affected miR-140 expression (Figure 7C, D). The addition of TGF-β during the ionomycin treatment resulted in a significant decrease (P ≤0.02) in miR-140 compared to ionomycin treatment alone, while the addition of ionomycin to the TGF-β treatment did not significantly affect the miR-140 expression level. All together, these results indicate that the presence of TGF-β interferes with the ionomycin-induced nuclear translocation of NFAT3 and ultimately miR-140 expression.