Bone Morphogenetic Protein-2 Stimulates Runx2 Acetylation*

Runx2/Cbfa1/Pebp2aA is a global regulator of osteogenesis and is crucial for regulating the expression of bone-specific genes. Runx2 is a major target of the bone morphogenetic protein (BMP) pathway. Genetic analysis has revealed that Runx2 is degraded through a Smurf-mediated ubiquitination pathway, and its activity is inhibited by HDAC4. Here, we demonstrate the molecular link between Smurf, HDACs and Runx2, in BMP signaling. BMP-2 signaling stimulates p300-mediated Runx2 acetylation, increasing transactivation activity and inhibiting Smurf1-mediated degradation of Runx2. HDAC4 and HDAC5 dea-cetylate Runx2, allowing the protein to undergo Smurf-mediated degradation. Inhibition of HDAC increases Runx2 acetylation, and potentiates BMP-2-stimulated osteoblast differentiation and increases bone formation. These results demonstrate that the level of Runx2 is controlled by a dynamic equilibrium of acetylation, deacetylation, and ubiquitination. These findings have important medical implications because BMPs and Runx2 are of tremendous interest with regard to the development of therapeutic agents against bone diseases.

Runx2/Cbfa1/Pebp2␣A is a transcription factor belonging to the Runx family, which contains three members: Runx1, Runx2, and Runx3 (1). The importance of Runx proteins in mammalian development has been clearly demonstrated by gene disruption studies (2). Mouse embryos with homozygous mutations in Runx1 showed normal morphogenesis and yolk sac-derived primitive hematopoiesis, they lacked fetal liver hematopoiesis, indicating that Runx1 is required for definitive hematopoiesis (3). Runx3 functions as a tumor suppressor of gastric cancer. Runx3-null gastric mucosa develops hyperplasia because of the activation of proliferation and suppression of apoptosis in epithelial cells. Runx3-null gastric epithelial cells are less sensitive to the growthinhibiting and apoptosis-inducing activities of transforming growth factor-␤ (TGF-␤). 2 Thus, Runx3 is a major growth regulator of gastric epithelial cells. Previously, we showed that the extent of the Runx3 acetylation is up-regulated by the TGF-␤ signaling pathway and downregulated by histone deacetylase (HDAC) activities (4).
Runx2 is a key regulator of chondroblast and osteoblast differentiation, and of bone development in vivo (5)(6)(7). Targeted disruption of this Runx2 in mice resulted in a complete lack of ossification due to the maturational arrest of osteoblasts (8 -10). Runx2 regulates the expression of major extracellular matrix genes (including alkaline phosphatase, osteopontin, osteocalcin, type I collagen, and type X collagen) expressed by chondroblasts and osteoblasts (5,11). Inactivation of one allele of the RUNX2 gene was identified as the cause of human cleidocranial dysplasia (CCD), an autosomal dominant bone disorder (12,13). On the other hand, constitutively active mutation of FGFR1 results in increased Runx2 expression and the premature suture closure that is the hallmark of human craniosynostosis (14).
Expression and activation of Runx2 is regulated by many bone-derived growth factors, including bone morphogenetic proteins (BMPs) (5,15). BMPs form a unique group of proteins within the TGF-␤ superfamily of genes, and play pivotal roles in the regulation of heart, neural, cartilage, and bone development (16,17). The Smad family of proteins has been identified as the downstream propagators of BMP signals (18). BMP-activated Smads (Smad1, -5, and -8) induce Runx2 gene expression (15), and Smads interact physically with the Runx2 protein to induce osteoblast differentiation (15,19,20).
The p300 protein functions primarily as a co-activator of transcription for a number of nuclear proteins including Runx family members (21)(22)(23)(24) and also functions as a histone acetyl transferase (HAT) (25). Acetylation of specific lysine residues on histone tails is believed to neutralize the positive charge, and to generate a more accessible chromatin structure for transcription factors (26). p300 is also capable of acetylating a number of non-histone proteins (27). For example, E2F1 is acetylated by p300, and the acetylation is reversed by histone deacetylase-1 (HDAC1), indicating that reversible acetylation is also a mechanism for regulation of non-histone proteins (28). The increase in half-life through acetylation appears to protect the protein from ubiquitination, since acetylation and ubiquitination occur only at lysine residues (29,30).
Protein ubiquitination at lysine residues plays a crucial role in defining substrate specificity and subsequent protein degradation by protea-some. Runx2 is degraded through a ubiquitin/proteasome-mediated pathway (31,32). Recently, Smurf1 (Smad ubiquitin regulatory factor 1) was identified as the E3 ubiquitin ligase responsible for ubiquitin-dependent Runx2 degradation (4,33). Overexpression of Smurf1 in osteoblast precursor cells inhibits BMP signaling and osteoblast differentiation (34). Consistent with these findings, targeted inactivation of Smurf1 resulted in enhanced bone formation (35).
In contrast, HDAC4-null mice displayed premature ossification of developing bones, mimicking the phenotype that results from constitutive Runx2 expression. Additionally, a remarkable similarity was observed between Runx2-null mutants and the HDAC4 gain-of-function phenotypes (8, 9, 36 -38). A recent study demonstrated that HDAC4 or-5 is required to inhibit Runx2 function (39), suggesting that HDAC4 is a physiological repressor of Runx2.
Although these results clearly demonstrate that Smurf and HDAC play opposing roles in BMP signaling of Runx2 regulation, the underlying molecular links have not been clearly defined. In this study, we investigated the molecular mechanism that mediates HDAC and Smurf1 effects on Runx2. Our finding that BMP-2 induces Runx2 acetylation furnishes for the first time an important molecular link between four factors involved in osteogenesis, BMP-2, Runx2, HDAC, and Smurf1, and provides one mechanism that assembles a great deal of previously unconnected data into one coherent model.

EXPERIMENTAL PROCEDURES
Cell Culture-All cell culture media and antibiotics were from Invitrogen (Carlsbad, CA). C2C12 and 293 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, antibiotics, and anti-mycotics, at 37°C, in 5% CO 2 . Runx2 Ϫ/Ϫ cell line (H1-127-21-2) was established from calvaria-like tissue of Runx2 knock-out mouse and maintained in ␣-MEM containing 10% fetal bovine serum, penicillin G (100 units/ml), and streptomycin (100 g/ml). C2C12 cells were used for the analysis, which required intact BMP signaling pathway. Runx2 Ϫ/Ϫ cell line was used when discrimination of Runx2-specific response from nonspecific response was required. 293 cells were used when high transfection efficiency was required and intact BMP signaling pathway was not required.
Plasmids and Antibodies-The Runx2 full-length type I isoform with MRIPV N-terminal sequence (Q08775-3) and its deletion mutants were tagged with Myc or HA, in a CMV promoter-derived mammalian expression vector (pCS4 -3Myc and pCS4 -3HA, respectively). Mutations affecting the Runx2 acetylation site were introduced by PCR and cloned into the pCS4 -3Myc vectors. The BMP receptor 1 (BMPR1) and HA-p300 expression vectors were obtained from Dr. K. Miyazono (Tokyo University) and Dr. M. Ewen (Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA), respectively. The expression plasmids for the Myc-HDAC series were cloned into pCS4 -3Myc. The FIGURE 1. BMP-2 increases Runx2 protein levels by a transcription-independent mechanism. A, in C2C12 cells, Myc-tagged Runx2 was expressed in the absence (Ϫ) or presence (ϩ) of BMP-2 (300 ng/ml). Levels of Runx2 were determined by immunoblotting (IB) using an anti-Myc antibody. B, in C2C12 cells, HA-tagged Runx2 was co-expressed with increasing concentrations of BMPR1. Levels of Runx2 and BMPR1 were determined by immunoblotting (IB) using anti-HA and anti-FLAG antibodies, respectively. C, following transfection with Myc-tagged Runx2 in the presence (Rx2ϩBMPRI) or absence (Rx2) of BMPRI expression, 293 cells were treated with cycloheximide (CHX, 40 g/ml) and cultured for the indicated times. Levels of Runx2 were determined by immunoblotting (IB) using an anti-Myc antibody. D, band intensities of Runx2 protein in C were quantitated and plotted against time to determine the half-life of Runx2 proteins in the presence (Rx2ϩBMPRI) or absence (Rx2) of BMPRI expression. E, C2C12 cells were transfected with p300, or treated with BMP-2, and the mRNA levels of Runx2 and GAPDH were determined using semi-quantitative RT-PCR. CMV promoter-driven exogenous Runx2 mRNA and endogenous Runx2 mRNA were amplified separately. GAPDH was amplified as a control. Following RT-PCR, PCR products were collected from various reaction cycles (from [25][26][27][28][29][30], and fragments were separated using agarose gel electrophoresis. Endogenous Runx2 mRNA was amplified using forward (5Ј-GCTGTTAGGCCTCGTATTCCTGTA-3Ј) and reverse (5Ј-CCGCTCGAGTCAATATGGC-CGCCA-3Ј) primers. The CMV promoter-driven exogenous Runx2 mRNA was amplified with a forward primer designed against the vector sequence (5Ј-AGCAAGCTTGATTTAGGT-GACACT-3Ј) and a reverse primer obtained from the Runx2 coding sequence (5Ј-CCGCTCGAGTCTTGGTTCCCGGGGACC-3Ј). The number of cycles performed for each PCR reaction is indicated. F, band intensities of endogenous Runx2 in the bottom of E were quantitated and plotted against the number of PCR cycles.
DNA Transfections and Reporter Assay-Transient transfections were performed using the calcium phosphate method for 293 cells or Lipofectamine Plus reagent (Invitrogen) for the Runx2 Ϫ/Ϫ cell line. Luciferase assays were performed using the Luciferase Reporter Assay kit (Promega), according to the manufacturer's recommendations. The pCMV ␤-gal plasmid was included as an internal control to determine the efficiency of transfection.
Immunoprecipitation and Immunoblotting-Following transfection, the C2C12 and 293 cells were lysed in ice-cold MLB buffer (25 mM HEPES (pH 7.5), 150 mM NaCl, 1% Nonidet P-40, 0.25% sodium deoxycholate, 10% glycerol, 25 mM NaF, 1 mM EDTA, 1 mM Na 3 VO 4 , 250 M phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin) and cleared by centrifugation. The resulting supernatants were immunoprecipitated with the appropriate primary and secondary antibodies, and protein A-or protein G-Sepharose beads. All incubations were conducted at 4°C. The Sepharose beads were washed extensively and bound proteins were separated using SDS-PAGE, and then transferred to polyvinylidene difluoride membranes. The membranes were incubated with the appropriate antibodies, washed, and then incubated with horseradish peroxidase-coupled secondary antibodies. After washing, the reactive proteins were visualized with an enhanced chemiluminescence (ECL) reagent (Amersham Biosciences). For HDAC inhibitor treatments, both C2C12 and 293 cells were cultured in fresh medium for 1 day following transfection, and then treated with the inhibitors (Tri- nM Trichostatin A. The cells were lysed in MLB buffer, and the protein extract was incubated with anti-Myc antibody. The antibody complex was then precipitated with protein G-Sepharose. Precipitates were washed four times with MLB buffer, separated using SDS-PAGE, fixed, enhanced with Amplify (Amersham Biosciences), and exposed to x-ray film at Ϫ70°C for 10 days.
Small Interfering RNA (siRNA) Transfection-One day prior to transfection, 293 cells (3 ϫ 10 5 cells per well) were plated in 6-well plates and grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum in the absence of antibiotics. Cells were transfected with pCS4 -3Myc-Runx2. Five hours post-transfection, a second transfection was performed with 20 M Smad1 siRNA and Smad5 siRNA (Invitrogen), HDAC4 siRNA, and HDAC5 siRNA (Santa Cruz Biotech-FIGURE 2. BMP-2-mediated Runx2 stabilization is correlated with Runx2 acetylation. A, Myctagged Runx2 was expressed in C2C12 cells with BMP-2, p300, or both proteins, as indicated. Levels of acetylated Runx2 were determined by immunoprecipitation (IP) using an anti-acetyl-lysine antibody (Ac-Lys), followed by immunoblotting (IB) with an anti-Myc antibody. Runx2 (Rx2) and p300 levels were determined by immunoblotting using anti-Myc and anti-HA antibodies, respectively. B, Myc-tagged Runx2 was expressed in 293 cells with BMPR1, p300, or both proteins, as indicated. Levels of acetylated Runx2, Runx2 (Rx2), and p300 were determined as described above. C, Myc-tagged Runx2 was expressed in 293 cells with BMPR1, p300, or both proteins, as indicated. Levels of acetylated Runx2 were determined by immunoprecipitation (IP) using an anti-Myc antibody, followed by immunoblotting (IB) with an anti-acetyl-lysine antibody. Runx2 (Rx2), p300, and BMPR1 were detected by immunoblotting using anti-Myc and anti-HA antibodies, respectively. D, Myc-tagged Runx2 was co-expressed with p300 in 293 cells and metabolically labeled with [ 3 H]acetate in the presence or absence of BMP-2. Upper panel, acetylated Runx2 was immunoprecipitated (IP) with an anti-Myc antibody and incorporation of [ 3 H]acetate was detected by fluorography. Lower panel, levels of Runx2 were determined by immunoblotting (IB) using an anti-Myc antibody. E, C2C12 cells were transfected with vector alone or the HA-tagged p300 expression plasmid. Acetylation of endogenous Runx2 was determined by immunoprecipitation (IP) using an anti-acetyl-lysine antibody followed by immunoblotting with an anti-Runx2 antibody. Levels of endogenous Runx2 (Rx2), p300, and tubulin were determined by immunoblotting. nology). pCS4 -3Myc-Runx2 and siRNA were transfected using Lipofectamine and Lipofectamine2000 (Invitrogen), respectively. A commercially available nonspecific random siRNA was used as a control. Cells were harvested 48 h after the second transfection, and subjected to immunoblot analysis and RT-PCR.
Mouse Calvarial Organ Culture-Neonatal calvarial bones were obtained from 4-day-old pups of ICR mice and analysis of new bone formation was performed as described by Ref. 45. The amount of new bone formation was assessed morphologically using the Bioquant Nova Prime software (Nashville, TN).

BMP-2 Increases Runx2 Protein Levels via a Transcription-independent Mechanism-
The pluripotent mesenchymal precursor cell line C2C12 provides a model system to study the early stage of osteoblast differentiation during bone formation. In this model, BMP-2 inhibits the differentiation of C2C12 into multinucleated myotubes and induces osteoblast differentiation (46). Runx2 is a major target of the BMP-2 signaling pathway at the transcriptional level (15). To address whether BMP-2 regulates Runx2 post-translationally as well, C2C12 cells were transfected with Myc-tagged Runx2, in the presence or absence of BMP-2, and the protein levels were compared by immunoblotting. The level of exogenously expressed Runx2 was much higher in BMP-2-treated cells (Fig. 1A). This result was confirmed by coexpression of HA-tagged Runx2 and increasing amounts of the BMP receptor 1 (BMPR1). Runx2 levels increased in proportion to the abundance of BMPR1 (Fig. 1B). The half-life of Runx2 was ϳ5.3 h in the absence of BMPR1 and extended to 50.2 h in the presence of BMPR1 (Fig. 1, C and D). To eliminate the possibility of BMP-2 mediated transcriptional activation of the transfected gene, we measured mRNA levels of exogenous and endogenous Runx2, using semi-quantitative RT-PCR. Although BMP-2 increased the level of endogenous Runx2 mRNA, the exogenous level was not affected (Fig. 1,  E and F). These results suggest that, in response to BMP-2 signaling, the accumulation of exogenous Runx2 is independent of the increase in transcription, but is associated with increased stability.
BMP-2-mediated Runx2 Acetylation and Stabilization-Because Runx2 and Runx3 are degraded via the ubiquitin-proteasome-mediated pathway (4,31,32), and acetylation can protect proteins from ubiquitination (4, 29), we examined whether BMP-2 induces Runx2 acetylation, thereby stabilizing the Runx2 protein. Immunoblot analysis of BMP-2transfected cells revealed that Runx2 is acetylated by BMP-2 treatment ( Fig. 2A). Because p300 is associated with TGF-␤ superfamily signaling pathways, and Runx1 and Runx3 are targets for p300 acetyltransferase activity (4,24,30), we examined whether p300 is associated with BMP-2-mediated Runx2 acetylation. p300-acetylated Runx2, and p300-mediated Runx2 acetylation increased in the presence of BMP-2 ( Fig. 2A, top  panel). Notably, the total amount of exogenous Runx2 was proportional to acetylation levels ( Fig. 2A, third panel). Similarly, overexpression of BMPR1 alone, or in combination with p300, resulted in an increase in FIGURE 3. Smad is required for p300-mediated Runx2 acetylation. A, Myc-tagged Runx2 and various Smads were coexpressed in 293 cells, as indicated. Levels of acetylated Runx2, Runx2 (Rx2), and Smads were determine as described above. B, 293 cells were transfected with siRNAs against Smad1 or Smad5 and the mRNA levels of endogenous Smad1 and Smad5 were measured by RT-PCR. C, band intensities of endogenous Smad1 and Smad5 in B were quantitated and plotted. D, Myc-tagged Runx2 was co-expressed with Smad1 (S1) and/or Smad5 (S5) siRNA in 293 cells. The level of acetylated Runx2 was determined as described above. E, Myc-tagged Runx2 was expressed in 293 cells with BMPRI, Smad5, or both proteins, as indicated. Levels of acetylated Runx2 were determined as described above. F, 293 cells were transfected with Runx2, p300, Smad1 (S1), or Smad5 (S5), or the indicated combinations of these expression plasmids, and the physical interaction between Runx2 and p300 was examined by immunoprecipitation (IP) followed by immunoblotting (IB). Levels of Runx2 (Rx2), Smads, and p300 protein were determined by immunoblotting.

BMP-2 Stimulates Runx2 Acetylation
Runx2 acetylation and an accumulation of the protein (Fig. 2, B and C). Transcription of exogenous Runx2 was unchanged by p300 and/or BMP-2 (Fig. 1E). Confirmation of Runx2 acetylation by p300 and BMP-2 was provided using 293 cells, in which Runx2 was metabolically labeled with [ 3 H]acetate. Acetylation of Runx2 increased when p300 was overexpressed and further increased in response to BMP-2 treatment (Fig. 2D).
To determine whether endogenous Runx2 is acetylated by p300, C2C12 cells were transfected with p300 expression plasmid and the acetylation status of Runx2 was assessed. Overexpression of p300 increased endogenous Runx2 acetylation and total Runx2 levels (Fig.  2E). These results demonstrate that the BMP-2 signaling pathway stimulates p300-mediated Runx2 acetylation, thereby promoting Runx2 accumulation.
Smad Is Required for p300-mediated Runx2 Acetylation-To understand how BMP-2 stimulates p300-mediated Runx2 acetylation, we transfected 293 cells with Smads and Runx2 expression plasmids and examined Runx2 acetylation. Smad2, -4, and -5 strongly increased, and Smad1 and -3 weakly increased, the level of Runx2 acetylation (Fig. 3A). Because Smad1 and Smad5 are known mediators of BMP-2 signaling, we examined the effect of Smad1 and Smad5 siRNA on p300-mediated Runx2 acetylation. siRNA knock-down of Smad1 and Smad5 was confirmed by RT-PCR (Fig. 3, B and C). The level of Runx2 acetylation was weakly decreased by Smad1 siRNA and strongly decreased by Smad5 siRNA. Combined Smad1 and Smad5 siRNA treatment further decreased Runx2 acetylation (Fig. 3D). Furthermore, coexpression of Smad5 and BMPRI synergistically increased Runx2 acetylation (Fig. 3E), suggesting that Smads mediate Runx2 acetylation in response to BMP-2. To determine the specific role of Smads in Runx2 acetylation, Smad1 and Smad5 were overexpressed in 293 cells. The physical interaction between Runx2 and p300 increased when Smad1 or Smad5 were overexpressed (Fig. 3F). Notably, Smad5 more effectively helped the interaction than Smad1. The correlation between the effect of Smad1 and Smad5 on Runx2 acetylation and the Runx2-p300 interaction suggests that BMP-2-activated Smads (mainly Smad5) stimulate p300-mediated Runx2 acetylation by facilitating the interaction between Runx2 and p300.
Acetylation Protects Runx2 from Smurf1-mediated Degradation-Runx2 contains ten lysine residues that are potential acetylation sites (Fig. 4A). To identify which residues are targets for acetylation, a series of Runx2 deletion mutants were co-expressed with p300, and cell lysates were immunoprecipitated using an anti-acetyl-lysine antibody. Acetylation was detected in the central and C-terminal regions of Runx2 (data not shown). To identify the acetylated residues, various Runx2 mutants with lysine-to-arginine (KR) substitutions were examined for p300-dependent acetylation. The Runx2-KR-225/230/350/351 mutant removed virtually all Runx2 acetylation by p300 (Fig. 4B). Using this mutant, we examined whether acetylation is responsible for stabilization of the protein. Runx2 had a half-life of ϳ5 h, but became very stable when coexpressed with p300 (half-life Ͼ 24 h) (Fig. 4C). In contrast, the Runx2-KR-225/230/350/351 mutant was extremely stable, even in the absence of p300 (Fig. 4C). These results indicate that the presence of unmodified lysine residues increases the susceptibility of the protein to degradation, and that acetylation of the lysine residues protects Runx2 from degradation.
Because Runx2 is degraded through the ubiquitin-mediated pathway and Smurf1 is the ubiquitin ligase for this pathway (33,34), we examined whether acetylation protects Runx2 from Smurf1-mediated degradation. Runx2 was rapidly degraded when Smurf1 is overexpressed (Fig.  4D). However, co-expression of p300 counteracted the Smurf1-mediated Runx2 degradation (Fig. 4D). In contrast, even in the absence of
Runx2 contains a PPXY motif that is required for interaction with the WW domain of Smurf1 (47) (Fig. 4A). To examine whether Smurf1 directly mediates Runx2 degradation, we constructed the deletion mutant Runx2-del-415-418, which lacks the PPXY motif, and the mutant Runx2-PR-415, which has a proline-to-arginine substitution at the critical amino acid 415. Both the Runx2-del-415-418 and Runx2-PR-415 mutants were extremely stable (half-life Ͼ24 h) (Fig. 4C), and were highly resistant to Smurf1-mediated degradation (Fig. 4D). These results are consistent with Smurf1-mediated degradation of Runx2. HDAC Deacetylates Runx2-It has been shown that Runx3 acetylation is down-regulated by HDACs (4), and Runx2 activity is suppressed by class II HDACs (39) and increased by targeted inactivation of HDAC4 (37). Therefore, we examined the effect of HDAC4 and -5 on Runx2. Co-immunoprecipitation and immunoblotting analysis revealed that HDAC4 and -5 physically interacted with Runx2, and decreased Runx2 protein levels (Fig. 5A). To eliminate the possibility of HDACmediated transcriptional inhibition of the transfected Runx2 gene, we measured mRNA levels of exogenous Runx2 and GAPDH, using semiquantitative RT-PCR. The level of exogenous Runx2 mRNA was not inhibited (Fig. 5B), suggesting that the HDAC-mediated decrease of Runx2 is independent of the decrease in transcription, but is associated with the decreased protein stability. Furthermore, HDAC4 and -5 effectively decreased p300-mediated Runx2 acetylation, and Runx2 protein abundance (Fig. 5C). Consistent with this observation, siRNA-mediated knock-down of HDAC4 or -5 resulted in an increase in Runx2 acetylation, and Runx2 abundance (Fig. 5, D-F). A similar result was observed when the HDACs were inhibited by Tricostatin A (TSA), CHAP27, and SCOP402 (Fig. 5G). These results demonstrate that Runx2 acetylation/ deacetylation is a dynamic process, and suggest that HDAC4 and -5 may be the effectors of Runx2 deacetylation.
Runx2 Acetylation Is Required for Its Transactivation Activity-To examine whether Runx2 acetylation affects its transactivation activity, the TGF-␤ responsive element-luciferase reporter plasmid (T␤RE-luc) and Runx2 expression plasmids were co-transfected into 293 cells. Transfected cells were treated with increasing concentrations of HDAC inhibitors (TSA, CHAP27, or SCOP402). The HDAC inhibitors increased the Runx2-dependent reporter activity further, in a dose-dependent manner (Fig. 6A). These findings indicate that the transactivation activity of Runx2 may be increased by inhibiting Runx2 deacetylation (i.e. resulting in a net increase in acetylated Runx2).
To investigate whether the increase in transactivation activity of Runx2 by acetylation is a result of increased protein abundance or of the acetyla-FIGURE 5. Runx2 is deacetylated by HDAC. A, physical interaction between Runx2 and class II HDACs. In 293 cells, HA-tagged Runx2 was coexpressed with or without Myc-tagged HDAC4 (H4) or HDAC5 (H5). The interaction between Runx2 and HDAC was analyzed by co-immunoprecipitation using an anti-HA antibody to precipitate Runx2. The abundance of Runx2 (Rx2) and HDACs was determined by immunoblotting using anti-HA and anti-Myc antibodies, respectively. B, in 293 cells, Myc-tagged Runx2 was coexpressed with HDAC4 or HDAC5 as indicated. The mRNA levels of CMV promoter-driven exogenous Runx2 and GAPDH were determined using semi-quantitative RT-PCR as described above. GAPDH was amplified as a control. Following quantitative RT-PCR, PCR products were collected from various reaction cycles (from [25][26][27][28][29][30], and fragments were separated using agarose gel electrophoresis. C, deacetylation of Runx2 by HDACs. In 293 cells, Myc-tagged Runx2 was co-expressed with HDAC4 (H4), HDAC5 (H5), and p300. The effect of p300 and HDACs on Runx2 acetylation and stability was analyzed by co-immunoprecipitation using an anti-acetylated-lysine antibody, and immunoblotting using an anti-Myc antibody to detect Runx2 and HDACs. Anti-tubulin was used to detect tubulin as a loading control. D, endogenous mRNA levels of HDAC4, HDAC5, Runx2, and GAPDH were measured by RT-PCR in HDAC4 or HDAC5 siRNA-treated cells. E, band intensities of endogenous HDAC4 and HDAC5 in D were quantitated and plotted. F, Myc-tagged Runx2 was coexpressed with siRNAs against HDAC4 (H4) or HDAC5 (H5) in 293 cells. The effect of HDAC4 and HDAC5 siRNA on Runx2 acetylation and protein levels were analyzed as described above. G, the effect of pharmacological inhibition of HDAC on Runx2. 293 cells were transfected with Myctagged Runx2 and treated with HDAC inhibitors (TSA, CHAP27, or SCOP402; 60 nM, 16 h). Runx2 acetylation and protein levels were analyzed as described above.
tion itself, the transactivation activities of Runx2 and Runx2-KR mutants were measured in the presence or absence of p300. Runx2 increased luciferase reporter activity by 4-fold, relative to controls, and co-transfection with p300 increased Runx2-dependent luciferase activity by 9-fold (Fig. 6B). Notably, mutation of the lysine residues targeted by p300 caused a significant decrease in Runx2 transactivation activity (Fig. 6B), although the stability of the mutant proteins was increased (Fig. 4, B and C). Furthermore, p300 failed to increase the activity of the mutant Runx2 proteins (Fig. 6B), although the KR mutants of Runx2 still interacted with p300 (Fig. 6C). These results suggest that Runx2 acetylation is required not only for stabilization, but also for transactivation activity.
HDAC Inhibitors Stimulate Osteoblast Differentiation in Vitro-Because Runx2 is a key regulator of osteoblast differentiation, and a major target of BMP-2 signaling, we investigated whether or not HDAC inhibitors, which increase Runx2 acetylation, could also induce osteoblast differentiation. Alkaline phosphatase (ALP) is an osteoblast-specific marker that is induced in the early stages of osteoblast differentiation (46). Treatment of C2C12 pluripotent mesenchymal precursor cells with each HDAC inhibitor alone caused a slight increase in ALP activity (data not shown). The increased levels of Runx2 potentiate BMP-2 activity in osteoblast differentiation (15). Based on this observation, C2C12 cells cultured with low concentrations of BMP-2 were treated with increasing concentrations of HDAC inhibitors. BMP-2 alone (30 ng/ml, one-tenth of the usual concentration) did not increase ALP activity. However, when combined with HDAC inhibitors, BMP-2 strongly increased ALP activity in a dose-dependent manner (Fig. 7A).
OC, another osteoblast-specific marker, is expressed at the osteoblast mineralization stage (48). To confirm that Runx2 acetylation could stimulate osteoblast differentiation, we examined the effect of HDAC inhibitors on the OC promoter and on the endogenous OC gene expression. To exclude the effects of HDAC inhibitors on endogenous Runx2, we transfected Runx2 expression plasmid and luciferase reporter plasmids in H1-127-1-2 cells, which were established from Runx2 Ϫ/Ϫ mouse calvaria, and treated the cells with the HDAC inhibitor, SCOP402. In the absence of Runx2, the HDAC inhibitor SCOP402 barely activated the OC promoter. Runx2 overexpression activated the OC promoter 3-fold in Runx2 Ϫ/Ϫ cells. In the presence of Runx2, HDAC inhibitor SCOP402 activated the OC promoter 4-fold over the activation observed for Runx2 alone (Fig. 7B). The disruption of a RUNX binding site (OSE2) in the OC promoter, completely abrogated OC promoter activation, either by HDAC inhibitor, or by Runx2 overexpression (Fig. 7B).
We examined whether or not HDAC inhibitors could affect Runx2dependent endogenous OC gene expression. In Runx2 Ϫ/Ϫ cells, basal OC expression was weakly detected by RT-PCR and the treatment of HDAC inhibitors weakly stimulated basal OC expression Transfection with Runx2 stimulated OC gene expression, and this was increased further by HDAC inhibitors (Fig. 7, C and D). A quantitative analysis of FIGURE 6. Acetylation/deacetylation controls Runx2 transactivation activity. A, the effect of HDAC inhibitors on Runx2-dependent transcription. 293 cells were transfected with pGL3-T␤REluc (0.05 g) and the Runx2 expression plasmid (Runx2, 0.1 g), or empty vector (vec, 0.1 g). Transfected cells were treated with increasing concentrations of HDAC inhibitors (TSA, CHAP27, or SCOP402; 0, 30, 60, 100 nM, 16 h). The effect of HDAC inhibitors on Runx2-dependent transactivation was measured using a luciferase reporter assay. The plasmid pCMV-␤gal (0.05 g) was included as an internal control for transfection efficiency and luciferase activities were normalized to ␤-galactosidase activities. Each experiment was performed in triplicate, and data from three independent experiments are presented. B, substitution of critical lysine residues with arginine abrogates the transactivation activity of Runx2. 293 cells were transfected with pGL3-T␤RE-luc (0.05 g) and wild-type Runx2, or the indicated mutants in the presence or absence of co-transfected p300 expression plasmid. Runx2 transactivation activities were determined using the luciferase reporter assay as described above. C, 293 cells were transfected with wild-type Runx2, or the indicated mutants in the presence or absence of p300 coexpression. Interaction of Runx2 and p300 was determined by co-immunoprecipitation and protein levels of Runx2, p300, and tubulin were detected by immunoblotting.
RT-PCR indicated that HDAC inhibitor-stimulated OC expression increased 3-5 times in the presence of Runx2 (Fig. 7D). This is consistent with the results of OC promoter activity analysis (Fig. 7B). Collectively, the increase in Runx2 acetylation by HDAC inhibitors strongly stimulated osteoblast differentiation.
OC Expression Is Specifically Enhanced by Runx2 Acetylation-Because OC expression can be regulated by other osteogenic transcription factors, such as Dlx5 (49), or Osterix (Osx) (50), we examined whether or not Runx2 is the specific target of HDAC inhibitors for activation of the OC promoter. The overexpression of Dlx5 or Osx weakly stimulated osteocalcin expression, and the treatment of HDAC inhibitors also weakly stimulated basal OC expression; however, the addition of HDAC inhibitors in combination of Dlx5 or Osx expression did not further stimulate OC expression in Runx2 Ϫ/Ϫ cells (Fig. 7E). In contrast, the combination of Runx2 overexpression and HDAC inhibitor treatment further enhanced OC mRNA expression in the cell (Fig. 7E). These results suggest that, among these three osteogenic transcription factors, Runx2 is the only target of HDAC inhibitors in osteoblast differentiation.
An HDAC Inhibitor Increases Bone Formation-We next examined the effect of HDAC inhibitors on bone formation using calvarial organ cultures from newborn mice. Simvastatin, a positive stimulator of bone formation, was used as a positive control (51). Calvaria were treated with SCOP402 or Simvastatin (Fig. 8A). The new bone areas were significantly increased in the presence of SCOP402 compared with that of vehicle-treated controls. The effect of SCOP402 is comparable to Simvastatin (51). Quantitative histomorphometric analysis is shown in Fig.  8B. These results indicate that up-regulation of Runx2 acetylation by HDAC inhibitors may not only stimulate osteoblast differentiation but also increases new bone formation.

DISCUSSION
BMPs play pivotal roles in the regulation of bone induction, maintenance and repair. Runx2 has been identified as a master gene controlling osteoblast differentiation, and as a major target of BMP-2. BMP-2 sig-FIGURE 7. HDAC inhibitors stimulate osteoblast differentiation. A, C2C12 cells were cultured for 3 days in the presence or absence of BMP-2 (30 ng/ml), and treated with increasing concentrations of HDAC inhibitors (0, 10, 30, 60 nM). Enzyme activity of the osteoblast-specific marker ALP was assayed as described by Katagiri et al. (46). B, Runx2 Ϫ/Ϫ cells (H1-127-21-2) were transfected with pGL2-1050 OC-luc (wild-type OSE2, WT OSE), or a RUNX binding site-mutated version (mutant OSE), with or without the Runx2 expression plasmid. The reporter activities were determined in the presence, or absence of the HDAC inhibitor SCOP402. C, Runx2 Ϫ/Ϫ cells were transfected with Runx2. After 24 h, the cells were treated with HDAC inhibitors for 3 days. RNA was then purified from the cells, and OC expression was determined by RT-PCR. The OC mRNA was amplified using forward (5Ј-GTGAATTCACCTAGCAGACACCATGAG-GAC-3Ј) and reverse (5Ј-ACGGATCCGTCTTCAAG-CCATACTGGTCTG-3Ј) primers. D, band intensities of PCR-amplified OC in C were quantitated and shown as a graph. E, Runx2 Ϫ/Ϫ cells were transfected with Runx2, Dlx5, or Osx expression plasmids. After 24 h, the cells were treated with HDAC inhibitors for 3 days. The levels of OC mRNA were measured by RT-PCR. naling converges on Runx2 through various pathways. First, BMP-2activated Smads increase transcription of the Runx2 gene (15). Secondly, BMP-2-activated Smads interact with the Runx2 protein and function synergistically to stimulate osteoblast differentiation (15,19,20). In this report, we demonstrate a third interaction, i.e. BMP-2 signaling resulting in Runx2 acetylation. Acetylation protects Runx2 from Smurf1-mediated degradation. Runx2 acetylation is mediated by the acetyltransferase activity of p300, and deacetylation is mediated by class II HDACs. Increases in acetylated Runx2 were detected when cells were treated with BMP-2 (Fig. 2), and acetylation or mutation of the lysine residues on Runx2, inhibited Smurf1-mediated Runx2 degradation (Fig.  4D). An increase in the acetylation of Runx2, via pharmacological inhibition of HDAC, potentiated BMP-2-mediated osteoblast differentiation, and increased bone formation (Figs. 7 and 8).
The acetylation of Runx2 was increased by treatment with BMP-2 and was further augmented by p300. Similarly, acetylation was increased by overexpression of BMPR1, Smads, or p300 and was further augmented by combinations of these factors. These results indicate that Runx2 acetylation is positively regulated by the BMP-2 signaling pathway, and BMP-2-activated Smads stimulate p300-mediated Runx2 acetylation by facilitating their physical interaction. However, it is unclear whether p300 is the only physiological effector of this acetylation, because P/CAF (p300-and CBP-associated factor) can also stimulate Runx2 acetylation (data not shown).
A previous report indicated that acetylation of chromatin is required for osteoblast differentiation (52). Undifferentiated cells showed low acetylation of histones H3 and H4, while differentiated osteoblasts showed higher acetylation levels. This change is coupled functionally to the chromatin remodeling events that mediate the developmental induction of osteoblast marker gene transcription, including that of OC, in bone cells (52). In this study, we suggest another function for acetylation in osteoblast differentiation. Acetylation activated the transactivation activity of Runx2, and increased its half-life, by suppressing Smurf1-mediated degradation. This enabled Runx2 to stimulate osteoblast differentiation. Our results cannot rule out the involvement of a general chromatin remodeling of osteocalcin gene promoter by histone acetylation (52) because OC expression was slightly increased by HDAC inhibitors treatment only (Fig. 7E). However, the importance of Runx2 acetylation is further supported by the results of our osteocalcin promoter activity assays, in which chromatin remodeling events were not involved (Fig. 7B). Our results also indicate that other osteogenic transcription factors, such as Dlx5 and Osx, are not influenced by acetylation events (Fig. 7E).
Our data indicate that increased acetylation and abundance are both important for Runx2 activity, because mutations of the lysine residues targeted by p300 stabilized the mutant proteins, but resulted in a decrease in Runx2 transactivation activity (Fig. 6B). This result suggests that the acetylation of the lysine residues not only inhibits Smurf1mediated ubiquitination, but also is required for transactivation activity. The addition of an acetyl group to a lysine residue could create a new surface for protein association. As the SH2 domain can interact with phospho-Tyr residues, 14-3-3 proteins can interact with phosphor-Ser/ Thr (53,54), whereas the bromodomain functions as a structural module specific for acetyl lysine-containing motifs (27). Therefore, an interesting question to be addressed in the future is whether or not acetylation of Runx2 creates a new surface for protein association, thereby facilitating the binding of the protein to other cell regulatory machineries.
Previously, Vega et al. (37) claimed that HDAC4 inhibits Runx2 activity by blocking Runx2 DNA binding, but detailed experimental data regarding Runx2 acetylation was not provided. However, our results clearly support a requirement for HDAC4 and -5 deacetylase activity to regulate acetylation, abundance and activity of Runx2. Moreover, several other groups have reported that HDAC inhibitors increased Runx2dependent activation of the osteocalcin promoter (55) and osteoblast maturation and differentiation (56,57). However, none of these reports discussed the possible acetylation/deacetylation of Runx2. Our results strongly suggest that Runx2-mediated osteogenic differentiation requires acetylation-mediated regulation of the Runx2 protein by HATs and HDACs. HDAC3 and HDAC6 have also been also shown to interact with Runx2 (55,58), suggesting that there are multiple HDACs targeting Runx2.
The proposed link between BMP-2 signaling and Runx2 acetylation is further supported by our finding that HDAC deacetylates Runx2 (Fig.  5). Treatment with HDAC inhibitors increased Runx2 acetylation (Fig.  5G), potentiated BMP-2 activity in induction of osteoblast differentiation marker (Fig. 7A) and stimulated bone formation (Fig. 8). These results strongly support our model that acetylation/deacetylation and ubiquitination are physiological mechanisms governing the abundance and activity of Runx2, which is critical for osteoblast differentiation.
Because of their osteogenic potential, BMPs and Runx2 show tremendous potential for development as therapeutic agents to heal bone fractures, prevent osteoporosis, and enhance bone formation around alloplastic materials implanted into bone. It is estimated that 100 million people worldwide are at risk for osteoporosis (59). For the treatment of osteoporosis, both inhibition of osteoclasts and activation of osteoblasts are required. Despite recent success with drugs that inhibit bone resorption, there is a clear need for osteoblast activators that will increase bone formation substantially. Mundy et al. (51) provided an important milestone in the development of osteoblast activators by screening chemicals that activate the BMP-2 promoter. They found that Lovastatin and Simvastatin could induce BMP-2 expression, thereby increasing bone formation (51). In the current study, we found that pharmacological inhibition of HDAC-dependent deacetylation enhances the osteogenic activity of BMP-2, by increasing Runx2 activity and stability. Our results provide a new theoretical basis for developing therapeutic agents against osteoporosis.