General Transcription Factor IIA-γ Increases Osteoblast-specific Osteocalcin Gene Expression via Activating Transcription Factor 4 and Runt-related Transcription Factor 2*

ATF4 (activating transcription factor 4) is an osteoblast-enriched transcription factor that regulates terminal osteoblast differentiation and bone formation. ATF4 knock-out mice have reduced bone mass (severe osteoporosis) throughout life. Runx2 (runt-related transcription factor 2) is a runt domain-containing transcription factor that is essential for bone formation during embryogenesis and postnatal life. In this study, we identified general transcription factor IIAγ (TFIIAγ) as a Runx2-interacting factor in a yeast two-hybrid screen. Immunoprecipitation assays confirmed that TFIIAγ interacts with Runx2 in osteoblasts and when coexpressed in COS-7 cells or using purified glutathione S-transferase fusion proteins. Chromatin immunoprecipitation assay of MC3T3-E1 (clone MC-4) preosteoblast cells showed that in intact cells TFIIAγ is recruited to the region of the osteocalcin promoter previously shown to bind Runx2 and ATF4. A small region of Runx2 (amino acids 258–286) was found to be required for TFIIAγ binding. Although TFIIAγ interacts with Runx2, it does not activate Runx2. Instead, TFIIAγ binds to and activates ATF4. Furthermore, TFIIAγ together with ATF4 and Runx2 stimulates osteocalcin promoter activity and endogenous mRNA expression. Small interfering RNA silencing of TFIIAγ markedly reduces levels of endogenous ATF4 protein and Ocn mRNA in osteoblastic cells. Overexpression of TFIIAγ increases levels of ATF4 protein. Finally, TFIIAγ significantly prevents ATF4 degradation. This study shows that a general transcription factor, TFIIAγ, facilitates osteoblast-specific gene expression through interactions with two important bone transcription factors ATF4 and Runx2.

Skeletal integrity requires a balance between bone-forming cells (osteoblasts) and bone-resorbing cells or osteoclasts. Imbalance between bone formation and resorption results in metabolic bone diseases such as osteoporosis. Multipotential mesenchymal cells proliferate and differentiate into osteoblasts that synthesize and deposit the mineralizing extracellular matrix of bone. Osteoblast activity is regulated by a number of growth factors and hormones, including bone morphogenetic proteins, insulin-like growth factor 1, basic fibroblast growth factor 2, parathyroid hormone, tumor necrosis factor-␣, and extracellular matrix signals (1)(2)(3)(4)(5)(6)(7)(8)(9). Runx2 is a runt domain-containing transcription factor identified as a transcriptional activator of osteoblast differentiation and the master gene for bone development in vitro and in vivo (10 -14). Runx2 knock-out mice die at birth and completely lack both skeletal ossification and mature osteoblasts (10,12). Runx2 haplo-insufficiency causes the skeletal disorder, cleidocranial dysplasia, a disease characterized by defective endochondral and intramembranous bone formation. Runx2 is expressed in mesenchymal condensations during early development at E11.5 and acts as an osteoblast differentiation factor (13).
One of the most striking characteristics of ATF4 protein is its very short half-life (30 -60 min) in many cell types (26). ATF4 is rapidly degraded via a ubiquitin/proteasomal pathway. This degradation requires the presence of the serine residue 219 in the context of DSGXXXS within the ATF4 molecule and its phosphorylation by an unknown kinase. This phosphorylation was shown to be required for subsequent recognition by the SCF ␤TrCP and degradation by the 26 S proteasome (27). Although Atf4 mRNA is ubiquitously expressed, ATF4 protein preferentially accumulates in osteoblasts (28). This accumulation is explained by a selective reduction of proteasomal degradation in osteoblasts. Indeed, inhibition of the ubiquitin/proteasomal pathway by MG115, which blocks the N-terminal threonine in the active site of ␤-subunit of 26 S proteasomal complex (29,30), led to ATF4 accumulation and induced Ocn mRNA expression in non-osteoblastic cells (28). These observations suggest that modulation of ATF4 stability constitutes an important step to control its protein level and activity and, ultimately, osteoblast-specific gene expression and bone formation.
Transcription factor IIA (TFIIA) is a general transcription factor consisting of three subunits designated TFIIA␣, TFIIA␤, and TFIIA␥ (31). TFIIA interacts with and stabilizes TFIID (also known as TBP, TATA box-binding protein) to DNA and activates transcription (32,33). Although TFIIA was classified as a general transcription factor when it was first identified, more and more evidence shows that this elusive factor may play an important role in the regulation of tissue-specific gene expression via interactions with tissue-or cell type-specific transcription factors (34 -36).
The Ocn promoter has been the major paradigm for unraveling the mechanisms mediating osteoblast-specific gene expression and defining a number of key transcription factors or cofactors (13,14,(23)(24)(25)(37)(38)(39)(40)(41). However, very few studies have focused on how tissue-specific transcription factors interface with general transcriptional initiation factors in osteoblasts. In this study, by using a combination of a yeast twohybrid system and pulldown assays as well as functional assays, we show that TFIIA␥, the smallest subunit (12 kDa) of TFIIA (42), interacts with both Runx2 and ATF4. TFIIA␥ delays ATF4 protein degradation and increases its activity. Together with ATF4 and Runx2, TFIIA␥ enhances osteoblast-specific Ocn gene expression.

EXPERIMENTAL PROCEDURES
Reagents-Tissue culture media were purchased from Invitrogen and fetal bovine serum from HyClone (Logan, UT). Other reagents were obtained from the following sources: antibodies against TFIIA-␣, TFIIA-␥, ATF4, Runx2, and horseradish peroxidase-conjugated mouse or goat IgG from Santa Cruz Biotechnology (Santa Cruz, CA), mouse monoclonal antibody against ␤-actin from Sigma, and GST antibody from Amersham Biosciences. All other chemicals were of analytical grade.
Cell Cultures-Mouse MC3T3-E1 subclone 4 (MC-4) cells were described previously (43,44) and maintained in ascorbic acid-free ␣-modified Eagle's medium, 10% fetal bovine serum (FBS), and 1% penicillin/streptomycin and were not used beyond passage 15 Yeast Two-hybrid Analysis-A yeast pLexA two-hybrid system (Clontech) was used to identify proteins that bind to mouse Runx2. A cDNA fragment encoding the aa-263-351 region of Runx2 was subcloned into the BamHI/XhoI sites of pLexA, creating an in-frame fusion with the DNA binding domain of the LexA gene that is controlled by the strong yeast ADH1 promoter. The resultant plasmid pLexA-Runx2 (aa 263-351) was then transformed into a yeast reporter strain (YM4271), and the transformed cells (1 ϫ 10 9 ) were mated for 24 h with cells (2.5 ϫ 10 8 ) of a pretransformed two-hybrid library made from human brain cDNA. The resultant mating mixture was spread on 20 ϫ 10-cm plates to select for expression of the LEU2 and lacZ reporter genes. Approximately 2 ϫ 10 6 colonies were screened. Sixty four positive colonies were isolated. The prey plasmids were extracted from the positive colonies and the cDNA inserts in the plasmids were amplified by PCR and sequenced. Of the 64 positive colonies, 5 are the full-length TFIIA␥ cDNAs, and the rest contained 16 different cDNAs.
DNA Constructs and Transfection-p657mOG2-luc, p657m-OG2OSE1mt-luc, p657mOG2OSE2mt-luc, p657mOG2-(OSE1 ϩ 2)mt-luc, p4OSE1-luc, p4OSE1mt-luc, p6OSE2-luc, p6OSE2mt-luc, pCMV/␤-galactosidase, pCMV/ATF4, pCMV/ Runx2, pCMV/FLAG-Runx2 and its deletion mutants (aa 1-330, aa 1-286, and aa-258), GST-Runx2 and GST-ATF4 fusion protein expression vectors were described previously (1,13,23,25,45). The full-length cDNA of human TFIIA-␥ was cloned by an RT-PCR strategy using total RNA from human Saos2 osteoblastic cells as a template and specific primers (forward, 5Ј-ATG GCA TAT CAG TTA TAC AGA AA-3Ј, and reverse, 5Ј-TTC TGT AGT ATT GGA GCC AGT A-3Ј). Digested PCR products were purified and subcloned into the NotI/BamHI sites of the pFLAG-5a expression vector (Sigma). Addition of a C-terminal FLAG sequence into the TFIIA-␥ cDNA facilitates monitoring of expression levels and immunoprecipitation using M2 antibody (Sigma). GST-TFIIA␥ fusion protein expression plasmid was constructed by subcloning the full-length TFIIA␥ cDNA into the glutathione S-transferase gene fusion vector pGEX-4T-1 (Amersham Biosciences) in correct reading frame. The accuracy of DNA sequences was verified by automatic sequencing. The size of expressed proteins was confirmed by Western blot analysis using specific antibodies. For expression and functional studies, cells were plated on 35-mm dishes at a density of 5 ϫ 10 4 cells/cm 2 . After 24 h, cells were transfected with the indicated plasmid DNAs (0.01 g of pRL-SV40, 0.25 g of test luciferase reporter, and 1.0 g of expression plasmids balanced as necessary with ␤-galactosidase expression plasmid such that the total DNA was constant) and Lipofectamine 2000 (Invitrogen) according to manufacturer's instructions. After 36 h, whole cell extracts were prepared and used for Western blot analysis or dual luciferase assay using the dual luciferase assay kit (Promega, Madison, WI) on a Veritas TM microplate luminometer (Turner Biosystem, Inc., Sunnyvale, CA). Firefly luciferase activity was normalized to Renilla luciferase activity for transfection efficiency.
RNA Isolation and Reverse Transcription (RT)-Total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. RT was performed using 2 g of denatured RNA and 100 pmol of random hexamers (Applied Biosystem, Foster, CA) in a total volume of 25 l containing 12.5 units of MultiScribe reverse transcriptase (Applied Biosystem, Foster, CA) according to the manufacturer's instructions.
Regular PCR-Regular PCR was performed on a 2720 Thermal Cycler (Applied Biosystem, Foster, CA), using 2.5 l of the cDNA (equivalent to 0.2 g of RNA) and AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA) in a 25-l reaction according to the manufacturer's instructions. The DNA sequences of primers used for PCR were as follows: mouse/rat TFIIA␥, 5Ј-ATG GCA TAT CAG TTA TAC AGA AAT ACA-3Ј (forward), 5Ј-GGT ATT TTT ACC ATC ACA GGC T-3Ј(reverse); mouse/rat Atf4, 5Ј-ATG GCT TGG CCA GTG CCT CAG A-3Ј (forward), 5Ј-GCT CTG GAG TGG AAG ACA GAA C-3Ј (reverse); mouse/rat Hprt, 5Ј-GTT GAG AGA TCA TCT CCA CC-3Ј (forward), 5Ј-AGC GAT GAT GAA CCA GGT TA-3Ј (reverse). For all primers the amplification was performed as follows: initial denaturation at 95°C for 30s followed by 31 cycles of 95°C for 15 s, 60°C for 30 s, 72°C for 30 s and extension at 72°C for 7 min. The amplified PCR products were run on a 1.2% agarose gel and visualized by ethidium bromide staining.
Immunoprecipitation-GST, GST-TFIIA␥, GST-ATF4, and GST-Runx2 fusion proteins were purified using the Bulk GST purification module kit (Amersham Biosciences) according the manufacturer's instructions. Whole cell extracts (500 g), nuclear extracts (200 g), or GST fusion proteins (1.0 g) were pre-cleaned twice with 50 l of protein A/G-agarose beads (Stratagene, La Jolla, CA) for 30 min followed by pelleting of beads. The protein A/G-agarose beads were blocked with 10 g/ml bovine serum albumin in 1ϫ phosphate-buffered saline for 1 h before use to reduce nonspecific binding of proteins. Five g of respective antibody was added and incubated for 2 h at 4°C with gentle rocking. The immune complexes were collected by addition of 30 l of protein A/G-agarose beads and incubation for 1 h at 4°C followed by centrifugation. Precipitates were washed five times with 1ϫ washing buffer (20 mM HEPES, pH 7.6, 50 mM KCl, 1 mM dithiothreitol, 0.25% Nonidet P-40, 5 mM NaF, 1 mM EGTA, 5 mM MgCl 2 , 0.25 mM phenylmethylsulfonyl fluoride), and the immunoprecipitated complexes were suspended in SDS sample buffer and analyzed by SDS-PAGE followed by Western blot analysis using the indicated antibodies.
ChIP Assays-ChIP assays were performed as described previously (41) using a protocol kindly provided by Dr. Dwight Towler (Washington University) (47). After sonication, the amount of chromatin was quantified using the PicoGreen double-stranded DNA quantitation assay (Molecular Probes) according to the manufacturer's instructions. The equivalent of 10 g of DNA was used as starting material (input) in each ChIP reaction with 2 g of the appropriate antibody (TFIIA␥, or control rabbit IgG). Fractions of the purified ChIP DNA (5%) or inputs (0.02-0.05%) were used for PCR analysis. The reaction was performed with AmpliTaq Gold DNA polymerase (Applied Biosystems) for 35 cycles of 60 s at 95°C, 90 s at 58°C, and 120 s at 68°C. PCR primer pairs were generated to detect DNA segments located near the Runx2-binding site at Ϫ137/Ϫ131 (primers P1 and P2), ATF4-binding site at Ϫ55/Ϫ48 (primers P3 and P4) in mouse osteocalcin gene 2 (mOG2) proximal promoter, or the Runx2-binding site located between Ϫ370 and Ϫ42 in the proximal mouse Runx2 promoter region (primers P5 and P6) (48), and the mOG2 gene region (ϩ177/ϩ311) (primers P7 and P8) (see Fig. 2A and Table 1). The PCR products were separated on 3% agarose gels and visualized with ultraviolet light. All ChIP assays were repeated at least three times.
siRNA-ROS17/2.8 osteoblast-like cells, which contain high levels of TFIIA␥ protein, were transfected with mouse TFIIA␥ siRNA kit (Santa Cruz Biotechnology) or negative control siRNA(low GC, catalog number 12935-200, Invitrogen) using Lipofectamine 2000 (Invitrogen) according the manufacturer's instruction. After 36 h, total RNA was harvested for quantitative real time RT-PCR analysis for TFIIA␥, Ocn, Bsp, Opn (osteopontin), and Atf4 mRNAs. A second set of mouse TFIIA␥ siRNAs (sense, AUG ACA ACA CUG UGC UAU AUU; antisense, UAU AGC ACA GUG UUG UCA UUU) was designed in the project laboratory and used to confirm the results using the first set of TFIIA␥ siRNA.
Statistical Analysis-Results were expressed as means Ϯ S.D. Students' t test was used to test for differences between two groups. Differences with a p Ͻ 0.05 was considered as statistically significant.

RESULTS
TFIIA␥ Interacts with Runx2 and ATF4-A yeast pLexA two-hybrid system (Clontech) was used to identify proteins that bind to mouse Runx2. cDNA fragments encoding several C-terminal regions of Runx2 were subcloned into the BamHI/ XhoI sites of pLexA, creating in-frame fusions with the DNA binding domain of the LexA gene that is controlled by the strong yeast ADH1 promoter. Preliminary experiments using relatively larger regions of Runx2 (aa 232-391, aa 232-428, and aa 232-517) as baits were not successful because of their ability to autoactivate the lacZ reporter gene in yeast. In contrast, by using the aa 263-351 region of Runx2 as a bait, we identified TFIIA␥, a general transcriptional factor involved in the initiation step of eukaryotic transcription, as a Runx2-interacting factor. A diagram and a picture of a positive colony are shown in Fig. S1. To verify the TFIIA␥-Runx2 interaction identified by yeast two-hybrid system, we conducted pulldown assays. COS-7 cells were transiently transfected with expression vectors for FLAG-TFIIA␥, Runx2, and ATF4 (a recently identified Runx2-interacting factor). After 36 h, whole cell extracts were prepared for immunoprecipitation (IP) assay using a TFIIA␥ antibody followed by Western blot analysis for Runx2 and ATF4. As seen in Fig. 1A (lane 2), Runx2 protein was present in a TFIIA␥ anti-FIGURE 1. Protein-protein interactions among TFIIA␥, Runx2, and ATF4. A, whole cell extracts from COS-7 cells overexpressing pFLAG-TFIIA␥, pCMV-Runx2, and pCMV-ATF4 were immunoprecipitated (IP) with normal IgG (lane 1) or TFIIA␥ antibody (lane 2) followed by Western blot (WB) analysis using Runx2 or ATF4 antibodies. In reciprocal IPs, the same extracts were immunoprecipitated with normal IgG (lanes 3 and 5), Runx2 antibody (lane 4), or ATF4 antibody (lane 6) followed by WB using M2 antibody. B, nuclear extracts from ROS17/2.8 cells were immunoprecipitated with normal IgG (lane 1) or TFIIA␥ antibody (lane 2) followed by WB using Runx2, ATF4, or Fra-1 antibodies. In reciprocal IPs, the same extracts were immunoprecipitated with normal IgG (lanes 3, 5 and 7), Runx2 antibody (lane 4), ATF4 antibody (lane 6), or Fra-1 antibody (lane 8) followed by WB using TFIIA␥ antibody. C, mixture of purified GST-TFIIA␥ and GST-Runx2 was immunoprecipitated by TFIIA␥ antibody followed by WB for Runx2 (lane 1). A mixture of purified GST-TFIIA␥ and GST-ATF4 was immunoprecipitated by TFIIA␥ antibody followed by WB for ATF4 (lane 2). A mixture of purified GST and GST-Runx2 was immunoprecipitated by Runx2 antibody followed by WB for GST (lane 3). A mixture of purified GST and GST-ATF4 was immunoprecipitated by ATF4 antibody followed by WB for GST (lane 4). In reciprocal IPs, a mixture of purified GST-TFIIA␥ and GST-Runx2 was immunoprecipitated by normal IgG (lane 5) or Runx2 antibody (lane 6) followed by WB for TFIIA␥. A mixture of purified GST-TFIIA␥ and GST-ATF4 was immunoprecipitated by ATF4 antibody (lane 7) followed by WB for TFIIA␥. D, nuclear extracts from ROS17/2.8 cells were mixed with equal amount of nuclear extracts from COS-7 cells overexpressing FLAG-Runx2(wt), FLAG-Runx2 (aa 1-330), FLAG-Runx2 (aa 1-286), and FLAG-Runx2 (aa 1-258), and immunoprecipitated with TFIIA␥ antibody followed by WB for Runx2 (M2 antibody). Experiments were repeated 2-3 times, and qualitatively identical results were obtained. FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 body immunoprecipitate. Interestingly, anti-TFIIA␥ antibody also immunoprecipitated ATF4. Reciprocal IPs showed that both Runx2 and ATF4 antibodies immunoprecipitated the FLAG-tagged TFIIA␥ (Fig. 1A, lanes 4 and 6). To determine whether TFIIA␥ can interact with Runx2 and ATF4 in osteoblasts, nuclear extracts from ROS17/2.8 cells that express high levels of Runx2, ATF4, and TFIIA␥ were immunoprecipitated with anti-TFIIA␥ antibody followed by Western blot analysis for Runx2, ATF4, or Fra-1 (a member of AP1 family). Results show that both Runx2 and ATF4 but not Fra-1 proteins were present in anti-TFIIA␥ immunoprecipitates (Fig. 1B, lane 2). Reciprocal IPs showed that antibodies against Runx2 or ATF4 but not Fra-1 immunoprecipitated TFIIA␥ in ROS17/2.8 cells (Fig. 1B, lanes 4, 6, and 8). Normal control IgG failed to significantly pull down Runx2, ATF4, or TFIIA␥ in either COS-7 cells or osteoblasts. Taken together, these studies confirm that TFIIA␥ interacts with Runx2 and ATF4 in osteoblasts or when coexpressed in COS-7 cells.

TFIIA␥ Interacts with ATF4 and Runx2
Although Runx2 and ATF4 interact in osteoblasts, IP assays using purified GST fusion proteins failed to show a direct phys-ical interaction between ATF4 and Runx2 (25), suggesting that accessory factors may be involved in their interactions. To determine whether TFIIA␥ can directly interact with Runx2 or ATF4 in the absence of other nuclear proteins, we mixed GST or GST-TFIIA␥ with GST-ATF4 or GST-Runx2 fusion proteins purified from Escherichia coli, followed by IP and Western blot analysis. As shown in Fig. 1C, both GST-Runx2 and GST-ATF4 proteins mixed with GST-TFIIA␥ were immunoprecipitated by anti-TFIIA␥ antibody (lanes 1 and 2). Anti-Runx2 or anti-ATF4 antibody was unable to immunoprecipitate GST protein mixed with GST-Runx2 (Fig. 1C, lane 3) or GST-ATF4 (lane 4). Reciprocal IPs show that GST-TFIIA␥ was immunoprecipitated by both anti-Runx2 or anti-ATF4 antibodies (Fig. 1C, lanes  6 and 7) but not by normal control IgG (lane 5). These results demonstrate that TFIIA␥ directly binds to both Runx2 and ATF4. As a first step to identify the TFIIA␥-binding domain, FLAG-Runx2 deletion mutant expression vectors (wild type aa 1-528, aa 1-330, aa 1-286, and aa 1-258) were transfected into COS-7 cells because of the high transfection efficiency. Nuclear extracts were prepared 36 h later, mixed with equal amounts of nuclear extracts of ROS17/2.8 (which contain large amounts of endogenous TFIIA␥), and immunoprecipitated using anti-TFIIA␥ antibody followed by Western blot analysis for Runx2 (M2 antibody). As shown in Fig. 1D, deletion of Runx2 from aa 528 to aa 286 did not reduce TFIIA␥ binding. However, further deletion from aa 286 to aa 258 completely abrogated TFIIA␥-Runx2 complex formation. These data clearly demonstrate the following: (i) endogenous TFIIA␥ can interact with overexpressed FLAG-Runx2 proteins in vitro; and (ii) the aa 258 -286 region of Runx2 is required for TFIIA␥ binding. Interestingly, this same region is required for ATF4-Runx2 interactions (25).
To determine whether, in intact cells, TFIIA␥ is associated with the endogenous osteocalcin gene 2 (mOG2) promoter region that has been shown to bind Runx2 and ATF4, we performed the chromatin immunoprecipitation (ChIP) assay using MC3T3-E1 (clone MC-4) preosteoblast cells. After shearing, soluble chromatin was immunoprecipitated with either an antibody against TFIIA␥ or control IgG. The positions and sequences of primers used for PCR analysis of ChIP DNAs are shown in Fig. 2A and Table 1. As shown in Fig. 2B, the PCR bands amplified with primers P1/P2 and P3/P4 and corresponding to ChIP DNAs immunoprecipitated with TFIIA␥ antibody revealed that TFIIA␥ specifically interacts with chromatin fragments of the proximal mOG2 promoter that contain Runx2-or ATF4-binding sites. Furthermore, TFIIA␥ antibody B, MC-4 cells were seeded at a density of 50,000 cells/cm 2 in 35-mm dishes, cultured in 10% FBS medium overnight, and cross-linked with formaldehyde for ChIP assays. IPs were conducted with TFIIA␥ antibody (Ab) or normal control IgG. PCR products were run on 3% agarose gel and stained with ethidium bromide. Purified input chromatin was used to perform parallel PCRs with the respective primer pairs. Experiments were repeated three times with similar results.

Oligonucleotide name Sequence
TGTAGGCGGTCTTCA AGCCAT also immunoprecipitated a Runx2-binding site-containing chromatin fragment of the proximal Runx2 promoter (primers P5/P6). In contrast, TFIIA␥ antibody failed to immunoprecipitate a chromatin fragment of mOG2 gene that contains no Runx2-or ATF4-binding sites (primers P7/P8). Taken together, these data show that TFIIA␥ is recruited to a chromatin fragment of the mOG2 promoter that was previously demonstrated to be bound by Runx2 and ATF4 in osteoblasts (13,22).

TFIIA␥ Stimulation of Endogenous Ocn mRNA Expression and the 657-bp mOG2 Promoter Activity Is Dependent upon the Presence of ATF4 and Runx2-ATF4
is an osteoblast-enriched protein that is required for late osteoblast differentiation (i.e.

Silencing of TFIIA␥ Markedly Reduces Levels of Endogenous Ocn and Bsp mRNAs and ATF4 Protein in Osteoblasts-
To determine whether TFIIA␥ is required for the endogenous Ocn mRNA expression in osteoblasts, we knocked down the endogenous TFIIA␥ transcripts by siRNA. ROS17/2.8 osteoblast-like cells, which express high levels of TFIIA␥ and Ocn and Bsp mRNAs, were transiently transfected with TFIIA␥ siRNA reagent from Santa Cruz Biotechnology according to the manufacturer's instructions. This siRNA is a pool of three specific 20 -25-nucleotide siRNA targeting both mouse and rat TFIIA␥. As shown in Fig. 6A, quantitative real time RT-PCR analysis showed that levels of TFIIA␥ mRNA were efficiently reduced by TFIIA␥ siRNA in a dose-dependent manner. The level of Ocn mRNA was reduced greater than 50% by TFIIA␥ siRNA (p Ͻ 0.01, control versus TFIIA␥ siRNA). Interestingly, Bsp mRNA, another ATF4 downstream target gene (22), was also reduced by 50% (p Ͻ 0.01, control versus TFIIA␥ siRNA). This inhibition was specific because levels of Opn and Atf4 mRNAs were not reduced by TFIIA␥ siRNA. In contrast, as shown in Fig. 6B, levels of all these mRNAs were not reduced by the negative control siRNA (Invitrogen). Although Atf4 mRNA was not altered by TFIIA␥ siRNA, the level of endogenous ATF4 protein was significantly reduced by silencing TFIIA␥ in osteoblasts (Fig. 6C). Similar results were obtained when a different set of TFIIA␥ siRNA was used (Fig. S2).
Overexpression of TFIIA␥ Increases the Levels of ATF4 Protein-The above studies clearly demonstrated that TFIIA␥ increased ATF4-dependent transcription activity and Ocn gene expression probably by targeting ATF4 protein. To further study the mechanism of this regulation, we determined the effect of TFIIA␥ overexpression on the levels of ATF4 protein. C3H10T1/2 cells, which express undetectable level of endogenous ATF4 protein (28), were transiently transfected with 1.0 g of ATF4 expression plasmid and increasing amounts of TFIIA␥ expression plasmid (0, 0.5, 1, and 2 g). After 36 h, cells were harvested for Western blot analysis. As shown in Fig. 7A, overexpression of TFIIA␥ in C3H10T1/2 cells increased the levels of ATF4 protein in a dose-dependent manner. This increase in ATF4 protein was specific because levels of Runx2 were not altered by TFIIA␥. TFIIA␥ similarly elevated levels of ATF4 protein in COS-7 cells (Fig. 7B). Next, we determined if TFIIA␥ could increase the levels of endogenous ATF4 proteins in osteoblasts. ROS17/2.8 cells were transiently transfected with indicated amount of TFIIA␥ expression vector. Western blot analysis shows that TFIIA␥ dose-dependently increased levels of endogenous ATF4 protein in ROS17/2.8 cells (Fig. 7C). Similar results were obtained in MC-4 cells (Fig. 7D). Interestingly, overexpression of TFIIA␥ did not increase the levels of Atf4 mRNA in all these cells examined (bottom, Fig. 7, A-D). Taken collectively, TFIIA␥ markedly increased levels of ATF4 proteins in osteoblasts and non-osteoblasts.
TFIIA␥ Increases ATF4 Protein Stability-Lassot et al. (51) recently showed that acetylase p300 markedly increased the levels of ATF4 protein and ATF4-dependent transcriptional activity by inhibiting ATF4 protein degradation via a proteasomal ubiquitin pathway. As an initial step to determine whether TFIIA␥ alters ATF4 protein stability, C3H10T1/2 cells were transiently transfected with ATF4 expression vector in the presence of ␤-galactosidase, TFIIA␥, or Runx2 expression vectors. After 36 h, cells were treated with 50 g/ml of protein synthesis inhibitor cycloheximide (CHX) (i.e. to completely block de novo protein synthesis) and harvested at different time points of CHX addition (0, 0.5, 1, and 3 h) followed by Western blot analysis for ATF4 and Runx2. This technique has been widely used to study protein stability (51). As shown in Fig.  8A, in the absence of TFIIA␥ overexpression, ATF4 protein was rapidly degraded and almost undetectable on Western blot by 3 h after CHX addition, which is consistent with a previous study (51). However, overexpression of TFIIA␥ greatly delayed the degradation process with the levels of ATF4 protein only slightly reduced by 3 h after CHX addition. In contrast, levels of Runx2 protein were not affected by TFIIA␥ (Fig. 8B).

DISCUSSION
This study identifies TFIIA␥ as a bridging molecule between Runx2, ATF4, and the transcription machinery in osteoblasts. Although Runx2 and ATF4 interact in osteoblasts or when coexpressed in COS-7 cells, IPs using purified GST fusion proteins were unable to demonstrate a direct physical interaction between ATF4 and Runx2 (25). Thus, accessory factors are likely involved in bridging these two molecules. Several lines of evidence support that TFIIA␥ may be a factor linking Runx2 and ATF4. Accumulating evidence establishes that ubiquitin-proteasome pathways control osteoblast differentiation and bone formation. For example, the proteasome inhibitors epoxomicin and proteasome inhibitor-1, when administered systemically to mice, strongly stimulated bone volume and bone formation rates by greater than 70% after only 5 days of treatment (52). Although the mechanism of this regulation remains unclear, critical bone transcription factors seem to be targets for the ubiquitin-proteasomal pathway. Zhao and co-workers (52,53) recently showed that Smurf1, an E3 ubiquitin-protein isopeptide ligase, accelerated Runx2 ubiquitin-proteasomal degradation and inhibited osteoblast differentiation and bone forma- tion in vitro and in vivo. Although Atf4 mRNA is ubiquitously expressed, in most cells ATF4 proteins are rapidly degraded via the ubiquitin-proteasome pathway with a half-life of 30 -60 min. However, this degradation pathway is less active in osteoblasts, thereby allowing ATF4 accumulation (28). Indeed, inhibition of the ubiquitin/proteasomal pathway by MG115, which blocks the N-terminal threonine in the active site of ␤-subunit of 26 S proteasomal complex (29,30), led to ATF4 accumulation and induced Ocn mRNA expression in non-osteoblastic cells (28). Similarly, silencing of ␤-TrCP1, an E3 ubiquitin-protein isopeptide ligase that interacts with ATF4, by RNA interference, resulted in ATF4 accumulation and increased Ocn expression. Thus, ATF4 is a major target of the uquibitin-proteasome pathway, and modulation of ATF4 stability may play a critical role in the regulation of osteoblast-specific gene expression. Because ␤-TrCP1 is present in osteoblasts (28), other factor(s) must be present in these cells to protect ATF4 from the proteasomal degradation that occurs in other cell types. Experiments from this study show that overexpression of TFIIA␥ dose-dependently increases ATF4 protein in osteoblasts (ROS17/2.8 and MC-4 cells) and non-osteoblasts (C3H10T1/2 and COS-7 cells) without altering Atf4 mRNA. Experiments using the protein synthesis inhibitor CHX further demonstrate that TFIIA␥ greatly inhibits ATF4 degradation. TFIIA␥ siRNA decreases ATF4 stability in osteoblasts. Lassot et al. (51) recently found that ATF4 is similarly stabilized by cofactor p300, a histone acetyltransferase. p300 inhibits ATF4 ubiquitination and degradation through interaction with the ATF4 N terminus. Interestingly, this stabilization does not require either the acetyltransferase activity of p300 or the serine residue 219 in the context of DSGXXXS within ATF4 molecule that is known to be required for ATF4 degradation via the SCF ␤TrCP and the 26 S proteasome (51).
TFIIA␥ stimulation of Ocn gene transcription is dependent on the presence of both ATF4 and Runx2. As a master regulator of osteoblast differentiation, Runx2 alone is sufficient to activate expression of many osteoblast-specific genes, including Ocn and Bsp, by direct binding to their promoters (13). In contrast, although ATF4 directly binds to the OSE1 site of the mouse Ocn gene and activates OSE1, it alone is not sufficient for activation of the endogenous Ocn gene or the 657-bp mOG2 promoter which contains sufficient information for the bonespecific expression of Ocn in vivo (54). Instead, ATF4 stimulation of Ocn is dependent on the presence of Runx2 as demonstrated by our recent study (25). ATF4 interacts with Runx2 and activates Runx2-dependent transcriptional activity. A recent study shows that SATB2, a nuclear matrix protein that directly interacts with both ATF4 and Runx2, activates osteoblast differentiation and controls craniofacial patterning in vivo (55). This study shows that although TFIIA␥ interacts with Runx2, it does not directly activate Runx2. Like ATF4, TFIIA␥ alone is not sufficient to activate transcription from either the Ocn gene or the 657-bp mOG2 promoter. In fact, even TFIIA␥ and ATF4 together are not sufficient for Ocn gene expression without the presence of Runx2 (Fig. 5). However, in the presence of both ATF4 and Runx2, TFIIA␥ greatly activates Ocn gene expression.
General transcription factors were originally defined as such because they were thought to be universally required for transcription. In eukaryotic cells, initiation of transcription is a complex process, which requires RNA polymerase II and many other basal transcription factors and/or cofactors, including TFIIA, TFIIB, TFIID (TBP or TATA box-binding protein), TFIIE, TFIIF, and TFIIH (56 -59). Binding of TBP to the TATA box is the first step, which is regulated by TFIIA. TFIIA enhances transcription by interacting with TBP and stabilizing its binding to DNA (32,33). More and more evidence shows that general transcription factors play unique roles in the regulation of tissue-specific gene expression under physiological and pathological conditions. For example, the androgen receptor , via its N-terminal AF1 domain, interacts with basal transcription factors TBP and TFIIF and activates tissue-specific transcription in target tissues and cells (60). Likewise, TAFII 17 (a component of the TFIID complex), via specific protein-protein interactions with the vitamin D receptor (VDR), increases osteoclast formation from osteoclast precursors in response to 1,25-dihydroxyvitamin D 3 in patients with Paget disease (61). In osteoblasts, bone transcription factors such as Runx2 and ATF4 directly bind to specific DNA sequences in their target gene promoters (i.e. OSE2 or NMP2 and OSE1, respectively) and activate osteoblast-specific gene expression, osteoblast differentiation, and bone formation (1, 10 -14, 24, 43). Obviously, cooperative interactions between osteoblast-specific transcription factors and basal (general) transcriptional machinery are essential for achieving maximal transcription of osteoblast-specific genes. However, little is known about these interactions. Experiments from this study demonstrate that TFIIA␥, which is expressed at high level in osteoblasts, facilitates osteoblastspecific gene expression via two mechanisms. 1) TFIIA␥ stabilizes ATF4 and increases the levels of ATF4 proteins. The increased levels of ATF4 further activate Runx2 activity and Ocn transcription (25). 2) Through its ability to directly interact with both ATF4 and Runx2, TFIIA␥ could recruit these two critical bone transcription factors to the basal transcriptional machinery and greatly enhance osteoblast-specific gene expression. In support of our observation, Guo and Stein (62) showed that Yin Yang-1 (YY1) regulates vitamin D enhancement of Ocn gene transcription by interfering with interactions of the VDR with both the VDR element and TFIIB. TFIIB interacts with both VDR and YY1 (63). Likewise, Newberry et al. (64) showed that TFIIF (RAP74 and RAP30) mediates Msx2 (a homeobox transcription factor required for craniofacial development) inhibition of Ocn promoter activity. Finally, a recent study showed that TFIIB could directly bind to the transactivation domain of Osterix, another important osteoblast transcription factor (65).
TFIIA consists of three subunits designated TFIIA␣, TFIIA␤, and TFIIA␥. TFIIA␣ and TFIIA␤ are produced by a specific proteolytic cleavage of the ␣␤ polypeptide that is encoded by TFIIA-L (31,33). TFIIA␥ is the smallest subunit with a molecular mass of 12 kDa (42). Although it is encoded by a distinct gene (TFIIA␥), TFIIA␥ shares a high degree of homology with TFIIA␣ and TFIIA␤. Interestingly, TFIIA␣ activates testis-specific gene expression via interactions with a tissue-specific partner, ACT (activator of CREM in testis) and CREM (34). Likewise, TFIIA␣ enhances human T-cell lymphotropic virus type 1 gene activation through interactions with the Tax protein, a factor associated with adult enhances human T-cell lymphotropic virus type 1 (HTLV-1) (35,66). It remains to be determined whether TFIIA␣ and TFIIA␤ can also interact with ATF4 and Runx2 and similarly activate osteoblast-specific gene expression.
It should be noted that although TFIIA␥ belongs to the family of general transcription factors, its expression seems to show some tissue or cell specificity. Osteoblastic cells (MC-4 cells and ROS17/2.8), C3H10T1/2 fibroblasts, and L1 preadipocytes express high levels of TFIIA␥ proteins. In contrast, the levels of TFIIA␥ protein were undetectable in F9 teratocarcinoma cells  and COS-7 on Western blots. The meaning of this observation remains unknown.
These findings suggest that TFIIA␥ is a critical factor regulating ATF4 stability and functions as a molecular linker between ATF4 and Runx2 and the basal transcriptional machinery. TFIIA␥ may play a unique role in the regulation of osteoblast-specific gene expression and ultimately osteoblast differentiation and bone formation. A working model is proposed in Fig. 9, which summarizes the role of TFIIA␥ in osteoblast-specific mOG2 gene expression. Future study aimed at identifying factors that affect levels and activity of TFIIA␥ will allow us to address the functional significance of TFIIA␥ in osteoblast function in greater detail.