Thyroid Hormone-independent Interaction between the Thyroid Hormone Receptor β2 Amino Terminus and Coactivators*

Thyroid hormone receptors (TRs) mediate hormone action by binding to DNA response elements (TREs) and either activating or repressing gene expression in the presence of ligand, T3. Coactivator recruitment to the AF-2 region of TR in the presence of T3 is central to this process. The different TR isoforms, TR-β1, TR-β2, and TR-α1, share strong homology in their DNA- and ligand-binding domains but differ in their amino-terminal domains. Because TR-β2 exhibits greater T3-independent activation on TREs than other TR isoforms, we wanted to determine whether coactivators bound to TR-β2 in the absence of ligand. Our results show that TR-β2, unlike TR-β1 or TR-α1, is able to bind certain coactivators (CBP, SRC-1, and pCIP) in the absence of T3 through a domain which maps to the amino-terminal half of its A/B domain. This interaction is specific for certain coactivators, as TR-β2 does not interact with other co-factors (p120 or the CBP-associated factor (pCAF)) in the absence of T3. The minimal TR-β2 domain for coactivator binding is aa 21–50, although aa 1–50 are required for the full functional response. Thus, isoform-specific regulation by TRs may involve T3-independent coactivator recruitment to the transcription complex via the AF-1 domain.

Thyroid hormone receptors (TRs) 1 belong to the superfamily of nuclear receptors and contain at least five discrete domains: 1) the amino-terminal A/B domain containing AF-1 function; 2) the DNA-binding or C domain, which is highly conserved among nuclear receptors; 3) the hinge region or D domain, where corepressors bind; 4) the ligand-binding or E domain; and 5) the carboxyl-terminal AF-2 or F domain (1). TR acts as a transcription factor on thyroid hormone response elements (TREs) in the absence and presence of its ligand, triiodothyronine (T 3 ) (Ref. 2, and for review, see Refs. [3][4][5]. On positively regulated TREs (e.g. growth hormone, malic enzyme, myosin heavy chain-␣), gene expression is repressed in the absence of T 3 and stimulated when T 3 binds to the TR (6 -8). In contrast, on negatively regulated genes (e.g. TSH ␣ und ␤ subunits, myosin heavy chain-␤), gene expression is activated in the absence of ligand and repressed in the presence of ligand (9 -11).
There are three known TR isoforms: TR-␣1, TR-␤1, and TR-␤2. A fourth isoform, ␣-2, does not bind T 3 and may inhibit the function of other TRs. The different isoforms of the TR are derived from two different genes, c-erbA-␣ and c-erbA-␤, found on different mammalian chromosomes. TR-␣1 and ␣2 are generated from the c-erbA-␣ locus by alternative RNA splicing of carboxyl-terminal exons (12)(13)(14), whereas the TR-␤ isoforms are derived from differential exon utilization of the c-erbA-␤ locus (15,16). TR-␤1 and TR-␤2 therefore differ only in their amino-terminal domains (A/B domains). Whereas TR-␣1 and TR-␤1 are expressed ubiquitously (19), TR-␤2 is expressed almost exclusively in hypothalamus (20) and pituitary (15) and, therefore, could play an important role in controlling the thyroid axis centrally (16). Within the TR-␤2 amino terminus are two distinct domains (n-terminal and c-terminal) that have been shown to mediate ligand-independent activation on positive and negative TREs, respectively (18,19).
Transcriptional regulation by TRs is modified by coactivating and corepressing proteins. Two corepressors, Nuclear receptor CoRepressor (NCoR) and Silencing Mediator of Retinoic and Thyroid hormone receptors (SMRT), have been shown to bind to the hinge region of the TR in the absence of ligand (3,4). These corepressors mediate ligand-independent repression on positive TREs, probably through deacetylation of histones (23)(24)(25). Binding of the ligand results in release of the corepressor and recruitment of coactivators.
Over the last years a number of coactivators have been described that interact with the TR, including the CREB Binding Protein, CBP (26), Steroid Receptor Coactivator-1, SRC-1 (5), the CBP Interacting Protein, pCIP (27)(28)(29), p120 (30), and P300/CBP Associated Factor, pCAF (28). These coactivators contain LXXLL motifs that bind to the AF2 domain of liganded TR (31). The majority of these proteins have been shown to contain intrinsic histone acetyltransferase activity (28,(32)(33)(34) and probably function as activators through this mechanism. The ligand-independent activation of transcription by the TR-␤2 isoform could be mediated by binding of these cofactors to the amino terminus (AF-1 domain) of the receptor. This could explain the greater ligand-independent activation of TR-␤2 compared with TR-␤1 in some transfection systems. We therefore investigated the interaction of coactivators with the A/B (AF-1) domain of TR isoforms.

EXPERIMENTAL PROCEDURES
Constructs Used in Transfection Assays-The TRE constructs contain two copies of an idealized TRE (DRϩ4, LYS, PAL) upstream of a minimal thymidine kinase promoter (109 bp of 5Ј-flanking DNA of the herpes simplex TK promoter fused to the luciferase reporter gene (35)). The cDNAs encoding the full-length coactivators CBP, SRC-1, pCIP (generous gift from Dr. W. Chin, Lilly Corp., Indianapolis, IN), p120, and pCAF (generous gift from Dr. Y. Nakatani, National Institutes of Health) were placed into the pSG5 expression vector. The cDNAs en-* This work was supported by grants from the National Institute of Health (to R. N. C., A. N. H., and F. E. W.) and by a research grant of the Deutschen Akademischen Austauschdienst (to C. O. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
coding human TR-␣1, TR-␤1, and TR-␤2 were inserted into the expression vector pSG5, which employs the SV40 early promoter (36). The human TR amino termini were obtained by polymerase chain reaction amplification of human full-length TR cDNAs (TR-␣1 and TR-␤1) or genomic DNA (TR-␤2) and ligated in-frame into an expression vector containing five copies of the GAL4 DNA binding domain. The TR-␤2 amino-terminal deletion constructs were made using polymerase chain reaction to introduce an EcoRI site at the 5Ј-end and an XbaI site at the 3Ј-end of the constructs. The amino-terminal cDNAs were cloned inframe with the GAL4 DNA as an EcoRI-XbaI fragment in the GAL4 vector (37). The reporter used for heterologous expression systems was UAS-TK fused upstream of the luciferase gene (38). The integrity of all constructs was confirmed by restriction endonuclease digestion and dideoxy sequencing.
Transfection Assays-Transient transfection studies were performed in JEG cells. Transfections were performed in 12-well plates on subconfluent cells, using the calcium-phosphate technique without glycerol shock. In the 12-well format, 1 g of reporter with 0.2 g of TR-construct and 0.33 g of coactivator-pSG5 per well were transfected. Sixteen hours after transfection, culture medium was replaced, and 10 nM T 3 was added as indicated. 36 -40 h after transfection, cells were harvested and assayed for luciferase activity. Luciferase activity is expressed as -fold stimulation compared with transfection with the empty vector alone.
GST Assays-The amino-terminal deletion constructs used in the transfection assays were removed from the GAL4 vector using restriction digest (EcoRI and XbaI) and ligated in-frame with GST in the pGEX4T2 (Amersham Pharmacia Biotech). Recombinant proteins were synthesized in JM 109 bacteria and purified on glutathione-Sepharose resin under nondenaturing conditions, and in the presence of protease inhibitors (Complete TM , Roche Molecular Biochemicals), GST proteins were analyzed on SDS-PAGE before use in the assay to ensure equivalence of preparations. 35 S-labeled co-activators (CBP, SRC-1, pCIP, p120, or pCAF) were generated in an in vitro transcription/translation system (TNT, Promega Biotech, Madison, WI). As a control, an unprogrammed translation with 35 S-methionine was employed. After a 30min exposure of the translated proteins to the indicated GST protein and extensive washing with NET (150 mM NaCl, 1 mM EDTA, and 0.5% Nonidet P-40) at 4°C, the proteins trapped by the resin were resolved on SDS-PAGE and detected by autoradiography. Relative binding of the coactivators was quantified by densitometry.

RESULTS
In the absence of ligand (T 3 ), TR-␤2 has an increased capacity to activate gene expression on negative TREs, when compared with the other TR isoforms (19). We first wanted to determine whether this ability of TR-␤2 is mediated by its unique amino terminus. As shown in Fig. 1A, transient transfection studies of the isolated amino terminus of TR-␤2 fused to the GAL4 DNA binding domain exhibits greater ligand-independent activation than the amino termini of TR-␣1 or TR-␤1 (7-fold versus 0.5-or 2.0-fold, respectively) on a GAL4 reporter (UAS-TK). Thus, the TR-␤2 amino terminus contains a T 3 - To investigate whether the amino terminus of TR-␤2 functionally interacts with coactivator proteins, we cotransfected the isolated TR amino termini-GAL4 constructs and cDNAs encoding different coactivators: CBP, SRC-1, pCIP, p120, and pCAF. As shown in Fig. 1B, cotransfection of CBP, SRC-1, or pCIP expression vectors with the amino terminus of TR-␤2 enhanced transcriptional activation 42-, 50-, and 43-fold, respectively. In contrast, cotransfection of these coactivators with the amino-terminal domains of the TR-␣1 or TR-␤1 did not increase reporter gene expression over base-line activity observed with a GAL4 "empty vector." Furthermore, cotransfecting p120 or pCAF with the TR-␤2 amino terminus did not augment its activity, indicating that the TR-␤2 amino terminus specifically interacts with certain coactivators (CBP, SRC-1, and pCIP).
To evaluate whether these coactivators bind to the TR-␤2 amino terminus in vitro, we next performed GST interaction assays. As shown in Fig. 2A, radiolabeled CBP, SRC-1, and pCIP bound avidly to the amino terminus of TR-␤2, but not to the TR-␣1 or TR-␤1 amino terminus. The coactivators pCAF and p120 did not bind to any of the GST amino-terminal fusion constructs. Fig. 2B confirms that equivalent amounts of GST proteins were used in the GST interaction assays. These data support the results from the transfection assays, suggesting a functional and structural T 3 -independent interaction between the TR-␤2 amino terminus and certain coactivators.
To address the question of whether this interaction is of importance in the context of the full-length receptor, we next performed transfection assays and GST-interaction assays with the full-length receptors. Fig. 3A demonstrated that cotransfection of the full-length receptors with either a CBP or SRC-1 expression vector results in a specific 7-12-fold stimulation by TR-␤2 in the absence of T 3 . Deletion of the aminoterminal amino acids 1-50 (construct TR-␤2⌬1-50) resulted in complete loss of the ligand-independent activation of TR-␤2. In contrast, CBP or SRC-1 cotransfection yielded a 20-fold T 3 -dependent stimulation of reporter gene expression regardless of the isoform tested, indicating that all three TR isoforms and the TR-␤2 deletion construct showed similar functional interaction with these coactivators and the AF-2 domain in the presence of T 3 . Specific ligand-independent activation of TR-␤2 in the presence of CBP or SRC-1 was also observed on other response elements, Lys and Pal element, respectively (Fig. 3B). Fig. 3C supports these findings by demonstrating that full-length TR-␤2 in the absence of T 3 bound to CBP, SRC-1, and pCIP 25-, 15-, and 21-fold above background, respectively, as quantified by densitometry. In the presence of T 3 , there was only a small increase in binding of CBP, SRC-1, and pCIP to TR-␤2 (26-fold, 27-fold, and 29-fold, respectively). In contrast, full-length TR-␤1 interacted much less well with the same coactivators in the absence of T 3 (0.5-, 3.1-, and 6.1-fold, respectively) versus in the presence of T 3 (25-fold, 11-fold, and 25-fold, respectively). This structural assay also showed specificity, as p120 and pCAF did not exhibit T 3 -independent interaction with TR-␤2.
To isolate the region of the TR-␤2 amino terminus that is important for ligand-independent interaction with coactivators, we constructed a number of deletion constructs of the TR-␤2 amino terminus (shown in Fig. 4A) fused downstream and in-frame with the GAL4 DNA-binding domain. Shown in Fig. 4B are results with GAL4 fusion constructs tested on the UAS-TK reporter. When cotransfected with CBP or SRC-1, only constructs which contain amino acids 1-50 (N-␤2, 1-75, and 1-50) were completely sufficient to mediate reporter gene activation. Constructs containing only amino acids 21 to 50 (21-120 and 21-87) stimulated reporter gene expression about 60 -80% of the full-length TR-␤2 amino terminus, whereas constructs lacking amino acids 1-50 (51-120 and 89 -120) were unable to stimulate reporter gene activity.
We next expressed these deletion constructs as GST fusion proteins and then employed them in GST-interaction assays. As shown in Fig. 4C, the fusion proteins containing amino acids 1 to 50 (N-␤2, 1-75, and 1-50) were able to bind 35 S-labeled CBP, SRC-1, and p-CIP as efficiently as the full-length TR-␤2 amino terminus. In contrast, proteins with a deletion of the first 50 amino acids were unable to bind to these coactivators (51-120 and 87-120), and proteins retaining amino acids 21-50 showed significant but reduced coactivator binding. Fig. 4D demonstrates that equal amounts of GST fusion proteins were used in the GST-interaction assays. DISCUSSION Members of the nuclear hormone receptor superfamily activate gene transcription by binding to their cognate response elements in the regulatory regions of target genes either as monomers, homodimers, or heterodimers with the retinoid X receptor (RXR). Depending on the target gene and nuclear receptor, transcriptional activity is either activated or repressed in the presence of ligand. Ligand-dependent activation of NRs occurs principally through the AF-2 domain of the ligand-binding domain. In the presence of ligand, amino acid residues in helices 3 and 12 allow for the formation of a groove which binds to the LXXLL motifs of coactivator molecules (38). Coactivator molecules include members of the p160 family (SRC-1, TIF II, and ACTR), RIP 140, TRIP100, p300/CBP, and p120 as well as members of the DRIP complex and a number of FIG. 2. The TR-␤2 amino terminus specifically interacts with a subset of nuclear receptor coactivators. A, the TR amino termini were expressed as GST fusion proteins (N-␣1, N-␤1, and N-␤2) and used to pull down S 35 -labeled coactivators (CBP, SRC-, pCIP, p120, and pCAF). Input represents 50% of radiolabeled coactivator used in the assay. B, GST fusion proteins of the TR amino termini resolved on an SDS-PAGE gel, stained with Coomassie Blue. In the first lane, a molecular weight marker is shown.

FIG. 3. Ligand-independent interaction of the full-length TR-␤-2 with certain coactivators.
A, JEG 3 cells were transiently transfected in the absence or presence of T 3 (ϮT 3 ) with 0.2 g of the indicated TR fused to the GAL4 DNA binding domain (TR-␣1, TR-␤1, TR-␤2, and TR-␤2 ⌬ 1-50) or an empty GAL4 vector control (vector) with 0.33 g of coactivator in pSG5 and 1 g of a DRϩ4 luciferase reporter. The data are expressed as -fold activity Ϯ S.E., where 1 represents the luciferase activity of the GAL4 empty vector alone. B, JEG 3 cells were transiently transfected as in Fig. 3A. Instead of a DRϩ4 luciferase reporter, a Lys or a Pal luciferase reporter, respectively, was employed. C, full-length TR-␤1 and TR-␤2 were expressed as GST fusion proteins and used to pull down S 35 -labeled coactivators (CBP, SRC-1, pCIP, p120, and pCAF) in the absence and presence of ligand (ϮT 3 ).
other proteins (39,40). Although the mechanism of action of the coactivator complex has not been fully ascertained, it is believed that transcriptional activation is mediated, at least in part, by histone acetylation by the coactivator complex, which is formed in response to ligand.
In addition to the AF-2 domain present in the LBD, NRs also possess an AF-1 domain in their amino-terminal region or A/B domain. The AF-1 function has been shown to be responsive to growth factors in context of the ER and to be a target for phosphorylation by MAP kinase in context of PPAR␥ (41,42). Indeed, the down-regulation of PPAR␥ transcriptional activity by MAP kinase is because of a decrease in ligand-binding by PPAR␥ because of intermolecular communication between the A/B and LBDs. Thus, the activity of an NR cannot be viewed in context of the AF-2 domain alone, as the AF-1 domain may influence AF-2 function either through structural alterations or by independently recruiting other proteins. Both of these functions are supported by recent studies which demonstrate that the androgen receptor (AR) AF-1 and AF-2 domains interact in mammalian cells (43) and that the ER␣ AF-1 domain can bind members of the p160 coactivator family while the ER␤ AF-1 domain can also bind p160 members when phosphorylated by MAP kinase (44).
The TR isoforms differ most prominently in their A/B domains, though limited function of these domains has been shown. We and others have demonstrated that the separate TR isoform A/B domains affect DNA binding (35,45). In addition, a region of the TR-␣1 A/B domain directly recruits members of the basal trancriptional machinery (45). Furthermore, the TR-␤2 amino terminus appears to possess a function which differentiates its ligand-independent activity from the other TR isoforms. Indeed, the TR-␤2 amino terminus has been shown to be constitutively active when fused to a heterologous DNAbinding domain (18), and its separate activity on negative TREs maps to another unique region in the A/B domain (19).
Unlike the majority of nuclear receptors, TR isoforms possess both ligand-independent and dependent functions which are mediated by their ligand-binding domains. In the absence of ligand, TR-␣1 and TR-␤1 isoforms repress transcription on positive TREs through the recruitment of nuclear corepressors and resulting histone deacetylase containing complexes (39). In contrast, the TR-␤2 isoform is a poor repressor on positive TREs, indicating that its unique A/B domain may confer a separate activity. This is further supported by the increased ligand-independent activity of the TR-␤2 isoform on negative TREs (19), suggesting that the A/B domain of this isoform may alter its ability to recruit either coactivators or corepressors.
In the present study we have demonstrated that the TR-␤2 amino terminus allows for the recruitment of members of the p160 family and CBP through a specific domain located between amino acids 1-50. Indeed, this recruitment can be demonstrated in direct in vitro GST pull-down assays as well as in functional studies in mammalian cells where further activation of this region is seen in the presence of either p160 family members or CBP. This region corresponds to the activation function previously mapped by Sjoberg et al. (18) and suggests that the constitutive function of this isoform may be related to its ability to interact with coactivators in the absence of ligand. Importantly, specificity is also demonstrated in that pCAF, which is known to interact with the DNA-binding domain of the NRs (22), is unable to enhance the activation function of the TR-␤2 amino terminus. As well, p120, another NR coactivator, is unable to augment the function of, or bind to, the TR-␤2 amino terminus.
To ensure that the interaction of p160 family members and CBP with the TR-␤2 amino terminus is not artifactual, we have also demonstrated, using in vitro GST pull-down assays, that the entire TR-␤2 isoform selectively binds to these coactivators in the absence of ligand. Furthermore, these binding studies are in agreement with the functional effects of these coactivators on the entire TR-␤2 isoform in transfection studies. These data suggest that the recruitment of coactivators by the TR-␤2 amino terminus impairs ligand-independent repression by this isoform by either preventing the recruitment of corepressors or more likely by altering the ratio histone acetylation/deacetylation, which ultimately determines the degree of transcriptional activation. In this model, the TR-␤2 isoform would recruit both corepressors (through the LBD) and coactivators (through the amino terminus) which would prevent silencing. Tissue-specific expression of coregulators in the pituitary or hypothalamus, where TR-␤2 action is paramount, would ultimately affect the ligand-independent activity of this isoform. Further studies in hypothalamic TRH neurons and pituitary thyrotrophs will determine the cofactor profile in these cells.