The p160 Coactivator PAS-B Motif Stabilizes Nuclear Receptor Binding and Contributes to Isoform-specific Regulation by Thyroid Hormone Receptors*

Thyroid hormone receptors (TRs) are hormone-regulated transcription factors that play multiple roles in vertebrate endocrinology and development. TRs are expressed as a series of distinct receptor isoforms that mediate different biological functions. The TRβ2 isoform is expressed primarily in the hypothalamus, pituitary, cochlea, and retina, and displays an enhanced response to hormone agonist relative to the other TR isoforms. We report here that the unusual transcriptional properties of TRβ2 parallel the ability of this isoform to bind p160 coactivators cooperatively through multiple contact surfaces; the more broadly expressed TRβ1 isoform, in contrast, utilizes a single contact mechanism. Intriguingly, the PAS-B domain in the p160 N terminus plays a previously unanticipated role in permitting TRβ2 to recruit coactivator at limiting triiodothyronine concentrations. The PAS-B sequences also play an important role in coactivator binding by estrogen receptor-α. We propose that the PAS-B domain of the p160 coactivators is an important modulator of coactivator recruitment for a specific subset of nuclear receptors, permitting stronger transcriptional activation at lower hormone concentrations than would otherwise occur, and allowing isoform-specific mRNA splicing to customize the hormone response in different tissues.

Notably the transcriptional properties of TR␤2 differ from those of the otherwise closely related TR␤1 isoform. The TR␤1 isoform represses classical target genes in the absence of hormone, and only becomes a transcriptional activator on binding to hormone agonist such as T 3 (30 -34). TR␤2 in contrast does not repress, but instead modestly activates target genes even in the absence of hormone, and displays an enhanced ability to activate target genes compared with TR␤1 over a wide range of T 3 concentrations (30 -35). We have suggested that the enhanced T 3 response of TR␤2 permits tissues that preferentially express this isoform to respond to lower concentrations of hormone than do tissues that exclusively express TR␣1 or TR␤1 (34).
We wished to more fully understand the molecular basis behind the unusual transcriptional properties of the TR␤2 isoform. We report here that the enhanced transcriptional response of TR␤2 is closely paralleled by the enhanced ability of this isoform to bind to the p160 family of coactivators: SRC1 (also known as CoA-1), GRIP1 (also known as SRC2 or CoA-2), and activator of thyroid hormone receptor (ACTR) (also known as SRC3, or CoA-3). TR␤2, unlike TR␤1, binds to the p160 proteins in the absence of T 3 due to a hormone-independent interaction between the TR␤2 N terminus and a glutamine-rich (Gln-rich) region in the C-terminal domain of these coactivators. However, the enhanced ability of TR␤2 to recruit p160 coactivators at limiting T 3 concentration can be observed even in the absence of the Gln-rich domain, a phenomenon that requires the PAS-B domain within the p160 N-terminal region. The PAS-B domain itself does not bind detectably to any TR isoform tested, but instead greatly elevates the ability of the LXXLL motifs in the same coactivator to bind to TR␤2 under low T 3 conditions. Notably the PAS-B domain is also required for estrogen receptor (ER)-␣ to efficiently bind p160 coactivator in response to estradiol (E2), but plays little or no role in p160 coactivator recruitment by TR␤1, TR␤0, or farnesoid X receptor. We propose that the TR␤2 isoform possesses an enhanced transcrip-tional response to T 3 due to its ability to preferentially recruit p160 coactivators through a synergistic array of protein-protein interactions not available to TR␤1. We further propose that the PAS-B domain of the p160s plays an important, and previously undetected role in stabilizing coactivator recruitment by a subset of nuclear receptors and their isoforms.

EXPERIMENTAL PROCEDURES
Molecular Clones-The pSG5 expression vectors containing avian or human TR␤0, TR␣1, TR␤1, or TR␤2, and the DR4 M-pTK-luciferase reporter construct were previously described (32, 34, 36 -39). Full-length or subdomains of the SRC1, GRIP1, ACTR, or DRIP205/TRAP220 coactivators were either obtained in, or generated by subcloning into a pSG5 or pCR3.1 expression vector backbone. Glutathione S-transferase (GST)coactivator and GST-receptor fusions were created in pGEX vectors by using standard recombinant DNA methodologies (32,34,36,37). Mutations were created by the QuikChange Site-directed Mutagenesis Kit using the protocol recommended by the manufacturer (Stratagene, La Jolla, CA). All mutations were subsequently confirmed by sequence analysis.
Transient Transfections-Transfections using CV-1 cells were performed by the Effectene protocol as recommended by the manufacturer (Qiagen, Valencia, CA), using 10 ng of pSG5-TR expression vector, 100 ng of reporter plasmid, 60 ng of pCH110 as an internal transfection control, and sufficient pUC18 to bring the total DNA concentration to 200 ng (34). After a 24-h incubation, the transfection medium was replaced with fresh, hormone-depleted medium, and either ethanol carrier alone, or 3,3Ј,5-triiodo-L-thyronine (T 3 ) (Sigma) was added. The cells were incubated for an additional 24 h, harvested, and lysed in 100 l of Triton Lysis Buffer (0.2% Triton X-100, 91 mM K 2 HPO 4 , and 9.2 mM KH 2 PO 4 ). Luciferase and ␤-galactosidase activities were measured as reported (34). Protein expression levels were analyzed by immunoblot (34).
Protein-Protein Interaction Assays-GST pulldown assays were adapted to a microplate format that enhanced the reproducibility and sensitivity of the methodology (40). Briefly, GSTcoactivator fusion proteins were synthesized in Escherichia coli strain BL-21 transformed by the corresponding pGEX vector. The bacteria were lysed, and the GST fusion proteins were recovered and purified by binding to a glutathione-agarose matrix. The pSG5-TR plasmids were synthesized as 35 S-radiolabeled proteins in vitro using a TNT Quick kit (Promega Corp., Madison WI). Each radiolabeled protein (typically 2-5 l of TNT reaction product per assay) was incubated at 4°C with the immobilized GST fusion protein of interest (ϳ10 -20 ng immobilized to 5 l of agarose matrix per reaction) in a total volume of 100 l of Binding Buffer A (40). The binding reactions were carried out in 96-well multiscreen filter plates (Millipore, Bedford, MA); any given comparison of TR␤2 to TR␤1 was performed in parallel in the same plate. After a 2-h incubation with rocking at 4°C, the filter wells were washed 3 times with 200 l of ice-cold wash buffer (40), and any radiolabeled proteins remaining bound to the immobilized GST fusion proteins were subsequently eluted with 50 l of 20 mM glutathione in 100 mM Tris-HCl, pH 7.8. The eluted proteins were resolved by SDS-PAGE and were visualized and quantified using a PhosphorImager/STORM system (GE Healthcare) and the GraphPad Prism 4 statistical/plotting package (La Jolla, CA). Results were reproducible over different days and with different preparations of protein.
Coimmunoprecipitations were performed by introducing 200 ng of pSG5-SRC1a and 200 ng of either pSG5-HA-tagged TR␤1 or pSG5-HA-tagged TR␤2 into HeLa cells by the Effectin-mediated transient transfection protocol and modification as described in Ref. 33. Approximately 1.2 ϫ 10 5 cells were used per assay. After 47 h, 100 nM T 3 was added, or not, and the cells were incubated for an additional 1 h. The cells were lysed and the coactivator was immunoprecipitated with anti-SRC1 antibodies (Affinity BioReagents, PA1-840). The immunoprecipitates were analyzed by SDS-PAGE and immunoblotted using antisera directed against hormone-binding domain sequences shared by TR␤1 and ␤2 (Affinity BioReagents, MA1-215). The bands were visualized and quantified by Flurochem8900 Imager (Alpha Innotech, San Leandro, CA).
Protease Resistance Assay-All steps, unless otherwise noted, were performed on ice. 35 S-Radiolabeled SRC1-(1-781) or SRC1-(595-781) was synthesized using the TNT system and bound to the immobilized GST-TR␤1 or -␤2 in Binding Buffer A in the presence of 625 nM T 3 , as described above. The protein complexes were washed 3 times in 50 mM Tris-Cl, pH 7.5, containing 625 nM T 3 , and were resuspended in 25 l of the same buffer for each protease concentration employed. Twenty-five microliters of serially diluted protease (either Endoproteinase Glu-c "V8," or elastase, Sigma) were added to each sample, and the samples were incubated 15 min at 20°C with rocking (41). The reactions were terminated with 25 l of concentrated SDS-PAGE sample buffer, boiled 5 min, resolved by SDS-PAGE, and the proteolytic degradation products were visualized and quantified by phosphorimager analysis.

TR␤2 Displays an Enhanced Transcriptional Response to T 3 Hormone Compared with the Other TR Isoforms Due to Its
Unique N-terminal Domain-To compare the transcriptional properties of the different human TR isoforms ( Fig. 1), we expressed each isoform in CV-1 cells together with a luciferase reporter containing a cognate direct repeat (DR)-4 response element (34). CV-1 cells lack endogenous TRs, and there was virtually no effect of T 3 on the DR4-luciferase reporter in the absence of ectopic TR expression (see Ref. 34, and data not shown). As anticipated, ectopic introduction of the human TR␤1 isoform into these cells repressed luciferase expression below basal level in the absence of T 3 , and induced luciferase expression in the presence of T 3 ( Fig. 2A). In contrast, introduction of the human TR␤2 isoform into the CV-1 cells did not repress the DR4-luciferase reporter, but instead induced a modest activation of luciferase expression even in the absence of hormone ( Fig. 2A). Significantly, TR␤2 also displayed enhanced transcriptional activity in the presence of T 3 , with TR␤2 inducing higher levels of luciferase activity than did TR␤1 at all subsaturating T 3 concentrations ( Fig. 2A). Stated reciprocally, human TR␤2 induced an equal reporter gene activation as did human TR␤1, but at a ϳ5-fold lower T 3 concen-tration: half-maximal stimulation for TR␤2 was ϳ0.2 nM T 3 compared with ϳ1.0 nM T 3 for TR␤1. TR␤1 and ␤2 were expressed at nearly equal levels in the transfected cells, and the enhanced transcriptional activation properties of TR␤2 over TR␤1 were also observed in Chinese hamster ovary and 293 cells, and over a range of TR expression vector inputs (see Ref. 34, and data not shown). The third major mammalian TR isoform, TR␣1, closely resembled TR␤1 in our transfection assay, repressing in the absence of hormone and activating only in response to T 3 ( Fig. 2A).
The TR␤1 and TR␤2 isoforms differ only in their N-terminal domains, which are derived from different exons (Fig. 1A). We therefore examined two truncated forms of TR␤ to determine if these N-terminal domains contributed positively or negatively to TR function. TR␤0 is a naturally occurring isoform found in birds, reptiles, and amphibians and encodes a receptor with a highly truncated N-terminal domain (42,43) (Fig. 1A). TR␤0 is believed to represent an ortholog of mammalian TR␤1, and avian TR␤0 displayed transcriptional properties nearly identical to those of human TR␤1 when introduced into the CV-1 cells, whereas avian TR␤2 mimicked the properties of human TR␤2 (Fig. 2B). A human TR␤2 bearing an artificial N-terminal truncation (denoted TR␤1⌬N; Fig. 2A, dashed curve) yielded results similar to avian TR␤0 ( Fig. 2A). We conclude that the TR␤2 N-terminal sequences contribute in a positive manner, and are essential for the enhanced T 3 transcriptional response of this isoform in diverse vertebrates.
The Enhanced Transcriptional Response of the TR␤2 Isoform Parallels an Enhanced Ability to Bind to p160 Coactivators in Vitro and in Vivo-TR␤1 and TR␤2 bind T 3 and release SMRT and N-CoR corepressors in a comparable fashion, suggesting these properties are not the basis for the enhanced transcriptional properties of TR␤2 (34). We therefore focused on the interactions of TR␤1 and TR␤2 with coactivators. We first tested the ability of a full-length p160 coactivator, SRC1a, to bind to either TR␤1 or TR␤2. TR␤2 displayed an elevated ability to bind to SRC1a in a GST pulldown assay compared with TR␤1, and this was observed both in the absence of hormone and over a range of T 3 concentrations (Fig. 3, A, and quantified in B). Mutational disruption of all of the LXXLL motifs in SRC1a prevented hormone-dependent binding to either TR␤1 or TR␤2, although this LXXAA SRC1 mutant retained a mod-  . Enhanced reporter gene activation by TR␤2. An expression construct for each TR isoform was introduced into CV-1 cells by lipofection, together with a DR4-luciferase reporter and a pCH100-lacZ internal control. After 24 h in hormone-depleted medium, the cells exposed to T 3 as indicated on the bottom of each panel, were incubated for an additional 24 h, harvested, and the relative luciferase levels (absolute luciferase/␤-galactosidase units) were calculated. Fold induction equals the relative luciferase levels observed in the presence of each receptor compared with that observed for an empty expression vector control. The data represent the mean Ϯ S.D. (error bars) of two or more independent experiments; error bars smaller than data point symbols may not be visible. The EC 50 values for TR␤2 differed from that of TR␤0, TR␤1, TR␤2, and ⌬N-TR␤2 at a p value Ͻ 0.001. A, human-derived TR isoforms. B, avian (Gallus/chicken)-derived TR isoforms. est, hormone independent binding to TR␤2 that was not observed for TR␤1 (data not shown). Analogous results were observed for the other two members of the p160 family: GRIP1 and ACTR (Fig. 3, C and D). The enhanced ability of TR␤2 versus TR␤1 to bind SRC1a at limiting T 3 could also be observed in vivo using a co-immunoprecipitation method (Fig.  3E). In contrast to the p160 coactivators, binding of DRIP205, a distinct receptor coactivator (44,45), was indistinguishable between TR␤1 and TR␤2 (Fig. 3F).
Our results suggested that an enhanced affinity for p160 coactivators was at least one mechanism underlying the enhanced transcriptional activity of TR␤2 relative to TR␤1. If true, sufficient p160 expression in cells might overcome this difference in coactivator affinity between TR␤1 and TR␤2; consistent with this concept, increasing ectopic expression of SRC1a in CV-1 cells first elevated, then equalized the transcriptional activity of TR␤1 to that of TR␤2 (Fig. 3G). This equalization of TR␤1 and TR␤2 activity was not observed using a p205 DRIP coactivator, which enhanced both TR␤1 and TR␤2 activity in parallel, consistent with the equal affinities for this coactivator seen in vitro (Fig. 3H).
Enhanced p160 Binding by TR␤2 Reflects Both Hormone-independent and Hormone-dependent Interactions with the Coactivator-An enhanced ability of TR␤2 to bind to coactivator compared with that of TR␤1 was also observed when an SRC-1a-(595-1441) construct containing only the central LXXLL motifs plus the Gln-rich region, and was observed both in the absence and presence of T 3 (Fig. 4A, and schematic in Fig.  1B). SRC1a contains a fourth LXXLL receptor-interaction motif at its extreme C terminus; an SRC-1a construct limited to the Gln-rich domain and this C-terminal LXXLL also displayed preferred binding to TR␤2 compared with TR␤1 (Fig. 4B). SRC1a constructs containing the Gln-rich domain, but lacking all LXXLLs, bound to the TR␤2 independent of T 3 status and failed to bind to TR␤1 (Fig. 4C). Conversely, SRC1a constructs restricted to the central LXXLL motifs, or to the C-terminal LXXLL, but lacking the Gln-rich region, demonstrated little or no binding to either TR␤2 or TR␤1 minus T 3 , and exhibited near equal binding to both isoforms plus T 3 (Fig. 4D, and data not shown). These results indicate that the Gln-rich coactivator domain is necessary and sufficient for hormone-independent SRC1a binding by TR␤2, and contributes to the elevated ability of TR␤2 to bind SRC1a in the presence of T 3 ; the latter, how- . Enhanced p160 coactivator recruitment by TR␤2. A, preferential binding of SRC1a by TR␤2 compared with TR␤1. Full-length, 35 S-radiolabeled SRC1a was incubated with immobilized, GST fusions of full-length human TR␤1 or TR␤2 in the presence of increasing [T 3 ] as indicated. The resulting coactivator-TR complexes were characterized by SDS-PAGE and phosphorimager analysis. A representative phosphorimager scan is presented. B, preferential binding of SRC1a by TR␤2 compared with TR␤1 quantification. A series of GST pulldowns were performed as in panel A and quantified by phosphorimager analysis. The coactivator bound to each receptor is expressed as a percent of input. Data represent the mean Ϯ S.D. (error bars) of at least two independent experiments. C, preferential binding of GRIP1 by TR␤2 compared with TR␤1. Protocol was as in panel B. D, preferential binding of ACTR by TR␤2 compared with TR␤1. Protocol was as in panel B. E, preferential coimmunoprecipitation (IP) of TR␤2 by SRC1a. SRC1a and either HA-tagged TR␤1 or HA-tagged TR␤2 were introduced into HeLa cells by lipofection of the corresponding expression vectors. After 48 h, 100 nM T 3 was added, or not, and the cells were incubated for 1 h more. The cells were lysed and the coactivator was immunoprecipitated with anti-SRC1 antibodies. The immunoprecipitates were analyzed by SDS-PAGE and immunoblot using anti-TR␤ antisera. F, equal binding of DRIP205 by TR␤2 and TR␤1. Protocol was as in panel B. G, equalization of the transcriptional properties of TR␤1 and TR␤2 by ectopic expression of SRC1a. An expression construct for either TR␤1 or TR␤2 was transfected into CV-1 cells together with the DR4-luciferase reporter, the pCH100-lacZ internal control, and increasing amounts of an expression vector for SRC1a. The cells were incubated ϪT 3 or ϩT 3 and were analyzed for relative luciferase expression as described in the legend to Fig. 2. H, failure to equalize the transcriptional properties of TR␤1 and TR␤2 by ectopic expression of DRIP205. An expression construct for each TR isoform was introduced into CV-1 cells by lipofection, together with a DR4-luciferase reporter, a pCH100-lacZ construct, and increasing amounts of an expression vector for DRIP205. The cells were incubated and analyzed for relative luciferase expression as described in the legend to Fig. 2. Symbols in each panel indicate statistical confidence that TR␤2 differs in EC 50 or B max from TR␤1 as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; #, curves are not statistically distinguishable. ever, also requires the presence of at least one LXXLL receptorinteraction motif in the coactivator. Similar results were observed with GRIP1 (data not shown).
The N-terminal Domain of SRC1 Selectively Enhances TR␤2 Binding Even in the Absence of the Gln-rich Domain-We made an unexpected observation when expanding these studies to an SRC1-(1-781) construct lacking the Gln-rich region, but retaining the SRC1 N-terminal and central LXXLL domains (Fig. 1B). Although the loss of the Gln-rich domain virtually eliminated the ability of the SRC1-(1-781) construct to bind to TR␤2 in the total absence of T 3 , this construct still retained a strong preference for binding to TR␤2 compared with TR␤1 at low and intermediate T 3 levels (Fig. 4E). The N-terminal domain of SRC1 was required for this phenomenon, given SRC1 constructs limited to the LXXLL motifs bound nearly equally to TR␤1 and TR␤2 (Figs. 4D and 5). Similar results were obtained with GRIP1 (Fig. 4F).
To better define this phenomenon by an internally controlled protocol, we mixed an 35 S-labeled SRC1-(1-781) con- . Identification of three domains on the p160 coactivators required for preferential recruitment by TR␤2. A, preferential TR␤2 binding to an SRC1a-(595-1441) construct containing the central LXXLL motifs, the Gln-rich domain, and the C-terminal LXXLL motif. Protocol was as described in the legend to Fig. 3B. B, preferential TR␤2 binding to an SRC1a-(891-1441) construct containing the central LXXLL motifs, the Gln-rich domain, and the C-terminal LXXLL motif. Protocol was as described in the legend to Fig. 3B, except a GST-SRC1-(891-1441) construct was incubated with 35 S-radiolabeled full-length TR␤1 or TR␤2, as indicated. Percent bound is presented, normalized to TR␤2 levels. C, preferential, hormone-independent TR␤2 binding to an SRC1a-(891-1441) construct containing the Gln-rich domain but lacking the C-terminal LXXLL motif. Protocol was as in panel B, except the C-terminal LXXLL in the GST-SRC1-(891-1441) construct was mutated to LXXAA. D, equal TR␤2 and TR␤1 binding to an SRC1a-(566 -891) construct limited to the central LXXLL motifs. Protocol was as in panel B. E, preferential TR␤2 binding to an SRC1a-(1-781) construct containing the N-terminal and central LXXLL domains, but lacking the Gln-rich region. Protocol was as in panel A. F, preferential TR␤2 binding to a GRIP1-(1-776) construct containing the N-terminal and central LXXLL domains, but lacking the Gln-rich region. Protocol was as in panel A. Statistical confidence symbols are as described in the legend to Fig. 3. struct (containing both the N-terminal and LXXLL domains, Fig. 1B) and an 35 S-labeled SRC1-(595-781) construct (containing only the LXXLL domain, Fig. 1B) together and assayed their ability to competitively bind to GST-TR␤1, GST-TR␤2, or a GST-⌬N-TR construct (Fig. 5). The TR␤2 N terminus greatly enhanced the preference of this receptor for SRC1-(1-781) versus SRC1-(595-781) under limiting hormone (Fig. 5A), the TR␤1 N terminus had a detectable, but much less of an effect (Fig. 5B), and a TR␤ mutant lacking the N-terminal domain displayed no preference (Fig. 5C). (13, 46 -48) (Fig. 1B). To map the specific domain responsible for the enhanced interaction with TR␤2 under limiting T 3 , we tested a series of N-terminal deletions (Fig. 1B) of the SRC1-(1-781) construct in the GST pulldown assay. Deletion of the basic helix-loop-helix domain had no observable effect on either TR␤1 or TR␤2 binding (Fig. 6A). Further truncation of the SRC1 N terminus, resulting in loss of the PAS-A domain, increased the overall binding of the SRC1 coactivator to both TR␤1 and TR␤2, but nonetheless, preserved the TR␤2 Ͼ TR␤1 binding phenotype in response to limiting T 3 (Fig. 6, B and C). Larger N-terminal truncations that deleted the PAS-B domain of SRC1 virtually eliminated the preferential binding of TR␤2 compared with TR␤1 (Fig. 6, D-F). Internal deletions restricted exclusively to PAS-A or to PAS-B mimicked the results seen with the sequential N-terminal deletions (Fig. 6, G versus H). Although the N terminus of SRC-1 contains two "cryptic" LXXLL sequences (Fig. 1B), these are not known to bind nuclear receptors and had no effect on the interaction of the coactivator with either TR isoform (Fig.  6). We conclude that the PAS-B domain, conserved in all p160 coactivators, is responsible for the preferential ability of TR␤2 to bind this coregulator in the absence of the Gln-rich domain.

The Enhanced Binding of TR␤2 by SRC1a Maps to a PAS-B Domain within the p160 Coactivator-The N terminus of the p160 coactivators contains a basic helix-loop-helix domain, a PAS-A and a PAS-B domain
It was conceivable that the PAS-B motif in SRC1 functioned through a direct interaction with TR␤2. However, an SRC1-(1-338) construct containing this motif failed to interact detectably with either TR␤2 or TR␤1 in a GST pulldown assay (data not shown). The PAS-B domain of SRC1 is known to interact with an LDELL/LLEQL motif region in the SRC1 C-terminal region (Fig. 1B, box), potentially generating intra-or intermolecular SRC1-SRC1 interactions (49); however, this region is absent from the SRC1-(1-781) construct used in our studies. Consistent with prior studies (49), we also failed to detect any interaction between the SRC1-(1-338) N-terminal domain and the central LXXLL motifs that are retained in the SRC1-(1-781) fragment (data not shown). Our evidence therefore suggests that the PAS-B SRC1a domain enhances TR␤2 binding under limiting T 3 concentrations through an indirect effect.
The PAS-B Domain Enhances the Ability of TR␤2 to Bind to LXXLL Motif Pairings Not Recognized by TR␤1-We next tested if the PAS-B motif worked by enhancing in some fashion the affinity of TR␤2 for the central LXXLL coactivator motifs known to mediate agonist-dependent nuclear receptor binding. We employed GRIP1 for these experiments due to the prior availability of the appropriate mutations. It has been proposed that receptor dimers interact with two LXXLL motifs within one p160 coactivator (50 -53). TRs reportedly prefer LXXLL-2 and LXXLL-3 (54 -57), and consistent with this observation, mutation of LXXLL-1 to an LXXAA sequence had little or no effect on the ability of full-length GRIP1 to bind to either TR␤1 or TR␤2 (Fig. 7A). In contrast, inactivation of either LXXLL-2 or LXXLL-3 by mutagenesis virtually eliminated the ability of GRIP1 to bind to TR␤1, but preserved substantial binding to TR␤2 (Fig. 7, B and C). These results suggest that TR␤2 is capa- ble of recognizing a different combination of LXXLL motifs than TR␤1.
Deletion of the GRIP1 PAS-B region strongly reduced the ability of TR␤2 to bind to the LXXLL-2 mutant in response to T 3 (Fig. 7D). Interestingly, removal of the GRIP1 C-terminal Gln-rich domain also interfered with the ability of TR␤2 to recognize the LXXLL-2 mutant (Fig. 7E). These results indicate that the N-and C-terminal regions of the coactivator operate together to stabilize TR␤2 binding to LXXLL motifs that are not recognized by TR␤1.
Partial Protease Degradation Suggests That SRC1a Can Assume a Different Conformation Once Bound to TR␤2 Compared with That When Bound to TR␤1-Our results suggest that TR␤2 interacts with p160 coactivators differently than does TR␤1. If so, this different mode of interaction might impose a different conformation on the coactivator. To examine this question, we bound radiolabeled SRC1a to either GST-TR␤1 or GST-TR␤2 under high T 3 concentrations and probed the coactivator conformation by use of a limited protease digestion, with the expectation that differences in coactivator conformation would manifest as differences in protease susceptibility (58). There were consistent differences in the elastase and V8 protease degradation patterns of the same SRC1a preparation when bound to TR␤2 than when bound to TR␤1 (Fig. 8A and data not shown). Different protease protection patterns were also observed using an SRC1-(1-781) construct (Fig. 8B), but not with an SRC-(595-781) construct (data not shown). These results are consistent with the SRC1a coactivator assuming a different conformation when bound to TR␤2 than when bound to TR␤1.
The N-terminal Domain of SRC1 Also Contributes to Efficient p160 Coactivator Binding by Estrogen Receptor-␣-To determine if the PAS-B sequences contribute to p160 binding not only for specific TR splice forms, but also for other classes of nuclear receptors, we examined the comparative ability of the SRC1-(1-781) and SRC1-(595-781) constructs to bind to ER␣. Notably the SRC1 construct containing the N-terminal domain FIGURE 7. Broadened specificity of TR␤2 for coactivator LXXLL motifs not recognized by TR␤1. The ability of GST-TR␤1 and GST-TR␤2 to bind to the 35 S-radiolabeled GRIP1 constructs indicated was assayed as described in the legend to Fig. 4E. A, TR binding of full-length GRIP1 in which LXXLL-1 was mutated to LXXAA. B, TR binding of full-length GRIP1 in which LXXLL-2 was mutated to LXXAA. C, TR binding of full-length GRIP1 in which LXXLL-3 was mutated to LXXAA. D, TR binding of GRIP1 in which LXXLL-2 was mutated to LXXAA and bearing an N-terminal deletion. E, TR binding of GRIP1 in which LXXLL-2 was mutated to LXXAA and bearing a deletion of the Q-rich domain. Statistical confidence symbols are as described in the legend to Fig. 3. bound much more strongly to ER␣ in response to estradiol than did the SRC1 construct lacking the N-terminal domain (Fig.  9A), whereas both SRC1 constructs bind equally well to the ⌬N-TR␤ construct (e.g. Fig. 5). The enhanced binding of the SRC-(1-781) construct to ER␣ required the PAS-B domain of the coactivator (Fig. 9B) and was not observed using a ER␣ construct lacking the receptor N-terminal domain, or a fulllength FXR construct (data not shown). We conclude that the PAS-B domain of SRC1a enhances the ability of the p160 coactivators to recruit a subset of nuclear receptors in response to hormone agonist.

DISCUSSION
Wild-type TR␤2 displays enhanced transcriptional activation properties not seen with other TR isoforms-Unlike the TR␤1 isoform, the unliganded TR␤2 isoform fails to repress transcription, and instead activates transcription even in the absence of hormone (33). TR␤2 also displays an elevated ability to activate positive response elements (and to repress negative response elements) in response to hormone compared with TR␤1 (29 -32, 34, 35). The divergent transcriptional properties of these two isoforms are not due to differences in their levels of expression or different affinities for T 3 but instead appear to reflect inherent differences in the ability of these two different isoforms to recruit certain coactivators (34).
In this report, we confirm that this enhanced transcriptional responsiveness is unique to TR␤2, and that TR␣1 closely parallels the TR␤1 response. Given the preferential expression of TR␤2 in the hypothalamus and pituitary, where it plays an important role in negative feedback regulation of circulating thyroid hormone levels, we have suggested that this enhanced responsiveness of TR␤2 to T 3 can help it sense and suppress surges of thyroid hormone before the less responsive, but more widely expressed TR␣1, TR␤0, and TR␤1 isoforms do so, thereby avoiding peripheral thyrotoxicity (34). Consistent with this proposal, TR mutations that selectively impair the enhanced T 3 response of TR␤2 result in Pituitary Resistance to Thyroid Hormone Syndrome, an endocrine disease characterized by central resistance but peripheral thyrotoxicity (34). The role the enhanced transcriptional response of TR␤2 might play in the other cell types in which this isoform is expressed, such as the retina and the inner ear, remains to be established.
The Unique N Terminus of TR␤2 Is Essential for the Enhanced Transcriptional Activation Properties of This Isoform-TR␤2 and TR␤1 are identical in sequence with the exception of their N-terminal domains, which are encoded by different exons in the two different isoforms. Significantly, the avian TR␤0 isoform, which possesses a severely truncated N-terminal domain, more closely resembles mammalian TR␤1 in its transcriptional properties, and is readily distinguished from mammalian or avian TR␤2. The same is true of an artificial, N-terminal truncation mutation of mammalian TR␤2. We conclude that the N terminus of TR␤2 contributes positively to the enhanced T 3 response seen for this isoform.
The TR␤2 N-terminal Domain, Unlike That of TR␤1, Interacts with and Helps Recruit p160 Coactivators in a Hormoneindependent Manner-In common with other nuclear receptors, the hormone-binding domains of TR␣1, TR␤0, TR␤1, and TR␤2 interact with the LXXLL motifs found in many coactivators, and this interaction requires hormone agonist. Interestingly, the N-terminal domain of TR␤2, but not of TR␤1 or TR␣1, can also interact in a hormone-independent manner with a Gln-rich region located near the C terminus of all three p160 coactivators (30,32). This coactivator interaction likely contributes to the in vivo ability of TR␤2 to activate target gene expression in the absence of T 3 (32). Furthermore, we report here that in the presence of T 3 these TR␤2 N-terminal domain/ coactivator contacts cooperate with the LXXLL coactivator contacts mediated by the hormone-binding domain of the receptor, yielding stronger p160 recruitment by TR␤2 than by TR␤1 under limiting T 3 concentrations.
Both in the CV-1 cells used here, and in gene knock-out studies, SRC1a appears to play a particularly important role in defining the transcriptional activity of TR␤2 (59,60). However, our current study does not exclude the possibility that coactivators outside of the p160 family also contribute to the unique hormone responsiveness of this isoform. CREB-binding protein, pCIP, and NRC, for example, display an enhanced interaction with TR␤2 in a phenomenon that requires the TR␤2 N-terminal domain, although probably by a mechanism different in its details from that of the p160 coactivators (30,31). It appears likely that the relative transcriptional properties of TR␤2 and TR␤1 will differ in different cell types, on different target genes, and perhaps in different species (31) depending on which coactivators predominate in each context.

A PAS-B Motif in the SRC1 Unexpectedly Modulates Isoformspecific Nuclear Receptor Binding, Strongly Enhancing the p160 Interaction with TR␤2 under Limiting T3 Conditions-The
Gln-rich region alone was insufficient to account for the enhanced ability of TR␤2 to recognize the p160 coactivators. Instead, enhanced recruitment of SRC1 by TR␤2 under limiting T 3 was also observed for coactivators containing only the LXXLL motifs and an N-terminal domain PAS-B motif. The PAS-B domain not only increased the ability of TR␤2 to bind to p160 coactivator under limiting T 3 , but also permitted TR␤2 to bind to GRIP1 and SRC1 derivatives in which LXXLL-2 or -3 motifs, crucial for TR␤1 binding, had been disrupted by mutation.
No direct interaction of the PAS-B domain with any portion of TR␤2 was detected. Instead, our results suggest that the PAS-B domain functions indirectly, most likely by enhancing the ability of LXXLLs in the center of the p160 coactivators to bind to TR␤2. We explored several means by which this might occur. The PAS-B domain is known to interact with a subset of LXXLL motifs, including an LDELL/LLEQL region within the C-terminal region of the full-length p160 coactivators (49,61). However, these motifs are absent from the SRC1-(1-781) construct, and therefore cannot contribute to the PAS-B phenotype we report here. Alternatively, it was possible that, despite the absence of the necessary flanking basic amino acids, the central, receptor-interaction LXXLL motifs retained in our SRC1-(1-781) constructs might interact with the PAS-B domain; however, no such contacts were observed in our hands or by other researchers (49).
Although we have been unable to define the specific mechanisms of action of the PAS-B and TR␤2 N-terminal domains in stabilizing the LXXLL/receptor interaction under limiting T 3 conditions, we strongly suspect that these domains are acting through global conformational effects on their respective proteins. In support of this conjecture, SRC1 appears to assume a different conformation (based on protease susceptibility) when bound to TR␤2 than when bound to TR␤1. Conversely, TR with the ␤2 N-terminal domain appears to assume a different conformation state (by similar protease probes) than does TR␤1 (data not shown). We suggest that the SRC1 and TR␤2 N termini confer conformations on their respective proteins that permit and help stabilize higher affinity interactions between them. Unfortunately little is currently known about the secondary or tertiary structure of these N-terminal nuclear receptor domains, which are believed to be relatively disordered in solution and to assume an induced-fit conformation only in response to contacts with other proteins (62,63).
The PAS-B Domain Also Contributes to the Ability of the p160 Coactivators to Bind to Other Nuclear Receptors, Including ER␣-In common with the TR␤2 N-terminal region, the N-terminal domains of estrogen, glucocorticoid, and androgen receptors interact with the Gln-rich region of the p160 coactivators (64 -67). We report here an additional parallel between TR␤2 and these steroid receptors, in that the PAS-B domain also contrib-utes to the ability of SRC1 to bind to ER␣, and this further requires the presence of the ER␣ N terminus. Further experiments will be required to define how wide a role the PAS-B domain plays in nuclear receptor function.
PAS domains serve as chemical and environmental sensors in a number of proteins in which they are found (68). For example, a recent report reveals that the PAS-B domain of hypoxia inducible factor 2␣ forms a large internal cavity, and that artificial compounds that dock to this cavity can modulate the interaction of hypoxia inducible factor 2 with its transcription factor partner, ARNT (69). If the PAS-B domain of the p160 coactivators can similarly bind and be regulated by small molecules, a cross-talk may exist by which the recruitment of the p160s to nuclear receptors can be controlled not just by receptor ligands, but also by coactivator ligands. These hypothetical coactivator ligands may be endocrine hormones, cellular metabolites, or yet some other regulatory molecule.