Impaired Micro-RNA Pathways Diminish Osteoclast Differentiation and Function*

Micro-RNAs (miRNAs) are important in regulating cell fate determination because many of their target mRNA transcripts are engaged in cell proliferation, differentiation, and apoptosis. DGCR8, Dicer, and Ago2 are essential factors for miRNA homeostasis. Here we show that these three factors have critical roles in osteoclast differentiation and function. Gene silencing of DGCR8, Dicer, or Ago2 by small interfering RNA revealed global inhibition of osteoclast transcription factor expression and function, decreased osteoclastogenesis, and decreased bone resorption in vitro. In vivo, CD11b+-cre/Dicer-null mice had mild osteopetrosis caused by decreased osteoclast number and bone resorption. These results suggest that miRNAs play important roles in differentiation and function of osteoclasts in vitro and in vivo. We found a novel mechanism mediating these results in which PU.1, miRNA-223, NFI-A, and the macrophage colony-stimulating factor receptor (M-CSFR) are closely linked through a positive feedback loop. PU.1 stimulates miRNA-223 expression, and this up-regulation is implicated in stimulating differentiation and function of osteoclasts through negative regulation of NFI-A levels. Down-regulation of NFI-A levels is important for expression of the M-CSFR, which is critical for osteoclast differentiation and function. NFI-A overexpression decreased osteoclast formation and function with down-regulation of M-CSFR levels. Forced expression of the M-CSFR in M-CSF-dependent bone marrow macrophages from Dicer-deficient mice rescued osteoclast differentiation with up-regulation of PU.1 levels. Our studies provide new molecular mechanisms controlling osteoclast differentiation and function by the miRNA system and specifically by miRNA-223, which regulates NFI-A and the M-CSFR levels.

The skeleton in vertebrates is comprised of bone and cartilage that include three specific cell types: osteoblasts and osteoclasts in bone and chondrocytes in cartilage (1). Bone metabolism is maintained by a well organized balance of old bone resorption by osteoclasts and new bone formation by osteoblasts (1,2). Osteoblasts are bone-forming cells derived from multipotent mesenchymal stem cells (1). In contrast, multipo-tent hematopoietic stem cells give rise to myeloid stem cells, which further differentiate to megakaryocytes, granulocytes, and monocyte-macrophages (2). Osteoclasts are derived from monocyte-macrophage precursors (2). The osteoclast is the primary bone-resorbing cell, and it is formed from fusion of precursors to form multinucleated osteoclasts (2). Osteoclastogenesis is regulated by exogenous hormones and cytokines that positively or negatively modulate osteoclast proliferation, survival, differentiation, and function (3). Two essential cytokines, macrophage colony-stimulating factor-1 (M-CSF) 2 and receptor activator of NFB ligand (RANKL), are produced by osteoblasts or activated T cells and are required for osteoclast differentiation, function, and survival (1)(2)(3). Both M-CSF, as in the op/op mouse, and M-CSF receptor deficiency are characterized by osteopetrosis and the lack of macrophages and osteoclasts (1,2).
Many transcription factors involved in osteoclastogenesis have been identified through studies in genetically engineered mice (4 -10). Mouse mutants lacking factors that function early in the lineage including PU.1, c-Fos, and NFB (p50 and p52) are osteopetrotic and lack either macrophages and osteoclasts or only osteoclasts (4 -7). In particular, c-Fos and NFB (p50 and p52) are required for the differentiation of monocyte precursors into osteoclasts (5)(6)(7), and c-Fos induces a second transcription factor, NFATc1, that is essential for osteoclastogenesis (8,9). Other transcription factors, such as MITF, are also involved in osteoclastogenesis (10). Although much is known about the transcriptional regulation that controls osteoclast differentiation, potential roles for post-transcriptional gene regulation are not as well defined.
Micro-RNAs (miRNAs) are single-stranded RNAs 19 -25 nucleotides in length that regulate several pathways including the development timing, hematopoiesis, organogenesis, apoptosis, cell proliferation, and tumorigenesis (11). The first known miRNA, the lin-4 small temporal RNA, was discovered in the nematode Caenorhabditis elegans (12). Since then, several groups identified hundreds of miRNAs in different orga-nisms (13)(14)(15). miRNA genes are transcribed by RNA polymerase II to generate the primary transcripts (pri-miRNAs) (16). pri-miRNAs are processed by the double-stranded RNA-binding protein DiGeorge syndrome critical region gene 8 (DGCR8) and the nuclear RNase III enzyme Drosha into stem-loopstructured miRNA precursors (pre-miRNAs) (16). pre-miRNAs are exported by the nuclear export factor exportin-5 into the cytoplasm (16) and are processed by the cytoplasmic RNase III enzyme Dicer into mature miRNAs (17). Mature miRNAs are incorporated into the multiprotein RNA-induced silencing complex (RISC) with Dicer and Argonaute proteins (16). Argonaute2 (Ago2) binds miRNA and mediates gene silencing (18). Binding of miRNA-RISC to a partially complementary mRNA results in silencing by the inhibition of translation or mRNA degradation (18). On the other hand, direct cleavage of mRNA needs perfect or nearly perfect complementarity between the miRNA and the its target mRNA (16). Some miRNAs have been implicated in mouse hematopoietic differentiation (19 -22). For example, miRNA-223, a mouse bone marrow-specific miRNA, is specifically expressed in myeloid lineage, and its expression is regulated by PU.1 and CAAT enhancer-binding proteins (19,20). miRNA-181a expression is up-regulated only in differentiated B cells (19). miRNA-150 expression is elevated during developmental stages of B and T cell maturation (21). In transgenic mice expressing mouse pre-miRNA-155 in a B cell-specific manner, pre-B cell proliferation was markedly stimulated in the spleen and bone marrow, indicating that miRNA-155 regulates mouse pre-B cell proliferation (22). miRNA-146 is up-regulated by proinflammatory mediators in human monocytes (23).
Recent studies have shown that miRNAs play critical roles in osteoblast differentiation and bone development (24 -27). We reported the first evidence that miRNA-223 is important for osteoclastogenesis (24). Expression of miRNA-26a is up-regulated in late osteoblast differentiation of human adipose tissuederived stem cells, and miRNA-26a modulates this process by targeting the SMAD1 transcription factor (25). miRNA-125b controls osteoblast differentiation in mouse mesenchymal stem cells, ST2, by regulating cell proliferation (26). Dicer regulates chondrocyte proliferation and differentiation during skeletal development (27). Despite accumulating evidence that posttranscriptional gene regulation by miRNAs is essential for osteoclast, osteoblast, and chondrocyte differentiation in vitro and in vivo study, their roles in osteoclast differentiation and function are largely unknown compared with osteoblasts and chondrocytes. To study the global functions of miRNAs in osteoclast differentiation and bone resorbing activity, we employed a retroviral system for delivery of DGCR8, Dicer1, or Ago2 small interfering RNA into osteoclast precursors and CD11b ϩ -cre/ Dicer-null mice and found that the miRNA pathway for generation of miRNAs including miRNA-223 is critical for osteoclast differentiation and function, and miRNAs may be implicated in regulation of the bone remodeling cycle by osteoclasts. Thus, miRNAs including miRNA-223 would be expected to be potential targets of therapeutic strategies applied to drug discovery in bone metabolic disorders with excessive osteoclast activity such as osteoporosis.

EXPERIMENTAL PROCEDURES
Construction of Retroviral Vectors-The mouse DGCR8, Dicer1, and Ago2 siRNA target sequences were designed with search of off target effect by Enhanced siDirect (RNAi Co., Ltd.) ( Table 1). The 63-mer sense and antisense strands of DNA oligonucleotides were synthesized by Integrated DNA Technologies. Construction of a retroviral vector for delivery of small interfering (siRNA) and generation of virus were performed using the pSilencer 5.1-H retroviral system (Ambion) according to the manufacturer's instructions. Scrambled siRNA retroviral vectors were obtained from Ambion. PU.1, M-CSFR, and NFI-A cDNA (Open Biosystems) were inserted into pMX retroviral vector (Cell Biolabs, Inc.) digested with BamHI and NotI restriction enzymes. Clones containing the cDNA insertion were verified by DNA sequencing.
Cell Culture and Infection-Subconfluent Plat-E packaging cells (Dr. Kitamura, University of Tokyo, Tokyo, Japan) were transfected with retroviral vectors using FuGENE (Roche Applied Science). After 48 h, the supernatant was collected, filtered through a 0.45-m syringe filter, and used to transduce target cells. Mouse primary M-CSF-dependent bone marrow macrophages (osteoclast precursors) were prepared from the femur and tibia of 4 -6-week-old C57BL/6J mice (Jackson Lab) and Dicer-null osteoclast precursors and incubated in Petri dishes (100-mm dish) (Fisher) in ␣MEM (Sigma) containing 10% FBS (Atlas) in the presence of conditioned medium (1:10) from a M-CSF-producing CMG14-12 cell line that was a gift from Dr. Teitelbaum (Washington University School of Medicine, St. Louis, MO) (28). After 3 days in culture, the cells were plated in tissue culture dishes (100-mm dish) (Fisher) in ␣MEM containing 10% FBS in the presence of M-CSF conditioned medium (1:10). After 24 h in culture, the cells were infected with retrovirus for 6 h in ␣MEM containing 10% FBS in the presence of M-CSF conditioned medium (1:10) and 4 g/ml polybren (Sigma). Infected cells were cultured for an additional 3 days for puromycin (Sigma) selection, and the cells were harvested for assay.
Real Time RT-PCR-Total RNA was isolated from infected cells using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Two micrograms of total RNA were reverse transcribed to cDNA using SuperScript II Reverse Transcriptase (Invitrogen). Real time PCR analysis was performed using the SYBR green PCR method according to the manufacturer's suggestions (Applied Biosystems). The primer sequences are as follows: mouse DGCR8, forward, 5Ј-AGT TCA ACC GTG AGA TGA AGC GGA-3Ј, and reverse, 5Ј-AGG ATG CAA ACC TCA GAC TTC CCA-3Ј; mouse Dicer1, forward, 5Ј-GCA GGC TTT TAC ACA CGC CT-3Ј, and reverse, 5Ј-GGG TCT TCA TAA AGG TGC TT-3Ј; mouse Ago2, forward, 5Ј-AGT TCT TGC TTC TCC TGC TCC G-3Ј,  miRNA RT-PCR and Real Time RT-PCR-Total RNA was isolated from osteoclast precursors or infected cells using the mirVana miRNA Isolation Kit (Ambion) that was used according to the manufacturer's protocol. RT-PCR and real time RT-PCR were performed using the mirVana qRT-PCR miRNA Detection Kit (Ambion) that was used according to the manufacturer's suggestions. miRNA-223, miRNA-155, and U6 primer were purchased from Ambion. In RT-PCR analysis, the PCR products were confirmed using a 20% polyacrylamide gel, and U6 small nuclear RNA was used as a loading control. In real time RT-PCR analysis, the PCR products were normalized to U6 levels for each reaction.
Osteoclast Formation Assay-Infected cells and Dicer-null cells were plated in 24-well plates (Corning) in ␣MEM containing 10% FBS in the presence of M-CSF-conditioned medium (1:20) and RANKL (50 ng/ml). After 5 days of culture, the cells were stained for TRAP activity. TRAP-positive multinucleated cells containing more than three nuclei were counted as osteoclasts under microscopic examination. Osteoclasts bone resorbing activity was measured using the OsteoLyse TM assay kit (Lonza) according to the manufacturer's protocol.
Bone Histomorphometry-Bone histomorphometry analysis of femurs from Dicer wt/wt and Dicer f/f littermates (3 months old) was conducted at the University of Alabama at Birmingham Center for Metabolic Bone Disease Core Laboratory. The femurs were decalcified with 20% EDTA in PBS, embedded in paraffin, sectioned, and stained with hematoxylin and eosin and TRAP.
Transient Transfection-Antisense inhibitors of miRNA-223 and negative control oligonucleotides were obtained from Ambion. For antisense inhibition of miRNAs, RAW264.7 cells, a murine osteoclast progenitor cell line, were plated in 24-well plates (40% confluence) (Corning) and transfected with each antisense miRNA inhibitors or negative control using the siPORT NeoX transfection reagent (Ambion) according to the manufacturer's instructions. Subsequently, the cells were incubated in ␣MEM containing 10% FBS in the presence of RANKL (100 ng/ml) for 5 days to induce TRAP-positive multinucleated osteoclasts, and then the cells were stained for TRAP. TRAPpositive multinucleated cells containing more than three nuclei were counted as osteoclasts under microscopic examination.
Micro-RNA Northern Blotting-Total RNA was isolated from infected cells using the mirVana miRNA Isolation Kit (Ambion) that was used according to the manufacturer's protocol. Micro-RNA Northern blot was performed using Micro-RNA Northern blot assay kit (Signosis) according to the manufacturers instructions. Biotin prelabeled micro-RNA-223 and sno234 probe were purchased from Signosis.

Micro-RNA Regulator Gene Silencing in Osteoclast Precursors-
Plasmids containing siRNA sequences to DGCR8, Dicer1, or Ago2 were transduced by retroviruses in osteoclast precursors. We measured mRNA and protein expression levels by real time RT-PCR and Western blotting. Real time RT-PCR demonstrated that siRNA to DGCR8, Dicer1, or Ago2 decreased miRNAs Control Osteoclast Differentiation and Function FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 mRNA expression levels by 66 -76% compared with scrambled siRNA at 48 h after infection (Fig. 1, A-C). The levels of the respective proteins were inhibited without effects on levels of the other factors (Fig. 1, D-F). Next, we confirmed that the retroviral vectors to DGCR8, Dicer1, and Ago2 were functional and inhibited miRNAs present during osteoclast differentiation. We have previously reported that miRNA-223 is expressed in RAW264.7 cells, mouse osteoclast precursor cell lines (24). In addition, miRNA microarray (data not shown) showed that miRNA-223 is expressed in osteoclast precursors, and miRNA-155 was barely detectable. Using miRNA RT-PCR and real time RT-PCR with primers specific to miRNA-223 and miRNA-155, we showed that miRNA-223 was expressed in osteoclast precursors in the experiments reported here and that miRNA-155 was barely detectable (Fig. 1G). miRNA RT-PCR and real time RT-PCR analysis showed that DGCR8, Dicer1, and Ago2 gene silencing by siRNA significantly suppressed expression of miRNA-223 (Fig. 1, H and I), suggesting that these siRNA vectors work in osteoclast precursors. The effect of Ago2 was surprising because it is generally believed that Ago2 proteins bind miRNAs and mediate downstream gene silencing (11). However, consistent with our data, miRNA biogenesis in hematopoietic cells from precursor-miRNAs was impaired in Ago2-deficient mouse (31). This and our data indicate that the Slicer endonuclease activity of Ago2 may be dispensable for hematopoiesis including osteoclast precursors. Our data show that DGCR8, Dicer1, and Ago2 are important in miRNA biogenesis in osteoclast precursors.
miRNAs Control Expression and Function of Transcription Factors Required in Osteoclastogenesis-It is not known whether miRNAs play a critical role in osteoclast differentia- , or Ago2 (F) protein expression levels by siRNA gene silencing was analyzed by immunoblotting (IB) using nuclear or whole cell extracts. Quantitative image analysis of these protein expression levels was normalized to GAPDH (% scr). scr, scrambled. miRNA RT-PCR (G and H) and quantitative real time RT-PCR (I) were performed to measure miRNA-223 and miRNA-155 expression levels. The PCR products were confirmed using a 20% polyacrylamide gel, with U6 small nuclear RNA as a loading control (G and H). PCR products were normalized to U6 level for each reaction (I). The data represent the means Ϯ S.D. of three experiments in triplicate.

miRNAs Control Osteoclast Differentiation and Function
tion and function. To elucidate the role of miRNAs in osteoclastogenesis, we utilized RANKL-induced differentiation of osteoclast precursors harboring reduced levels of DGCR8, Dicer1 or Ago2 protein. Analysis of expression of the transcription factors important for osteoclastogenesis such as PU.1, MITF, c-Fos, and NFATc1 (1, 2) revealed increased levels by day 3 during RANKL-stimulated osteoclastogenesis in cells expressing scrambled siRNA (Fig. 2). In DGCR8, Dicer1, or Ago2 siRNA-expressing cells, in contrast, the expression levels of these proteins by Western blotting did not reach normal levels during differentiation with RANKL stimulation, although there was slight up-regulation by day 3 (Fig. 2).
The transcription factors required in osteoclastogenesis shown in Fig. 2 are recruited to a variety of osteoclast-specific marker gene promoters such as TRAP, MMP-9, Ctsk, CTR, and integrin ␤ 3 and NFATc1 itself for the formation and activity of multinucleated osteoclasts (33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43). We examined whether PU.1, MITF, c-Fos, or NFATc1 are recruited to the NFATc1 or TRAP promoter. ChIP analysis showed that NFATc1 and c-Fos were recruited to the NFATc1 promoter by 24 and 48 h after RANKL stimulation in scrambled siRNA-expressing cells (Fig.  3, A and B). NFATc1 was also recruited to the TRAP promoter by 24 and 48 h after RANKL stimulation in scrambled siRNAexpressing cells (Fig. 3C). Low levels of c-Fos were already bound to the TRAP promoter in osteoclast precursors without RANKL stimulation (0 h), and increased binding was observed at 24 and 48 h in scrambled siRNAexpressing cells (Fig. 3D). PU.1 and MITF were more strongly recruited to the TRAP promoter in the same binding pattern as c-Fos in scrambled siRNA-expressing cells (Fig. 3,  E and F). The binding of these factors to the NFATc1 or TRAP promoter was strikingly attenuated by a loss of DGCR8, Dicer1, or Ago2 protein (Fig. 3).
Consistent with the results of the ChIP analysis in Fig. 3, the mRNA expression levels of the osteoclastspecific markers were markedly inhibited during osteoclast differentiation with DGCR8, Dicer1, or Ago2 siRNA gene silencing (Fig. 4). In particular, Ctsk, CTR, and integrin ␤ 3 mRNA expression levels had been suppressed by day 1 after RANKL stimulation in DGCR8, Dicer1, or Ago2 siRNA-expressing cells compared with scrambled siRNA-expressing cells (Fig. 4, C-E). These data of Chip assay and real time RT-PCR analysis can be attributed to inhibition of expression of the transcription factor proteins important for osteoclastogenesis by loss of DGCR8, Dicer1, or Ago2 protein during osteoclast differentiation as shown by Fig. 2.
Inhibition of Osteoclast Differentiation and Function-Consistent with data of Figs. 2-4, DGCR8, Dicer1, or Ago2 siRNA gene silencing inhibited TRAP-positive multinucleated osteoclast number and bone resorbing activity compared with scrambled siRNA-expressing cells (Fig. 5). These data provide direct evidence that the miRNA pathway for the generation of miRNAs play critical roles in osteoclast differentiation and function in vitro.
Dicer-null Mice Show Mild Osteopetrosis-We generated Dicer wt/wt and Dicer f/f mice using CD11b ϩ -cre mice to examine whether micro-RNAs are involved in differentiation and function of osteoclasts in vivo as well as in vitro. CD11b is expressed during osteoclast differentiation from mononucleated early progenitor cells to mature multinucleated osteoclasts (30). First of all, Dicer protein levels were reduced as shown by Western blot analysis of Dicer-null osteoclast precursors (Fig.  6A). Next, to confirm whether a functional Dicer KO is generated, Dicer-null osteoclast precursors were analyzed for their ability to produce mature miRNA-223 by miRNA real time RT-PCR assay. miRNA-223 was detected in Dicer wt/w osteoclast precursors, but its expression levels were decreased by 58% in Dicer-null osteoclast precursors (data not shown). We next studied the skeletal tissues of Dicer wt/wt and Dicer f/f mice. Histological examination using TRAP stain revealed reduced numbers of TRAP-positive osteoclasts in growth plate and cancel-   (Fig. 6C). Osteoblast number (Ob.N/BS) and osteoblast surface (Ob.S/BS) were not different between Dicer wt/wt and Dicer f/f mice (Fig. 6C). These data provide evidence that the impaired miRNA pathway causes aberrant osteoclast differentiation and function in vivo as well as in vitro.    generation of miRNAs is important for osteoclast differentiation and function in vitro and in vivo. However, the mechanism by which miRNAs control osteoclastogenesis and bone resorbing activity remains unknown. Our miRNA microarray showed that miRNA-223 expression is sustained during osteoclast differentiation (data not shown). Therefore, we pursued the function of miRNA-223 that is needed in osteoclastogenesis. A bioinformatic search indicated a putative target site for miRNA-223 in the NFI-A 3Ј-untranslated region that is highly conserved among mammals. Recently, it was reported that NFI-A negatively controls expression of the M-CSFR by unknown mechanisms in the NB4 cell line, derived from the bone marrow of a patient with acute promyelocytic leukemia (44). M-CSFR is expressed in osteoclast precursors through osteoclasts and is critical for differentiation, function, and survival of osteoclasts (1,2). Consistent with these reports, NFI-A was up-regulated, and M-CSFR was down-regulated in DGCR8, Dicer1, or Ago siRNA-expressing cells and Dicer-null osteoclast precursors (Fig. 7, A-D). On the other hand, NFI-A expression was not detectable in scrambled siRNA-expressing cells and Dicer wt/wt osteoclast precursors (Fig. 7, A-D) that expressed miRNA-223 (Fig. 7, G-I). We next studied whether down-regulation of NFI-A expression is needed in osteoclasto-genesis and bone resorbing activity. Overexpression of NFI-A in osteoclast precursors significantly suppressed TRAP-positive osteoclast formation compared with control cells (58% down) (Fig. 7, E and F) and bone resorption (data not shown). Expression of M-CSFR and PU.1 were down-regulated in NFI-A-overexpressing cells (Fig. 7G). In addition, pri-miRNA-223 and mature miRNA-223 expression levels were also decreased in this study (data not shown). However, ChIP analysis failed to detect the mechanism of a negative effect of NFI-A on M-CSFR expression, even though two putative NFI-A-binding sites are presented on the M-CSFR promoter (data not shown).

NFI-A Down-regulation by miRNA-223 Is Implicated in
We next examined whether overexpression of M-CSFR rescues osteoclast formation and function in Dicer-null osteoclast precursors in response to M-CSF and RANKL. Dicer-null osteoclast precursors prepared from the Dicer f/f mice had impaired osteoclastogenesis compared with control cells (Fig. 7, H and I).
Forced expression of M-CSFR in Dicer-null osteoclast precursors significantly rescued TRAP-positive osteoclast formation compared with that of Dicer-null osteoclast precursors (Fig. 7,  H and I). Moreover, we found that PU.1 expression is up-regulated by M-CSFR overexpression in rescued cells (Fig. 7J), even though its expression is impaired the miRNA pathway is defective as shown Fig. 2 FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 4673
levels were down-regulated, although pri-miRNA-223 expression levels were up-regulated in this study (data not shown). In addition, NFI-A expression levels were not altered (Fig. 7J). These results demonstrate that the action of miRNA-223 to suppress NFI-A levels in osteoclast precursors is a prerequisite for induction of M-CSFR expression and osteoclastogenesis.
To further investigate whether miRNA-223 is essential for osteoclastogenesis, we performed osteoclast formation assays using antisense miRNA-223 oligonucleotides. Inhibition of miRNA-223 induced down-regulation of TRAPpositive osteoclast formation compared with control in RAW264.7 cells (Fig. 8, A and B). We previously reported that TRAP-positive osteoclast formation was not affected in pre-miRNA-223 siRNA-expressing RAW264.7 cells in which the miRNA-223 was decreased by 66% (24). There were no alterations of NFI-A expression level by this level of miRNA-223 decrease (data not shown). However, in the present studies, the miRNA-223 decrease was 92.0% (data not shown), and NFI-A levels were up-regulated (Fig. 8C). In addition, unlike osteoclast precursors wherein NFI-A is not expressed, NFI-A is moderately expressed in RAW264.7 cells (Fig. 8C). Thereby, we reasoned that the direct suppression of mature miRNAs by antisense oligonucleotides may be able to inhibit more effectively than that of pre-miRNA-223 by siRNA including the target sequence of stem-loop structure. We next analyzed NFI-A and M-CSFR expression levels in RAW264.7 cells with antisense miRNA-223 oligonucleotides. Western blot analysis showed up-regulation of NFI-A levels and a negative effect of NFI-A on M-CSFR levels (Fig. 8C), suggesting that miRNA-223 expression is critically important for osteoclastogenesis through M-CSFR expression. Moreover, we found that overexpression of PU.1

miRNAs Control Osteoclast Differentiation and Function
in osteoclast precursors induced expression of miRNA-223 compared with control cells (Fig. 8D). These data are in agreement with the report of two putative PU.1-binding sites present in the mouse miRNA-223 promoter (20).

DISCUSSION
In this study, we used DGCR8, Dicer1, and Ago2 siRNA gene silencing and Dicer-null mice to clarify the importance of the miRNA pathway for generation of miRNAs in osteoclast differentiation and function in vitro and in vivo. The first discovery was that loss of Ago2 protein as well as loss of DGCR8 or Dicer1 proteins attenuated expression levels of miRNAs in osteoclast precursors (Fig. 1, H and I). Ago2 is known to be important in regulation of gene silencing (11). Our data are in agreement with a recent report using Ago2 Ϫ/Ϫ hematopoietic bone marrow cells concluding that the Slicer endonuclease activity is dispensable for hematopoiesis and that Ago2 is an important regulator of miRNA homeostasis (31). Ago2 could be very specialized member of the Argonaute family with an important function within RISC in the regulation of miRNA homeostasis.
PU.1, MITF, c-Fos, and NFATc1 are crucial transcription factors in osteoclastogenesis (1,2), and these transcription factors bind to the TRAP, MMP9, Ctsk, CTR, or integrin ␤ 3 pro-moters for expression of osteoclast-specific markers in activated osteoclasts (33)(34)(35)(36)(37)(38)(39)(40)(41). Expression of these molecular markers is essential for osteoclast function such as bone resorption to maintain bone metabolism (1,2). However, this sequence of events was extremely reduced by impaired miRNA homeostasis. The question became why the levels of these transcription factor proteins were not increased by loss of DGCR8, Dicer1 or Ago2 protein (Fig. 2). We demonstrated that one mechanism by which this occurs is PU.1 down-regulation by the negative control of NFI-A on M-CSFR levels (Fig. 7) because M-CSF induces PU.1 expression through M-CSFR in the hematopoietic myeloid-osteoclast lineage (1-3). As has been mentioned, we were not able to detect NFI-A binding on the M-CSFR promoter by ChIP assay, even though two putative binding sites are present (data not shown), suggesting that this effect may be indirect.
In studies in vitro, we demonstrated that the impaired miRNA pathway causes aberrant osteoclast differentiation and function caused by inhibition of expression and activity of transcription factors required in osteoclastogenesis and consequently reduced osteoclast-specific gene expression (Figs. 2-5). Moreover, RANK expression was also down-regulated in DGCR8, Dicer1, or Ago siRNA-expressing cells and Dicer-null  FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 osteoclast precursors (data not shown). The RANKL-RANK pathway is pivotal in osteoclast differentiation, activity, and survival (1)(2)(3). RANK is the receptor for RANKL (1,2), and RANK is expressed in osteoclast precursors and mature osteoclasts, where it can be up-regulated by PU.1 and M-CSF stimulation (45,46). Mice lacking RANK are osteopetrotic (47), demonstrating that RANK is essential for osteoclast differentiation. This could be one of the major causes of aberrant osteoclast differentiation by impaired miRNA homeostasis. These data strongly suggest that miRNAs play a critical role in osteoclast differentiation and function in vitro. Moreover, using the cre-loxP system, we showed that CD11b ϩ -cre/Dicer f/f mice have mild osteopetrosis caused by reduced osteoclast numbers and bone resorbing activity without affecting osteoblast number or activity (Fig. 6). This suggests that the miRNAs are involved in bone metabolism, at least a part, by controlling differentiation and function of osteoclasts in vivo. Unexpectedly, miRNA real time RT-PCR analysis revealed the moderate expression levels, not low expression levels, of miRNA-223 in Dicer-null osteoclast precursors (data not shown) as well as Dicer1 siRNA gene silencing (Fig. 1I). Recent studies have demonstrated that overexpression of each Ago protein (Ago1-Ago4) increases the abundance of mature miRNAs compared with Dicer overexpression (48). In addition, Ago2 is a key regulator of miRNAs biogenesis rather that the Slicer endonuclease activity (31), suggesting that Ago2 would be a more important factor than Dicer for miRNA homeostasis in vivo. Moreover, it has been reported that RISC including Dicer and Ago2 more efficiently cleaves target mRNAs by using pre-miRNAs than the duplex miRNAs that do not have the stem-loop structure (49). For these reasons, we will need to generate Ago2 or DGCR8 osteoclast precursor-specific null mice to further investigate miRNA function and bone metabolism in vivo in a setting with lower levels of the miRNAs including miRNA-223 and perhaps more severe osteopetrosis.

miRNAs Control Osteoclast Differentiation and Function
We have shown previously that miRNA-223 expression is important for osteoclastogenesis. However, we were not able to provide detailed molecular mechanisms. First of all, we showed that osteoclast differentiation of RAW264.7 cells was significantly suppressed by antisense to miRNA-223 (Fig. 8, A and B). Moreover, this antisense miRNA markedly suppressed osteoclast bone resorbing activity (data not shown), indicating that miRNA-223 may be important for regulation of differentiation and function of osteoclasts in vitro. However, overexpression of pre-miRNA-223 also suppressed TRAP-positive osteoclast formation (24), suggesting that appropriate expression levels of miRNA-223 would be needed in osteoclastogenesis. Although all the conserved miRNA-223 target mRNAs for osteoclastogenesis or its function have not been identified, we found that NFI-A suppression by miRNA-223 is important (Fig. 8, A-C). This suggests that expression of miRNA-223 is essential for osteoclast differentiation and function by overcoming a constitutive inhibition induced by NFI-A. In addition, we found that PU.1 induces miRNA-223 expression in osteoclast precursors (Fig. 8D). This appears to be important for down-regulation of NFI-A levels by miRNA-223. In fact, PU.1 levels are up-regulated during osteoclastogenesis by M-CSF and RANKL (1,2). In summary, we propose a mechanism of miRNA-223 action associated with osteoclast differentiation and function (Fig. 9). In osteoclast precursors, PU.1 induced by M-CSF stimulates production of pri-miRNA-223 and RANK. Importantly, expression of miRNA-155, a target of PU.1 (50), is not induced by M-CSF and RANKL stimulation during osteoclastogenesis. 3 Next, pri-miRNA223 is processed into mature miRNA-223 from pre-miRNA-223 by RNase III enzymes including Dicer, and miRNA-223 down-regulates the NFI-A levels required for up-regulating of M-CSFR levels in cells, which in turn feed forward to increase expression of transcription factors such as PU.1, MITF, and c-Fos. These transcription factors are induced by M-CSF and RANKL through the up-regulated M-CSFR and RANK, and consequently cells differentiate into activated osteoclasts by up-regulated osteoclast-specific markers.
Our results and those of others in aggregate indicate that the functions of DGCR8, Dicer, and Ago2 are necessary to maintain bone metabolism, including the differentiation and function of osteoclasts. Understanding the role of post-transcriptional gene regulation by miRNAs and identification of target mRNAs negatively regulated by miRNAs in normal bone biology is critical to understanding bone metabolism disorders. Identification of osteoclast-specific miRNAs may provide potential tar-gets for drug development in bone metabolic disorders with abnormal osteoclast activity.  FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7