Histone Deacetylase 7 Associates With Runx2 and Represses Its Activity During Osteoblast Maturation in a Deacetylation-Independent Manner


  • The authors state that they have no conflicts of interest.


HDAC7 associates with Runx2 and represses Runx2 transcriptional activity in a deacetylase-independent manner. HDAC7 suppression accelerates osteoblast maturation. Thus, HDAC7 is a novel Runx2 co-repressor that regulates osteoblast differentiation.

Introduction: Runx2 is a key regulator of gene expression in osteoblasts and can activate or repress transcription depending on interactions with various co-factors. Based on previous observations that several histone deacetylases (HDACs) repress Runx2 activity and that HDAC inhibitors accelerate osteoblast differentiation in vitro, we hypothesized that additional HDACs may also affect Runx2 activity.

Materials and Methods: A panel of HDACs was screened for repressors of Runx2 activity. Immunofluorescence, co-immunoprecipitation, GST-pulldowns, and chromatin immunoprecipitations were used to characterize the interactions between Runx2 and HDAC7. Expression of osteoblast markers was examined in a C2C12 cell osteoblast differentiation model in which HDAC7 levels were reduced by RNAi.

Results: Runx2 activity was repressed by HDAC7 but not by HDAC9, HDRP, HDAC10, or HDAC11. HDAC7 and Runx2 were found co-localized in nuclei and associated with Runx2-responsive promoter elements in osseous cells. A carboxy-terminal domain of Runx2 associated with multiple regions of HDAC7. Although direct interactions with Runx2 were confined to the carboxy terminus of HDAC7, this region was dispensable for repression. In contrast, the amino terminus of HDAC7 bound Runx2 indirectly and was necessary and sufficient for transcriptional repression. Treatment with HDAC inhibitors did not decrease inhibition by HDAC7, indicating that HDAC7 repressed Runx2 by deacetylation-independent mechanism(s). Suppression of HDAC7 expression in C2C12 multipotent cells by RNAi accelerated their BMP2-dependent osteoblast differentiation program. Consistent with this observation, BMP2 decreased nuclear localization of HDAC7.

Conclusions: These results establish HDAC7 as a regulator of Runx2's transcriptional activity and suggest that HDAC7 may be an important regulator of the timing and/ or rate of osteoblast maturation.


Runx2 is an essential regulator of bone formation. The crucial importance of Runx2 is revealed by the phenotypes associated with alterations to its expression. Heterozygosity at the Runx2 locus results in the bone disorder cleidocranial dysplasia,(1) which is characterized by short stature, dental defects, and reduced or absent clavicles, whereas homozygous Runx2 knockout mice die at birth and completely lack mineralized bone.(2,3) Runx2 binds DNA and acts as both a transcriptional activator and repressor. Runx2 induces transcription by recruiting co-activators such as the p300 histone acetyltransferase.(4) Transcriptional repression by Runx2 is mediated by associations with co-repressors including mSin3a,(5) groucho/TLE,(6) YAP,(7,8) and histone deacetylases (HDACs).(9–11) HDAC3 and HDAC6 were shown to bind, respectively, to amino-terminal and carboxy-terminal repression domains and to repress Runx2-mediated transcription by HDAC-inhibitor sensitive mechanisms.(9,11) HDAC4 inhibits Runx2 transcriptional activity in a different manner; it binds to the Runt domain and interferes with DNA binding.(10) HDAC4 and HDAC5 also negatively regulate Runx2 activity by deacetylating lysines in the Runx2 protein, leading to ubiquitin-mediated proteolysis.(12)

Histone deacetylases remove acetyl groups from histone core proteins, resulting in a less transcriptionally active chromatin state, and nonhistone substrates. There are 18 mammalian HDAC family members. They are divided into four subclasses based on homology to prototypic yeast deacetylase proteins.(13,14) HDAC7 is a member of class IIa. HDACs in this class (i.e., HDAC4, 5, 7, and 9) exhibit tissue-restricted expression patterns(15–23) and are actively shuttled between the nuclear and cytoplasmic compartments.(24–29) Phosphorylation of conserved serines in their amino termini leads to interaction with 14-3-3 proteins and export from the nucleus, therein attenuating their repressive activities.(25,26,29,30) Class IIa HDACs contain similar overall domain structures, possessing an amino-terminal domain of roughly 450–500 amino acids that includes a nuclear localization sequence and that mediates protein–protein interactions and a similarly sized carboxy-terminal domain comprised of the deacetylase catalytic domain and a nuclear export sequence. Despite possessing what seem to be functional deacetylase catalytic domains, class IIa histone deacetylases have not been shown to repress transcription by directly deacetylating histones. Rather, they are thought to function by recruiting repression complexes composed of class I HDACs and co-repressor proteins such as SMRT, N-CoR, B-CoR, and mSin3a, where the actual deacetylation of chromatin is most likely conducted by the class I HDACs.(18,19,31–33) The amino termini of class IIa HDACs have also been shown to possess deacetylation-independent repression activities including recruitment of CtBP(24,34) and HP1(35) co-repressor proteins, directly inhibiting the ability of transcription factors to bind DNA(10) or sequestering transcription factors into inactive subnuclear bodies.(36)

Histone deacetylases play important roles in bone formation, because alterations to HDAC expression or activity have significant effects on osteoblast maturation. Suppression of HDAC3 or HDAC1 expression by RNA interference accelerated the course of osteoblastic differentiation.(9,37) Histone deacetylase inhibitors accelerate osteoblast maturation in vitro in similar manners.(38–40) In light of the known interactions between Runx2 and a small number of histone deacetylases and of the dramatic effect that HDAC inhibitors exert on osteoblast differentiation, we sought to determine whether additional HDAC proteins modulate Runx2 transcriptional activity. We found that HDAC7 associates with Runx2 and functions as a transcriptional co-repressor of Runx2 activity in osteoblasts.


Cell culture

C2C12 and COS cells were grown in DMEM containing 10% FBS, 200 mM l-glutamine, 50 units/ml penicillin, and 50 μg/ml streptomycin. MC3T3-E1, UMR-106, and ROS17/2.8 cells were cultured in MEM supplemented with 10% FBS, 50 units/ml penicillin, 50 μg/ml streptomycin, and 1% nonessential amino acids.


Drs Victoria Richon (HDAC9 and HDRP; Merck, Boston, MA, USA), Tso-Pang Yao (HDAC10; Duke University), and Scott Hiebert (pcDNA3.1-human HDAC7-FLAG; Vanderbilt University) kindly provided the indicated FLAG-tagged HDAC expression plasmids. The HDAC11 expression clone was purchased from Open Biosystems and subcloned into a pcDNA3.1-FLAG vector (cloning details are available on request). pCMV5-HA-Runx2 (MASNS isoform),(41) Gal-Runx2 1–383 and 383–513,(11) and GST-Runx2(9) expression plasmids were previously described. HDAC7 deletion constructs were generated by PCR, cloned into pcDNA3.1-FLAG vector, and sequenced to verify the integrity of the coding sequence. HDAC7 point mutations were generated with the Quikchange site-directed mutagenesis kit (Stratagene).

Reporter assays

C2C12 cells were transfected with Lipofectamine (Invitrogen) in 12-well plates with 200 ng of p6OSE2-luc, 100 ng pRL-null, 300 ng pCMV5-HA-Runx2, and 450 ng of HDAC expression plasmids as indicated. pcDNA3.1 was added to transfections to maintain a uniform amount of total DNA per transfection. Luciferase activity was measured 24 h after transfection using the dual-luciferase assay system (Promega). Each transfection was performed in triplicate and normalized to Renilla-luciferase activity. In some experiments, trichostatin A (TSA) or valproic acid (VPA) was added 4–5 h after transfection and left on the cells until analysis.


Cell lysates were prepared by rinsing cultures with PBS before lysing the cells on ice for 5 min in modified RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate, 1 mM EDTA) supplemented with complete protease inhibitor tablets (Roche). Crude lysates were sonicated and cleared by centrifugation at 12,000 rpm at 4°C. Total protein was quantified using the detergent-compatible protein assay (BioRad). For detection of endogenous HDAC7 protein, 40 μg of total protein was resolved by SDS-PAGE and transferred to Immobilon-P membranes (Millipore). Membranes were blotted with antibodies against HDAC7 (12174; Abcam), Runx2 (clone M70; Santa Cruz), FLAG (clone M2; Sigma-Aldrich), actin (clone I-19; Santa Cruz), histone H3 (9715; Cell Signaling), acetylated-histone H3 (06-599; Upstate/Millipore), Gal4 (clone SC-577; Santa Cruz), and horseradish peroxidase (HRP)-conjugated secondary antibodies. Proteins were visualized using ECL-Plus chemiluminescent substrate (G.E. Health Systems).


Cells were grown on glass coverslips, rinsed in PBS, fixed in 4% paraformaldehyde for 30 min, permeabilized with PBS + 0.3% TritonX-100 for 5 min, blocked for 30 min in immunofluorescence buffer (3% BSA, 20 mM MgCl2, 0.3% Tween20 in PBS), and incubated with primary antibodies in immunofluorescence buffer. Cells were washed three times with PBS + 0.1% TritonX-100, incubated 30 min with Alexa-conjugated secondary antibodies at 1:800 (Invitrogen), washed, and mounted in 90% glycerol/0.4% N-propyl-gallate. Images were obtained using an Olympus Fluoview 500 confocal microscope and processed using Adobe Photoshop.


For immunoprecipitation of endogenous proteins, ROS17/2.8 nuclear extracts were prepared from two to three confluent 10-cm plates per reaction. Cells were washed twice and collected in PBS by scraping and centrifugation. The cell pellet was resuspended in Iso-Hi buffer (10 mM Tris-HCl, pH 7.8, 140 mM NaCl, 1.5 mM MgCl2, 0.5% NP-40) and incubated for 10 min on ice. Nuclei were collected by centrifugation at 5000 rpm for 5 min at 4°C. Isolated nuclei were resuspended in lysis buffer (0.5% SDS, 0.25% sodium deoxycholate, 0.5% TritonX-100, 1 mM EDTA in PBS supplemented with protease inhibitors), sonicated, and precleared by centrifugation at 12,000 rpm at 4°C. Lysates were cleared by incubation with protein A/G-plus agarose beads (Santa Cruz) for 30 min at 4°C and subjected to immunoprecipitation overnight with 0.75 μg normal rabbit IgG or anti-Runx2 antibody. Immunoprecipitates were collected by addition of protein A/G plus-agarose, washed three times with lysis buffer, boiled in SDS-PAGE loading buffer, separated by PAGE, and detected by immunoblotting as described above. For immunoprecipitation of exogenous proteins, COS cells in 60-mm dishes were transfected with Lipofectamine (Invitrogen) with 2 μg each of the expression plasmids as indicated and brought to 4 μg total DNA with pcDNA3 as needed. Twenty-four hours after transfection, whole cell lysates were prepared in lysis buffer as above and immunoprecipitated with FLAG-M2-Agarose (Sigma).

Biotinylated oligonucleotide pulldown

ROS17/2.8 cell nuclear lysates were prepared as for immunoprecipitation. Double-stranded DNA fragments containing the osteocalcin promoter Runx2 binding site B or mutant Runx2 binding site have been previously reported(42) and were generated by annealing complementary oligonucleotides, one of which contained a 5′-biotin modification, and used for affinity purification of ROS cell nuclear proteins as described.(43)

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) analyses were performed on ROS17/2.8 cell lysates as previously reported(4) using rabbit polyclonal antibodies against Runx2, HDAC7, or normal rabbit IgG. Primers against the rat osteocalcin promoter were previously described.(4) Primers for PCR against other Runx2 binding sites were as follows: bone sialoprotein (BSP) forward, 5′-ACCATCTTCTCCTCAGCCTTC-3′; BSP reverse, 5′-TTCTCCTCTCAGCAGCCAAT-3′; matrix metalloproteinase-13 (MMP13) forward, 5′-TACTAGCCACCGCTTCTTCC-3′; MMP13 reverse, 5′-AGCAGTGCCTGGAGTCTCTC-3′; Runx2 forward, 5′-AACCTTCTGAATGCCAGGAA-3′; Runx2 reverse, 5′-CTGGAGAGACAGAATCATGTGG-3′.

GST pulldowns

For GST pulldown experiments, [35S]trans-labeled proteins were synthesized using the T7 TnT System (Promega). GST fusion proteins were isolated by sonicating E. coli cultures and incubating overnight with glutathione sepharose, followed by three washes of PBS. The yield and integrity of GST-fusion proteins were verified by SDS-PAGE and Coomassie staining. TnT proteins and GST-fusion proteins were preblocked separately in binding buffer (50 mM Tris, pH 7.6, either 150 nM [low stringency], 250 nM [moderate stringency], or 500 mM [high stringency] NaCl, 1 mM EDTA, 0.5% NP-40, 5 mM DTT, 0.5% BSA) for 60 min at 4°C and co-incubated overnight. The bound complexes were washed twice with binding buffer and once with binding buffer lacking BSA. Bound complexes were resolved on SDS-PAGE, fixed, incubated with autoradiographic enhancer (Amplify buffer; G.E. Health Systems), dried, and visualized by autoradiography.

HDAC7-suppressed cells

Short-hairpins RNAs against mouse HDAC7 were cloned into the pSilencer5.1U6-Retro vector (Ambion). Retroviruses were produced in 293T cells and used to infect C2C12 cells. After puromycin selection, HDAC7 expression was determined by real-time PCR and immunoblotting. For osteoblast differentiation experiments, confluent cultures of stably transduced C2C12 cells were grown in differentiation medium (DMEM supplemented with 10% FBS, 50 mg/ml ascorbic acid, 2 mM β-glycerolphosphate, and 300 ng/ml BMP2). Medium was replaced every 2–3 days. Alkaline phosphatase activity was measured by incubating cell lysates in 0.6 M 2-amino-2-methyl-1-propanol, 2.4 mM MgCl2, and 9.6 mM p-nitrophenyl phosphate at 37°C for 30 min. Reactions were stopped by addition of 2 N NaOH. The activity was determined as absorbance at 410 nm and normalized to protein content. RNA was harvested using Trizol reagent according to the manufacturer's directions and analyzed by real-time PCR as previously described.(9)


HDAC7 inhibits Runx2 transcriptional activity

In previous studies, we found that HDACs 3, 4, and 6 can bind and inhibit transcriptional activation by Runx2.(9,11) In light of these observations, we tested the ability of additional HDACs and the histone deacetylase–related protein, HDRP, to modulate the ability of Runx2 on the p6OSE2-luc reporter construct in C2C12 multipotent progenitor cells (Fig. 1A). Runx2 activated the reporter ∼20-fold, whereas co-transfection with HDAC7 reduced this activation by ∼50% to near 10-fold above the reporter alone (p ≤ 0.05). In contrast, HDAC9, HDRP, HDAC10, and HDAC11 did not significantly or reproducibly affect activation by Runx2. None of the HDAC proteins had significant effects on the reporter in the absence of Runx2 (data not shown). Transfection of increasing amounts of HDAC7-FLAG plasmid resulted in progressively less activation of the reporter by Runx2, indicating that HDAC7 is a concentration-dependent repressor of Runx2 (Fig. 1B). We observed similar inhibition of Runx2's activation of the p6OSE2-luc reporter by HDAC7 in COS, MC3T3-E1, and ROS17/2.8 cells (data not shown). HDAC7 also repressed Runx2-dependent activation of the mouse osteocalcin gene 2 (mOG2) promoter by ∼50% in MC3T3-E1 cells (Fig. 1C). Thus, HDAC7, like HDAC3, 4, and 6, is capable of inhibiting Runx2-mediated gene activation.

Figure FIG. 1..

Identification of HDAC7 as a co-repressor of Runx2 in the osteoblast lineage. (A) HDAC7 inhibits transcriptional activation by Runx2. C2C12 cells were co-transfected with the p6OSE2-luc and pRL-null reporter plasmids and expression plasmids encoding Runx2 and HDAC proteins as indicated. Firefly luciferase readings were normalized to Renilla luciferase levels and are shown with the activity from the control vector defined as basal (1-fold activation). ap ≤ 0.05 by Student's t-test. (B) Repression of Runx2 is dependent on HDAC7 levels. C2C12 cells were transfected as in A but with 300 ng HA-Runx2 plasmid and increasing amounts of HDAC7-FLAG expression vector as indicated. Luciferase activity was determined as above. (C) HDAC7 inhibits transcriptional activation by Runx2 of the osteocalcin promoter. MC3T3 cells were co-transfected with the mOG2-luc and pRL-null reporter plasmids and expression plasmids encoding Runx2 and HDAC proteins as indicated. Luciferase activity was determined as described above. ap = 0.002 by Student's t-test. (D) HDAC7 is expressed in osteoblast progenitor, osteosarcoma, and mouse primary osteoblast cells. Cell lysates from the indicated cell lines were blotted with antibodies for HDAC7 and actin.

HDAC7 is expressed in osteoblast-related cell lines

HDAC7 is expressed in vascular endothelial cells, heart, placenta, lung, and thymus, and at lower levels in several other tissues.(15,18,19) Its expression status in osteoblasts has not been reported. Because Runx2 is required for osteoblast development and maturation, we assessed HDAC7 expression in osseous cells to determine whether the two proteins could possibly interact in a natural setting. Immunoblotting revealed that HDAC7 protein is abundant in osteoblast-related cell lines, such as multipotent progenitor cells (C3H10T1/2 and C2C12 cells) and committed pre-osteoblasts (MC3T3-E1), as well as in mouse primary calvarial osteoblasts (Fig. 1D). Osteosarcoma-derived cell lines (ROS17/2.8 and UMR-106) expressed less HDAC7. These data show that HDAC7 and Runx2 could potentially interact during osteoblast development.

HDAC7 physically associates with Runx2 in the nucleus

The distribution of HDAC7 between the nucleus and cytoplasm varies depending on the cell line, but when in the nucleus, HDAC7 is found in punctate foci.(18,31) To determine HDAC7 distributions patterns in osseous cells, we examined the localization of endogenous HDAC7 and Runx2 in ROS17/2.8 cells and in primary osteoblasts. In each of these cell types, the majority of HDAC7 was cytoplasmic; however, discrete foci were detected in nuclei (Figs. 2A and 2B). Runx2 was found in nuclear matrix–associated foci as previously reported(44,45) and co-localized with HDAC7 in numerous foci within the nuclei of ROS17/2.8 cells and primary calvarial osteoblasts (Figs. 2A and 2B). We observed a similar distribution of overexpressed FLAG-tagged HDAC7 in these and other osseous cells (Table 1). Interestingly, co-transfection of Runx2 increased the nuclear localization of HDAC7. Thus, ∼30% more C2C12 cells expressed HDAC7-FLAG primarily in their nuclei when Runx2 was co-overexpressed (Fig. 2C). Moreover, ∼10% more cells expressed HDAC7-FLAG in both nuclear and cytoplasmic compartments when Runx2 was present. This was associated with a concomitant 45% decrease in the number of cells in which HDAC7-FLAG was mainly present in the cytoplasm. These data were confirmed by immunoblotting of cytoplasmic and nuclear fractions (Fig. 2D). Overexpression of Runx2 also enhanced nuclear localization of endogenous HDAC7 (Fig. 2E). To further show interactions between endogenous Runx2 and HDAC7 proteins, we immunoprecipitated ROS17/2.8 cell nuclear extracts with a polyclonal antibody against Runx2 or with normal rabbit IgG as a negative control. Immunoblotting showed that HDAC7 and Runx2 were both detected specifically in the anti-Runx2 immunoprecipitates (Fig. 2F). Together, these data showed that Runx2 and HDAC7 proteins interact with one another in the nucleus.

Table Table 1.. Predominant Localization of HDAC7-FLAG
original image
Figure FIG. 2..

HDAC7 physically interacts with Runx2 in vivo. (A and B) HDAC7 and Runx2 co-localize in subnuclear foci. Endogenous Runx2 and HDAC7 were visualized in ROS17/2.8 cells (A) and mouse primary calvarial osteoblasts (B) and examined by confocal microscopy. Overlap between the HDAC7 staining and Runx2 staining are shown as yellow in the Merge panels. (C) Runx2 enhances nuclear localization of HDAC7-FLAG. C2C12 cells were transfected with HDAC7-FLAG and empty vector or HA-Runx2 and immunostained for FLAG and HA. For co-transfected cells, only those cells clearly positive for both FLAG and HA were counted. The experiment was repeated three times with similar results. Data from one representative experiment are presented. (D) C2C12 cells were transfected with HDAC7-FLAG and HA-Runx2 as indicated, separated into cytoplasmic “C” and nuclear “N” fractions, resolved by SDS-PAGE, and examined by immuoblotting against FLAG. Blotting against AKT and histone H3 were used as markers, respectively, for the cytoplasmic and nuclear fractions and to show protein loading. (E) Runx2 enhances endogenous HDAC7 nuclear localization in ROS 17/2.8 cells. Cells were transfected with pcDNA3-Runx2 and examined by immunofluorescence microscopy against Runx2 (left) and HDAC7 (right). Compare HDAC7 distribution in the cell plainly overexpressing Runx2 (in top right quadrant) with remaining cells that do not appear to overexpress Runx2. (F) HDAC7 co-immunoprecipitates with Runx2. Nuclear extracts of ROS17/2.8 cells were immunoprecipitated with normal rabbit immunoglobulin or a rabbit polyclonal antibody against Runx2. Precipitated material and a fraction of the input nuclear proteins were resolved by PAGE and visualized by immunoblotting with antibodies against HDAC7 and Runx2. The Runx2 band in the “input” lane is underexposed to avoid overexposing the band in the anti-Runx2 immunoprecipitation lane. (G) HDAC7 associates with Runx2 responsive promoters. ROS cell nuclear proteins were affinity purified with biotinylated oligonucleotides containing a wildtype or mutated Runx2-binding site from the osteocalcin promoter coupled to streptavidin-agarose beads. After washing, the bound materials were resolved by PAGE and visualized by immunoblotting as indicated. (H) HDAC7 associates with Runx2-responsive elements in endogenous promoters. Chromatin immunoprecipitation of ROS cell lysates were performed using normal rabbit immunoglobulin or antibodies against Runx2 (Santa Cruz; clone M70), HDAC7 Ab1 (Abcam 12175), or HDAC7 Ab2 (Abcam 12174) and subjected to PCR amplification of Runx2-responsive regions of the indicated promoters. Note that HDAC7 Ab2 was only examined with the OCN promoter sequence. N/D, not determined; Input, 0.5% of the immunoprecipitated chromatin; Water, a no-template control.

HDAC7 associates with Runx2-responsive promoters in vivo

Having found that HDAC7 inhibits Runx2-mediated transcriptional activation and co-localizes with Runx2 in the nucleus, we asked if HDAC7 is associated with the promoter of Runx2 target genes. We incubated ROS nuclear extracts with biotinylated 25-bp-long oligonucleotides that encode a fragment of the osteocalcin promoter containing a well-characterized Runx2 binding site or with a mutant Runx2 site. Oligonucleotides were bound to streptavidin-agarose beads to collect complexes. Immunoblotting revealed that significant amounts of Runx2 and HDAC7 were associated with the wildtype oligonucleotide (Fig. 2G). HDAC7's interaction with this sequence is dependent on Runx2 as mutation of the Runx2 binding site severely reduced binding of both proteins (Fig. 2G). These data indicate that HDAC7 interacts with Runx2–DNA complexes and are consistent with a role for HDAC7 in regulating transcription of Runx2 target genes. Finally, ChIP analysis was used to measure HDAC7 association with Runx2 at native Runx2-responsive promoters in vivo. Antibodies against Runx2 and HDAC7 immunoprecipitated regions of the osteocalcin, BSP, MMP13, and Runx2 gene promoters that contain defined Runx2 binding sites (Fig. 2H). Collectively, these data show that HDAC7 associates with Runx2 at Runx2-regulated promoter elements.

Carboxy termini of Runx2 and HDAC7 interact with each other

HDAC3, 4, and 6 were previously shown to bind to different regions of Runx2; the amino terminus, Runt domain, and carboxy terminus, respectively.(9–11) To define the interacting regions of Runx2 and HDAC7, we tested a series of peptide fragments of each protein (Fig. 3A) for their ability to bind to each other. We first performed GST-pulldown experiments in which in vitro synthesized HDAC7 was incubated with GST-Runx2 deletion constructs purified from E. coli. Under conditions of high stringency (500 mM NaCl), HDAC7 interacted most strongly with the carboxy terminus of Runx2 (amino acids 383–513; Fig. 3B), although weaker interactions with other regions of Runx2 could sometimes be detected, particularly under reduced stringency (Fig. 3B and data not shown). The HDAC7 protein consists of a carboxy-terminal deacetylase catalytic domain (amino acids 472–912) and an amino-terminal regulatory domain (amino acids 1–478; Fig. 3A). When we assayed the abilities of HDAC7 amino- and carboxy-terminal peptides to bind the carboxy terminus of Runx2, only the HDAC7 carboxy terminus interacted with the Runx2 carboxy terminus in vitro (Fig. 3C).

Figure FIG. 3..

Mapping the interaction domains of HDAC7 and Runx2 in vitro. (A) Schematic diagrams of the Runx2 and HDAC7 proteins, indicating the location of selected features of each protein and summarizing the results from Figs. 4 and 5. (B) HDAC7 interacts most strongly with the Runx2 C-terminal fragments in vitro. (Top) Autoradiograph of 10% input of [35S]labeled, in vitro synthesized HDAC7-FLAG protein and the fraction bound to GST-Runx2 fusion peptides. (Bottom) Coomassie staining of input GST-Runx2 peptides purified from E. coli by binding to glutathione-sepharose showing relative abundance of each. *Appropriate GST-fusion polypeptides. (C) The C terminus of HDAC7 binds the Runx2 C terminus in vitro. Autoradiograph of GST-pulldown showing 10% of the input of in vitro synthesized full-length HDAC7, N-terminal and C-terminal HDAC7 polypeptides, and the fraction bound to a GST-Runx2 C-terminal fragment. Equivalent amounts of GST and GST-Runx2 383–513 were determined by SDS-PAGE and Coomassie staining (data not shown).

To examine this interaction in vivo, FLAG-tagged HDAC7 was co-expressed in COS cells with Gal-tagged Runx2 amino acids 1–383 and 383–513 (Fig. 4). Gal-Runx2 383–513 was readily detected in FLAG-HDAC7 immunoprecipitates, but Gal-Runx2 1–383 was not (Fig. 4A). Thus, Runx2 amino acids 383–513 are sufficient to bind HDAC7 in vitro and in vivo. Additional experiments using Runx2 mutant polypeptides containing single amino acid point mutations within the Runx2 nuclear matrix targeting signal (NMTS) strongly reduced the interaction with HDAC7, suggesting that the Runx2 NMTS mediates the binding to HDAC7 (data not shown).

Figure FIG. 4..

In vivo interactions between HDAC7 and Runx2. (A) Both the amino and carboxy termini of HDAC7 bind the Runx2 C terminus in vivo. COS cells were co-transfected with Gal-Runx2 and HDAC7-FLAG expression plasmids, lysed, and immunoprecipitated with anti-FLAG agarose beads. Immunoblotting of the precipitates reveals binding of full-length, amino-terminal and carboxy-terminal fragments of HDAC7 specifically to the carboxy terminus of Gal-Runx2 (top). (Bottom) Ten percent of the whole cell lysates was immunoblotted against Gal or FLAG to verify expression of Gal-Runx2 and HDAC7-FLAG polypeptides. (B) Multiple regions of HDAC7 interact with a Gal-Runx2 C-terminal fragment in vivo. Immunoprecipitations were performed as in A using additional HDAC7-FLAG peptides.

We also defined the HDAC7 domains responsible of in vivo interactions. Both HDAC7 1-478-FLAG (amino terminus) and HDAC7 472–912-FLAG (carboxy terminus) exhibited robust binding to Gal-Runx2 383-513, but not Gal-Runx2 1-383 (Fig. 4A). Thus, in contrast to the in vitro GST pulldowns, both halves of HDAC7 interacted with the Runx2 carboxy terminus in vivo. To better define the region(s) of the HDAC7 amino terminus that associate with Runx2, we next tested interactions between Gal-Runx2 383–513 and a series of smaller HDAC7 amino-terminal fragments. Shown in Fig. 4B, all fragments of the HDAC7 amino terminus bound Gal-Runx2 with the exception of amino acids 1–76. Notably, HDAC7 1–98 did interact with Runx2, indicating that the region between amino acids 76 and 98 is sufficient to interact with Runx2. HDAC7 99–478 and 230–478 also co-immunoprecipitated with Gal-Runx2, indicating that amino acids 76–98 do not make up the only interaction domain for Runx2 in the HDAC7 amino terminus.

Amino terminus of HDAC7 is necessary and sufficient to inhibit Runx2 activity

To understand the mechanism by which HDAC7 represses Runx2, we determined which region(s) of HDAC7 is important for repression (Fig. 5A). We found that amino acids 1–478 are necessary and sufficient to inhibit Runx2 activation of the p6OSE2-luc reporter. In contrast, the carboxy-terminal amino acids 472–912, which include the deacetylase domain, did not possess any repressive activity in these assays; in fact, the HDAC7 472–912 peptide consistently increased activation of the reporter by Runx2 (Fig. 5A and data not shown). A series of smaller HDAC7 amino-terminal peptides reproducibly reduced activation by Runx2, although only the repression produced by HDAC7 amino acids 1–351 reached the level of statistical significance (Fig. 5A). Consistent with the observation that the 472–912 fragment did not repress Runx2, an HDAC7 protein containing an inactivating point mutation (H690F) that abolishes in vitro deacetylase activity(31) repressed Runx2 transcriptional activity comparably to wildtype HDAC7 (Fig. 5B). Thus, the amino-terminal portion of HDAC7 is necessary and sufficient for repression of Runx2.

Figure FIG. 5..

HDAC7 inhibits Runx2's transcriptional activity through a deacetylase-independent mechanism. (A) The HDAC7 N terminus inhibits activation of p6OSE2-luc by Runx2. C2C12 cells were co-transfected with p6OSE2-luc and pRL-null reporter plasmids and expression plasmids for HA-Runx2 and HDAC7-FLAG peptides. None of the HDAC7 peptides significantly affected the reporters in the absence of Runx2 (data not shown). Firefly luciferase was normalized to Renilla luciferase and is shown relative to activity from pcDNA3 alone. ap ≤ 0.05; bp ≤ 0.01 vs. Runx2 + pcDNA3 by Student's t-test. (B) The HDAC7 deacetylase catalytic domain function is dispensable for inhibition of Runx2 activity. Assays were performed as in A using either wildtype HDAC7 or HDAC7 with a single point mutation (H690F). ap ≤ 0.05; bp ≤ 0.01 vs. Runx2 by Student's t-test. (C and D) Repression by HDAC7 is insensitive to HDAC inhibitors. (C) Repression of Runx2 activity by HDAC7 is not inhibited by TSA. Transfections were performed as in A, and the cells were treated with either 20 nM TSA or the DMSO vehicle for 16–20 h before harvest. The data series with and without TSA was analyzed independently, because TSA treatment caused a significant increase in the activity of the pRL-null reporter. ap ≤ 0.01, bp ≤ 0.001 by Student's t-test. (D) Increased levels of acetylated histone H3 after TSA treatment. C2C12 cells were treated with TSA in parallel to the experiment shown in C. Protein extracts were subjected to SDS-PAGE and immunoblotting against total and acetylated histone H3.

To further characterize the mechanism whereby HDAC7 inhibits Runx2, we measured repression of Runx2 activity by HDAC7 in C2C12 cells treated with the broad-spectrum HDAC inhibitor TSA. As previously reported,(40) activation of the reporter by Runx2 was enhanced by TSA, presumably reflecting TSA relief of endogenous HDACs (Fig. 5C). However, TSA did not reduce the ability of HDAC7 to repress activation by Runx2, indicating that HDAC7 represses Runx2 by a mechanism that does not depend on deacetylation. Similar results were obtained with the HDAC inhibitor valproic acid (data not shown). To verify the activity of the TSA in these experiments, parallel cultures were treated with TSA or the DMSO vehicle and subjected to immunoblotting against total histone H3 and acetylated histone H3. Figure 5D shows that overnight treatment with TSA increased the level of acetylated histone H3, consistent with reduced HDAC activity in the TSA-treated cells. We conclude that HDAC7 represses Runx2 through a mechanism that does not require deacetylase catalytic activity.

HDAC7 suppression accelerates osteoblast maturation

In light of Runx2's importance during osteoblast maturation, we hypothesized that reducing HDAC7 levels will enhance Runx2 activity and osteoblast maturation. To test this hypothesis, we generated a set of retroviral expression vectors that encode short hairpin RNAs (shRNAs) against HDAC7 or a control shRNA. We stably transduced C2C12 multipotent progenitor cells with the HDAC7 and control shRNA retroviruses. Cells transduced with HDAC7 shRNA expressed ∼40% less HDAC7 mRNA relative to control shRNA cells (data not shown) and showed a strong reduction in HDAC7 protein expression (Fig. 6A). We used these cells for our subsequent osteoblast differentiation assays.

Figure FIG. 6..

Accelerated osteoblast differentiation in HDAC7-suppressed C2C12 cells. (A) HDAC7 protein is reduced in HDAC7 shRNA-containing C2C12 cells. Cell lysates from control shRNA and HDAC7 shRNA lines were immunoblotted against HDAC7 or actin as a loading control. (B) Alkaline phosphatase activity was measured after 0–4 days in differentiation media containing 300 ng/ml BMP2. Alkaline phosphatase activity was normalized to protein content and is shown relative to the control shRNA cells at day 0. Ap ≤ 0.02, Bp ≤ 0.05 by Student's t-test. (C) Expression of osteoblast markers is enhanced in HDAC7-suppressed cells. Real-time PCR was used to compare mRNA expression in HDAC7-suppressed (dark bars) and control shRNA expressing (white bars) C2C12 cells treated with 300 ng/ml BMP2 for 3 and 7 days.

BMP2 induces C2C12 cells to undergo a program of osteoblast differentiation.(46) We measured alkaline phosphatase activity as an initial measure of osteoblast maturation in control and HDAC7-suppressed cells. After being cultured for up to 4 days in BMP2-containing differentiation medium, the HDAC7-suppressed cells exhibited significantly increased alkaline phosphatase activity compared with the control cells (Fig. 6B). Real-time PCR was used to examine the relative mRNA expression of various genes expressed in the osteoblast lineage after 3 and 7 days in differentiation medium (Fig. 6C). Expression of osteoblast genes alkaline phosphatase (ALP), osteocalcin (OC), osteopontin (OP), Runx2, osterix, BSP, type I collagen a1 (COL1a1), and estrogen receptor (ER)-α were accelerated or enhanced at one or both time points, whereas expression of LRP5 was not significantly affected by HDAC7 suppression cells. The osteoclast inductive signal RANKL was elevated whereas osteoprotegerin (OPG) was reduced. These data establish that many markers of osteoblast maturation are expressed earlier and/or at higher levels in HDAC7-suppressed cells.

BMP2 decreases HDAC7 nuclear localization

Pro-differentiation signals relieve MEF2 from repression by HDAC4, 5, and 7 in myoblast precursors and relieve Runx2 from repression by HDAC4 and 5 in osteoblast precursors by promoting nuclear export of the HDACs.(12,24,47) We thus hypothesized that the osteogenic factor BMP2 might affect HDAC7 localization in a similar manner during osteogenesis. Within 5 min, BMP2 reduced the proportion of HDAC7-FLAG (Fig. 7A) and endogenous HDAC7 (Fig. 7B) in the nucleus of C2C12 cells. This effect was transient because HDAC7 localization returned to basal conditions after 30 min. Densitometry of the bands in Fig. 7B indicates a reduction in nuclear HDAC7 levels of ∼21%. These data show that BMP2 transiently decreases the level of HDAC7 in the nucleus of osteogenic cells.

Figure FIG. 7..

BMP2 causes rapid decrease in amount of HDAC7 in the nucleus. (A) C2C12 cells were transfected with HDAC7-FLAG, treated with 300 ng/ml BMP2 for the indicated time, and fixed and processed for immunofluorescence against FLAG. The predominant localization of FLAG immunofluorescence in each transfected cell was scored with the investigator blind to the experimental treatment of each slide. The experiment was repeated three times with similar results. Data from one representative experiment are presented. (B) BMP2 decreases the level of endogenous HDAC7 in the nucleus. C2C12 cells were treated with 300 ng/ml BMP2 for 15 min before being separated into cytoplasmic “C” and nuclear “N” fractions. Akt and Histone H3 serve as markers for the cytoplasmic and nuclear fractions.


This study establishes HDAC7 as a transcriptional co-repressor of Runx2 and as a regulator of osteoblast maturation. Several recent studies found that both class I and class II histone deacetylases repress Runx2 activity.(9–12) HDAC7 differs from these other HDACs because its catalytic domain is not required for repression. Thus, HDAC7 represses Runx2 by a mechanism distinct from the other HDACs. Data within this report also indicate that only certain class II HDACs block Runx2 transcriptional activation, because HDAC7, but not HDAC9, HDAC10, HDAC11, or HDRP, inhibited Runx2.

Reduced expression of HDAC7 resulted in increased alkaline phosphatase activity and altered expression of numerous genes in C2C12 cells induced to undergo osteoblast-like differentiation by BMP2. Collectively, the panel of markers examined suggests that reduction of HDAC7 levels accelerates or enhances osteoblast development. Genes involved various stages of osteoblast development such as osteoblast specification, extracellular matrix maturation, and mineralization are affected. Interestingly RANKL and ERα were increased, and OPG was downregulated in the HDAC7-suppressed cells, suggesting that HDAC7 may regulate bone remodeling and signaling between osteoblasts and osteoclasts in vivo. It is important to recognize that it is possible and indeed likely that HDAC7 regulates the activity of several transcription factors, in addition to Runx2, during bone formation. It remains unclear whether altered expression of any particular gene in HDAC7-suppressed cells reflects regulation by HDAC7 directly, indirectly, or perhaps by a combination of both, although the detection by ChIP of HDAC7 at Runx2-responsive elements in the osteocalcin, BSP, osteopontin, and MMP13 promoters indicates that these genes are directly regulated by Runx2–HDAC7 complexes. The significance of HDAC7 to skeletogenesis in vivo has not been determined because the mouse knockout of HDAC7 results in lethality at embryonic day 9.5, before skeletal formation, from deregulated expression of MMP10 in the vascular endothelium.(48) Thus, a conditional knockout will be required to examine the consequences of loss of HDAC7 expression in vivo. Collectively, these data indicate that HDAC7 is an important regulator of osteoblast function.

The onset of Runx2 expression in osseous cells does not directly correlate to the expression of Runx2 target genes,(49,50) suggesting that their transcription is subject to additional levels of regulation. Our observations with HDAC7 suggest a scenario where Runx2 expression “primes” osteoblast genes as competent for expression but that HDAC7 directly participates in repressing their expression until the appropriate time and place. BMP2 transiently decreases HDAC7 nuclear localization, suggesting one mechanism whereby HDAC7-mediated repression of Runx2, and likely that of other transcription factors, is relieved during osteoblast maturation. Similar scenarios are seen the muscle lineage, where myogenic stimuli relieve MEF2 from class II HDAC–mediated repression and allow differentiation to proceed.(24,47)

The nuclear-cytoplasmic distribution of class IIa HDACs is regulated through phosphorylation of the HDAC amino terminus by kinases including the calcium/calmodulin protein kinase (CaMK) II(27,29,30,51,52) and protein kinase D (PKD).(53–55) The transient redistribution of HDAC7 elicited by BMP2 suggests that HDAC7's repressive activity on Runx2 may be highly dynamic, perhaps allowing HDAC7 to precisely modulate the expression of Runx2 target genes. The mechanism by which BMP2 reduces HDAC7 nuclear localization remains to be determined. It will also be important to learn if other osteogenic stimuli affect the subcellular distribution of HDAC7.

HDAC7's repressive activity mapped to the noncatalytic amino terminus. Experiments with a series of smaller amino-terminal peptides showed that multiple regions of the HDAC7 amino terminus participate in binding Runx2. Because this region co-immunoprecipitated in vivo with Runx2 but exhibited limited direct interaction in vitro, it is likely that HDAC7's in vivo association with Runx2 and its repressive activity are mediated through other proteins as part of a larger complex with additional co-repressor proteins. Several studies showed that deacetylation of histone substrates by class II HDACs is dependent on recruitment of class I deacetylases, HDAC1 and HDAC3, which catalyze the deacetylation reaction.(18,19,31–33) Significantly, our results with TSA and valproic acid argue that HDAC7 represses Runx2 through a deacetylation-independent mechanism. This observation distinguishes HDAC7 from HDACs 3, 4, and 5, which repress Runx2 through TSA-sensitive mechanisms.(9,12)

The amino termini of class IIa HDACs repress transcription by at least six deacetylase-independent mechanisms that are not fully understood. For example, HDRP, a naturally occurring variant of HDAC9, lacks the deacetylase catalytic domain yet is a strong co-repressor of MEF2-mediated transcription.(52,56,57) Second, caspase-3 cleaves HDAC4 into an amino-terminal fragment that translocates to the nucleus and strongly represses MEF2.(58,59) TSA did not block the repressive activity of the HDAC4 N-terminal cleavage product and only partially relieved repression by full-length HDAC4.(59) Third, the amino termini of HDAC4, HDAC5, HDAC9, and mouse HDAC7, repress MEF2 through a conserved motif that recruits the CtBP co-repressor protein.(24,34) However, it is unclear whether the human HDAC7 protein used in our studies can bind CtBP as it lacks much of the conserved CtBP binding motif. Fourth, HDAC4, HDAC5 and HDRP associate with HP1, an adaptor protein that silences chromatin by recruiting histone methyltransferases.(35) Fifth, the amino terminus of HDAC4 inhibits MEF2 by sequestering it into distinct bodies within the nucleus.(36) HDAC4 was reported to inhibit DNA-binding by Runx2.(10) We have not noted an overt redistribution of Runx2 immunofluorescence caused by HDAC7 overexpression, but it is possible that HDAC7 sequesters Runx2 into inactive nuclear bodies in an analogous fashion. We consider it unlikely that HDAC7 functions by impairing DNA binding by Runx2, because it does not interact with the DNA-binding domain of Runx2. Finally, HDAC4 and HDAC5 can directly deacetylate the Runx2 protein, targeting it for ubiquitin-mediated proteolysis.(12) HDAC7 does not affect Runx2 stability in our hands (data not shown), thus we feel that is an unlikely mechanism. We have not yet examined the potential roles of other co-repressors in HDAC7's repression of Runx2. The ability of multiple subregions of the HDAC7 amino terminus to interact with Runx2 in vivo and to weakly inhibit transcription may indicate that multiple repressive mechanisms, perhaps each involving different binding partners, contribute to HDAC7's full repressive capability.

Runx2 and HDAC7 were each previously found in nuclear matrix-associated foci,(18,31,45,60) but it was not known whether the foci of Runx2 and HDAC7 proteins represented overlapping or distinct populations. Our data show that Runx2 and HDAC7 co-localize in subnuclear foci. Transcriptional regulation by Runx2 depends on its assembly into foci that associate with the nuclear matrix. Mice carrying a Runx2 allele lacking the nuclear matrix targeting signal exhibit phenotypes indistinguishable from null animals,(61) and expression of Runx2 proteins containing a single point mutation in the NMTS can inhibit the osteolytic and invasive properties of breast cancer cells.(62) Likewise, HDAC4, HDAC5, and HDAC7 are found in matrix-associated bodies known as MAD-bodies, for matrix-associated deacetylase bodies.(31,32) Whether the HDAC7-Runx2 foci are indeed associated with the nuclear matrix, and the biological significance of these interactions remain intriguing questions.

This study establishes HDAC7 as a novel Runx2 co-repressor. Understanding the mechanisms by which HDAC7 and other HDACs repress Runx2 function, their significance in the context of larger protein complexes at native gene promoters and the regulatory inputs that coordinate their activity will be significant areas for future studies to increase our understanding of skeletal development, maintenance, and pathological states.


The authors thank Xiaodong Li, Luke Hoeppner, Ulrike Mödder, Matthew Drake, and Sundeep Khosla for technical assistance and advice. We acknowledge the use of a confocal microscope made available through an NCRR Shared Instrumentation Grant 1 S10 RR16851. This study was supported by NIH RO1 Grants AR48147 and AR050074, NIH Institutional Training Grants AR050938 and CA09138, and the Minnesota Medical Foundation.