Phosphorylation of Thyroid Hormone Receptor-associated Nuclear Receptor Corepressor Holocomplex by the DNA-dependent Protein Kinase Enhances Its Histone Deacetylase Activity*

It is well documented that unliganded thyroid hormone receptor (TR) functions as a transcriptional repressor of specific cellular target genes by acting in concert with a corepressor complex harboring histone deacetylase (HDAC) activity. To fully explore the cofactors that interact with the transcriptionally repressive form of TR, we biochemically isolated a multiprotein complex that assembles on a TR·retinoid X receptor (RXR) heterodimer in HeLa nuclear extracts and identified its polypeptide components by mass spectrometry. A subset of TR·RXR-associated polypeptides included NCoR, SMRT, TBL1, and HDAC3, which represent the core components of a previously described NCoR/SMRT corepressor complex. We also identified several polypeptides that constitute a DNA-dependent protein kinase (DNA-PK) enzyme complex, a regulator of DNA repair, recombination, and transcription. These polypeptides included the catalytic subunit DNA-PKcs, the regulatory subunits Ku70 and Ku86, and the poly(ADP-ribose) polymerase 1. Density gradient fractionation and immunoprecipitation analyses provided evidence for the existence of a high molecular weight TR·RXR·corepressor holocomplex containing both NCoR/SMRT and DNA-PK complexes. Chromatin immunoprecipitation studies confirmed that unliganded TR·RXR recruits both complexes to the triiodothyronine-responsive region of growth hormone gene in vivo. Interestingly, DNA-PKcs, a member of the phosphatidylinositol 3-kinase family, was found to phosphorylate HDAC3 when the purified TR·RXR·corepressor holocomplex was incubated with ATP. This phosphorylation was accompanied by a significant enhancement of the HDAC activity of this complex. Collectively, our results indicated that DNA-PK promotes the establishment of a repressive chromatin at a TR target promoter by enhancing the HDAC activity of the receptor-bound NCoR/SMRT corepressor complex.

The thyroid hormone receptors (TRs), 2 ␣ and ␤, are hormoneinducible transcription factors. They mediate the physiological actions of thyroid hormone, which regulates the growth, development, and metabolism of a wide variety of tissues in higher organisms (1). At the target gene promoter, TR ␣ or ␤ interacts with a distinct DNA sequence, termed thyroid hormone response element (TRE), either as a homodimer or a heterodimer with retinoid X receptor (RXR) (2,3). The DNA-bound receptor then functions either as a transcriptional activator or a repressor, depending on its hormonal status, the host cell, and the promoter context (3,4). In the absence of thyroid hormone, TR generally functions as a silencer of basal level transcription from the target promoter (5)(6)(7). Ligand binding to TR releases transcriptional silencing and leads to the activation of target gene expression. Investigations into the molecular basis of this ligand-dependent switch have led to the identification of distinct cellular coregulatory factors, corepressors and coactivators, which mediate the transcriptional response by the receptor (8 -10). In the unliganded state, the receptor exists in a conformation that allows it to interact with a corepressor, which mediates transcriptional repression (8 -12). Ligand binding triggers a dramatic change in receptor conformation, resulting in the displacement of the corepressor and recruitment of a coactivator, which mediates transcriptional activation.
Several laboratories have characterized the corepressors that interact with TR (8,13). Two distinct but structurally related corepressors, nuclear receptor corepressor (NCoR) and silencing mediator of retinoid and thyroid receptors (SMRT), have been identified (11,12). The NCoR and SMRT polypeptides harbor multiple independent repression domains, which contribute to the overall repression function of these corepressors (11,12). NCoR or SMRT uses its carboxyl-terminal receptor interaction domain to interact directly with the ligand-binding domain of unliganded TR but does not interact with the ligandoccupied receptor. Within the receptor interaction domain of the corepressor, two corepressor-nuclear receptor boxes con-taining (I/L)XX(I/V)I signature motifs mediate interaction with the ligand-binding domain (14). It has been demonstrated that the loss of interaction with NCoR or SMRT impairs transcriptional silencing by TR, indicating that these factors are bona fide corepressors (15).
Biochemical characterization of cellular NCoR and SMRT revealed that these proteins exist in large multiprotein complexes (16 -18). The core complexes of NCoR and SMRT are essentially the same and composed of NCoR/SMRT, histone deacetylase 3 (HDAC3), a WD40-repeat protein TBL1, and the G protein suppressor 2. Interestingly, there are also reports that NCoR or SMRT associates with a complex containing Sin3, HDAC1, and HDAC2 (19,20). Despite significant progress in the isolation and functional characterization of NCoR/SMRT complexes, the identities of the coregulatory proteins that interact with unliganded TR⅐RXR have not been fully investigated. It is unclear whether TR⅐RXR needs to interact with regulatory factors in addition to the NCoR/ SMRT corepressor to function as a transcriptional repressor. In this report, we describe the biochemical isolation of an unliganded TR⅐RXR⅐corepressor holocomplex containing two distinct protein complexes. Although one of these complexes represents the previously described NCoR/SMRT core complex, the other one is a DNA-dependent protein kinase (DNA-PK) enzyme complex, containing the catalytic subunit of DNA-PK, its regulatory subunits Ku70 and Ku86, and the poly(ADP-ribose) polymerase 1 (PARP1). Using chromatin immunoprecipitation, we demonstrated that DNA-PK is indeed recruited by the repressive form of TR⅐RXR heterodimer at a natural target promoter in vivo. Most importantly, our studies revealed that DNA-PK targets the HDAC3 component of the NCoR/SMRT corepressor complex for phosphorylation, and this modification enhances the histone deacetylase activity of the complex. We propose that the DNA-PK complex contributes to TR-mediated transcriptional repression by increasing histone deacetylation and thereby influencing chromatin structure and function.
Isolation of TR⅐RXR-interacting Polypeptides from HeLa Nuclear Extracts Using Receptor Affinity Chromatography-Nuclear extracts (NE) of HeLa cells were prepared as described previously (21). The bacterial expression constructs encoding GST-hTR␤ and His 6 -tagged hRXR␣ (full-length) have been described previously (21). Briefly, GST-TR (0.5 g) was incubated with glutathione (GSH)-Sepharose beads (50 l, 1:1 slurry) for 2 h at 4°C. The beads were washed four times with a wash buffer containing 20 mM Tris-HCl, pH 7.9, 100 mM NaCl, 15% glycerol, and 0.02% IGEPAL CA-630 (Sigma). The immobilized TR was incubated with purified His 6 -hRXR␣ (2 g) for 1 h at room temperature followed by repeated washing with the wash buffer. The bead-bound TR⅐RXR heterodimer was then incubated with HeLa cell NE (2.5 mg, 2.5 g/l) for 2 h at 4°C on an end-to-end shaker. The beads were subsequently washed four times with the wash buffer. For identification of polypeptides by mass spectrometry, the bound proteins were dissociated under denaturing condition with 0.2% sodium N-lauroyl sarcosinate. For analyses of the TR⅐RXR holocomplex by glycerol density gradient fractionation or immunoprecipitation, the bead-bound complex is eluted under native conditions using 20 mM GSH.
Protein Identification by Mass Spectrometry-Gel-resolved proteins were digested with trypsin and batch-purified on a reversed-phase micro-tip, and resulting peptide pools were individually analyzed by matrix-assisted laser desorption/ionization reflectron time-of-flight mass spectrometry for peptide mass fingerprinting, as described (22,23). Selected peptide ions (m/z) were taken to search a "non-redundant" protein data base (NCBI, Bethesda, MD) utilizing the Peptide Search algorithm (Applied Biosystems, Foster City, CA).
Glycerol Density Gradient Centrifugation-Protein complexes bound to TR⅐RXR heterodimer were eluted using reduced GSH and fractionated on a 15-60% glycerol gradient based on a previously published procedure (24).
Immunoprecipitation-A rabbit polyclonal antibody that recognized both NCoR and SMRT (10 g) was covalently coupled to protein-A-Sepharose, using the chemical cross-linking agent dimethyl pimelimidate, as described previously (25). The immobilized antibody was incubated with purified TR⅐RXR holocomplex for 3 h at 4°C. Following repeated washings with wash buffer, the bound proteins were eluted with SDS sample buffer and analyzed by Western blotting using antibodies against NCoR/SMRT and DNA-PK.
Phosphorylation of TR⅐RXR Holocomplex and HDAC3-The TR⅐RXR heterodimer-bound protein complexes were isolated from HeLa nuclear extract as detailed above, except that phosphatase inhibitor cocktails (I and II) were used during receptor immobilization and incubation with nuclear extract. After the final wash, the beads were suspended in 50 l of kinase buffer (20 mM HEPES, pH 7.9, 10 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 10% glycerol, 100 mM KCl, 0.02% IGEPAL, and 1 mM dithiothreitol). The purified complex was then phosphorylated using a mixture of ␥ 32 P-labeled ATP (3.6 fmol of 3000 Ci/mmol) and unlabeled ATP (50 M) in the presence or absence of the PI3K inhibitor wortmannin (1 M) as previously described (26).
For immunoprecipitation of phosphorylated proteins, 32 Plabeled TR⅐RXR holocomplex was suspended in radioimmune precipitation assay buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1.0% IGEPAL, 0.5% deoxycholate, 0.1% SDS) and incubated with protein-A-Sepharose beads bound to an antibody against GST or NCoR or DNA-PKcs or HDAC3 or control IgG (10 g) for 2 h at 4°C. Following extensive washing with radioimmune precipitation assay buffer, the bound proteins were eluted by boiling with SDS sample buffer and analyzed by SDS-PAGE and autoradiography.
For the expression of FLAG epitope-tagged human HDAC3, cDNA encoding the full-length HDAC3 was cloned into the NdeI and BamH1 sites of pET-15b vector (Novagen) downstream of a 3ϫFLAG sequence (inserted at the NCo1 site). The 3ϫFLAG-HDAC3 fusion protein was expressed in bacteria as described previously (21). FLAG-HDAC3 or GST (0.5 g) immobilized on either M2 beads or GSH beads was subjected to a phosphorylation reaction using 5.0 units of DNA-PK enzyme (Promega, Madison, WI). The proteins were denatured in SDSsample buffer, resolved on an SDS-PAGE, stained with Coomassie Blue, and visualized by autoradiography.
Histone Deacetylase Assay-Histone deacetylase assay was carried out using a HDAC assay kit as per manufacturer's instruction (Upstate Biotechnology). Briefly, an H4 synthetic peptide (residues 2-24) was acetylated in the presence of [ 3 H]acetyl-CoA and p300/CBP-associated factor, a histone acetyltransferase. HDAC activity was assayed by incubating streptavidin-bound radiolabeled H4 peptide with bead-bound purified TR/RXR holocomplexes. Wortmannin (1 M) and trichostatin A (1 M) were used where indicated.
ChIP Assay-The ChIP assays were carried out as described previously (27).
Cell Culture and Transfection-HeLa cells were grown in Dulbecco's modified Eagle's medium with 5% fetal calf serum containing penicillin and streptomycin. 48 h prior to transfection, medium was replaced with DMEM containing 5% T 3 -stripped serum (28). Cells were transfected with an expression plasmid pCI-hTR␤ (0.025 or 0.05 g), a reporter plasmid pGL DRϩ4-luciferase (0.5 g), and an internal control plasmid Renilla luciferase (0.01 g) using the Lipofectamine-2000 reagent (Life Technology). Wortmannin (0.5 M) or NU7026 was added where indicated. The TRE-driven firefly luciferase and the internal control Renilla luciferase activities were measured by the instruction provided with the dual luciferase assay system (Promega).

RESULTS
Unliganded TR⅐RXR Heterodimer Interacts with the Components of NCoR/SMRT and DNA-PK Complexes-To identify the polypeptides that interact with TR⅐RXR heterodimer, we employed protein affinity chromatography, using the preformed heterodimers as bait. Human TR␤ and RXR␣ were expressed in Escherichia coli as recombinant proteins fused to GST and polyhistidine tags, respectively. Recombinant GST-TR was first immobilized on GSH beads and then incubated with purified His-RXR to form the TR⅐RXR heterodimers. As shown in Fig. 1, the immobilized TR⅐RXR heterodimers contained stoichiometrically equivalent amounts of each receptor partner (lane 4). In addition, multiple proteolytic degradation products of each receptor were also retained on the beads. The GSH beads bound to either GST alone (control) or hormone-free TR⅐RXR (test) were then incubated with or without HeLa nuclear extracts. The beads were washed extensively, and the bound proteins were eluted with a buffer containing 0.2% N-lauroyl sarcosine. The fractions eluted from the control and test beads were analyzed by SDS-PAGE, and the polypeptide components were visualized by silver staining. We noted that the control GST beads non-specifically retained a number of proteins (NS1-5) from HeLa nuclear extracts (lane 2). The same polypeptides were also retained by the TR⅐RXR beads (lane 4). Strikingly, about fifteen additional polypeptides with relative molecular masses ranging from 50 to 400 kDa were detected in the fraction eluted from the GST-TR⅐RXR column but were absent in the fraction eluted from the control beads (lane 4). These polypeptides, therefore, interacted specifically with unliganded TR⅐RXR heterodimers and represented potential constituents of one or more corepressor complex.
To determine the identities of the TR⅐RXR-associated polypeptides, we performed large scale purification using TR⅐RXR affinity beads, excised bands containing individual polypeptides from SDS-PAGE, and subjected these bands to in-gel tryptic digestion followed by mass spectrometry analysis as described under "Experimental Procedures." Peptide sequences were obtained from several of these polypeptides, and their identities were established (Table 1). Consistent with previously published reports, several signature components of the NCoR/SMRT complex such as nuclear receptor corepressor 1 (NCoR), nuclear receptor corepressor 2 (SMRT), transducin ␤-like 1 (TBL1), G-protein pathway suppressor 2, and histone deacetylase 3 (HDAC3) were found to be associated with TR⅐RXR. Interestingly, we also identified four known components of the DNA-PK complex, namely, DNA-dependent protein kinase catalytic subunit (DNA-PKcs), PARP1, ATP-dependent DNA helicase II (Ku86), and thyroid autoantigen (Ku70), as proteins that were retained specifically by TR⅐RXR. The mass spectrometric analysis also identified specific association of three other proteins, SMC1, DNA-ligase III, and HSP70, with the TR⅐RXR heterodimer (Table 1). Additionally, we obtained seven mass spectrometric identifications that represented minor polypeptides in the SDS-PAGE. These polypeptides are nucleolin (NCBI JH0148), KIAA1380 protein (BAA92618), RAD50 (XP_034865), zinc finger CW type with coiled-coil domain 2 (NP_078933), thyroid receptor interacting protein 8 (Q15652), eukaryotic translation elongation factor 1␣ (XP_088468), and eukaryotic translation elongation factor 1␥ (NP_001395).
The TR⅐RXR-associated polypeptides, therefore, included the components of at least two distinct complexes, the NCoR/ SMRT corepressor complex, and the DNA-PK enzyme complex. It is notable that the approach used here did not result in the isolation of any known coactivator, thereby indicating the highly specific nature of our affinity chromatography. The identification of known corepressors such as NCoR and SMRT also gave us confidence that the polypeptides retained by the TR⅐RXR beads may represent molecules with relevant functions in the receptor-mediated repression pathway.
Binding of T 3 to TR Inhibits the Interaction of Both NCoR/ SMRT and DNA-PK Complexes with TR⅐RXR Heterodimer-Binding of T 3 to TR is reported to trigger the dissociation of NCoR/SMRT from the receptor (11,12). We, therefore, examined whether the retention of the components of NCoR/SMRT and DNA-PK complexes is influenced by the hormonal status of TR. Immobilized TR⅐RXR was incubated with HeLa nuclear extracts in the presence and absence of T 3 . The receptor-bound proteins were then eluted as described above in Fig. 1 and analyzed by Western blotting using antibodies against NCoR/SMRT and DNA-PKcs. As shown in Fig. 2A, addition of T 3 markedly reduced the association of both NCoR/SMRT and DNA-PKcs with TR⅐RXR. Binding of T 3 to TR⅐RXR, however, had a significantly greater impact on the retention of NCoR/SMRT than that of DNA-PKcs. Whereas treatment with T 3 resulted in Ͼ90% reduction in the binding of NCoR/SMRT to TR⅐RXR, the binding of DNA-PKcs was reduced to ϳ40%.
In a parallel experiment, we investigated whether T 3 is able to dissociate NCoR/SMRT and DNA-PK complexes bound to TR⅐RXR. In this experiment, immobilized TR⅐RXR heterodimers FIGURE 2. Thyroid hormone inhibits the interaction of NCoR and DNA-PK complexes with TR⅐RXR heterodimer. A, TR⅐RXR heterodimers were assembled on GSH beads as described under "Experimental Procedures." The heterodimers were incubated with 1 M T 3 for 10 min and incubated with HeLa NE containing 1 M T 3 or appropriate amount of vehicle (0.1 M NaOH). The heterodimer-bound polypeptides were dissociated with 0.2% sodium Sarkosyl and analyzed by Western blotting using specific antibodies against NCoR, DNA-PKcs, and GST (affinity tag of recombinant TR). Note that the NCoR antibody recognizes both NCoR and SMRT. B, T 3 -dependent release of NCoR/SMRT and DNA-PK complexes from preformed TR⅐RXR⅐corepressor holocomplexes. The bead-bound TR⅐RXR holocomplexes were isolated as described in Fig. 1. A portion of the beads was incubated with T 3 (1 M) for 15 min, and the released proteins were collected. The remaining portion was similarly incubated with control buffer containing vehicle, and the bead-bound polypeptides were eluted with 0.2% sodium sarcosyl ("Bound" fraction). Equivalent amounts of T 3 -eluted and bead-bound fractions (50 g) were analyzed by Western blotting using antibodies against various components of the NCoR and DNA-PK complexes as indicated. C, "Bound" represents the total bead-bound (15 l) HDAC activity, and "eluted" represents the amounts of HDAC activity recovered in the same amount of beads by T 3 elution. HDAC activity was measured as described under "Experimental Procedures." Three independent sets of the same experiment were performed and the HDAC activities of a representative experiment are shown.

TABLE 1 Identification of TR⅐RXR-associated polypeptides by mass spectrometry
For protein mass spectrometric analysis, a large-scale isolation of TR⅐RXR-associated polypeptides was performed, and the polypeptides were resolved in an 8% SDS-PAGE and stained with Coomassie R250. Proteins were designated according to their apparent molecular sizes. The specific polypeptide bands from lane 4 were excised, and the protein identities were determined by mass spectrometric analysis as described under "Experimental Procedures." The peptide sequences derived from indicated protein bands and their identities are shown. The number of actual peptides identified for each of the 13 polypeptides listed ranged from 4 to 14. Here we have listed only the sequences representing the top two peptides that displayed the tallest peak heights in the mass spectrum.

Band
Protein identity NCBI no. Peptide sequence were incubated with HeLa nuclear extracts to assemble the receptor-protein complexes in the absence of T 3 . After thorough washings, the TR⅐RXR-bound polypeptides were incubated with buffer in the presence or absence of 1 M thyroid hormone. The supernatant containing the eluted proteins were analyzed by Western blotting using antibodies against various components of the NCoR/SMRT and DNA-PK complexes. None of the polypeptide components of either complex was detected in the eluted fraction in the absence of T 3 (data not shown). As expected, treatment with T 3 resulted in the dissociation of NCoR, HDAC3, and TBL-1, the known components of the NCoR/SMRT core complex (right lane, Fig. 2B). In agreement with results from Fig. 2A, we found that the components of the DNA-PK complex, namely, DNA-PKcs, PARP1, and Ku70, were also released from the TR⅐RXR heterodimer upon T 3 treatment (right lane, Fig. 2B). However, the hormone-dependent release of DNA-PK components from TR⅐RXR was relatively less efficient than that of the components of NCoR/SMRT complex. When we examined the HDAC activity of the T 3 -eluted fraction, we found that as much as 60% of the TR⅐RXR-bound HDAC activity was recovered in this fraction, and the HDAC inhibitor trichostatin-A (TSA) blocked this activity (Fig. 2C). The simultaneous hormone-induced release of both NCoR and DNA-PK complexes from TR⅐RXR hinted at a joint involvement of these complexes in the maintenance of the transcriptionally silent state of the receptor and raised the possibility that these two complexes might be physically and functionally linked to each other.

Evidence for the Existence of a TR⅐RXR Holocomplex Containing the NCoR/SMRT Core Complex and the DNA-PK Enzyme
Complex-It is conceivable that TR⅐RXR interacts with the NCoR/SMRT and DNA-PK complexes independently of each other and forms distinct complexes. Alternatively, the constituents of the NCoR/SMRT and the DNA-PK complexes may coexist in a single holocomplex bound to TR/RXR. To distinguish between these possibilities, we eluted TR⅐RXR and its associated polypeptides from the affinity column under native conditions (GSH elution). We then subjected the eluted polypeptides to glycerol density gradient fractionation and performed an initial analysis of each fraction by Western blotting using antibodies against TR and RXR. The TR and RXR polypeptides resolved into two distinct regions in the gradient, representing complexes of distinct molecular sizes. A lighter complex I, comigrating with free TR⅐RXR heterodimer, was detected in fractions 3-7 (Fig. 3A, top  panel). A heavier complex II, containing TR and RXR, was present in fractions 9 -11 (middle panels). When the blot was probed with an antibody against NCoR/SMRT, we found that the fractions 9 -11 also contained the corepressor. The complex II, therefore, potentially represented a TR⅐RXR⅐corepressor complex (Fig. 3A,  bottom panel). FIGURE 3. Glycerol gradient sedimentation analysis of the TR⅐RXR⅐ corepressor holocomplex. A, TR⅐RXR⅐corepressor holocomplex was first assembled from HeLa NE as described in legends to Fig. 1. The bead-bound proteins were eluted under native condition with reduced GSH (15 mM). The purified proteins were loaded onto a 15-60% discontinuous glycerol gradient (GG), resolved by centrifugation, and subjected to fractionation. Proteins in each fraction were precipitated by 20% trichloroacetic acid and analyzed by Western blotting using indicated antibodies. The top panel shows the sedimentation profile of purified TR⅐RXR heterodimer contained in fractions 3-7. Bottom panels: TR⅐RXR forms high molecular weight protein complex(es) upon incubation with HeLa NE. As shown by Western blotting, TR, RXR, and NCoR comigrate with the complex(es) in fractions 9 -11. B, comigration of NCoR and DNA-PK complexes during glycerol gradient centrifugation. Fractions 9 -11 containing the high molecular weight TR⅐RXR-bound protein complex(es) were pooled, precipitated with 20% trichloroacetic acid, and analyzed for the presence of various components of NCoR and DNA-PK complexes by Western blotting using indicated antibodies. The "bound" represents 10% of the total TR⅐RXR-associated proteins before density gradient analysis. C, fractions 9 -11 containing the high molecular weight TR⅐RXR complex(es) were pooled and concentrated by Centricon filters (10,000 molecular weight cut off, Amicon Bioseparations) and assayed for HDAC activity using a synthetic [ 3 H]histone-4 peptide (residues 2-24) as substrate. The HDAC activity in lane 1 corresponds to the total activity that was loaded onto the gradient, and the activity in lane 3 corresponds to the HDAC activity that was recovered in the high molecular weight TR⅐RXR fractions. Where indicated, HDAC inhibitor TSA (1 M) was used. The HDAC activities are representative of the results of three independent GG analyses.
To analyze the molecular composition of complex II, we pooled the fractions 9 -11 and performed Western blot analysis using antibodies specific for individual components of both NCoR and DNA-PK complexes. We confirmed that the signature components of the core NCoR/SMRT corepressor complex, namely NCoR, HDAC3 and TBL-1, are present in complex II (Fig. 3B). The signature components of the DNA-PK complex, namely, DNA-PK CS and PARP1, were also detected in this high molecular size complex (Fig. 3B). As much as 40% of the TSA-sensitive HDAC activity that was loaded on the density gradient was recovered in the fractions containing complex II, indicating that it represents a bona fide TR⅐RXR⅐corepressor complex (Fig. 3C). These results are consistent with our hypothesis that the NCoR/SMRT and DNA-PK complexes coexist in a large holocomplex that interacts with TR⅐RXR.
To further verify the simultaneous association of the NCoR and DNA-PK complexes with TR⅐RXR to form a holocomplex, we performed immunoprecipitation experiments. The receptor⅐ corepressor complex was first assembled by incubating TR⅐RXR with HeLa nuclear extracts as described in Fig. 1. TR⅐RXR and the associated polypeptides were then eluted from the GSH beads under native condition (GSH elution). The eluted proteins were immunoprecipitated using an NCoR/SMRT antibody or preimmune serum. The immunoprecipitates were analyzed by Western blotting using TR, RXR, NCoR/SMRT, and DNA-PK antibodies. As expected, immunoblotting confirmed the presence of TR and RXR (data not shown) and NCoR/SMRT in the immunoprecipitate (Fig. 4, upper panel). Probing with DNA-PK antibody revealed that DNA-PK is also present in the immunoprecipitate (Fig. 4,  lower panel). Coimmunoprecipitation of NCoR/SMRT and DNA-PK with TR⅐RXR heterodimer strengthened our view that the components of the NCoR/SMRT and DNA-PK complexes coexist in a large multisubunit holocomplex assembled on unliganded TR⅐RXR.

NCoR/SMRT and DNA-PK Complexes Simultaneously
Occupy a TRE in Vivo-We next performed ChIP to examine whether the components of NCoR and DNA-PK complexes are recruited at the regulatory regions of an endogenous TR-regulated gene. The rat growth hormone (GH) promoter contains a well characterized thyroid hormone response element (TRE) located between Ϫ189 and Ϫ162 with respect to the transcriptional start site (Fig. 5A). In GH 3 pituitary tumor cells, the GH gene is actively silenced in the absence of T 3 and activated in the presence of T 3 . These cells are known to express three TR isoforms, TR␣, TR␤1, and TR␤2, and three RXR isoforms, RXR␣, RXR␤, and RXR␥ (29,30). At first, we determined the identity of the TR and RXR isoforms that occupy the TRE of the GH promoter in the absence or presence of T 3 . As shown in Fig. 5B, FIGURE 4. Coimmunoprecipitation of TR⅐RXR-linked NCoR and DNA-PK complexes. The TR⅐RXR holocomplexes were assembled on affinity beads and eluted with 15 mM reduced glutathione as described under "Experimental Procedures." The purified complexes were immunoprecipitated using control IgG (preimmune-serum) and an antibody against NCoR/SMRT. The IgG was covalently coupled to protein-A-Sepharose. The immunoprecipitated (IP) proteins were dissociated by boiling in SDS sample buffer and analyzed by Western blotting using antibodies against NCoR/SMRT (top panel) and DNA-PK (bottom panel). "Input" represents 5% of TR⅐RXR holocomplexes that were used for the immunoprecipitation reaction. The results are representative of two independent sets of experiments. . "Ϫ" and "ϩ" represent treatments with vehicle and T 3 , respectively. The lower panel shows the results of ChIP analyses carried out using antibodies against RXR␣, RXR␤, and RXR␥. For a given amount of template DNA, we performed a series of PCR reactions varying the cycle number between 25 and 32 cycles to make sure that the assay remains within the linear range. The data in the upper and lower panels represent 28 and 30 PCR cycles, respectively. The input lanes represent soluble chromatin (1%) that was reverse cross-linked and amplified under the same PCR conditions. C, GH 3 cells were treated with vehicle or T 3 , and ChIP assays were performed with antibodies against NCoR/SMRT, HDAC1, HDAC3, TBL1, and control IgG (upper panel) and an antibody against the DNA-PKcs (lower panel). The data in the upper and lower panels represent 29 and 30 PCR cycles, respectively. The ChIP results are representative of two independent sets of experiments.
TR␣ is the predominant TR isoform that occupies the GH TRE in GH3 cells. We also detected occupancy of this TRE by TR␤1, although to a lesser extent than TR␣. No significant binding of TR␤2 was seen at this TRE. We noted occupancy of the GH TRE by both RXR␣ and RXR␤. Very little RXR␥ was seen at this site. Neither TR nor RXR showed any interaction with the distal promoter regions of GH promoter (data not shown).
We observed that the TR and RXR isoforms remained bound to the GH promoter irrespective of thyroid hormone (Fig. 5B). As expected, NCoR/SMRT, HDAC3, and TBL1, the components of the NCoR/SMRT corepressor complex, were bound to the GH TRE in the absence of T 3 . No significant recruitment of HDAC1 was seen under these conditions. Addition of T 3 led to the release of the components of NCoR/SMRT complex from the TRE site (Fig. 5C, top panel). Most interestingly, we found that DNA-PKcs was also recruited at the GH TRE region in the absence of T 3 . Like the NCoR/SMRT complex, DNA-PKcs was also released from the promoter site upon T 3 treatment. Similar T 3 -sensitive GH promoter occupancy by Ku70, another component of the DNA-PK complex, was also noted (data not shown). The results of our in vivo ChIP experiments are, therefore, generally consistent with the results of our in vitro biochemical studies. These results also provided strong support to our hypothesis that a TR⅐RXR⅐corepressor holocomplex containing the components of both DNA-PK and NCoR/SMRT complexes binds to cellular target genes in vivo and mediates transcriptional repression. Hormone binding to TR is accompanied by the dissociation of NCoR and DNA-PK complexes from TRE-bound heterodimers, thereby releasing the repression.
DNA-PK Phosphorylates HDAC3 in a TR⅐RXR⅐Corepressor Holocomplex-The presence of DNA-PKcs in the TR⅐RXR⅐ corepressor holocomplex raised the possibility that this kinase may phosphorylate one or more components of the holocomplex complex to modulate the corepressor function. To test this possibility, we first assembled a TR⅐RXR⅐corepressor holocomplex in vitro, and then incubated the purified complex with [␥-32 P]ATP. As shown in Fig. 6A, a kinase activity present in the purified TR⅐RXR⅐corepressor holocomplex phosphorylated several polypeptide components of this complex (Fig. 6A, left  panel). Prominent among these phosphorylated polypeptides were two bands of approximate molecular sizes 400 and 50 kDa, respectively. A comparison with the Western blotting profile of DNA-PKcs indicated that the 400-kDa phosphorylated band is likely to represent this kinase, which is known to autophosphorylate itself. Interestingly, the 50-kDa phosphorylated band comigrated with HDAC3. None of these bands were phosphorylated in the presence of wortmannin, a potent inhibitor of DNA-PK, confirming that the phosphorylation events are catalyzed by this enzyme, a component of the purified TR⅐RXR⅐corepressor holocomplex.
To test whether the phosphorylated polypeptides of molecular sizes 400 kDa and 50 kDa actually represented DNA-PK and HDAC3, respectively, we carried out immunoprecipitation of the 32 P-labeled TR⅐RXR holocomplex using antibodies specific for these proteins. Following in vitro phosphorylation reaction, the holocomplex was denatured and then immunoprecipitated with antibodies against GST (for GST-TR), DNA-PK, HDAC3, or control IgG. The anti-DNA-PK antibody precipitated the 32 P-labeled 400-kDa polypeptide, confirming that it indeed represented DNA-PKcs (data not shown). Most importantly, as shown in Fig. 6A, right panel, the anti-HDAC3 antibody precipitated the 32 P-labeled 50-kDa polypeptide, indicating that it represented HDAC3. These results provided strong evidence that the DNA-PKcs present in the TR⅐RXR⅐corepressor complex targets the HDAC3 subunit of the complex for phosphorylation.
To demonstrate directly that HDAC3 is a target of phosphorylation by DNA-PKcs, we expressed HDAC3 as a FLAGtagged polypeptide in bacteria and immobilized the recombinant protein on beads linked to anti-FLAG antibody. The FLAG-HDAC3 or control GST protein was then incubated with a purified preparation of catalytically active DNA-PKcs (commercially available from Promega) in the presence of [␥-32 P]ATP. We found that HDAC3, but not GST, was effi-FIGURE 6. DNA-PK-dependent phosphorylation of HDAC3 in TR⅐RXR holocomplex. A, TR⅐RXR holocomplex was formed upon incubation with HeLa NE as described in Fig. 1. The bead-bound proteins were suspended in a kinase buffer and phosphorylated in the presence of ␥-32 P-labeled ATP as described under "Experimental Procedures." After extensive washings, the protein complexes were dissociated by boiling with SDS sample buffer and resolved in an 8% SDS-PAGE. The radiolabeled protein bands were detected by autoradiography. Left panel represents the total TR⅐RXR-associated polypeptides that were phosphorylated in the absence (lane 1) and presence (lane 2) of 1 M wortmannin (WM), a DNA-PK inhibitor. Right panel: after phosphorylation reaction, 32 P-labeled TR⅐RXR-associated protein complexes were dissociated from the GSH-agarose beads using 15 mM reduced GSH, suspended in radioimmune precipitation assay buffer and immunoprecipitated with antibodies against GST (for GST-TR), NCoR, DNA-PK, HDAC3 and control IgG as described under "Experimental Procedures." The immunoprecipitates were resolved in a SDS-PAGE, and the radiolabeled proteins were detected by autoradiography. NS denotes a nonspecific band. 50 KD denotes the HDAC3 protein. B, a purified preparation of DNA-PK specifically phosphorylates HDAC3. FLAG-HDAC3 immobilized on beads linked to anti-FLAG antibody, and GST bound to GSH beads (0.5 g each) were subjected to a phosphorylation reaction using 5 units of a purified preparation of DNA-PK (Promega) and ␥-32 P-labeled ATP as described under "Experimental Procedures." Following extensive washings, the bead-bound proteins were dissociated by boiling with SDS sample buffer, resolved in an 8% SDS-PAGE and stained by Coomassie Blue (upper panel). The gel was then dried and subjected to autoradiography (lower panel). 1 M wortmannin was used where indicated. The results are representative of two independent sets of experiments.
ciently phosphorylated by DNA-PKcs (Fig. 6B). As expected, this phosphorylation was strongly inhibited by wortmannin. These results provided direct evidence that HDAC3 is a substrate of phosphorylation by DNA-PK.
DNA-PK-dependent Phosphorylation Modulates HDAC Activity of TR⅐RXR⅐Corepressor Holocomplex-Because HDAC3 is a target of phosphorylation by DNA-PK in the TR⅐RXR⅐corepressor holocomplex, we examined whether this phosphorylation event has any influence on the histone deacetylation function of the complex. To examine this possibility, we assessed the HDAC activity of the purified TR⅐RXR⅐corepressor holocomplex with or without incubation with ATP. In the absence of added ATP, the complex exhibited basal level HDAC activity (Fig. 7, upper panel). Addition of ATP led to a greater than 2-fold enhancement in the HDAC activity of this complex. As shown in Fig. 6, HDAC3 is phosphorylated by DNA-PK under these conditions. Addition of wortmannin, which blocks DNA-PK activity and prevents HDAC3 phosphorylation, also blocked the ATP-induced increase in HDAC activity of the holocomplex (Fig. 7, upper panel). It should be noted, however, that the holocomplex retained the basal level of HDAC activity in the presence of wortmannin, indicating that this basal activ-ity is independent of DNA-PK function. It is also important to point out that the DNA-PK-dependent phosphorylation of HDAC3 did not alter the amount of NCoR or HDAC3 present in the holocomplex (Fig. 7, lower panel). This result indicated that phosphorylation of HDAC3 did not affect its interaction with NCoR or cause any dissociation from the holocomplex. Taken together, these results indicated that one of the functions of DNA-PK in the context of the TR⅐RXR⅐corepressor holocomplex is to phosphorylate its HDAC3 component to significantly enhance its overall histone modifying activity.
Inhibition of DNA-PK Activity Diminishes the Ability of TR to Function as a Transcriptional Repressor-We next examined whether the increased HDAC activity promoted by DNA-PK contributed to TR⅐RXR-mediated transcriptional silencing in vivo. In transient transfection experiments, we tested the ability of TR⅐RXR to repress a TRE-linked luciferase reporter gene in the presence or absence of the DNA-PK inhibitor wortmannin. In the absence of T 3 , increasing expression of hTR␤ led to a robust receptor-mediated silencing of basal transcription from the target reporter gene in HeLa cells (Fig. 8, lanes 1 and 3). When DNA-PK was functionally blocked by wortmannin, the silencing activity of TR was markedly reduced in comparison with the untreated control (compare lanes 1 and 2 and lanes 3 and 4). Whereas 60 -71% TR-mediated repression of basal transcription was observed in the absence of wortmannin, only 37-41% repression occurred in the presence of wortmannin. We also tested the effects of a highly selective DNA-PK inhibitor NU7026 on TR-mediated transcriptional silencing. NU7026 acts as a potent and specific inhibitor of DNA-PK with an IC 50 of 0.23 M. It does not affect the activity of other PI3K-related kinases. Our results suggest that blockade of DNA-PK activity by this inhibitor suppressed TR-mediated silencing by as much as 60% (lanes 3 and 4, Fig. 8B). Collectively, these results are consistent with an in vivo modulatory role of DNA-PK in TR-mediated transcriptional silencing.

DISCUSSION
The present study is aimed at the biochemical identification and functional characterization of transcriptional coregulatory proteins recruited by the repressive form of the TR⅐RXR heterodimer. We provide several lines of evidence in support of the concept that unliganded TR⅐RXR interacts simultaneously with two distinct protein complexes, the NCoR/SMRT corepressor complex and the DNA-PK enzyme complex, to assemble a single, large holocomplex. First, mass spectrometry analysis of polypeptides associated with TR⅐RXR identified signature components of both DNA-PK (DNA-PKcs, Ku86, Ku70, and PARP1) and NCoR/ SMRT (NCoR/SMRT, TBL1, and HDAC3) complexes. Second, components of both complexes were simultaneously released from the TR⅐RXR heterodimer upon treatment with T 3 , indicating a possible direct association between them. Third, a TR⅐RXR complex of high molecular size comigrated with the components of both NCoR/SMRT and DNA-PK complexes during density gradient fractionation. Fourth, immunoprecipitation of a TR⅐RXR⅐corepressor complex using a NCoR/SMRT antibody coprecipitated DNA-PK. Previous studies reported that the DNA-PK complex binds to single-stranded DNA ends (31,32). In our experimental conditions, however, purified TR⅐RXR heterodimers bound to the DNA-PK complex in the absence of any added DNA. Moreover, we did not see any significant reduction in the retention of DNA-PK on TR⅐RXR even after extensive treatment of HeLa nuclear extracts and purified TR⅐RXR complexes with DNase I (data not shown), ruling out any contribution of DNA-dependent interactions to the binding of these proteins by TR⅐RXR. Collectively, these biochemical experiments are consistent with the concept that in the unliganded state TR⅐RXR recruits both NCoR/SMRT and DNA-PK complexes, which are held together via protein-protein interactions between certain of their components.
The results of these in vitro experiments found support from our in vivo analysis of T 3 -regulated growth hormone gene expression in GH 3 pituitary tumor cells using chromatin immunoprecipitation. Both TR and RXR occupied the growth hormone promoter region containing a well characterized TRE, irrespective of thyroid hormone status of these cells. In the absence of T 3 , the signature components of both NCoR/SMRT and DNA-PK com-plexes were found to bind to this region, consistent with the findings of our in vitro experiments that these complexes are associated with the transcriptionally repressive form of TR⅐RXR. Binding of T 3 to TR, which triggers the transition of this receptor from the repressive to an active form, was accompanied by the dissociation of the components of the NCoR/SMRT complex from the TRE-bound heterodimer. Interestingly, hormone binding to TR also released the DNA-PK complex from the TRE site. These results provided strong evidence not only for the recruitment of a DNA-PK complex by unliganded TR⅐RXR at a natural target gene but also hinted at its close interaction with the components of the NCoR/SMRT complex.
The DNA-PK core complex is composed of four subunits: DNA-PKcs, PARP1, Ku86, and Ku70 (32,33). The DNA-PKcs, a serine/threonine protein kinase, belongs to a family of PI3Krelated protein kinases that also includes ATM (ataxia-telangiectasia, mutated), ATR (ATM and Rad3-related), mammalian target of rapamycin, and transformation/transcription domain-associated protein. The members of this family are reported to be involved in double-stranded DNA break repair and V(D)J recombination (32,33). Autophosphorylation of DNA-PKcs activates the enzyme and is critical for its DNA repair function (34 -36). The DNA-PKcs is also known to have a role in transcriptional regulation due to its ability to phosphorylate and regulate the activity of sequence-specific transcription factors and their cofactors. It was observed that DNA-PK and other PI3Ks such as ATM and ATR target the tumor suppressor p53 and its negative regulator hdm2/mdm2 to modulate the p53-mediated transcriptional pathway (37,38). Recently Ko and Chin (39) reported that TRBP (thyroid hormone receptor-binding protein), a nuclear receptor coactivator, exhibited reduced coactivation function for GR in DNA-PK-deficient cells in response to dexamethasone, indicating a role of this kinase in regulating coactivator function.
Direct association of a DNA-PK complex with a nuclear receptor was reported by Sartorius et al. (40). They observed that the DNA-binding domain of progesterone receptor is able to interact with an intact DNA-PK complex containing DNA-PKcs, PARP1, Ku86, and Ku70 present in HeLa nuclear extracts (40). Weigel et al. previously reported that transcriptionally active progesterone receptor is phosphorylated by DNA-PK (41). Studies by Giffin et al. (42) indicated that DNA-PK interacts with glucocorticoid receptor and phosphorylates it at the Ser-527 residue situated between the DNA-binding domain and ligand-binding domain. Although these studies collectively suggested that DNA-PK is able to interact directly with certain nuclear receptors to modify their transcriptional activities, the precise mechanism by which the phosphorylation by this kinase affects receptor function remained unclear.
An important finding of this report is that the DNA-PK modulates the activity of HDAC3, a key component of the NCoR/SMRT corepressor complex, to regulate chromatin function and transcription. We did not detect any significant phosphorylation of either TR or RXR when the purified TR⅐RXR⅐corepressor holocomplex was incubated with radiolabeled ATP (data not shown). We, however, observed that the HDAC3 subunit of the NCoR/SMRT complex is phosphorylated under these conditions, and this phosphorylation is sensitive to wortmannin, an FIGURE 8. Transcriptional repression by TR is impaired in the presence of DNA-PK inhibitors. A, HeLa cells were grown in T 3 -free medium with or without 0.5 M wortmannin (WM). The cells were transiently transfected with a reporter vector, TRE-luciferase (0.5 g), a TR expression vector pCI-hTR␤ (0.025 and 0.05 g), a RXR expression vector pCI-hRXR␣ (0.025 g), and an internal control vector expressing Renilla luciferase (0.01 g). 48 h after transfection, cells were harvested and assayed for luciferase activity as described previously (59). The basal reporter activity in the absence of transfected pCI-hTR␤ was taken as 100%, and the relative values for test samples were obtained. B, transient transfections were carried out in HeLa cells as described above with the following modifications: 1 M NU7026 was used instead of wortmannin, and slightly different amounts of pCI-hTR␤ (0.05 and 0.1 g) were used. Three independent sets of each experiment were carried out, and the results are represented as mean value Ϯ S.E.
inhibitor of DNA-PK and other PI3Ks. Because mass spectrometry analysis identified DNA-PKcs as the only PI3K present in the purified TR⅐RXR holocomplex, we propose that DNA-PK is the kinase that phosphorylates HDAC3. This conclusion is further supported by the observation that purified, catalytically active DNA-PK is able to phosphorylate HDAC3.
The PI3K-mediated phosphorylation of HDAC3 represents a novel post-translational modification of the class I HDAC enzymes. Previous studies documented that both HDAC1 and HDAC2 exist as phosphoproteins in the nuclear extracts. It was also determined that these phosphorylation events were mediated by casein kinase II (43,44). HDAC3 is also phosphorylated upon incubation with casein kinase II in vitro (45). It is conceivable that the close contacts between the components of the DNA-PK and NCoR/SMRT complexes in the holocomplex allow the DNA-PKcs to target HDAC3 for phosphorylation. It is of interest to note that sequence analysis of human HDAC3 protein by MotifScan (scansite.mit.edu/cgi-bin/motifscan_seq), a phosphorylation site prediction program, pointed to amino acids Thr-390 and Ser-405 as potential phosphorylation sites for DNA-PK. It, however, remains to be seen whether these amino acids represent actual sites of phosphorylation by DNA-PK.
Previous studies indicated that increased phosphorylation of the class I HDACs promotes their histone deacetylase activities (43)(44)(45). Consistent with these reports, we observed that DNA-PK-mediated phosphorylation of the HDAC3 component of the TR⅐RXR⅐corepressor holocomplex significantly enhanced the HDAC enzyme activity of the purified complex (Fig. 7). It is well established that the NCoR/SMRT corepressor complex containing HDAC3 is recruited by promoter-bound unliganded TR⅐RXR and that it promotes transcriptional silencing of target genes via local deacetylation of histones. The deacetylation function of HDAC3 removes the acetyl groups from critical lysine and arginine groups of histones to create a compact chromatin structure that is conducive to efficient transcriptional repression. We propose that PI3K signaling via DNA-PK plays a critical role in this process by phosphorylating and thereby enhancing the catalytic activity of HDAC3. If the DNA-PK activity were critical for maintaining a robust deacetylase function of the NCoR/SMRT corepressor complex, then one would predict that its blockade would lead to a suppression of TR-mediated transcriptional silencing in vivo. Indeed, we noted that the ability of TR to function as a transcriptional repressor in a cell-based transfection assay was markedly reduced in the presence of wortmannin, an inhibitor of DNA-PK.
The factors that regulate the activity DNA-PK within the TR⅐RXR⅐corepressor holocomplex are unknown. Earlier reports indicated that the DNA-PKcs must be physically associated with DNA to be fully active (31). Although binding to DNA ends is a well known mode of activation of this kinase, other mechanisms, including regulation by distinct kinases and protein-protein interaction, have been shown to promote DNA-PK activity (46 -48). The study by Ko and Chin has shown that the coactivator TRBP can stimulate the activity of the DNA-PK complex in the absence of any added DNA (39). Our studies with the purified TR⅐RXR holocomplex suggest that the DNA-PK activity may derive from the protein-protein interactions of the DNA-PKcs with other polypeptides within this complex.
The functional roles, if any, of the Ku subunits and PARP1 in TR-mediated silencing are presently unclear. PARP-1, the best characterized member of a family of enzymes that catalyzes the transfer of ADP-ribose from NADϩ to target proteins, is involved in multiple DNA repair pathways (49). ADP ribosylation of histones by PARP1 promotes their dissociation to cause a local DNA relaxation that allows the assembly of repair proteins and the initiation of the repair process (50). Cross covalent modifications, i.e. phosphorylation of PARP1 by DNA-PK and ADP-ribosylation of DNA-PK by PARP-1, have been found to be critical for the function of the DNA repair machinery (51,52). There is now increasing evidence that PARP1 plays a critical role in transcriptional regulation. Miyamoto et al. (53) reported that overexpression of PARP1 strongly inhibits T 3 -dependent transactivation by TR in transient transfection experiments. Ju et al. (54) have shown that PARP1 is a stable and integral component of the TLE corepressor complex, which is involved in the repression of MASH1 gene by transcription factor HES1 during neural cell differentiation. A recent study indicated that a complex containing DNA topoisomerase II␤ and PARP1 is recruited to target promoters during hormone-mediated transactivation by the nuclear receptors (55). Topoisomerase II␤ induces transient, site-specific double-stranded DNA breaks and PARP1 mediates nucleosome-specific modifications in histone H1, leading to local changes in chromatin architecture. Although our study did not address the recruitment of topoisomerase II␤ during TR-mediated transcriptional repression, we did note that DNA ligase III and structural maintenance of chromosome 1 are associated with the repressive form of TR⅐RXR (Table 1). These proteins are known to play critical functions in DNA break and repair events (56 -58). Further studies are clearly necessary to understand how the molecules involved in DNA break and repair pathways contribute to TRmediated transcriptional activation and repression mechanisms.
In summary, our study provides compelling evidence for the existence of a TR⅐RXR⅐corepressor holocomplex endowed with multiple enzymatic activities such as HDAC3 (histone deacetylase), DNA-PK (kinase), and PARP1 (ADP-ribosylase). Coexistence of several modifying activities in a single large complex offers the potential of regulation by multiple signaling mechanisms. It is possible that DNA-PKcs can target additional components of the holocomplex beside HDAC3 for phosphorylation. Similarly, PARP1 can modify components of the holocomplex via ADP ribosylation. Additionally, these activities may also alter chromatin structure and function by modifying histones. When recruited to a target gene promoter, the corepressor holocomplex will thus provide a highly versatile and efficient enzymatic machinery to locally alter chromatin structure that modulates the transcriptional function of TR. Future studies will analyze the nature of these chemical modifications and determine their functional roles in TR-mediated gene repression.