Mechanisms That Mediate Negative Regulation of the Thyroid-stimulating Hormone α Gene by the Thyroid Hormone Receptor*

A group of transcriptional cofactors for nuclear hormone receptors, referred to as corepressors (CoRs) and coactivators (CoAs), has been shown to induce transcriptional silencing and hormone-induced activation, respectively, of genes that contain positive hormone response elements. Transcriptional silencing by CoRs involves the recruitment of histone deacetylases (HDACs), whereas ligand-dependent activation is associated with the recruitment of CoAs, which possess or recruit histone acetyltransferases (HATs). In a reciprocal manner, negatively regulated genes are stimulated by nuclear receptors in the absence of ligand and are repressed in response to ligand binding to receptors. We show here that negative regulation of the thyroid-stimulating hormone α (TSHα) promoter by the thyroid hormone receptor (TR) involves a novel mechanism in which the recruitment of CoRs by TR is associated with transcriptional stimulation and histone acetylation. Expression of excess HDAC reverses the stimulation mediated by the TR·CoR complex, consistent with a pivotal role for acetylation in this event. Addition of the ligand, 3,5,3′-triiodothyronine (T3), induces transcriptional repression of the TSHα promoter and is associated with the loss of histone acetylation. T3-dependent repression is blocked by phosphorylation of cAMP response element binding protein, or by inhibition of HDAC, indicating that receptor action is subverted by maneuvers that stimulate histone acetylation of the target gene. We propose that negative regulation of a subset of genes by TR involves the active exchange of CoRs and CoAs with intrinsic promoter regulatory elements that normally strongly induce histone acetylation and transcriptional activation.

A group of transcriptional cofactors for nuclear hormone receptors, referred to as corepressors (CoRs) and coactivators (CoAs), has been shown to induce transcriptional silencing and hormone-induced activation, respectively, of genes that contain positive hormone response elements. Transcriptional silencing by CoRs involves the recruitment of histone deacetylases (HDACs), whereas ligand-dependent activation is associated with the recruitment of CoAs, which possess or recruit histone acetyltransferases (HATs). In a reciprocal manner, negatively regulated genes are stimulated by nuclear receptors in the absence of ligand and are repressed in response to ligand binding to receptors. We show here that negative regulation of the thyroid-stimulating hormone ␣ (TSH␣) promoter by the thyroid hormone receptor (TR) involves a novel mechanism in which the recruitment of CoRs by TR is associated with transcriptional stimulation and histone acetylation. Expression of excess HDAC reverses the stimulation mediated by the TR⅐CoR complex, consistent with a pivotal role for acetylation in this event. Addition of the ligand, 3,5,3-triiodothyronine (T3), induces transcriptional repression of the TSH␣ promoter and is associated with the loss of histone acetylation. T3-dependent repression is blocked by phosphorylation of cAMP response element binding protein, or by inhibition of HDAC, indicating that receptor action is subverted by maneuvers that stimulate histone acetylation of the target gene. We propose that negative regulation of a subset of genes by TR involves the active exchange of CoRs and CoAs with intrinsic promoter regulatory elements that normally strongly induce histone acetylation and transcriptional activation.
Nuclear receptors are transcription factors that regulate the expression of a wide array of target genes. An intriguing aspect of their action is that, within the same cell, some target genes are stimulated whereas others are repressed. This phenomenon is most readily explained by differences inherent in the regulatory elements of these target genes.
In the case of positively regulated genes, the mechanism of transcriptional control is relatively well understood. These genes contain hormone response elements that bind nuclear receptors, usually as homo-or heterodimers. For receptors such as the thyroid hormone receptors (TRs) 1 or the retinoic acid receptors, receptor binding in the absence of ligand is associated with transcriptional repression. This property of nuclear receptors involves the recruitment of corepressors (CoRs) like silencing mediator for retinoid and thyroid hormone receptors (SMRT) (1,2) and nuclear receptor corepressor (NCoR) (3)(4)(5). These CoRs in turn assemble a repression complex that includes Sin3 and histone deacetylases (HDACs), among other proteins (6 -8). Transcriptional silencing by this receptor-assembled complex is thought to involve chromatin remodeling caused by histone deacetylation.
In contrast to positively regulated genes, the mechanism by which nuclear receptors control the transcription of negatively regulated genes is less well understood. Proposed mechanisms include competition of nuclear receptors with other transcription factor-binding sites, receptor binding to so-called negative regulatory elements, direct interactions of nuclear receptors with transcription factors like AP1 or NFB, and competition for transcriptional cofactors like CBP (24).
In the case of the thyroid hormone receptor, negative regulation of gene expression in response to its ligand, 3,5,3Ј-triiodothyronine (T3), is an essential part of its physiological action. For example, the hypothalamic thyrotropin-releasing hormone (TRH) and the pituitary thyroid-stimulating hormone (TSH) ␣and ␤-subunit genes are inhibited by T3 as a physiologic feedback mechanism for modulating circulating thyroid hormone levels. These genes are stimulated in the absence of T3, and the addition of hormone induces rapid and strong transcriptional repression (25). Previous studies have localized hormone-responsive regions to the proximal promoter regions of these negatively regulated genes (26 -29). However, the nature of the response elements and the mechanism of ligand-dependent repression remain poorly defined. Recently, we reported (30) the unexpected finding that CoRs are involved in the basal activation of the TSH and TRH genes by unliganded TRs. TR mutations that prevent its ability to interact with CoRs eliminated stimulation of these promoters in the absence of ligand, and overexpression of NCoR or SMRT enhanced rather than suppressed the basal activity of these genes.
In this report, we describe a potential mechanism for transcriptional control of genes that are negatively regulated by TR. In this model, TR retains the fundamental features of interactions with CoRs and CoAs, but the functional consequences of these interactions are reversed in comparison to positively regulated genes. We demonstrate that the negatively regulated TSH␣ gene is unusually sensitive to the state of histone acetylation and appears to be controlled by partitioning of HDACs and HATs between TR and other transcription factors that bind to the TSH␣ promoter. We propose that a subset of negatively regulated genes are controlled by two-step mechanism as follows: 1) the unliganded TR recruits CoRs and withdraws HDAC from the basal promoter to cause activation; 2) T3 binding to the TR dissociates the CoRs/HDACs and recruits CoAs to restrict access of HATs to CREB and other components of the basal promoter, thereby causing ligand-dependent repression. The specificity of this effect is encoded by the promoter elements and transcription factors that bind to negatively regulated genes, as opposite effects are seen with positively regulated promoters studied under identical conditions.

MATERIALS AND METHODS
Plasmid Constructions-The mutant hTR␤1 cDNAs were prepared by oligonucleotide-directed mutagenesis, subcloned into pCMX, and verified by DNA sequencing as described previously (27,31). The numbering of the amino acid residues of TR␤ is based on a consensus nomenclature (32). The pCMX-NCoR expression vector was provided by M. G. Rosenfeld (University of California, San Diego, CA) (3), and pCMX-ACTR expression vector was provided by R. M. Evans (Salk Institute, San Diego, CA) (15). The HD1 (HDAC-1) cDNA was provided by C. A. Hassig (Harvard University, Cambridge, MA) (33); SRC-1 cDNA was provided by B. W. O'Malley (Baylor College of Medicine, Houston, TX) (9); F-SRC1 and TRAM-1 cDNAs were provided by A. Takeshita and W. W. Chin (Brigham and Womens Hospital, Boston) (16,34); and CBP was provided by R. H. Goodman (Vollum Institute, Portland, OR) (35). The CREB mutant (S119A) was provided by J. M. Leiden (University of Chicago, Chicago) (36). These cDNAs were transferred into pCMX. Gal4-CREB and Gal4-CREB (S119A) were provided by M. Z. Gilman (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY) (37). Full-length TR␤ was fused downstream of the VP16 activation domain in frame to create VP16-TR, and the TR interaction domains of NCoR were similarly linked to VP16 to generate VP16-NCoR (38) in pAASV. The expression vector for the catalytic subunit of the cAMPdependent protein kinase (RSV-Cat-␣) was provided by R. A. Maurer (University of Iowa, Iowa City, IA) (39).
The plasmid TREp-tk-Luc contains two copies of a palindromic TRE upstream of the thymidine kinase promoter (tk109) in the pA3 luciferase vector (40). DR4-tk-Luc and F2-tk-Luc were provided by P. M. Yen (National Institutes of Health, Bethesda) (41). TSH␣-Luc and the 5Ј deletion constructs created from it have been described (27). The Gal4 reporter plasmid, UAS-tk-Luc, contains two copies of the Gal4 recognition sequence (UAS) upstream of tk109, and UAS-E1BTATA-Luc (42) contains five copies of UAS upstream of E1BTATA in pA3-Luc.
Transient Expression Assays-TSA-201 cells, a clone of human embryonic kidney 293 cells (43), were grown in Opti-MEM (Life Technol-ogies, Inc.) with 4% Dowex resin-stripped fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml) and were transfected by the calcium phosphate method as described previously (40). The total amount of expression plasmid DNA was kept constant in the different experimental groups by adding corresponding amounts of the same plasmids without cDNA inserts. After exposure to the calcium phosphate-DNA precipitate for 8 h, Opti-MEM with 1% Dowex resinstripped fetal bovine serum was added, in the absence or presence of 1 M T3. Ethanol or 300 nM trichostatin A (Wako Pure Chemical, Osaka, Japan) was added in some of the experiments. Cells were harvested after 40 h for measurements of luciferase activity (44). Results are expressed as the mean Ϯ S.E. from at least three transfections, each performed in triplicate.
Chromatin Immunoprecipitation (CHIP) Assays-CHIP assays were performed as described previously (45,46) with the following modifications. TSA-201 cells were transfected with 5 g of Ϫ300 TSH␣-Luc and 10 g of the indicated expression plasmids. After 36 h, 1% formaldehyde was added to the culture media, and cells were incubated at room temperature for 15 min with mild shaking. Cells were collected and resuspended in lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1) with 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, pepstatin A, and aprotinin. After sonication for 25 s, lysates were cleared by centrifugation. An aliquot of the lysate (20 l) was removed as a control, and the remainder was diluted 10-fold with a solution containing 0.01% SDS, 1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.1, and 167 mM NaCl, protease inhibitors, and anti-acetylated histone H3 or H4 antibodies (Upstate Biotechnology, NY). After incubation with the antibodies overnight at 4°C, histone-DNA complexes were immunoprecipitated using protein A-agarose beads. Precipitants were sequentially washed twice with dilution buffer, followed by washing once with dilution buffer containing 500 mM NaCl. After the final wash, 500 l of elution buffer (1% SDS and 0.1 M NaHCO 3 ) was added, and beads were rotated at room temperature for 30 min. Subsequently, 5 M NaCl was added and incubated at 65°C for 4 h to reverse the formaldehyde cross-linking. DNA was recovered by proteinase K treatment, phenol/ chloroform extraction, and ethanol precipitation. Pellets were resuspended in water and subjected to PCR amplification using following primers: 5Ј CAGGATGTTATGTGTATGGCTC for the TSH␣ promoter and 3Ј CTTTATGTTTTTGGCGTCTTC for the luciferase gene. The 32 Plabeled PCR product (400 bp) was separated by polyacrylamide gel electrophoresis and quantitated using a PhosphorImager (Storm 860, Molecular Dynamics, CA).

Corepressors Mediate Basal Activation and Coactivators Mediate T3-induced Suppression of the TSH␣ Gene-Previous
studies have implicated CoRs in the basal activation of negatively regulated genes, such as the TSH␣, TSH␤, and TRH genes (30). Specifically, TR mutations (e.g. P214R) that impair interactions with CoRs reduce basal stimulation by unliganded TR but do not affect T3-dependent repression. In view of the apparently paradoxical effect of the CoRs on basal stimulation by the unliganded TR, we considered the possibility that CoAs might be involved in the T3-dependent repression of negatively regulated genes. For this purpose, the E457A mutant of TR␤ is informative. It binds T3 with near normal affinity, but it selectively lacks binding to CoAs, and it retains interactions with CoRs (47). In parallel, we examined the effects of another CoR mutant (TR␤ AHT), which contains three amino acid substitutions in the CoR interaction domain of TR but retains the ability to bind CoAs (3). The AHT mutation exhibits more complete elimination of CoR binding than the P214R mutation in TR␤ (data not shown).
By using the positively regulated promoter, TREp-tk-Luc, wild-type TR and the E457A mutant exhibited strong transcriptional silencing in the absence of T3, whereas the silencing ability of the AHT CoR mutant was abolished (Fig. 1A). Wildtype TR and the AHT mutant induced strong transcriptional stimulation in the presence of T3, but the T3-induced stimulation by the E457A mutant was reduced. These controls confirm the expected properties of these TR mutants in the context of positively regulated genes. By using the negatively regulated promoter, TSH␣-Luc, wild-type TR and the E457A mutant exhibited a comparable degree of basal activation in the absence of T3. However, basal activation was abolished using the AHT CoR mutant (Fig. 1B), confirming our previous observation using the P214R CoR mutant (30). Of note, T3-induced repression (below the basal level) was impaired using the E457A TR mutant but not with the AHT mutant. These results are consistent with a role for CoRs in basal activation, and CoAs in T3-induced repression, of negatively regulated genes.
Histone Deacetylase Plays a Pivotal Role in TR Regulation of the TSH␣ Gene-One potential mechanism for basal activation of the TSH␣ promoter by unliganded TR could involve partitioning of NCoR, SMRT, or other components of the repressor complex. Previously, we attempted to reverse TR-dependent activation of the TSH␣ promoter by overexpression of NCoR or SMRT (30). Unexpectedly, CoRs enhanced, rather than attenuated, basal activation by unliganded TR. This finding indicates that basal activation of these negatively regulated promoters is not caused by titration of limited quantities of CoRs. Recently, it was shown that the CoR⅐Sin3 complex also recruits histone deacetylases (HDACs) (6 -8), which are important for inducing alterations in histone structure and perhaps other transcription factors. We therefore hypothesized that the stimulatory effect of excess CoRs or Sin3 (data not shown) might involve the recruitment of HDAC by the TR⅐CoR complex.
To test this idea, the TSH␣ promoter was transfected with TR along with various combinations of CoRs and HDAC1. As found previously, cotransfection of NCoR enhanced basal activation by unliganded TR in comparison to the expression vector alone ( Fig. 2A). In contrast to NCoR, expression of HDAC1 reduced TR-mediated basal stimulation. Since HDAC1 did not alter basal expression in the absence of TR, this effect is TRspecific. Strikingly, coexpression of HDAC1 eliminated the NCoR-mediated enhancement of basal activation by unliganded TR. A similar effect of HDAC1 was seen in the presence of SMRT and Sin3 (data not shown). Of note, T3-dependent repression was maintained in the presence of NCoR or HDAC1. These results suggest that the TR⅐CoR complex may deplete limiting amounts of HDAC1 in the absence of T3, leading to stimulation of the TSH␣ promoter. In the presence of T3, when CoR complex dissociates from TR, overexpression of these factors does not appear to affect T3-mediated suppression of the TSH␣ promoter.
Coactivators Enhance the T3-induced Suppression of the TSH␣ Gene-In view of the finding that the E457A mutation in TR␤ (which selectively alters TR binding to CoAs) appears to abrogate T3-dependent repression, we tested the effect of coexpression of CoAs on negative regulation by TR. Expression of these CoAs (SRC1, FSRC1, glucocorticoid receptor interacting protein 1, TRAM1, and ACTR) enhanced T3-induced stimulation of positively regulated genes (Fig. 2B). Each of the CoAs tested enhanced the degree of T3-induced gene suppression by 2-3-fold. However, they had little or no effect on basal activation by unliganded TR. Thus, a role for CoAs in T3-mediated repression of the TSH␣ promoter is supported by mutations in TR that eliminate CoA binding and by overexpression of CoAs.
cAMP Response Elements (CREs) and the Basal Promoter Are Important for Negative Regulation of the TSH␣ Promoter by TR-The regions of the TSH␣ promoter that are involved in TR-mediated repression have been examined previously (27,48). These studies have suggested a role for the CREs as well as basal elements downstream of the TATA box. However, because deletion of the CREs greatly reduces the activity of the TSH␣ promoter, it is difficult to study repression by T3 after they have been removed. The finding that unliganded TR and NCoR enhance activity of this promoter provides an alternative strategy for localizing regulatory elements that are affected by the TR.
Various 5Ј deletion constructs of the TSH␣ promoter were therefore examined for loss of activation by unliganded TR, as well as alterations in T3-dependent repression. The activity of the promoter constructs in the absence of TR was normalized to 100% to allow the effect of TR to be presented on a comparable scale. As shown in Fig. 3A, stimulation of the TSH␣ promoter by unliganded TR was relatively constant until deletion of the CREs (between Ϫ156 and Ϫ100 bp), after which the degree of basal stimulation by the TR decreased by 80%. However, a small degree of TR-dependent basal activation was consistently seen using the minimal promoter constructs, including Ϫ30 TSH␣-Luc.
Because CoRs enhance the degree of activation by unliganded TR, NCoR was cotransfected with TR to more clearly examine promoter regions involved in basal activation (Fig. 3B). NCoR strongly stimulated the TR-mediated enhancement of Ϫ156 TSH␣-Luc. However, significant enhancement by NCoR was also observed using shorter versions of the promoter, such as Ϫ100 or Ϫ30 TSH␣-Luc, but not with the backbone plasmid, pA3-Luc. T3 suppression was 60-fold using Ϫ156␣-Luc and 9-fold using Ϫ100␣-Luc. Taken together, these findings support the idea that the CRE region, which resides between Ϫ156 to Ϫ100 bp of the TSH␣ promoter, is involved in negative regulation by TR. However, the basal promoter region (Ϫ30 to ϩ44 bp) also appears to retain features of a negatively regulated promoter, independent of the CRE.
Evidence That TR Is Not Tightly Bound to the Promoters of Negatively Regulated Genes-The TSH␣ promoter contains weak binding sites for TR, when assessed by gel mobility shift assays (27,48). However, site-directed mutagenesis of these putative binding sites fails to alter T3-dependent repression, casting doubt on their functional importance (48). To explore this issue further, we used two novel experimental approaches to detect whether TR is bound to the promoter of the TSH␣ gene. In the first approach, full-length wild-type TR␤ was fused to the transcriptional activation domain of VP16 to generate a constitutively active receptor, even in the absence of T3 (Fig.  4A). As controls, several positively regulated genes containing known thyroid hormone response elements (TREs), such as DR4, F2, and TREp, were tested for responsiveness to this constitutively active version of TR␤. The VP16-TR construct strongly stimulated each of these reporter genes. VP16 alone had no effect, indicating a specific requirement for the TR portion of the fusion gene. Addition of T3 stimulated transcription of these promoters further, probably by recruiting additional CoAs to the TR-LBD portion of this chimeric protein (data not shown). In contrast, UAS-tk-Luc, which contains Gal4-binding sites (UAS) and lacks a TRE, was not activated by VP16-TR. In the case of TSH␣ promoter, no activation by the constitutively active VP16-TR construct was observed, consistent with the idea that it does not contain high affinity TRbinding sites.
In an independent approach, a modified mammalian twohybrid assay was used to detect whether TR is bound to DNA (Fig. 4B). For example, the TR interaction domains of either NCoR (residues 1552-2453) or retinoid X receptor (residues 199 -462) were used to seek TR bound to various promoters. When the native TR was cotransfected with the positively regulated reporters, VP16-NCoR induced transcriptional acti-vation in the absence of T3, reflecting the TR-NCoR interaction (Fig. 4B). Similarly, VP16-NCoR stimulated the activity of Gal4-TR, using the UAS-tk-Luc reporter gene. In contrast, VP16-NCoR did not stimulate the activity of the TSH␣ promoter in the presence of native TR. Similar results were seen when VP16-retinoid X receptor was used as the interacting protein (data not shown). These results indicate that TR is not bound to the TSH␣ gene in a manner that is detected either by direct receptor binding to the promoter or by TR interactions with other binding proteins.
CREB Is Involved in Negative Regulation of the TSH␣ Promoter by T3-There are two consensus CREs between Ϫ156 and Ϫ100 bp of TSH␣ promoter (49). Since there is a marked difference in the degree of negative regulation by TR after deletion of the CREs, we tested the possibility that the CREbinding protein, CREB, might be a target of TR action. A dominant negative mutant of CREB (S119A), which is not capable of phosphorylation (36), was used to assess further the functional significance of CREB for TR regulation of the TSH␣ gene. Expression of the CREB S119A mutant completely blocked basal activation by TR and NCoR. T3-dependent repression was reduced from 12-to 1.8-fold in the presence of the CREB S119A mutant (Fig. 5A). This finding suggests that occupancy of the CREs by the inactive form of CREB precludes effective negative regulation of the promoter by TR.
CBP is a target protein of both phosphorylated CREB (35)

FIG. 2. Role of CoRs and CoAs in control of the TSH␣ gene by TR. A,
CoRs enhance, and HDAC reverses, basal stimulation of the TSH␣ gene by TR. The NCoR and/or HDAC expression plasmids (500 ng) were cotransfected into TSA-201 cells together with 100 ng of the negatively regulated reporter gene, TSH␣ (Ϫ846)-Luc, with or without 100 ng of TR, in the absence or presence of T3. B, CoAs enhance T3-dependent suppression of the TSH␣ gene. The indicated CoA expression plasmids (500 ng) were cotransfected into TSA-201 cells together with 100 ng of the negatively regulated reporter gene, TSH␣ (Ϫ846)-Luc, with or without 100 ng of TR, in the absence or presence of T3. The numbers above each bar indicate fold repression mediated by TR (ϪT3/ϩT3). and nuclear hormone receptors (50,51). We therefore examined the effects of CBP and catalytic subunit of protein kinase A (PKA) on negative regulation by T3. Coexpression of CBP attenuated the negative regulation of Ϫ156␣-Luc by T3, decreasing repression from 12-to 3-fold. In contrast, CBP enhances T3-stimulated activity of positively regulated genes (data not shown) (50). Interestingly, activation of the TSH␣ promoter by PKA completely eliminated the T3-dependent suppression of the promoter by T3. This effect of PKA is selective for the negatively regulated gene as T3-induced stimulation of positively regulated genes was preserved, or enhanced, in the presence of PKA (data not shown). In contrast, using the Ϫ100␣-Luc construct, which does not contain CREs, there was little or no effect of the CREB S119A mutant, CBP, or PKA (data not shown). These results support a role for activated CREB in TR regulation of the TSH␣ gene.
The possibility that CREB is involved in negative regulation by TR was also examined using Gal4-CREB and the UAS-E1BTATA-Luc reporter, which contains five Gal4-binding sites upstream of the minimal adenovirus E1B promoter (Fig. 5B). TR and T3 exhibit little effect on this promoter in the presence of the control, Gal4-DBD. The addition of Gal4-CREB induces basal activity and confers modest repression (3.9-fold) in the presence of TR and T3. In contrast, there was no effect of T3 using the Gal4-CREB S119A mutant. Similar to the native TSH␣ promoter, cotransfection of CBP or PKA with Gal4-CREB markedly stimulated its activity. In the context of Gal4-CREB, PKA or coexpression of CBP eliminated T3-induced repression by native TR. These results indicate that CREB is a functional target of TR and that it is sufficient to confer T3-dependent repression. In addition, stimulation of CREB activity by PKA, which is expected to generate a CREB⅐CBP complex, inhibits T3-mediated repression.
Negative Regulation of the TSH␣ Gene and CREB Are Sensitive to a Histone Deacetylase Inhibitor, Trichostatin A-In light of the data that coexpression of HDAC1 reverses TR-NCoR-mediated basal activation of the TSH␣ promoter and that CREB-CBP might be target for T3-dependent repression, we used a histone deacetylase inhibitor, trichostatin A (TSA) (52), to assess further the role of histone deacetylase in TSH␣ gene regulation. The effects of TSA were initially tested using the Gal4-CREB constructs (Fig. 6A). TSA markedly stimulated the activity of Gal4-CREB in comparison to its effects on transcription mediated by Gal4-DBD or the Gal4-CREB S119A mutant, suggesting that inhibition of HDAC enhances the ability of CREB-CBP to activate transcription. The sensitivity to TSA was also examined using various deletion mutants of the TSH␣ promoter (Fig. 6B). The Ϫ156␣-Luc construct, which FIG. 4. Detection of TR bound to promoters using mammalian two-hybrid assays. The format of the modified mammalian two-hybrid experiment is shown at the top of each figure. A, the VP16 or VP16-TR expression plasmids (300 ng) were transfected into TSA-201 cells together with 100 ng of the indicated luciferase reporter genes in the absence of T3. B, the VP16 or VP16-NCoR expression plasmids (300 ng) were cotransfected, in the absence of T3, with 50 ng of native TR and the indicated reporter genes (100 ng) or with 50 ng of Gal4-TR and UAS-tk-Luc (100 ng).

FIG. 5. CREB is involved in negative regulation of the TSH␣ gene by T3.
A, the TSH␣ (Ϫ156) reporter plasmids (100 ng) were cotransfected into TSA-201 cells together with 100 ng of TR, with or without the mutant CREB S119A, CBP, and PKA, in the absence and presence of T3. B, the indicated Gal4 expression plasmids (50 ng) and 100 ng of the Gal4-responsive reporter gene, UAS-E1BTATA-Luc, were cotransfected with 100 ng of TR, with or without CBP and PKA, in the absence and presence of T3. The numbers above each bar indicate fold repression mediated by TR (ϪT3/ϩT3). contains the CREs, was stimulated about 200-fold by treatment with TSA. The shorter TSH␣ promoter constructs (Ϫ100␣-Luc, Ϫ30␣-Luc), which lack the CREs, were also stimulated 20 -30fold by TSA, but the degree enhancement was greatly reduced. Cotransfection of the CREB S119A mutant with Ϫ156␣-Luc markedly decreased its induction by TSA, likely because of a dominant negative effect to block the binding of CREB. These results indicate that inhibition of HDAC activity enhances the activity of CREB-mediated transcription, suggesting a dynamic equilibrium between the processes of histone acetylation and deacetylation.
We next examined whether inhibition of HDAC by TSA would alter the ability of T3 to repress transcription of the TSH␣ promoter. As shown in Fig. 6C, treatment with TSA markedly increased the basal activity of the TSH␣ promoter in the presence of unliganded TR (note the different scale on the y axis). TSA also abolished the negative regulation by T3, suggesting that deacetylation is an essential feature of negative regulation of the TSH␣ gene by T3.
Effects of Unliganded TR and T3 on Histone Acetylation Associated with the TSH␣ Gene-The transfection assays shown above suggest that unliganded TR can compete for HDACs that otherwise repress TSH␣ promoter activity. Chromatin immunoprecipitation (CHIP) assays (45,46) were performed to examine the direct effects of TR on histone acetylation associated with TSH␣ promoter. In this assay, it is possible to assess the acetylation status of specific genes under different treatment conditions. The TSH␣ (Ϫ300/ϩ44) plasmid was transfected into TSA201 cells along with TR in the absence or presence of T3. Extracts from the transfections were immunoprecipitated with anti-acetylated histone H4 antibody, and TSH␣ promoter sequences that coimmunoprecipitate were detected by PCR amplification. The total amount of transfected plasmid in the cell was also determined by PCR using the aliquots of extracts before immunoprecipitation. The immunoprecipitated results were corrected for recovery using the total PCR products for each treatment (Fig. 7). As a control, transfection of PKA and CBP increased acetylated histone H4 associated with the TSH␣ promoter. To a lesser degree, treatment with TSA also increased histone H4 acetylation. In the absence of T3, TR increased acetylated H4 by 3-fold. Cotransfection of NCoR enhanced H4 acetylation by unliganded TR. These effects were reversed by the addition of T3. Similar results were obtained using anti-histone H3 antibody (data not shown). Thus, the state of acetylated histone H3 and H4 associated with the TSH␣ promoter reflects the transcriptional effects seen in transient expression assays. FIG. 7. Regulation of acetylated histone H4 associated with the TSH␣ gene. TSH␣(Ϫ300)-Luc (5 g) was transfected into TSA-201 cells together with 10 g of the indicated expression vectors in the absence or presence of T3. Lane 1 (Control) represents a control without transfection. Acetylated histone H4 associated with the TSH␣ promoter was determined using a CHIP assay as described under "Materials and Methods." Lane 2 (ϩTSA) was treated with 300 nM TSA. The total amount of transfected TSH␣ DNA was determined by ethanol precipitation of cell lysates followed by PCR amplification of promoter sequences. The 32 P-labeled PCR product was separated by polyacrylamide gel electrophoresis and quantitated using an image analyzer. The results shown in the top panel are expressed as the relative amount of PCR products coimmunoprecipitated with acetylated H4 corrected by the amount of total PCR products.

DISCUSSION
In this report, we make the following observations about T3-dependent regulation of the TSH␣ gene by TR. 1) The property of negative regulation is promoter-dependent as other positively regulated promoters are stimulated by T3 under the same experimental conditions. 2) Direct interaction of TR with the TSH␣ promoter is not necessary for its control, implicating protein-protein interactions as a primary mechanism of its regulation. 3) Ligand-independent stimulation of the promoter involves CoRs and is associated with increased histone acetylation. 4) Ligand-dependent inhibition of the gene involves CoAs and is associated with decreased histone acetylation. 5) The CRE is a critical regulatory element that contributes to the reversed pattern of regulation by CoRs and CoAs, and its properties can be replicated using Gal4-CREB and a heterologous promoter. 6) In addition to the CRE, the basal TSH␣ promoter is intrinsically subject to negative regulation. 7) Shifting the balance of histone acetylation, by inhibition of histone deacetylation with TSA, or by treatment with PKA, prevents T3-dependent repression. Taken together, these data support a mechanism in which the TSH␣ gene is controlled by TR-mediated partitioning of HATs and HDACs.
In previous studies, we demonstrated that proximal TSH␣ promoter is sufficient for negative regulation by T3 (27,48). Although in vitro translated or recombinant TR binds to several sites within this region, these binding sites are not high affinity, at least in comparison to positive TREs (48). In addition, mutagenesis of the low affinity TR-binding sites did not alter negative regulation by T3. These experiments raised the possibility that negative regulation of this promoter by TR might be mediated by TR interactions with promoter-bound proteins, as opposed to TR binding directly to DNA sequences. This concept is supported by a recent finding that the DNA binding domain of the TR is not required for negative regulation of this promoter (30). When the ligand binding domain of TR was linked to Gal4, the Gal4-TR LBD fusion protein was sufficient to mimic negative regulation by the native TR. Specifically, Gal4-TR induced activation in the absence of ligand, and the addition of T3 resulted in ligand-dependent repression. In this report, we provide additional evidence for the absence of direct binding of the TR to the TSH␣ promoter by using oneand two-hybrid assays to identify potential TR-binding sites. No TR-binding sites were detected using either VP16-TR in the one-hybrid assay or using VP16-NCoR and TR in the twohybrid assay. Taken together, these observations provide several independent lines of evidence to support the idea that negative regulation of this promoter does not involve direct interactions with DNA. These findings therefore share certain mechanistic elements in common with glucocorticoid repression of genes that are regulated by AP-1 or NFB (24).
The absence of direct DNA binding of the TR raises interesting questions concerning how the receptor acts to inhibit transcription and how specificity for certain promoters is achieved. Clearly, the effects of unliganded and T3-bound TR must be restricted to certain genes, even if the receptor is not tightly bound to DNA. Parallel experiments with positively regulated promoters demonstrate that TR affects them in a manner that is diametrically opposed to the effects on the TSH␣ promoter. Thus, the nature of the promoter and the transcription factors that bind to it are sufficient to encode positive or negative regulation by TR. What are the key regulatory elements for negative regulation of the TSH␣ promoter? Unfortunately, the apparent importance of protein-protein interactions in this process makes it difficult to localize these sequences by traditional mutagenesis (48). However, we were able to take advantage of the newly discovered feature that unliganded TR and NCoR enhance basal activation of the TSH␣ promoter to investigate sequences involved in this process, as well as T3-mediated repression. These experiments confirm previous findings that both the CREs (Ϫ156 to Ϫ100) and the proximal promoter elements (Ϫ30 to ϩ44) play a role in the action of TR (27,48). For example, deletion of the CREs in the TSH␣ promoter greatly reduces stimulation by unliganded TR and T3-induced repression. However, it is notable that features of negative regulation remain, even after deletion of the CREs, confirming previous studies in other cell lines (27). This feature of negative regulation by TR appears to be intrinsic to the proximal promoter, since T3-dependent repression is retained when it is linked to other enhancers (data not shown) (48). It seems likely that this region contains transcription factors that recruit HDAC. This idea is supported by the activation of this region of the promoter by the HDAC1 inhibitor, TSA. Several repressors have been reported to recruit HDAC to the promoter, including YY1 (53), Ume6 (54), retinoblastoma protein (45,55,56), and PLZF (57), among others. At present, it is not known whether these factors interact with the TSH␣ promoter. Recently, MeCP2, which binds to methylated CpG in genomic chromatin (58), was reported to bind to HDAC (59). The minimal TSH␣ gene contains several CpG sequences, and it is possible that MeCP2, or general transcription factors that bind to the TSH␣ promoter, are involved in the recruitment of HDAC. Thus, the configuration of regulatory elements, perhaps relative to positions of nucleosomes, and the nature of the transcription factors that are recruited to the promoter may have an important influence on whether TR acts in a positive or negative manner.
Several lines of evidence point to a critical role for transcription factor CREB in negative regulation of the TSH␣ promoter by TR. In addition to mutations of the CRE, which reveal a partial loss of TR regulation, maneuvers that alter the function of CREB have profound effects on negative regulation. The dominant negative CREB mutant, S119A, which cannot be activated by PKA, blocked stimulation by TR and repression by T3. Cotransfection of CBP, a coactivator of CREB, attenuated negative regulation, and maximal activation of CREB by coexpression of PKA abolished T3-mediated repression. These findings suggest that part of the mechanism of T3-mediated repression involves reversal of the activation state of CREB. Several aspects of negative regulation can be recapitulated using Gal4-CREB, in the context of a heterologous promoter. The activity of Gal4-CREB is stimulated by unliganded TR and NCoR, and these effects are reversed by T3. It is clear that, in the context of a native gene, CREB is not sufficient to dictate negative regulation by TR, as it binds to numerous other cellular genes that are not negatively regulated. However, these experiments underscore its potential importance as a target of TR action, perhaps because both TR and CREB interact with additional proteins involved in the process of protein acetylation. These results echo the findings of Kamei et al. (50) who reported that CBP is a limiting factor for nuclear receptor inhibition of AP-1. In this respect, it is notable that many genes that are negatively regulated by TR contain AP-1-binding sites. These include the collagenase (60 -62), prolactin (63), c-fos (61), and TSH␤ (64) promoters. The TRH promoter appears to contain a composite site for both CREB and AP-1 (65). In the GHF-1/ pit-1 gene, the CRE has been reported to be a target for negative regulation by TR through protein-protein interactions (66). Thus, genes that contain AP-1/CRE sites may be poised for negative regulation by nuclear receptors, perhaps reflecting competition for CoAs like CBP.
Although additional studies will be necessary to define exactly how TR interacts with transcription factors that bind to CREB and to the basal region of the TSH␣ gene, it is apparent that TR interactions with CoRs and CoAs are involved in this process. Previously, we reported (30) the paradoxical observation that CoRs are involved in basal stimulation of negatively regulated genes by unliganded TR. These findings were supported by the fact that a TR mutation (P214R) that selectively disrupts interactions with CoRs prevented basal stimulation by TR. This result was confirmed in this study using a different mutation in TR (AHT) that more completely disrupts TR interactions with CoRs. Initially, we postulated that stimulation of the TSH␣ promoter by unliganded TR might be accounted for by competition for a limiting amount of CoRs that were bound to other factors on the promoter. However, when excess CoRs were cotransfected with TR, we found that they enhanced, rather than reversed basal activation (30). In view of recent findings that CoRs, like NCoR and SMRT, form a complex with Sin3 and HDAC to mediate transcriptional repression via histone deacetylation (6 -8), we hypothesized that the ability of CoRs to enhance further basal activity might reflect the recruitment of a limiting amount of HDAC by the TR⅐CoR complex. Consistent with this idea, cotransfection of excess HDAC eliminated the basal activation by unliganded TR and CoRs. Supporting this result, inhibition of HDAC activity by TSA abolished the T3-induced repression by TR. These findings strongly implicate HDAC as a pivotal factor that controls TR regulation of the TSH␣ gene. It is notable that this promoter is very strongly induced by TSA. This feature may relate the ability of TR to negatively regulate this gene, in that relatively subtle alterations in the balance of acetylation mediated by CREB-CBP versus HDAC can induce marked changes in the activity of this gene.
Given the involvement of CoRs/HDAC and CREB/CBP in negative regulation, we propose a two-step model for TR control of the TSH␣ promoter (Fig. 8). In the absence of ligand, TR recruits CoRs, which in turn form a complex that includes HDAC. The withdrawal of HDAC from other target sites, such as the basal promoter, results in a net increase in histone acetylation, and consequently, increased basal transcription. Consistent with this step, CoR mutants such as P214R and AHT, only impair TR-mediated stimulation, without altering the degree of T3 suppression below the basal activity. The second step of negative regulation, involving T3-induced repression, is proposed to involve both CoRs and CoAs. The ligand-dependent release of CoRs and HDAC may facilitate deacetylation of the promoter. The ability of the HDAC inhib-itor to block T3-mediated repression is consistent with this idea. In addition, T3-induced recruitment of CoAs may restrict their access to other transcription factors like CREB. Supporting this possibility, a CoA mutant, E457A, was able to reverse CoR-mediated stimulation, but it did not exhibit T3-dependent suppression below basal activity. CoAs like SRC1 appear to be involved in this step, since their addition selectively enhances the degree of T3-dependent repression. In addition, CBP appears an important factor in this process as treatment with PKA, or overexpression of CBP, abrogates T3-dependent repression. In the case of TSH␣ gene, it is likely that phospho-CREB is a critical factor that participates in the binding of CBP. This novel mechanism for negative regulation, involving TR-mediated alterations in histone acetylation, likely applies to other nuclear receptors and genes that are repressed, rather than stimulated by nuclear receptor ligands. FIG. 8. Two-step model of negative regulation of the TSH gene by T3. Left, basal expression of TSH gene. In the basal state, transcription factors recruit histone deacetylases (HDACs), which cause transcriptional repression. On the other hand, CREB binding to the CRE(s) recruits CBP and associated HATs to the promoter. The basal transcription of the gene is therefore regulated by the degree of acetylation of histones and other factors, which is determined by a balance between HDAC and HAT. Middle, in the presence of unliganded TR. Direct interactions of the TR with promoter regulatory elements may not be required to control negatively regulated genes. The TR-binding site or protein partner has not been identified in the case of the TSH␣ gene. In the absence of T3, TR binds CoRs, such as NCoR, SMRT, and Sin3, which recruit HDAC. This TR⅐CoR complex sequesters HDAC from the promoter, resulting in increased histone acetylation and transcriptional activation. Right, in the presence of T3-bound TR. In the presence of T3, TR dissociates CoRs, making HDAC available to the promoter. Liganded TR also binds to CoAs, which compete CBP away from the CREB on the promoter. Both of these events result in net deacetylation and transcriptional repression. RXR, retinoid X receptor; GTFs, general transcription factors.