Regulation of Human Chondrocyte Function through Direct Inhibition of Cartilage Master Regulator SOX9 by MicroRNA-145 (miRNA-145)*

Background: SOX9 is essential for cartilage. Results: miR-145 directly targets SOX9, and increased miR-145 levels reduce expression of SOX9 and the extracellular matrix genes critical to cartilage function. Conclusion: SOX9 is subject to significant post-transcriptional regulation by miR-145 in human chondrocytes. Significance: Our data give new insights into the mechanisms regulating SOX9 and identify miR-145 as a new target for cartilage repair. Articular cartilage enables weight bearing and near friction-free movement in the joints. Critical to its function is the production of a specialized, mechanocompetent extracellular matrix controlled by master regulator transcription factor SOX9. Mutations in SOX9 cause campomelic dysplasia, a haploinsufficiency disorder resulting in severe skeletal defects and dwarfism. Although much is understood about how SOX9 regulates cartilage matrix synthesis and hence joint function, how this master regulator is itself regulated remains largely unknown. Here we identify a specific microRNA, miR-145, as a direct regulator of SOX9 in normal healthy human articular chondrocytes. We show that miR-145 directly represses SOX9 expression in human cells through a unique binding site in its 3′-UTR not conserved in mice. Modulation of miR-145 induced profound changes in the human chondrocyte phenotype. Specifically, increased miR-145 levels cause greatly reduced expression of critical cartilage extracellular matrix genes (COL2A1 and aggrecan) and tissue-specific microRNAs (miR-675 and miR-140) and increased levels of the hypertrophic markers RUNX2 and MMP13, characteristic of changes occurring in osteoarthritis. We propose miR-145 as an important regulator of human chondrocyte function and a new target for cartilage repair.

Cartilage master regulator transcription factor SOX9 is essential for cartilage development in mice (1), whereas heterozygous mutations in the human SOX9 gene cause the severe skeletal malformation syndrome campomelic dysplasia (2). Decreased SOX9 expression underlies the loss of the differentiated phenotype of human articular chondrocytes during subculture in vitro and represents the major shortcoming of current cell-based cartilage repair therapies (3,4). In addition, SOX9 levels are suppressed in human osteoarthritic cartilage (5), and this most likely contributes to the phenotypic instabil-ity observed in osteoarthritis. Transcription factor SOX9 has been shown to be essential for expression of the differentiated chondrocyte phenotype by directly activating genes coding for key cartilage extracellular matrix components such as Col2a1 (6) and aggrecan (7). However, less is known about the mechanisms controlling expression of SOX9 itself, although we have previously shown that SOX9 is regulated by hypoxia via transcription factor HIF-2␣ 2 in human articular chondrocytes (4).
MicroRNAs (miRNAs) are small non-coding RNA molecules of 21-23 nucleotides that control gene expression at the posttranscriptional level and have been shown to play a vital role in a wide variety of biological processes, whereas dysregulated expression of miRNAs is found in many pathological conditions (8) including osteoarthritis (9). The most studied miRNA in cartilage is miR-140, the level of which is relatively high in normal articular cartilage but reduced in osteoarthritic tissue (10). miR-140 knock-out mice are viable but develop osteoarthritic-like changes with age (11). Recent work in our laboratory has uncovered another microRNA, miR-675, which was shown to control expression of key cartilage matrix gene COL2A1 in human articular chondrocytes (12). Interestingly, expression of the primary miRNA transcripts for both miR-140 and miR-675 is regulated by SOX9 (12,13), thus further extending the known influence of this master regulator on cartilage function.
Here we show that miR-145, an miRNA whose main role to date has been liked to vascular smooth muscle cell maintenance (14,15) and human embryonic stem cell differentiation (16), can directly target SOX9 in normal human articular chondrocytes through binding a specific site in its 3Ј-UTR, which interestingly is not conserved in mice. Levels of miR-145 inversely correlate with protein and mRNA levels of SOX9 during human articular chondrocyte dedifferentiation. miR-145 overexpression in articular chondrocytes obtained from non-arthritic patients greatly reduced the levels of SOX9, together with the critical cartilage matrix components COL2A1 and aggrecan, whereas undesirable hypertrophy markers were significantly increased. Thus, we have identified miR-145 as a direct regulator of master transcription factor SOX9 with critical effects on human articular chondrocyte function, and we propose miR-145 as a promising target for cartilage repair.

EXPERIMENTAL PROCEDURES
Harvest of Human Articular Cartilage and Chondrocyte Culture-Healthy articular cartilage was obtained from patients after informed consent and following local ethics committee guidelines. Cartilage was harvested from the femoral condyle and tibial plateau following amputation due to sarcomas not involving the joint space. In total, tissue was obtained from 13 donors (average age of 38 years) (for full details, see Table 1). Cartilage specimens were collected on the day of surgery and cut into small pieces (1-2 mm 3 ). Diced cartilage was placed in 1.5 mg/ml collagenase type 2 (Worthington) with DMEM containing 10% fetal calf serum (FCS) (Biosera) and incubated at 37°C for 18 h with shaking. Isolated human articular chondrocytes were then passed through a cell strainer, pelleted, and washed twice with medium. Cells were seeded at a density of 8 ϫ 10 3 cells/cm 2 in DMEM with 10% FCS. Passage 0 cells were subcultured after 5-7 days of isolation and either used directly (P1 cells) or passaged once more (P2 cells) before experiments.
Transient Transfections and Luciferase Assays-First or second passage human chondrocytes were seeded at 2.4 ϫ 10 5 cells per 3.5-cm dish and transfected 1 day after seeding. Transfection with control or miR-145 inhibitors or precursors (Ambion) was carried out at a final concentration of 100 or 5 nM, respectively, using Lipofectamine 2000 (Invitrogen) for 4 h in Opti-MEM I. Following transfection, medium was changed with preequilibrated DMEM (in 20 or 1% oxygen) containing 10% fetal calf serum. Cells were incubated in each oxygen tension for a further 24 -40 h before termination of the experiment.
Plasmid DNA was similarly transfected with Lipofectamine 2000 into human articular chondrocytes (8 ϫ 10 4 cells/well in a 12-well plate). After 24 h, cells were lysed, and luciferase activity was determined with the Dual-Glo luciferase assay system (Promega).
Centrifugation was carried out at 14,000 ϫ g at 4°C for 10 min. Supernatant was mixed with 900 ml of ice-cold NT2 buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 0.05% Nonidet P-40) containing freshly added 200 units/ml RNaseOUT (Invitrogen), 0.5% vanadyl ribonucleoside (Invitrogen), 1 mM DTT, 15 mM EDTA, and 50 ml of mouse anti-human Ago2 (clone 2E12-1C9, Abnova, Taipei City, Taiwan)-coated Sepharose G beads (Abcam, Cambridge, UK). Incubation was carried out overnight at 4°C on a rocking platform. On the following day, beads were washed five times with ice-cold NT2 buffer and separated into two portions: one for RNA isolation to identify miRNAs and another portion for Western blotting to check for successful immunoprecipitation of Ago2. Mouse IgG1 isotype control (Abcam) was used as a negative control for the immunoprecipitation procedure. RNA (including miRNAs) from the IgG or Ago2 immunoprecipitations was isolated using TRIzol and glycogen (Invitrogen) as a carrier in the ethanol precipitation step. miR-145 detection was performed using a TaqMan specific assay as described above.
Statistics-Statistical significance was evaluated by twotailed Student's t test. All data are shown as mean Ϯ S.E. p Ͻ 0.05 was considered statistically significant.

Levels of miR-145 Inversely Correlate with SOX9 Expression during Human Articular Chondrocyte Dedifferentiation-A
TaqMan low density array was performed on RNA extracted from human articular chondrocytes from three different patients (showing no signs of joint disease). MicroRNAs potentially relevant to the differentiated chondrocyte phenotype were identified as those whose expression was altered with dedifferentiation through serial passage (data not shown). From this preliminary list, miR-145 was selected for validation using an individual TaqMan assay for the mature miRNA form because SOX9 mRNA was predicted to be targeted by this miRNA by the TargetScan software. miR-145 levels increased with loss of the differentiated chondrocyte phenotype, whereas SOX9 expression was decreased at the RNA and, more notably, at the protein level (Fig. 1). These experimental data, together with bioinformatic analysis, suggested that miR-145 could be directly involved in the repression of SOX9 levels during dedifferentiation. In humans (and larger animals), the avascular cartilage is hypoxic, and we have previously shown that hypoxia promotes the differentiated articular chondrocyte phenotype through HIF-2␣-mediated induction of SOX9 (Lafont et al. (4). Human chondrocytes showed increased levels of SOX9 upon exposure to hypoxia, whereas miR-145 expression remained unchanged (Fig. 1). Nevertheless, given that hypoxia is a normal physiological condition of chondrocytes in human articular cartilage, we performed subsequent experiments in hypoxia (in addition to normoxia, 20% oxygen).
miR-145 Negatively Regulates Endogenous SOX9 Expression in Human Articular Chondrocytes-To assess whether miR-145 was regulating SOX9, the mature 23-nucleotide miRNA was inhibited (with antisense chemically modified RNA-based inhibitors) or overexpressed in transient transfection experiments in human articular chondrocytes obtained from a range of donors following amputations due to sarcomas not involving the joint (Table 1). Transfection efficiency was high for pre-miRNA precursors but much lower for the miRNA inhibitors, with no negative effect on viability in either case (data not shown). The miRNA precursors used to overexpress the mature miRNAs are small, partially double-stranded RNAs that mimic endogenous precursor miRNAs. Mature miRNAs function as a part of the RNA-induced silencing complex, which is

miR-145 Directly Targets SOX9 in Human Chondrocytes
the effector complex where miRNAs regulate the translation/ stability of their target mRNAs. Association of the mature overexpressed miRNA with the main RNA-induced silencing complex component Argonaute2 was found to be 6-fold higher than endogenous levels and thus represents an unphysiological level of overexpression ( Fig. 2A). Such miR-145 overexpression greatly decreased the levels of SOX9 both at the mRNA and particularly at the protein level (Fig. 2, B and C). Such results are consistent with a common mechanism of miRNA-mediated regulation were translational repression mediated by the miRNA is accompanied by partial degradation of the target mRNA (17). Interestingly, overexpression of miR-145 appeared to have a greater impact in hypoxic conditions and largely prevented the increase in SOX9 expression induced by hypoxia. miR-145 inhibition resulted in increased SOX9 protein levels ( Fig. 2D), without significant changes in SOX9 mRNA levels (data not shown). We attribute the subtlety of the observed effect to the low transfection efficiency obtained for these molecules.
miR-145 Directly Targets SOX9 in Human Articular Chondrocytes Specifically through Nucleotides 266 -288 in 3Ј-UTR-miR-145 was predicted to target two sites in the 3Ј-UTR of SOX9, located at positions 266 -288 and 1386 -1408 (Fig. 3A). Reporter constructs containing full-length SOX9 3Ј-UTR were made in which the sites were deleted individually or in combination. Expression of the 266 -288 mutant reporter (site 1 deletion) was 2-fold higher than that of the wild type; however, deletion of the 1386 -1408 binding site (site 2) did not affect luciferase activity (Fig. 3B). Simultaneous deletion of both binding sites had no further effect, suggesting that the region 266 -288 (site 1) is the one required for the observed miRNA-mediated repression. Furthermore, miR-145 overexpression reduced by 2-fold the expression of a luciferase reporter containing wild type SOX9 3Ј-UTR. This effect was completely lost by the deletion of the putative 266 -288 miR-145 binding site, thus confirming site 1 as the relevant one for the observed miRNA-mediated repression (Fig. 3C).

miR-145 Regulates Expression of Key Cartilage Genes and microRNAs in Human Articular
Chondrocytes-To further investigate the functional significance of miR-145, we checked the effect of its overexpression on the production of the key cartilage-specific matrix molecules by human articular chon-drocytes obtained from six different patients. Levels of direct SOX9 target genes COL2A1 and aggrecan, both essential to normal cartilage function, were significantly reduced by miR-145 overexpression under both normoxic and hypoxic conditions (Fig. 4, A and B). Hypertrophy markers RUNX2 and MMP13 were both decreased by hypoxia but subsequently increased by miR-145 overexpression (Fig. 4, C and D, respectively). We have previously shown that expression of primary miR-675 transcript H19 is dependent on SOX9 and that miR-675 overexpression rescues COL2A1 levels in SOX9-depleted cells, indicating that regulation of COL2A1 by SOX9 is mediated, at least in part, by miR-675 (12). In the present study, overexpression of miR-145 markedly reduced both H19 and miR-675 levels in human chondrocytes, probably indirectly through SOX9 down-regulation (Fig. 5, A and B). However, miR-675 overexpression was not enough to restore COL2A1 levels after miR-145 overexpression (data not shown), indicating that miR-145 may be affecting COL2A1 levels through pathways other than the SOX9 3 H19 3 miR-675 axis. Interestingly, expression of the other known SOX9-dependent miRNA in cartilage (miR-140) was also reduced by miR-145 overexpression (Fig. 5C), although levels of aggrecanase ADAMTS5, a proposed target of miR-140 were not significantly altered (Fig. 5D). As a form of control, we observed that miR-21-3P, a non-SOX9-dependent miRNA not implicated in cartilage function, was unaffected by manipulation of miR-145 (data not shown).

DISCUSSION
SOX9 is essential for normal cartilage development and function (1,18). Much is known about the downstream targets of SOX9. In addition to direct activation of all the major tissuespecific matrix genes that give the tissue its function (loadbearing and articulation) (19,20), recent work has shown that SOX9 also activates transcription of two key microRNAs in human cartilage (miR-140 and miR-675), which mediate important functions in the tissue (11,12). Due to this extensive regulation of the chondrocyte phenotype by SOX9, identifying upstream regulators of this gene is critical both to our understanding of cartilage homeostasis and for developing new therapeutic approaches to cartilage repair. Although we have previously shown that transcription factor HIF-2␣ up-regulates SOX9 in hypoxia (4), relatively little is known about the allimportant upstream regulators of this gene. Here we show for the first time in human articular chondrocytes that expression of SOX9 is subject to post-transcriptional regulation by an miRNA (miR-145) binding to a specific site in its 3Ј-UTR.
In the present study, of the two predicted miR-145 binding sites in the human SOX9 3Ј-UTR, only one (which is not conserved in mice) was shown to be targeted by miR-145. Deletion of the second putative binding site (which is conserved in mice) from luciferase constructs harboring full-length human SOX9 3Ј-UTR had no effect on luciferase expression in human chondrocytes. Interestingly, a most recent study reports an effect of miR-145 on Sox9 levels in a murine cell line (C3H10T1/2) undergoing growth factor-induced chondrogenic differentiation (21). Although Yang et al. (21) observed altered luciferase expression in response to modulation of miR-145 using con- Human articular chondrocytes were isolated from a total of 13 different donors, with an average age of 38 years. Non-diseased cartilage was harvested from the knee following amputations due to soft tissue and osteosarcomas not involving the joint space. 1  53  Female  2  65  Male  3  16  Male  4  23  Male  5  62  Female  6  53  Male  7  35  Male  8  23  Male  9  11  Male  10  16  Male  11  66  Female  12  47  Male  13 22 Male JANUARY 6, 2012 • VOLUME 287 • NUMBER 2 structs containing isolated copies of this conserved predicted miR-145 binding site, assays were not performed in the context of full-length Sox9 3Ј-UTR. Therefore, binding in the mouse system still needs to be assessed under more endogenous conditions, i.e. using full-length Sox9 3Ј-UTR sequences. Nevertheless, if Sox9 proves to be directly targeted by miR-145 in mice (albeit by a different site than in humans), then tissuespecific deletion studies will be most informative to delineate the role of this microRNA in cartilage in vivo including investigation of its potential role in osteoarthritis using suitable mouse models.

miR-145 Directly Targets SOX9 in Human Chondrocytes
In humans, the avascular cartilage is hypoxic, and we have previously shown that hypoxia promotes the differentiated articular chondrocyte phenotype through HIF-2␣-mediated induction of SOX9 (4). Interestingly, overexpression of miR-145 appeared to have a greater impact in hypoxic conditions and largely prevented the increase in SOX9 expression induced by hypoxia. Nevertheless, we found that levels of HIF-2␣ remain unchanged after miR-145 overexpression (data not shown), and steady-state levels of miR-145 were not altered in hypoxia. Therefore, it is likely that the hypoxia-induced regulation of SOX9 via HIF-2␣ occurs independently of miR-145. On the other hand, miRNA activity may alter without changes in their steady-state levels, e.g. by regulation of the accessibility to their target mRNA by RNA-binding proteins in response to different stimuli (22,23). Whether hypoxia affects the capacity of miR-145 to modulate gene expression remains to be investigated.
The mechanisms maintaining articular chondrocyte homeostasis are perturbed in osteoarthritis, in which chondrocytes recapitulate certain aspects of endochondral ossification including increased expression of transcription factor RUNX2, which promotes hypertrophy, leading to elevated levels of type II collagen-degrading metalloproteinase MMP13 (24,25). Interestingly, in the present study, these two genes (RUNX2 and MMP13) were both decreased by hypoxia but subsequently increased by miR-145 overexpression. The normal healthy articular chondrocyte phenotype is maintained with a low level    JANUARY 6, 2012 • VOLUME 287 • NUMBER 2 JOURNAL OF BIOLOGICAL CHEMISTRY 921 of RUNX2 and relatively high SOX9 expression (26). Interestingly, Sox9 has been shown to negatively regulate Runx2 via direct activation of transcription factor Bapx2 in proliferating murine chondrocytes (27), and SOX9 expression is known to decrease prior to chondrocyte hypertrophy in the growth plate (28,29). Furthermore, SOX9 expression levels are reduced in osteoarthritic cartilage when compared with age-matched (non-diseased) controls (29). Hattori et al. (30) performed some very informative experiments by overexpressing Sox9 specifically in hypertrophic chondrocytes in mice (using a BAC-Col10a1 promoter) and demonstrated that Sox9 is a negative regulator of endochondral ossification, and hypertrophic markers such as Mmp13 were down-regulated in Sox9-overexpressing hypertrophic chondrocytes. Hence, it is reasonable to speculate that maintenance of SOX9 levels in articular chondrocytes, at least partially mediated by miR-145, is pivotal not just to maintenance of the articular phenotype but to prevention of the hypertrophy-related changes that occur in osteoarthritis. Interestingly, in profiling experiments measuring levels of a total of 723 miRNAs in human articular chondrocytes, miR-145 was one of only three miRNAs found to be significantly up-regulated in cultured osteoarthritic chondrocytes when compared with non-diseased controls (31). The extent to which miR-145 levels are altered in arthritic tissue itself remains to be determined, in particular in the early localized lesions, which may be more informative of initiating events.

miR-145 Directly Targets SOX9 in Human Chondrocytes
We report that miR-145 levels are abundant in human articular chondrocytes where SOX9 is specifically expressed and has an essential role in tissue function and homeostasis. However, miR-145 has been shown to be expressed in various mouse tissues (although cartilage was not tested), being highly abundant in fat and the aorta, and loss of miR-145 induced structural modifications specifically in the proximal aorta, where the aorta emerges from the heart (15). This is intriguing because the aortic valves in the heart express Sox9, which when deleted in the mouse leads to calcification of the aortic valves, the most prominent form of valvular disease (32). Significantly increased expression of Runx2 and other osteogenic markers was detected in these Sox9-deficient aortic valves. Thus, SOX9 appears to play a strikingly similar role in maintaining cellular phenotype and preventing hypertrophy and calcification in two very different tissues, articular cartilage and aortic heart valves, and the possible role of miR-145 in heart valve function and pathology may be an important area for future investigation.
SOX9 has been targeted in studies aimed at treatment of hyaline articular cartilage damage, including Sox9 cDNA transfected in mesenchymal stem cells (33) or viral transduction with SOX9 in passaged osteoarthritic articular chondrocytes (34). Through upstream miRNA targeting, here we offer a way of manipulating SOX9 expression without resorting to gene therapy and its associated hazards. Furthermore, the relative ease by which miRNAs can be manipulated pharmacologically

miR-145 Directly Targets SOX9 in Human Chondrocytes
provides distinct therapeutic opportunities (35). There are several tools available to selectively target miRNA pathways in vivo including both synthetic mimics and inhibitors. Anti-miR inhibitors are modified antisense oligonucleotides with full or partial complementarity of a mature miRNA that can reduce the endogenous levels of an miRNA. Currently there are over 70 clinical trials worldwide (completed or actively recruiting) based on miRNA manipulation to treat a range of conditions including various cancers and cardiovascular diseases; however, there are none to date for arthritis (source: www.clinicaltrials.gov). MicroRNAs normally simultaneously regulate the expression of several targets, and experiments to determine other direct targets of miR-145 are currently being carried out in our laboratory. From a therapeutic viewpoint (with regard to both efficacy and safety), it will be important to identify all such directly targeted pathways before proceeding with clinical trials.
In summary, we show that cartilage master regulator SOX9 is subject to significant post-transcriptional regulation by a specific microRNA (miR-145), which is highly expressed in human articular chondrocytes. This novel microRNA-mediated regulation involves a network involving complex interactions between microRNAs and key transcription factors and exerts profound phenotypic effects in human chondrocytes (Fig. 6). The extent to which these phenotypic changes are a direct consequence of miR-145 targeting of SOX9 remains to be determined. Importantly, our data suggest that manipulation of SOX9-targeting miR-145 in chondrocytes could provide a much needed successful treatment for cartilage injuries and potentially for the chronic disease of osteoarthritis, a condition for which there is still no disease-modifying treatment. miR-145 directly targets (inhibits) essential master regulator SOX9, which exerts profound phenotypic effects by up-regulation of two key cartilagespecific microRNAs in addition to directly targeting the essential matrix genes.