The expression and function of microRNAs in chondrogenesis and osteoarthritis

Authors


Abstract

Objective

To use an in vitro model of chondrogenesis to identify microRNAs (miRNAs) with a functional role in cartilage homeostasis.

Methods

The expression of miRNAs was measured in the ATDC5 cell model of chondrogenesis using microarray and was verified using quantitative reverse transcription–polymerase chain reaction. MicroRNA expression was localized by in situ hybridization. Predicted miRNA target genes were validated using 3′-untranslated region-Luc reporter plasmids containing either wild-type sequences or mutants of the miRNA target sequence. Signaling through the Smad pathway was measured using a (CAGA)12-Luc reporter.

Results

The expression of several miRNAs was regulated during chondrogenesis. These included 39 miRNAs that are coexpressed with miRNA-140 (miR-140), which is known to be involved in cartilage homeostasis and osteoarthritis (OA). Of these miRNAs, miR-455 resides within an intron of COL27A1 that encodes a cartilage collagen. When human OA cartilage was compared with cartilage obtained from patients with femoral neck fractures, the expression of both miR-140-5p and miR-455-3p was increased in OA cartilage. In situ hybridization showed miR-455-3p expression in the developing limbs of chicks and mice and in human OA cartilage. The expression of miR-455-3p was regulated by transforming growth factor β (TGFβ) ligands, and miRNA regulated TGFβ signaling. ACVR2B, SMAD2, and CHRDL1 were direct targets of miR-455-3p and may mediate its functional impact on TGFβ signaling.

Conclusion

MicroRNA-455 is expressed during chondrogenesis and in adult articular cartilage, where it can regulate TGFβ signaling, suppressing the Smad2/3 pathway. Diminished signaling through this pathway during the aging process and in OA chondrocytes is known to contribute to cartilage destruction. We propose that the increased expression of miR-455 in OA exacerbates this process and contributes to disease pathology.

Osteoarthritis (OA) is a degenerative joint disease characterized by degradation of articular cartilage, thickening of subchondral bone, and formation of osteophytes (1). The etiology of OA is complex, with the contribution of genetic, developmental, biochemical, and biomechanical factors. Chondrocytes are the only cells in cartilage and are responsible for the synthesis and turnover of extracellular matrix (ECM), which is crucial to tissue function.

During development, mesenchymal cells aggregate and differentiate into chondrocytes, which undergo a series of differentiation events: proliferation, hypertrophy, terminal differentiation, mineralization, and programmed cell death. Blood vessels penetrate the calcified matrix, bringing in osteoblasts that build new bone. The cartilage model grows by rounds of chondrocyte cell division accompanied by the secretion of ECM in growth plates. Chondrocytes in articular cartilage are constrained from completing this program, allowing maintenance of a functional cartilage layer (2).

Articular chondrocytes must enact a pattern of gene expression in order to achieve tissue homeostasis in response to signals from growth factors, mechanical load, or changes to the ECM (1). Transcription profiling demonstrates that chondrocyte gene expression is significantly altered in OA (3). One facet of this aberrant gene expression is the replay of chondrocyte differentiation with expression of genes associated with chondrocyte hypertrophy (e.g., matrix metalloproteinase 13, type X collagen) (2). The mechanism of transcriptional control of chondrogenesis is known in some detail; however, mechanisms leading to altered gene expression in OA are less well understood (1, 2).

Small noncoding RNAs (19–24 nucleotides long) known as microRNAs (miRNAs) are important regulators of gene expression (4). MicroRNAs are first transcribed as primary transcripts (pri-miRNA) and processed to ∼70-nt stem–loop structures (pre-miRNA). Pre-miRNA is processed by the ribonuclease Dicer, forming 2 short complementary RNA molecules, one of which is integrated into the RNA-induced silencing complex (RISC), after which miRNAs base pair with their complementary messenger RNA (mRNA) molecules, usually in the 3′-untranslated region (3′-UTR) (5). In mammals, the RISC functions to suppress translation, generally leading to decreased levels of steady-state mRNA.

MicroRNAs are necessary for normal skeletal development. The conditional knockout of Dicer in cartilage leads to decreased chondrocyte proliferation and accelerated hypertrophy, with consequent compromised skeletal growth (6). A specific miRNA, miRNA-140 (miR-140), was shown to be expressed only in cartilaginous tissues in developing zebrafish (7). We investigated the expression and potential targets of mouse miR-140 (8), which is specifically expressed in the cartilage tissue of mouse embryos during the development of both long and flat bones. We identified and validated histone deacetylase 4 (HDAC-4), a corepressor of RUNX-2 and myocyte-specific enhancer factor 2 (MEF-2) that is essential for chondrocyte hypertrophy and bone development, as a target of miR-140. We proposed that miR-140 functions in the developing skeleton to promote differentiation by functionally suppressing HDAC-4.

We also reported that miR-140 targets Cxcl12 (9) and Smad3 (10), both of which are implicated in chondrocyte differentiation. The function of miR-140 in vivo has been demonstrated by Miyaki et al (11), who used targeted deletion to create a miR-140–null mouse. This mouse has a mild developmental phenotype in the skeleton, potentially via reduced growth plate chondrocyte proliferation, but displays a premature OA phenotype driven at least in part by an increase in Adamts5 expression, with Adamts5 shown as a direct target of miR-140. A transgenic mouse overexpressing miR-140 in cartilage displayed no skeletal phenotype during development but was protected in an antigen-induced arthritis model. Nakamura et al (12) also showed skeletal abnormalities in a miR-140–null mouse, with accelerated hypertrophic chondrocyte differentiation. Dnpep was identified as a target of miR-140, with an increase in this aspartyl aminopeptidase in the miR-140–null mouse leading to reduced bone morphogenetic protein (BMP) signaling. MicroRNA-140 thus plays a key role in cartilage homeostasis and OA.

MicroRNA profiling in human cartilage has been performed, and miRNA targets have been identified with relevance to OA. This has led to the identification of miR-9 impacting on interleukin-1 (IL-1)–stimulated matrix metalloproteinase (MMP) expression (13) and miR-22 as a regulator of peroxisome proliferator–activated receptor α and BMP-7 signaling (14). Studies have also shown that miR-27a (15) and miR-27b (16) regulate MMP-13 expression in human OA chondrocytes. MicroRNA-34a has been reported to modulate chondrocyte apoptosis (17). The profile of miRNA expression is also altered between differentiated and dedifferentiated adult articular chondrocytes (18, 19) or in mesenchymal stem cells as they differentiate into chondrocytes (20, 21). Key miRNAs identified in these systems regulate matrix genes or signaling pathways pertinent to OA. MicroRNA-1 regulates aggrecan expression in a human chondrocyte-like cell line (22); miR-29a and miR-29b directly target COL2A1 (21), while miR-675 indirectly regulates COL2A1 expression in articular chondrocytes (23). BMP signaling regulates the expression of miR-199a, which targets Smad1 and regulates early chondrogenesis by reducing the expression of, for example, Col2a1 and Sox9 in a BMP-driven model (24). A number of miRNAs have also been identified as regulators of osteoblastogenesis, including miR-29, miR-141, miR-200a, miR-206, miR-210, and miR-2861 (for review, see ref.25).

ATDC5 cells are a murine embryonic carcinoma line that can differentiate in vitro through chondrogenesis, and regulation of known markers of the stages of differentiation (e.g., type II collagen and type X collagen) has been shown to mirror the in vivo process (26). We explored whether key miRNAs regulated in chondrogenesis were also regulated in OA cartilage. We profiled the expression of miRNAs in the ATDC5 cell model and measured expression of key miRNAs in human cartilage. Using this approach, we identified groups of miRNAs that may function cooperatively and demonstrated that in common with miR-140, miR-455 regulates and is regulated by Smad signaling. We hypothesize that these miRNAs regulate the switch from Smad2/3 signaling to Smad1/5/8 signaling in endochondral ossification and contribute to the alteration of transforming growth factor β (TGFβ) signaling in OA cartilage.

MATERIALS AND METHODS

Cell culture and RNA purification.

SW-1353, C3H10T1/2, and 3T3 cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) containing 10% (volume/volume) fetal bovine serum (Sigma), 2 mM glutamine, 100 IU/ml penicillin, and 100 μg/ml streptomycin. ATDC5 cells were maintained at 37°C, in an atmosphere of 5% CO2, in DMEM/Ham's F-12 medium containing 5% (v/v) fetal calf serum, 2 mM glutamine, 100 IU/ml penicillin, 100 μg/ml streptomycin, 5 ng/ml sodium selenite, and 10 μg/ml human transferrin. For the assays, cells were seeded at 6 × 104/well of a 6-well plate in the above-described medium containing 10 μg/ml bovine pancreatic insulin, and the medium was changed every second day. After 21 days, the medium was changed to α-minimal essential medium with the same supplements, and the atmosphere was changed to 3% CO2. On day 42, the cells were fixed in methanol and stained for detection of glycosaminoglycan, using 0.1% (weight/volume) Alcian blue in 0.1M HCl overnight at room temperature. At selected time points, cells were scraped into TRIzol reagent (Invitrogen) for RNA purification.

Collection of human cartilage and RNA purification.

Human articular cartilage was obtained from the femoral heads of patients undergoing total hip replacement surgery at the Norfolk and Norwich University Hospital. Cartilage from patients with OA (n = 10; 5 women and 5 men, ages 37–86 years) was compared with cartilage from patients with a fracture to the neck of the femur (n = 10; 5 women and 5 men, ages 68–94 years). OA was diagnosed using clinical history and examination coupled with radiographic findings; the gross pathology was confirmed at the time of joint removal. The patients with fractures had no known history of joint disease, and their cartilage was free of lesions; 80% of these patients underwent surgery within 36 hours of sustaining a fracture. This study was performed with ethics committee approval, and all patients provided informed consent. Cartilage was chopped into 2–5-mm pieces and snap-frozen in liquid nitrogen within 15–30 minutes of surgery. Cartilage was ground under liquid nitrogen using the a Spex CertiPrep 6750 Freezer Mill. RNA was purified using MirVana (Ambion) and reverse transcribed, and miRNA expression was measured using a TaqMan low-density array (Life Technologies) or, for miR-455-3p, the individual assay described below.

Profiling of miRNA and mRNA expression.

For ATDC5 cells, RNA samples were analyzed on an Agilent 2100 bioanalyzer and a NanoDrop spectrophotometer (Thermo Scientific). For each time point, RNA from 6 culture replicates was pooled for array. Samples were labeled using the miRCURY Hy3/Hy5 power labeling kit and hybridized on a miRCURY LNA Array v.10.0 (Exiqon). Signal was corrected for background and normalized using the global lowess regression algorithm. For transcriptomic analysis, samples were hybridized on an Illumina MouseWG-6 whole genome array (Cambridge Genomic Services). Signal was corrected for background and normalized by quantile normalization using the R package lumi.

Quantitative reverse transcription–polymerase chain reaction (RT-PCR).

Complementary DNA was synthesized from RNA using SuperScript II reverse transcriptase (Invitrogen) and either random hexamers or miRNA-specific primers according to the manufacturer's instructions. Complementary DNA was stored at −20°C. The relative quantitation of gene expression was performed using an ABI Prism 7700 Sequence Detection System (Applied Biosystems), following the manufacturer's protocol.

In situ hybridization.

Whole-mount in situ hybridization of mouse embryos and isolated tissues was performed as previously described in (8). Embryos were treated with proteinase K, and endogenous alkaline phosphatase activity was blocked by pretreatment of tissues with 2 mM levamisole. Hybridizations were performed at 50°C overnight in hybridization mix containing 100 pmoles of double-labeled locked nucleic acid (LNA) oligonucleotides (Exiqon). The nitroblue tetrazolium (NBT)/BCIP staining reaction was carried out at room temperature, after which the embryos were fixed in 4% paraformaldehyde and stored in phosphate buffered saline (PBS) at 4°C. Embryos were then blocked in 3% agar and serially sectioned (100 μm) using a Lancer series 1000 vibratome. Long bones from mouse embryos (stage E18) were stained in an identical manner, paraffin-embedded, and sectioned (10 μm). Sections were then counterstained with hematoxylin and eosin.

Whole-mount in situ hybridization for chick embryos was performed as previously described (27). Embryos were fixed in 4% paraformaldehyde, dehydrated in methanol, rehydrated, and treated with proteinase K. Hybridization with double-labeled LNA probes (Exiqon) was performed at 50°C overnight. After NBT/BCIP color development, embryos were embedded in OCT compound and sectioned on a cryostat.

Transient transfection.

The 3′-UTR of potential target mRNAs was amplified by PCR and subcloned into Ambion pMIR-Report vector. Mutation of the miRNA seed sequence was achieved using QuikChange (Agilent). The positive control contains a concatamer of 3 copies of the reverse complement of the mature miRNA sequence downstream of the luciferase gene in pMIR-Report. SW-1353 or 3T3 cells were plated at 2 × 104/well in a 24-well plate and grown overnight to ∼80% confluency. Cells were transiently transfected with 200 ng of luciferase reporter plasmid, 50 ng β-galactosidase expression plasmid (Promega), and 30–50 nM miRNA mimic or control (AllStars; Qiagen) using Lipofectamine 2000 according to the manufacturer's instructions (Invitrogen), and incubated for 48 hours. For growth factor induction, use of the p(CAGA)12-luc plasmid was as previously described (10). Cells were serum starved for 24 hours posttransfection and treated with TGFβ1 (4 ng/ml)/TGFβ3 (4 ng/ml) or activin A 20 ng/ml (R&D Systems) for 3 hours. For luciferase assay, cells were washed with ice-cold PBS, lysed in 1× Reporter Lysis Buffer, and assayed according to the manufacturer's instructions (Promega). β-galatosidase assays were performed using a Beta-Glo assay kit according to the manufacturer's instructions (Promega). Data are presented as relative light units normalized to β-galactosidase.

Cluster analysis.

Hierarchical cluster analysis and visualization were performed using Cluster and TreeView.

RESULTS

MicroRNA profiling during chondrogenic differentiation in the ATDC5 cell model.

ATDC5 cells were differentiated over a 42-day time course, followed by Alcian blue staining for the detection of glycosaminoglycan accumulation (data not shown). The expression of all mRNAs was profiled using an Illumina microarray in pooled RNA samples from 6 replicate wells at each time point. Known markers of chondrocyte differentiation were appropriately regulated with an early increase in Col2a1 expression and a later increase in Col10a1 expression (data not shown).

Our group previously demonstrated that miR-140-5p, which is selectively expressed in cartilage, targets at least HDAC-4, CXCL12, and Smad3, all of which are implicated in chondrocyte differentiation (8–10). We thus profiled the expression of all miRNAs in the same pooled RNA samples as those described above. Figure 1 shows the average expression of miRNAs in 7 individual groups, based on two-way unsupervised hierarchical clustering of the data, which demonstrated regulation of miRNAs during chondrocyte differentiation. Groups 1a and 1b are subclades but clearly have different expression patterns; the results in groups 3a and 3b were similar.

Figure 1.

Expression of microRNAs in the ATDC5 cell model of chondrogenesis. For each time point, RNA from 6 culture replicates was pooled, labeled, and hybridized. The experiment used a dual-label approach comparing each test sample with a common reference sample. Signal was corrected for background and normalized using the global lowess regression algorithm. Hierarchical cluster analysis and visualization were performed to generate heatmaps. The average expression of microRNAs in each cluster is plotted.

Thirty-nine miRNAs grouped with a pattern of expression similar to that of miR-140, although several of these were within genomic clusters and were potentially coregulated, collapsing this group to include 23 miRNA loci (Figure 1, group 3b). Expression of these miRNAs increased across the time course of differentiation. The expression of miR-140-5p and miR-455-3p (miR-455*, but shown to be the guide strand on www.miRBase.org) was validated using quantitative RT-PCR in triplicate samples from each time point (Figures 2A and B).

Figure 2.

Expression of microRNA-455-3p (miR-455-3p), miR-140-5p, Col27a1, and Wwp2 in the ATDC5 cell model. Microarray data were validated using quantitative reverse transcription–polymerase chain reaction in the individual replicate samples (n = 6). The expression of both miR-455-3p (A) and miR-140-5p (B) increased over the course of chondrogenesis. Expression of the genes in which the microRNAs are encoded, Col27a1 (for miR-455) (C) and Wwp2 (for miR-140) (D), was induced earlier and returned to noninduced levels on day 42. Values are the mean ± SEM.

For the miR-140–containing group, 7 miRNAs are located within the introns of protein-coding genes (miR-99a, ENSMUST0000114231; miR-140, Wwp2; miR-149, Gpc1; miR-338, Aatk; miR-455, Col27a1; miR-466/467, Sfmbt2; miR-676, Eda), with the remainder in intergenic regions. Of these genes, the expression of Col27a1 (containing miR-455), Wwp2 (containing miR-140), and Gpc1 (containing miR-149) was clearly regulated across the ATDC5 cell model in the parallel mRNA microarray experiment, while Aatk was expressed but not regulated (data not shown). Sfmtb2 and Eda were not detected by the probes on the array, and ENSMUST00000114231 was not on the array. The expression of Wwp2 and Col27a1 was validated by quantitative RT-PCR (Figures 2C and D), showing the general coregulation of the genes and miRNA. The earlier decrease in mRNA expression compared with that of miRNA may reflect differences in RNA stability between the the 2 RNA species.

Localization of miR-455-3p in chick and mouse development.

We have shown that the expression of miR-140 is restricted to the developing mouse skeleton (8). Because miR-455 resides in an intron of COL27A1, a collagen expressed in cartilage, we examined expression of miR-455-3p in the development of chick and mouse embryos (Figures 3A and B). In chick embryo, miR-455 was expressed in the skeleton of the developing limbs. Expression was first detected on day 6.5 (Hamburger-Hamilton stage 30; approximately equivalent to stage E12.5 in the mouse) with strong expression in developing long bones and developing digits (Figure 3A, parts iii–vi). Later in the course of development (days 7.5–8, approximately equivalent to stages E14.5–16.5 in the mouse), expression became more restricted to developing joints (Figure 3A, parts vii–xii), with staining in cartilage and perichondrium (Figure 3A, parts ix and xii). In the mouse embryo, expression in developing long bones was less apparent; however, in the stage E18 embryo, whole-mount staining and sectioning of isolated joints showed clear expression in the growth plate and perichondrium (Figure 3B, parts ix and x). Expression was also seen in the interdigital region (Figure 3B, parts iii–vi) and in the sutures of the developing skull (Figure 3B, parts vii and viii).

Figure 3.

Expression of microRNA-455 (miR-455) during chick and mouse development. A, In situ hybridization was performed on whole-mount chick embryos (i–iii), dissected limbs (iv, v, vii, viii, x, and xi), and sections through these at different stages of development (vi, ix, and xii). Expression was not detected at Hamburger-Hamilton stage 21 or stage 27 (i and ii) but was seen from Hamburger-Hamilton stage 30, approximately day 6.5, onward (iii–xii). Sections through limb buds showed staining in the perichondrium (vi, ix, and xii). B, In mouse embryos (i and ii), expression was seen in the interdigital regions (iii–vi) and in the developing joint in the growth plate and perichondrium (ix). Sectioning showed staining in and around the cartilage (x). Staining of the newborn calvaria showed strong staining in the developing sutures (vii and viii). hl = hindlimb; fl = forelimb.

Expression of miRNAs in human articular cartilage.

The miRNA fraction was purified from human articular cartilage obtained during total hip replacement for either OA or fracture to the neck of the femur. Similar samples have previously been used and validated as controls in profiling studies (29, 30). Measurement of miR-140-5p was taken from a TaqMan low-density array used to profile the expression of 365 miRNAs in these samples (Young DA: unpublished observations), with miR-455-3p measured using a separate TaqMan assay in the same samples. The data were normalized using a recently described method based on the mean expression value of all expressed miRNAs in a sample (31), and the results for miR-140-5p and miR-455-3p are shown in Figure 4A. Both miRNAs were expressed at higher levels in the OA samples compared with the fracture controls. We also localized expression of miR-455-3p in adult articular cartilage from OA knees using in situ hybridization. As shown in Figure 4B, expression was predominantly in the intermediate zone.

Figure 4.

Expression of microRNA-455 (miR-455) in adult human articular cartilage. A, Human articular cartilage samples obtained from the femoral heads of patients with osteoarthritis (OA; n = 10) were compared with those from patients undergoing hip replacement following fracture of the femoral neck (NOF; n = 10). RNA was purified, reverse transcribed, and assayed by quantitative reverse transcription–polymerase chain reaction for miR-455-3p and miR-140-5p. Bars represent the means. ∗∗ = P < 0.01; ∗∗∗ = P < 0.001, by Mann-Whitney U test. B, In situ hybridization of human articular cartilage from an OA knee was performed, showing staining predominantly in the intermediate zone. MEV = mean expression value.

Regulation and function of miR-455-3p.

We previously showed that miR-140 regulates Smad3 expression and could regulate TGFβ-induced signaling (10). The expression of miR-455-3p was induced by TGFβ1, TGFβ3, and activin A in human SW-1353 chondrosarcoma cells (Figures 5A and B) and in murine C3H10T1/2 cells (results not shown). MicroRNA-455 mimic diminished Smad-dependent signaling (induced by treatment with either TGFβ1 or activin A) to a (CAGA)12-Luc construct in a manner similar to that of miR-140 (Figures 5C and D), although the effect of miR-455 was more significant.

Figure 5.

MicroRNA-455 (miR-455) regulates and is regulated by Smad2/3 signaling. A and B, Human SW-1353 chondrosarcoma cells were serum-starved for 24 hours before the addition of transforming growth factor β1 (TGFβ1) or TGFβ3 (5 ng/ml) (A) or activin A (20 ng/ml) (B), and miR-455-3p was measured by quantitative reverse transcription–polymerase chain reaction (n = 3). C and D, Cells were transfected with the Smad2/3-responsive reporter (CAGA)12-Luc in the presence of miR-140 mimic, miR-455 mimic, or scrambled control at 50 nM before the addition of TGFβ1 (5 ng/ml) (C), or activin A (20 ng/ml) (D) for 6 hours (n = 3). Relative light units were normalized to β-galactosidase activity from a cotransfected expression construct. Bars show the mean ± SEM. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus time 0 (A and B) and versus scrambled (C and D), by t-test.

Using the miRNA body map (http://www.mirnabodymap.org), we predicted SMAD2, ACVR2B, CHRDL1 as targets for miR-455-3p with a potential impact on TGFβ signaling. We validated these as direct targets of miR-455-3p by cloning the 3′-UTR downstream of the luciferase gene in the pMIR-Report vector and showed that a miR-455 mimic reduced luciferase activity, while mutation of the seed sequence for miR-455 in the 3′-UTR abolished these effects (Figures 6A–C). For CHRDL1 and ACVR2B, the scrambled small interfering RNA showed some nonspecific effects on the wild-type construct compared with the mutant.

Figure 6.

MiicroRNA-455-3p targets components of the TGFβ signaling pathway. A–C, Cells (3T3) were transfected with the 3′-untranslated region of SMAD2 (n = 18) (A), ACVR2B (n = 12) (B), or CHRDL1 (n = 18) (C) cloned into pMIR-Report with or without control small interfering RNA or miR-455 mimic and incubated for 24 hours. Relative light units were normalized to β-galactosidase activity from a cotransfected expression construct. Values are the mean ± SEM. D, Overview of miR-455 and miR-140 impact on TGFβ signaling in cartilage. scr = scrambled (see Figure 5 for other definitions).

DISCUSSION

MicroRNA-140 is selectively expressed in cartilage in the developing skeleton (7, 8), during chondrocyte differentiation, and in human articular cartilage (14). Here, we attempted to identify additional microRNAs with functions in cartilage development and OA.

MicroRNA microarrays from time points across ATDC5 cell differentiation identified 7 clusters of co-expressed miRNAs. Groups 1 and 2 showed decreased expression of miRNAs in the induced cultures compared with control, with group 1b showing this only on days 5–26. Groups 3, 4, and 5 showed increased expression of miRNAs in the induced cultures compared with control, with group 3a showing this increase from day 15 onward, group 3b from day 10 onward, and group 4 from day 5 to day 26. Group 5 showed an alternating pattern of expression across the time course.

A comparison of miRNAs expressed in human mesenchymal stem cells (MSCs) in the setting of chondrogenic differentiation (20) or mouse MSCs differentiated via culture on polyhydroxyalkanoates (21) showed incomplete overlap with our data. This is difficult to interpret, because no time course of differentiation was presented, and each model likely measured different facets of chondrocyte differentiation.

A number of miRNAs regulated across the ATDC5 model have been described as having a role in osteoblast differentiation (20, 24, 25). Although no studies have compared miRNA expression during osteogenic, adipogenic, and chondrogenic differentiation in the same starting population of precursor cells, one study examined both osteogenic and adipogenic differentiation (32). In that study, approximately half of the miRNAs regulated during differentiation were common to both adipogenesis and osteogenesis and may have a role in the process of differentiation per se rather than differentiation to a specific lineage.

Group 1a contains miR-146a, miR-155, and miR-125b, all of which are regulated by inflammation mediators (e.g., IL-1, tumor necrosis factor α, and lipopolysaccharide [LPS]) and have a role in regulating inflammation/innate immunity (33–35). MicroRNA-125b inhibits osteoblastic differentiation from mouse ST2 cells (36) and is negatively regulated by BMP-2 treatment in C2C12 cells (24). Group 1b contains miR-29, which promotes osteogenesis, regulating collagen genes and inhibitors of osteoblast differentiation and chondrogenesis (25). Group 2 contains miR-199a, a BMP-2–responsive miRNA that regulates chondrogenesis via suppression of Smad1 (24).

In groups 3a and 3b, miR-466, miR-467, miR-669, and miR-297 are part of a genomic cluster and are potentially coregulated. Mmu-miR-99a and mmu-let-7c-1 also form a cluster. Group 5 contains miR-675, which is processed from a longer noncoding RNA called H19 and regulated by SOX9 during chondrocyte differentiation in vitro (23).

The expression of miRNAs in group 3b increased with differentiation and hypertrophy; these miRNAs included miR-140 (miR-140-5p), and miR-140* (miR-140-3p), the passenger strand. Although the passenger strand is generally considered to be nonfunctional, miR-140* was reportedly induced by nicotine and targeted the 3′-UTR of the dynamin 1 gene in the nervous system (37). MicroRNA-455-3p and miR-455-5p were also expressed in this group. MicroRNA-140 and miR-455 are both located within introns of protein-coding genes (Wwp2 and Col27a1, respectively), and these were regulated across the ATDC5 cell model, with kinetics similar to those of the miRNAs. Type XXVII collagen, the product of Col27a1, is a cartilage collagen (38). Although there are no predicted miR-455 target sites in Col27a, it is possible that miR-455 could indirectly regulate Col27a1 expression, e.g., via effects on TGFβ signaling (see below).

Whole-mount in situ hybridization showed miR-455 expression in the developing long bones of chicks. With time, expression became more restricted to developing joints, with expression observed in the cartilage and perichondrium. There was evidence of expression in muscle, in line with a report that miR-455 was expressed in myotubes treated with the proinflammatory cytokine TWEAK (39). In situ hybridization in mouse embryos confirmed expression in long bones and joints. We also observed expression in the sutures of the developing skull and in the interdigital region of the developing mouse paw. These processes involve apoptosis, and miR-455 may regulate apoptosis in these tissues. Col27a1 is expressed in the cartilage anlagens of the developing skeleton, most prominently in the proliferative zone of the growth plate, and in adult mice staining is seen in articular cartilage (38, 40, 41). Expression of miR-455 has been reported during the differentiation of brown adipocytes (42), and miR-455 may also have a role in innate immunity, because heat-killed Candida albicans and LPS induced its expression in macrophages (43) in an NF-κB–dependent manner. We could not detect induction of miR-455 following stimulation of human articular chondrocytes with a variety of Toll-like receptor ligands (data not shown).

We have previously shown that miR-140 targets Smad3 expression and regulates Smad-dependent TGFβ signaling (10). Here, we demonstrated that miR-455 abrogates Smad-dependent signaling and validated 3 direct targets of miR-455-3p: Smad2, activin receptor 2B, and chordin-like 1. The expression of miR-455 is induced by TGFβ1, TGFβ3, and activin. The Smad signaling pathway has been shown to regulate the maturation of some miRNAs by the Drosha complex (44), but pri-miR-455 is also induced by TGFβ1 (data not shown), and both Col27a1 and Wwp2 are induced by TGFβ1 (data not shown), suggesting that transcriptional induction is the likely mechanism.

The ATDC5 cell model relies on insulin to induce chondrogenesis; however, TGFβ expression is regulated by and regulates differentiation (45). In the growth plate, TGFβ signaling through Smad2/3 blocks chondrocyte terminal differentiation; conversely, BMP signaling through Smad1/5/8 promotes this process. The common mediator Smad4 is required for both pathways, and where this is limiting, signaling can be regulated through competition for Smad4 (46). TGFβ can signal through Smad1/5/8 in chondrocytes via the activin receptor–like kinase 1 (ALK-1) receptor rather than ALK-5. This has led to the elegant hypothesis that a change in the ratio of ALK-5 to ALK-1 with age shifts signaling toward the ALK-1–mediated Smad1/5/8 pathway, with differentiation to a catabolic phenotype contributing to cartilage destruction.

We therefore suggest that terminal differentiation is regulated by miRNAs, with miR-140 and miR-455 decreasing Smad2/3 and consequently decreasing TGFβ signaling and promoting Smad1/5/8-dependent BMP signaling via increasing availability of Smad4. MicroRNA-199a* regulates early chondrogenesis by targeting Smad1 (24), which fits with our data showing the miR-199a* expression decreased during late chondrogenesis in the ATDC5 cell model. Similarly, miR-21 has been shown to target BMP receptor type II (47), and this miRNA is repressed during late chondrogenesis. We have shown that activin receptor type IIB (ACVR2B) is a direct target of miR-455-3p, and recently, miR-210 was shown to target ACVR1B to promote osteoblastic differentiation (25). Activin signals via the Smad2/3 pathway, so down-regulation of these targets would decrease Smad2/3 signaling and potentially enhance Smad1/5/8 signaling. Chordin-like 1 is a BMP antagonist that, while not implicated in OA, has been shown to have an impact on MSC proliferation.

The expression of miR-140 and miR-455 was increased in OA cartilage compared with control cartilage. This finding contrasts with published data for miR-140–null mice and human OA cartilage (11, 49) and is likely explained by differences in the human cartilage samples used in each study. In our comparison of cartilage obtained from the hips of patients with OA and control cartilage obtained from patients with a fracture of the femoral head, ADAMTS5 expression (a demonstrated target of miR-140) was always decreased in OA (29). The average expression of DNPEP (another miR-140 target identified in null mouse studies) was also decreased in our OA samples compared with control, although the difference was not significant (50). The kinetics of miRNA expression across disease initiation and progression will be important.

Pursuing our hypothesis, a change in miR-140/miR-455 expression would lead to altered TGFβ/activin A signaling through the Smad2/3 pathway (Figure 6D). TGFβ signaling is important for the maintenance of articular cartilage, and decreased TGFβ signaling through the Smad2/3 pathway leads to OA-like changes in the joint (46). This provides a potential mechanistic link between miRNAs that regulate the Smad pathway and the pathology of OA.

In conclusion, miRNA-455 is expressed during chondrogenesis in adult articular cartilage and is differentially expressed in OA. It has the potential to alter TGFβ signaling, thereby modulating cartilage homeostasis.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Clark had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Swingler, Boot-Handford, Hajihosseini, Dalmay, Young, Clark.

Acquisition of data. Swingler, Wheeler, Carmont, Elliott, Barter, Abu-Elmagd, Donell, Münsterberg, Clark.

Analysis and interpretation of data. Swingler, Wheeler, Elliott, Abu-Elmagd, Donell, Münsterberg, Young, Clark.

Acknowledgements

We would like to thank Eran Hornstein (Weizmann Institute of Science, Israel) for his open discussion in the early part of this study.

Ancillary