miR-29 Modulates Wnt Signaling in Human Osteoblasts through a Positive Feedback Loop*

Differentiation of human mesenchymal stem cells into osteoblasts is controlled by extracellular cues. Canonical Wnt signaling is particularly important for maintenance of bone mass in humans. Post-transcriptional regulation of gene expression, mediated by microRNAs, plays an essential role in the control of osteoblast differentiation. Here, we find that miR-29a is necessary for human osteoblast differentiation, and miR-29a is increased during differentiation in the mesenchymal precursor cell line hFOB1.19 and in primary cultures of human osteoblasts. Furthermore, the promoter of the expressed sequence tag containing the human miR-29a gene is induced by canonical Wnt signaling. This effect is mediated, at least in part, by two T-cell factor/LEF-binding sites within the proximal promoter. Furthermore, we show that the negative regulators of Wnt signaling, Dikkopf-1 (Dkk1), Kremen2, and secreted frizzled related protein 2 (sFRP2), are direct targets of miR-29a. Endogenous protein levels for these Wnt antagonists are increased in cells transfected with synthetic miR-29a inhibitor. In contrast, transfection with miR-29a mimic decreases expression of these antagonists and potentiates Wnt signaling. Overall, we demonstrate that miR-29 and Wnt signaling are involved in a regulatory circuit that can modulate osteoblast differentiation. Specifically, canonical Wnt signaling induces miR-29a transcription. The subsequent down-regulation of key Wnt signaling antagonists, Dkk1, Kremen2, and sFRP2, by miR-29a potentiates Wnt signaling, contributing to a gene expression program important for osteoblast differentiation. This novel regulatory circuit provides additional insight into how microRNAs interact with signaling molecules during osteoblast differentiation, allowing for fine-tuning of intricate cellular processes.

Human mesenchymal stem cells have the potential to differentiate into multiple cell lineages, including osteoblast, adipocyte, chondrocyte, and neural (1)(2)(3)(4). Differentiation into a specific lineage depends on the extracellular cues received by the cell. Bone-anabolic agents such as bone morphogenetic protein (BMP)-2, parathyroid hormone, and canonical Wnts promote osteoblast differentiation through the stimulation of distinct second messenger pathways (5). In particular, canonical Wnt signaling is crucial for the regulation of bone mass in humans and for osteoblast differentiation (6 -10).
The binding of a canonical Wnt member (i.e. Wnt3a) to a Frizzled receptor and a co-receptor, LRP5/6, initiates a signaling cascade that results in the release of cytoplasmic ␤-catenin from an inhibitory complex consisting of Axin and glycogen synthase kinase (GSK)-3. Upon dephosphorylation and release, ␤-catenin can translocate to the nucleus, where it can interact with the T-cell factor/lymphoid enhancer factor-1 (TCF 2 / LEF1) family of transcription factors and activate transcription of genes necessary for osteoblast differentiation (11). Wnt signaling can be attenuated by several classes of negative regulators. For example, Dkk1 is a soluble factor that acts in conjunction with Kremen2, a decoy receptor, to inhibit Wnt signaling by preventing the binding of Wnt proteins to the LRP5/6 coreceptor (12). There is a marked increase in bone mineral density in Dkk1 haploinsufficient mice and in Kremen-null mice (13,14). Furthermore, the chromosomal region that contains the DKK1 locus has been linked to reduced bone mass in young osteoporotic men (15). In addition, the secreted frizzled-related protein (sFRP) family of proteins also binds extracellular Wnts, to prevent their binding to cell surface receptors. Wnts and their negative regulators have critical roles in bone development and/or maintenance, and their expression is modulated during the course of osteoblastic differentiation (16 -18).
Canonical Wnt signaling increases bone mass by a number of mechanisms (19). During early differentiation, Wnt signaling promotes mesenchymal stem cells proliferation (20 -22). Canonical Wnt signaling then drives the differentiation of osteochondral progenitors toward the osteoblastic lineage (23). In addition, Wnt signaling inhibits osteoblast and osteocyte apoptosis (19). However, it appears that canonical Wnt signaling regulates osteoblast differentiation in a dose-and time-dependent manner and that pathway components must be tightly regulated for proper differentiation. For example, low doses of LiCl or Wnt3a stimulate proliferation of human bone marrow-derived mesenchymal stem cells, but higher doses actually inhibit proliferation and initiate osteoblastic differentiation (20,22). Indeed, the expression of two negative regulators of Wnt signaling, sFRP2 and Dkk1, is decreased in mature osteoblasts, providing a potential mechanism for increased Wnt signaling in more differentiated cells (18). * This work was supported, in whole or in part, by National Institutes of Health There is increasing evidence that post-transcriptional regulation of gene expression, mediated by microRNAs (miRNAs), plays an important role in the control of osteoblastic differentiation (24 -28). miRNAs are 21-23-nucleotide, noncoding RNAs that negatively regulate gene expression at the post-transcriptional level (29). Mature miRNAs modulate gene expression by base pairing to complementary sequences within an mRNA target. Through their association with the RNA-induced silencing complex, miRNAs can facilitate degradation of the bound transcript or inhibit its translation. Because a single miRNA can have many targets, it is possible that one miRNA could regulate families of structural molecules or could regulate distinct signaling molecules within a particular pathway (30 -33). For example, we and others have demonstrated that the miR-29 family, consisting of miR-29a, miR-29c, miR-29b1, and miR-29b2, can down-regulate the expression of fibrillar collagens (i.e. COL1A1, COL3A1, and COL4A2) and the expression of osteonectin/SPARC (secreted protein acidic and rich in cysteine), which regulates collagen fibrillogenesis (24,25,34,35).
In addition, miR-29 is important for murine osteoblast differentiation (24,25). The expression of miR-29 family members increases during the progression of osteoblastic differentiation in primary cultures of murine calvarial osteoblasts. We demonstrated that expression of miR-29a and miR-29c is rapidly induced by canonical Wnt signaling in these osteoblasts (24). However, the mechanisms mediating this up-regulation and its consequences remain unexplored.
miR-29c and miR-29b2 are transcribed in tandem, in the same primary miRNA (pri-miRNA). This pri-miRNA is found within the last exon of an expressed sequence tag (EST) on human chromosome 1 (36). The pri-miRNA is processed to yield the two distinct miRNAs, miR-29c and miR-29b2. An interaction between the transcription factor Myc and the putative promoter region of the miR-29c-b2 gene was recently demonstrated, and Myc is thought to down-regulate expression of the gene in lymphoma cells (36). Similarly, in muscle cells, the transcription factor YY1 interacts with the miR-29c-b2 gene promoter and represses its expression (33). Like miR-29c and miR-29b2, miR-29a and miR-29b1 are transcribed within the same pri-miRNA. However, mapping of the pri-miRNA structure demonstrated that the miR-29a-b1 pri-miRNA is contained within an intron of a spliced EST on chromosome 7. The transcription factor Myc has also been shown to interact with the putative promoter region of this EST (36).
Because canonical Wnt signaling rapidly induces miR-29 levels, we hypothesized that Wnt signaling induces miR-29 transcription. In addition, given that both miR-29 and Wnt signaling are important for osteoblastic differentiation and that the expression of some negative regulators of Wnt signaling is decreased during osteogenesis, we hypothesized that miR-29 may negatively regulate inhibitors of Wnt signaling. We chose to test these hypotheses in human osteoblasts, focusing on the promoter for the miR-29a-b1 gene, because it is less well studied and because osteoblasts appear to have higher levels of miR-29a than miR-29c (24). Furthermore, we determined whether miR-29a could decrease the expression of selected negative regulators of Wnt signaling.

EXPERIMENTAL PROCEDURES
Cell Culture-The human fetal osteoblastic 1.19 cell line (hFOB) was obtained from American Type Culture Collection (ATCC, Manassas, VA). The hFOB cell line carries a temperature-sensitive SV40 T antigen transgene. Culture at the permissive temperature of 33.5°C allows these cells to proliferate, and culture at the restrictive temperature of 39.5°C slows proliferation and promotes differentiation (1). hFOB cells were maintained at 33.5°C in complete medium consisting of 1:1 Dulbecco's modified Eagle's medium/Ham's F-12 medium without phenol red (Invitrogen), supplemented with 10% fetal bovine serum (Atlas Biologicals, Fort Collins, CO), 0.3 mg/ml G418/ geneticin (Calbiochem), and 1ϫ penicillin/streptomycin (Invitrogen). For in vitro osteoblastic differentiation, confluent cultures of hFOB cells were maintained in complete medium with the addition of differentiation cocktail as follows: 100 g/l ascorbic acid, 10 Ϫ8 M menadione, 5 mM ␤-glycerol phosphate, and 10 Ϫ7 M 1-25(OH) 2 -vitamin D 3 (all from Sigma) (1).
Primary cultures of human osteoblasts were derived from bone chips obtained from healthy individuals undergoing routine orthopedic procedures (37)(38)(39)(40)(41)(42). Isolation of osteoblasts from these de-identified specimens was deemed exempt by the University of Connecticut Health Center Institutional Review Board. Experiments were performed on cells isolated from both males and females. Cells were maintained in 1:1 Dulbecco's modified Eagle's medium/Ham's F-12 medium without phenol red, supplemented with 10% nonheat-inactivated fetal bovine serum and 1ϫ penicillin/streptomycin. At confluence, culture medium was supplemented with 100 g/l ascorbic acid and 5 mM ␤-glycerol phosphate, to induce osteoblast differentiation. Primary human osteoblasts were cultured for up to 4 weeks, with twice weekly change of medium. Alkaline phosphatase staining was performed with a kit from Sigma.
Constructs-PCR and appropriate primer sets were used to amplify the first 2 kb and several deletion mutants of the human miR-29a putative promoter region (EU154353) using human genomic DNA as a template (36). Promoter fragments were subcloned into the luciferase reporter pGL4.10 (Promega, Madison, WI) using KpnI and EcoRV restriction enzyme sites. Site-directed mutagenesis of putative TCF/LEF-1-binding sites was performed using overlap extension (43).
Transfections-hFOB cells were plated at 10,000 cells/cm 2 . After 24 h, FuGENE 6 (Roche Applied Science) (FuGENE/DNA ratio 3 l to 1 g) was used to co-transfect luciferase constructs and a constitutively expressed ␤-galactosidase construct (Clontech; GenBank TM accession number U02451), as a control for transfection efficiency. 24 h post-transfection, cultures were either maintained in complete medium at 33.5°C ("proliferative conditions") or in osteoblastic differentiation medium at 39.5°C ("differentiation conditions") for 48 h. Cultures were

miR-29/Wnt Feedback Loop in Osteoblastic Differentiation
serum-deprived for 24 h prior to harvest in 1ϫ Reporter Lysis Buffer (Promega), according to the manufacturer's instructions. All transfection experiments were performed at least three times, using n ϭ 4 or 6 for each experiment, and at least two preparations of each plasmid were tested.
Co-transfection of hFOB cells with luciferase-UTR constructs, ␤-galactosidase expression construct, and 70-nm anti-miR miRNA inhibitors (Ambion, catalog no. AM1700) or pre-miR miRNA precursor mimics (Ambion, catalog no. AM17100) was accomplished using X-tremeGENE reagent (X-tremeGENE/nucleic acid ratio 5 l to 1 g; Roche Applied Science). 24 h post-transfection, cells were serumdeprived for 24 h and then harvested. Experiments utilizing miRNA inhibitors were performed at 39.5°C, except for Fig.  8B. This experiment was performed at 33.5°C, in proliferating conditions, because recombinant Wnt3a was incapable of stimulating TOPFlash activity at 39.5°C (data not shown). Experiments utilizing miRNA mimics were performed at 33.5°C, except for the experiment shown in Fig. 1, D and E.
This was performed in differentiating conditions because ALP and osteocalcin were not expressed in proliferating conditions (supplemental Fig. 1).
For some experiments, hFOB cells were transfected with the Wnt signaling luciferase reporter construct TOPFlash or its negative control, FOPFlash (Addgene, Cambridge, MA). Luciferase reporter and ␤-galactosidase expression constructs were co-transfected using FuGENE 6 or X-tremeGENE as described above. Cells were maintained at 33.5°C for 48 h posttransfection. Cells were serum-deprived for 24 h prior to treatment with 5-20 mM NaCl or LiCl (Sigma) or 0 -100 ng/ml recombinant human Wnt3a (R & D Systems, Minneapolis, MN) for up to 6 h. TOPFlash and FOPFlash were normalized to ␤-galactosidase. Data are represented as normalized TOPFlash/FOPFlash values.
Equal aliquots of cell lysate were used to determine luciferase activity (luciferase assay system, Promega) and ␤-galactosidase activity (Galacton chemiluminescent assay system, Tropix, Bedford, MA). To control for transfection efficiency, luciferase activity was normalized to that of ␤-galactosidase.
Western Blotting-hFOB cells were cultured for 3 days postconfluence under proliferative or osteoblastic differentiation conditions. In selected experiments, hFOB cells were trans-FIGURE 1. miR-29a is necessary for osteoblast differentiation. A, relative quantity of miR-29a normalized to 5 S rRNA. *, significantly different from proliferative conditions, p Ͻ 0.05. hFOB cells were transiently transfected with miR-29a inhibitor and then cultured for 3 days under osteoblast differentiation conditions. Relative quantity (RQ) of ALP (B) and osteocalcin mRNA expression (C) was normalized to 18 S rRNA. hFOB cells were transiently transfected with miR-29a mimic and then cultured for 3 days under osteoblast differentiation conditions. Relative quantity of ALP (D) and osteocalcin (E) mRNA expression was normalized to 18 S rRNA. *, significantly different from the scrambled negative control p Ͻ 0.05.  AUGUST 13, 2010 • VOLUME 285 • NUMBER 33 fected with 50 -150 nM miR-29a miRNA inhibitor or negative control inhibitor (scrambled) using Oligofectamine (Invitrogen) (anti-miR miRNA inhibitors, Ambion) and then cultured under osteoblastic differentiation conditions for 3 days. Cells were serum-deprived for the last 24 h of culture.

miR-29/Wnt Feedback Loop in Osteoblastic Differentiation
Cell layers were lysed in 1ϫ sample buffer (62.5 mM Tris, pH 6.8, 10% glycerol, and 2% SDS) and homogenized. Protein content was measured using the Bio-Rad DC protein assay kit, according to the manufacturer's instructions. Equal amounts of cell layer were subjected to electrophoresis through a 10.5% SDS-polyacrylamide gel under reducing conditions and transferred to a polyvinylidene difluoride membrane (Millipore, Billerica, MA). Membranes were blocked with 5% nonfat milk powder in PBST (phosphate-buffered saline, 0.1% Tween) and probed with goat anti-human sFRP2 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), mouse anti-human Dkk1 (Millipore), mouse anti-human Kremen2 (Abcam, Cambridge, MA), mouse anti-human active (de-phosphorylated at Ser-37 or Thr-41) ␤-catenin (Millipore), or mouse anti-human total ␤-catenin (Millipore) primary antibody. Goat anti-mouse or rabbit antigoat horseradish peroxidase-conjugated secondary antibody (Sigma) was used, as appropriate. Bands were visualized by chemiluminescence (Cell Signaling, Danvers, MA) and fluorography. Blots were stripped and re-probed with goat anti-rabbit actin antibody (Sigma). Relative band intensities in scanned images were analyzed with ImageJ software (rsbweb.nih.gov). Western blot experiments were performed at least twice, with n ϭ 3-6 in each experiment.
Quantitative Real Time PCR-RNA was isolated from cells using TRIzol (Invitrogen) and quantified spectroscopically.
Equal amounts of RNA were treated with RQ1 DNase I (Promega) prior to analysis, to exclude any signal from DNA contamination. To quantify mRNA levels in total RNA, DNased RNA was reverse-transcribed with Moloney murine leukemia virus-reverse transcriptase (Invitrogen), followed by qPCR with iQ SYBR Green Supermix (Bio-Rad) in a MiQ qPCR cycler (Bio-Rad). The primer sets used are shown in supplemental Table 1. RNA levels were determined using standard curves and were normalized to 18 S rRNA. To quantify miRNA levels in total RNA, the mirVana qRT-PCR miRNA detection kit and primers (Ambion) were used, according to the manufacturer's instructions. 50-ng aliquots were used for miR-29a and 5 S detection. RNA levels were determined using standard curves, and miR-29 levels were normalized to 5 S rRNA (relative quantity). qRT-PCR experiments were performed at least twice, with n ϭ 3 for each experiment, and each sample was analyzed in duplicate.
Data Analysis-Data are presented as mean Ϯ S.E. Data were analyzed by one-way analysis of variance with Bonferroni posthoc test or by Student's t test (KaleidaGraph, Synergy Software, Reading, PA).

miR-29a Expression in Differentiating Human Osteoblasts-
In confluent cultures of hFOB mesenchymal progenitor cells and in proliferating cultures maintained for 3 days post-confluence in the absence of differentiation mixture, there was little ALP activity, which is a marker of early osteoblastogenesis (44,45). However, in cells cultured for 3 days post-confluence in osteoblastic differentiation medium, ALP staining was readily apparent (supplemental Fig. 1A). qRT-PCR confirmed the induction of ALP mRNA by the osteoblastic culture conditions, which was accompanied by increased expression of transcripts for Runx2, a transcription factor critical for osteoblastic differentiation, and osteocalcin, a mature osteoblast marker (supplemental Fig. 1, B-D) (41, 46 -48). Importantly, miR-29a levels were also increased ϳ3-fold in these differentiating osteoblasts (Fig. 1A), similar to our observations of miR-29 expression in mouse osteoblasts (24).
To determine the function of miR-29a in human osteoblastic differentiation, hFOBs were transiently transfected with miR-29a inhibitor and then cultured under osteoblastic differentiation conditions. qRT-PCR showed that when miR-29a function is blocked, ALP (Fig. 1B) and osteocalcin (Fig. 1C) mRNA expression was significantly decreased compared with the

miR-29/Wnt Feedback Loop in Osteoblastic Differentiation
scrambled control inhibitor. When cells were transfected with miR-29a mimic, ALP mRNA levels (Fig. 1D) were unaffected, but osteocalcin (Fig. 1E) mRNA expression was significantly increased compared with the scrambled control mimic. miR-29 is not predicted to directly interact with either ALP or osteocalcin mRNA. These data suggest that miR-29a promotes osteoblastic differentiation and that loss of miR-29a function suppresses this process.
Next, we quantified miR-29a levels in differentiating cultures of primary human osteoblasts. We tested cells derived from bone chips obtained from two individuals, a 62-year-old female and a 54-year-old male (Fig. 2). Osteoblastic cells were grown to confluence and then cultured in the presence of ascorbic acid and ␤-glycerol phosphate for up to 4 weeks, promoting osteoblastic differentiation. qRT-PCR showed that, for the most part, Runx2 mRNA levels peaked in the early weeks of the differentiation process and were subsequently down-regulated (Fig.  2) (46). In samples derived from females, osteocalcin mRNA peaked at 4 weeks of culture ( Fig. 2 and data not shown), whereas in the male sample, osteocalcin mRNA peaked after only 1 week of culture. In all samples tested, miR-29a expression gradually increased with time in culture, with highest levels after 3-4 weeks, suggesting that miR-29a may be a marker of differentiation. Because all three markers of osteoblastic differentiation are increased in both the hFOB cell line and primary human osteoblasts, we proceeded to use the hFOB model system to investigate miR-29a promoter activity and the effects of miR-29a on Wnt signaling.
Canonical Wnt Signaling Up-regulates miR-29a in Human Osteoblasts-Treatment of cells with LiCl mimics canonical Wnt signaling by inhibiting GSK3 phosphorylation of ␤-catenin (49,50). To confirm this effect in our system, we treated hFOBs with 5-20 mM LiCl for 3 h and examined relative levels of active (de-phosphorylated) ␤-catenin using Western blot analysis (Fig.  3A). 10 mM LiCl increased active ␤-catenin compared with the NaCl control, whereas 5 and 20 mM LiCl did not. To document that LiCl activates canonical Wnt signaling in hFOBs, we used the Wnt signaling reporter construct TOPFlash. In this TOPFlash reporter, transcription of the luciferase gene is driven by 7 TCF/LEF-binding sites, upstream of a minimal promoter. Its negative control, FOPFlash, is a similar construct in which the TCF/LEF-binding sites are mutated. Treatment of transiently transfected hFOB cells with 10 mM LiCl for 6 h increased TOPFlash activity 4-fold, suggesting accumulation of nuclear ␤-catenin and activation of canonical Wnt signaling (Fig. 3B). Endogenous miR-29a expression was significantly increased in hFOBs treated with 10 mM LiCl for 1, 3, or 6 h, with the greatest fold increase apparent after 1 h (Fig. 3C). Furthermore, both 25 and 50 ng/ml recombinant Wnt3a significantly increased miR-29a expression after 3 h (Fig. 3D). A time course  AUGUST 13, 2010 • VOLUME 285 • NUMBER 33

JOURNAL OF BIOLOGICAL CHEMISTRY 25225
study demonstrated that 50 ng/ml Wnt3a induced miR-29a expression at 1 and 3 h but not at 6 h of treatment (Fig. 3E). Together, these data showed canonical Wnt signaling induces mature miR-29a expression in human osteoblasts.
miR-29a Promoter Is Activated by Wnt Signaling-The transcription start site for the EST containing the miR-29a-b1 coding region was previously defined (36). PCR and appropriate primer sets were used to amplify the genomic region Ϫ2159 to Ϫ1 bp from this transcription start site. This ϳ2.1-kb promoter fragment and a series of 5Ј deletion mutants were subcloned into a promoterless luciferase reporter construct (Fig. 4A). When transiently transfected into hFOB cells, the activity of the Ϫ2159 to Ϫ1 promoter-luciferase construct was significantly greater than that of the promoterless control. However, transcriptional activity was greatly increased with subsequent 5Ј deletions of the promoter, down to Ϫ688 bp. These data suggest the presence of negative regulatory elements between bp Ϫ2159 and Ϫ688 (Fig. 4B). The activity of the construct containing Ϫ142 to Ϫ1 of promoter region was not significantly different from the promoterless control. Thus, the minimal promoter for the miR-29a-b1 containing EST may lie within the Ϫ688 to Ϫ1 base region. Analysis of the 2159-bp promoter region using the CHIP Bioinformatics Mapper suggested the presence of three putative binding sites for TCF/LEF, the transcription factors activated by nuclear ␤-catenin during canonical Wnt signaling (51,52). Because the Ϫ982 to Ϫ1 promoter region showed high basal activity and had two putative TCF/ LEF sites, we used transient transfection of this promoter construct to determine whether the miR-29a promoter was responsive to canonical Wnt signaling. Similar to endogenous miR-29a, miR-29a promoter-reporter gene expression was significantly increased in cells treated with 50 ng/ml Wnt3a after 6 h (Fig. 4C). Treatment of hFOBs with 10 mM LiCl for 3 h also increased reporter gene expression (Fig. 4D). A time course study showed that treatment of cells with 10 mM LiCl for more than 3 h did not enhance reporter activity ( Fig. 4E and data not shown). These data suggested that canonical Wnt signaling can rapidly induce miR-29a promoter activity.
To determine whether the putative TCF/LEF-binding sites are responsible for this effect, we generated constructs with mutations at these sites, within the context of the Ϫ982 to Ϫ1 promoter region (Fig. 5A). Shown in Fig. 5B is the sequence of the putative TCF/LEF-binding sites within the miR-29 promoter (WT sequence) and the mutants we generated, which are predicted to abolish TCF/LEF binding. Basal level activity of the single TCF/LEF mutants was similar to that of the WT construct; however, the double mutant (mut/mut) had significantly greater reporter gene expression, suggesting that these mutations relieved TCF/LEF-mediated transcriptional repression (Fig. 5, C and D). Treatment of the wild type promoter construct with LiCl caused a significant increase in promoter activity, although this effect was abolished when either one or both TCF/LEF-binding sites were mutated (Fig. 5C). Similar effects were observed in transfected cells treated with 50 ng/ml Wnt3a (Fig. 5D). These data suggested that both of the TCF/LEF sites are necessary for the induction of miR-29a promoter activity by canonical Wnt signaling.

miR-29a Targets Negative Regulators of Wnt Signaling-Ca-
nonical Wnt signaling and miR-29a expression promote osteoblast differentiation (Fig. 1) (53). Furthermore, miR-29 is increased during osteoblastic differentiation, whereas selected negative regulators of Wnt signaling are decreased (Fig. 1) (18,24). To test the hypothesis that miR-29a targets antagonists of canonical Wnt signaling, we used RNAHybrid to identify potential targets of miR-29a that are known inhibitors of canonical Wnt signaling (54). Based on the potential for interaction of miR-29a with the 3Ј-UTR of each transcript, three potential targets, Dkk1, Kremen2, and sFRP2, emerged from FIGURE 5. Canonical Wnt signaling induces miR-29a promoter activity through two Tcf/Lef sites. A, diagram of the Ϫ982 to Ϫ1 miR-29a promoter and single or double mutants (mut) of the two predicted Tcf/Lef-binding sites. A black box indicates a wild type TCF/LEF site, and ϫ indicates a mutated site. B, consensus sequences for the Tcf/Lef-binding site, the wild type sequences in the endogenous Ϫ982 to Ϫ1 miR-29a promoter, and the sequences of the mutated binding sites (underlined). Activity of wild type and mutant Ϫ982 to Ϫ1 base constructs in transiently transfected hFOB cells treated with 10 mM NaCl or LiCl for 3 h (C) or with vehicle or 50 ng/ml recombinant human Wnt3a for 6 h (D). *, significantly different from NaCl or vehicle p Ͻ 0.05; #, significantly different from WT/WT, p Ͻ 0.05.

miR-29/Wnt Feedback Loop in Osteoblastic Differentiation
our search. Transcripts for Kremen2 and Dkk1 had one potential miR-29a-binding site, and the sFRP2 transcript had two potential binding sites in the 3Ј-UTR (Fig. 6A). We first determined whether the endogenous mRNA and protein levels of these Wnt antagonists were regulated during osteoblastic differentiation in hFOB cells. qRT-PCR and Western blot analysis showed that, compared with proliferating cells, Dkk1 and sFRP2 mRNA and protein levels were significantly decreased in hFOB cells undergoing osteoblastic differentiation (Fig. 6, B  and D, E and G). Although Kremen2 transcripts were not decreased in osteogenic cells, Kremen2 protein levels were significantly reduced, indicating repression of translation (Fig. 6, C  and F).
To determine whether the 3Ј-UTR of these selected Wnt antagonists could differentially regulate gene expression, we generated luciferase reporter constructs in which the 3Ј-UTR for Dkk1, Kremen2, or sFRP2 was cloned into the multiple cloning site of the pMIR-REPORT vector (Fig. 6A). The inserted 3Ј-UTR acts as the 3Ј-UTR for the luciferase gene, modulating luciferase transcript stability and/or translation. Constructs were transiently transfected into hFOB cells, and cells were then cultured under proliferative or osteogenic conditions. The 3Ј-UTR of each Wnt antagonist, Dkk1, Kremen2, and sFRP2, mediated a significant decrease in gene expression in osteogenic cells compared with proliferative cells (Fig. 7A).
Next, we determined whether miR-29a could directly regulate the 3Ј-UTRs of the selected Wnt antagonists. Luciferase 3Ј-UTR constructs were transiently transfected into hFOB cells along with scrambled or miR-29a inhibitors or mimics (Fig. 7, B and C). Inhibitortreated cells were cultured under osteogenic conditions, where endogenous miR-29a expression is high. Transfection with miR-29a inhibitor relieved 3Ј-UTR-mediated repression of luciferase activity of Dkk1, Kremen2, and sFRP2 3ЈUTR constructs. Cells transfected with the miRNA mimic were cultured under proliferating conditions, where miR-29a expression is low. Transfection with the miR-29a mimic decreased luciferase activity of Dkk1, Kremen2, and sFRP2 3ЈUTR constructs. Together, these data suggested that miR-29a acts directly on the 3Ј-UTRs of these Wnt signaling antagonists, to negatively regulate their expression.
We confirmed this by examining the endogenous protein levels of Dkk1, Kremen2, and sFRP2 when miR-29a function is blocked by transfection with the miR-29a inhibitor. Western blot analysis of transfected hFOB cells grown under osteogenic conditions showed that blocking miR-29a function significantly increased Dkk1, Kremen2, and sFRP2 protein expression compared with the scrambled control ( Fig. 7D-F). These data demonstrated that miR-29a negatively regulates endogenous levels of Dkk1, Kremen2, and sFRP2 protein during osteogenic differentiation in human cells.
Because miR-29 negatively regulates Wnt antagonists, we hypothesized that miR-29 positively affects Wnt signaling activity. To test this hypothesis, hFOB cells were transiently transfected with the Wnt signaling reporter TOPFlash or negative control FOPFlash, along with miR-29a mimic or scrambled control. In cells transfected with the scrambled control, treatment with Wnt3a caused a modest increase in TOPFlash activity, as seen in the past. However, in cells treated with miR-29a mimic, Wnt3a increased TOPFlash activity by ϳ4-fold, indicating that Wnt signaling was potentiated (Fig. 8A). In the converse experiment, when cells were transfected with miR-29a inhibitor, Wnt3a was unable to significantly increase TOPFlash activity (Fig. 8B). These data demonstrate that miR-29a can enhance Wnt signaling activity, likely due, at least in part, to its inhibition of Wnt antagonists.

DISCUSSION
Both canonical Wnt signaling and miR-29 promote osteoblast differentiation, through a variety of mechanisms (24,25,55). Here, we provide evidence for a novel positive feedback loop involving canonical Wnt signaling and miR-29, likely contributing to human osteoblast differentiation (Fig. 8C). Specifically, canonical Wnt signaling induces miR-29a transcription.
The subsequent down-regulation of key Wnt signaling antagonists, Dkk1, Kremen2, and sFRP2, by miR-29a can potentiate Wnt signaling and contribute to a gene expression program important for osteoblast differentiation. miRNAs orchestrate osteoblast differentiation through their regulation of diverse signaling molecules and pathways (56). For example, miR-210 acts on transcripts for activin A receptor type 1B, downregulating the receptor and promoting osteoblastic differentiation (57). In ST2 cells, BMP-2 induces expression of miR-2861, and miR-2861 down-regulates histone deacetylase (HDAC) 5. Because HDAC5 enhances degradation of Runx2, the induction of miR-2861 promotes differentiation. Furthermore, mice treated systemically with a miR-2861 inhibitor develop osteopenia (58).
Importantly, miRNAs can also inhibit osteoblast differentiation. Transgenic mice expressing miR-206 in osteoblasts develop a low bone mass phenotype due to impaired osteoblast differentiation, because miR-206 targets connexin43, which propagates signaling between osteoblasts to promote differentiation (27). Furthermore, miR-204 and -211 target Runx2 in bone marrow stromal cells. Because Runx2 is required for the transcription of genes important for osteoblast differentiation, overexpression of miR-204 represses osteoblast differentiation and promotes adipogenic differentiation (28). In addition, miR-26a inhibits osteoblast differentiation of human adipose-derived stem cells by down-regulating SMAD1, an important downstream mediator of bone morphogenetic protein signaling (59). miR-141 and miR-200a were recently found to inhibit osteoblast differentiation by targeting Dlx5 (distal-less homeobox 5), a transcription factor expressed in pre-osteoblasts (60). These studies highlight the wide range of signaling pathways modulated by miRNAs in osteoblasts.
There is a rapidly growing understanding of the mechanisms controlling the post-transcriptional processing of miRNAs, but there is less known about the regulation of miRNA gene transcription (61,62). miRNAs can be transcribed by RNA polymerase II or III, and transcription can be FIGURE 7. miR-29a negatively regulates Wnt antagonists. A, activity of luciferase-3Ј-UTR constructs in transiently transfected hFOB cells, under proliferating or differentiating conditions. *, significantly different from proliferative conditions, p Ͻ 0.05. B, activity of luciferase-3Ј-UTR constructs in hFOB cells, transiently co-transfected with 70 nm miRNA inhibitor and cultured under differentiating conditions. *, significantly different from scrambled control. C, activity of luciferase-3Ј-UTR constructs in hFOB cells, transiently co-transfected with 70 nm miRNA mimic and cultured under proliferative conditions. *, significantly different from scrambled control p Ͻ 0.05. Protein levels for Dkk1 (D), Kremen2 (E), and sFRP2 (F) in differentiating hFOBs treated with 50 -150 nm scrambled or miR-29a inhibitor. Protein expression was normalized to actin. *, significantly different from scrambled control, p Ͻ 0.05.
Here, we show that the proximal promoter region of the miR-29a-b1 containing EST has strong promoter activity when transfected into human osteoblasts. The 982-bp miR-29a promoter region contains two putative TCF/LEF sites. When both sites were mutated, there was a significant increase in basal promoter activity (Fig. 5). This is likely due to the repressor function of TCF/LEF. When the accumulation of nuclear ␤-catenin is limited, TCF/LEF can repress transcription through its interaction with the Groucho/TLE (Transducinlike Enhancer of Split) family of repressor proteins (71)(72)(73). Interaction of the co-activator ␤-catenin with TCF/LEF results in the displacement of Groucho/TLE and allows the induction of gene transcription (74). Interestingly, mutation of only one TCF/LEF site was not sufficient to relieve repression of the basal promoter activity. However, mutation of a single TCF/ LEF site was sufficient to abolish induction of the promoter by Wnt signaling. This suggests that both sites are needed to induce transcription stimulated by Wnt.
We show that miR-29 promotes osteoblastic differentiation, and its ability to down-regulate the expression of key Wnt signaling inhibitors likely contributes to this effect by promoting Wnt signaling. Like bone, Wnt signaling is important in cardiomyocyte differentiation. During myogenesis, miR-29c expression is de-repressed, promoting myogenic differentiation (33). In cardiomyocytes, sFRP2 inhibits differentiation by inhibiting transcriptional activation of Wnt3a (75). It is possible that targeting of sFRP2 RNA by miR-29 is one mechanism contributing to myogenic differentiation. Because miR-29 is expressed in a wide array of tissues, it is likely that other miR-29 regulatory networks could exist (76 -78).
miRNAs provide a mechanism for fine-tuning intricate cellular processes. In combination with a variety of transcription factors and signaling molecules, miRNAs help to control the complex program of osteoblast differentiation. Our data demonstrate that miR-29 and Wnt signaling are involved in a regulatory circuit that modulates osteoblast differentiation and likely impacts bone remodeling and the maintenance of bone mass. This novel regulatory circuit provides additional insight into how miRNAs interact with signaling molecules during osteoblast differentiation and has implications for the treatment of bone loss diseases.