|Home | About | Journals | Submit | Contact Us | Français|
Estrogen receptors (ERs) are hormone-regulated transcription factors that regulate key aspects of reproduction and development. ERs are unusual in that they do not typically repress transcription in the absence of hormone but instead possess otherwise cryptic repressive functions that are revealed upon binding to certain hormone antagonists. The roles of corepressors in the control of these aspects of ER function are complex and incompletely understood. We report here that ERs recruit SMRT through an unusual mode of interaction involving multiple contact surfaces. Two surfaces of SMRT, located at the N- and C-terminal domains, contribute to the recruitment of the corepressor to ERs in vitro and are crucial for the corepressor modulation of ER transcriptional activity in cells. These corepressor surfaces contact the DNA binding domain of the receptor, rather than the hormone binding domain previously elucidated for other corepressor/nuclear receptor interactions, and are modulated by the ER's recognition of cognate DNA binding sites. Several additional nuclear receptors, and at least one other corepressor, N-CoR, share aspects of this novel mode of corepressor recruitment. Our results highlight a molecular mechanism that helps explain several previously paradoxical aspects of ER-mediated transcriptional antagonism, which may have a broader significance for an understanding of target gene repression by other nuclear receptors.
Key aspects of vertebrate reproduction, development, and physiology are controlled by nuclear receptors: transcription factors that regulate target gene expression in response to small, hydrophobic ligands (8, 34, 38). The nuclear receptor family includes endocrine receptors such as the estrogen receptors (ERs), thyroid hormone receptors (TRs), and retinoic acid receptors (RARs) (3, 7, 76). Additional members of this family respond to intermediates in lipid metabolism, such as the peroxisome-proliferator-activated receptors (PPARs), farnesoid X receptors (FXRs), and liver X receptors (LXRs), or to xenobiotics such as the pregnane X receptors (37, 39, 66). Yet others have no known ligand, such as COUP-TF (44). Defects in nuclear receptor function play causal or contributory roles in a wide variety of developmental, endocrine, and neoplastic diseases (4, 8, 31, 41, 49, 61, 65).
Many nuclear receptors can both repress and activate target gene expression. This transcriptional dualism reflects the ability of these receptors to recruit alternative auxiliary proteins, denoted corepressors and coactivators, that mediate the specific molecular events necessary for target gene regulation (10, 15, 28, 36, 51). Coactivators include acetyltransferases or methyltransferases that insert activation marks in chromatin, chromatin remodeling activities that alter the accessibility of chromatin, and components of the mediator complex that help recruit the general transcriptional machinery (10, 15, 28, 36, 51). Corepressors characteristically exert the opposite effects (10, 15, 28, 36, 51).
Two corepressors play key roles in transcriptional repression by nuclear receptors: silencing mediator of retinoic acid and thyroid hormone receptors (SMRT) and its paralog, nuclear corepressor (N-CoR) (24, 38, 42, 48). The N-terminal and central domains of both N-CoR and SMRT are studded with docking surfaces that help recruit additional corepressor components such as TBL1, TBLR1, GPS2, and a variety of histone deacetylases (24, 38, 42, 48). Conversely, the N-CoR and SMRT C-terminal domains contain CoRNR motifs that are known to tether these corepressors to their nuclear receptor partners (6, 20, 32, 45, 71). Molecular events that regulate the CoRNR motif/nuclear receptor interaction determine the recruitment or release of the entire corepressor complex.
Each CoRNR box forms an extended α-helix that binds to a docking surface derived from portions of the nuclear receptor's hormone binding domain (HBD) (20, 45, 74). This docking surface is accessible in the unliganded nuclear receptor due to a permissive positioning of receptor helix 12 (10, 48). Hormone agonists induce a reorientation of helix 12 in the nuclear receptor that blocks the corepressor docking surface, releasing the SMRT or N-CoR complex and forming a new docking site for LXXLL motifs found in many coactivators (10, 48). Antagonists, conversely, are believed to induce a nuclear receptor conformation that further stabilizes corepressor binding and destabilizes coactivator binding (2, 14, 17, 52, 58). Additional mechanisms, such as corepressor phosphorylation, can also have an impact, positive or negative, on the corepressor/nuclear receptor interaction (47).
However, these known corepressor/nuclear receptor interactions fail to adequately account for all aspects of corepressor function. This is particularly evident in the case of ERα. SMRT and N-CoR are recruited to ERα target genes in response to antagonists in vivo, and the ratio of corepressors to coactivators helps to determine the transcriptional response to selective estrogen modulators (SERMs) (21-24, 27, 55-57, 60, 62, 64, 72). Changes in corepressor function have been implicated in the resistance of certain reproductive cancers to hormone therapy (9, 13). Despite this strong biological evidence for corepressor function in antagonism-mediated ERα repression, most studies have found that the physical interaction in vitro between ERα and the corepressor is relatively weak and fails to respond to estrogen agonists or antagonists in the expected fashion (e.g., see reference 79). In fact, structures derived from the ERα HBD raise questions about the accessibility of the corepressor docking surface in the presence of these different ligands (33).
Notably, SMRT and N-CoR are extremely large proteins, and due to practical limitations, most previously reported assays utilized protein constructs limited to the C-terminal receptor interaction domain (cRID) of the corepressor. We report here that ERα interacts strongly with a distinct receptor interaction domain located within the N-terminal domains (nRIDs) of these corepressors. Furthermore, both the nRID and cRID differ from most previously characterized modes of receptor docking by interacting with the DNA binding domain of ERα. The ability of SMRT to function with ERα requires the combined contributions of both nRID and cRID. Notably, an additional subset of nuclear receptors, including TRα1, TRβ1, and PPARα, also recognize the nRID, although with less efficiency than does ERα. The nRID therefore represents an alternative mechanism by which corepressors can tether to specific members of the nuclear receptor family and may also serve to modulate corepressor function once recruited to these receptors.
The pSG5-hERα and pSG5-Myc-SMRTα(1-2470) clones were created by using PCR and standard recombinant DNA methodologies (1). Similarly, subdomains of ERα and SMRTα were amplified by high-fidelity PCR and subcloned into a pSG5-HA vector using KpnI and SacI restriction sites or HindIII and XhoI restriction sites, respectively. pSG5-Myc-SMRT(1-1968) was created by the cleavage and religation of the parental pSG5-Myc-SMRT(1-2470) construct (12, 25). pGEX-MP-SMRTα(1-289) and pGEX-MP-ERα were created by high-fidelity PCR and subcloning into pGEX-MPa using XbaI and XhoI restriction sites. pSG5-Myc-SMRT(310-2470) and pSG5-Myc-SMRT(428-2470) were created by introducing a HindIII restriction site at position 310 or 428 by PCR; subsequent cleavage by HindIII and religation resulted in the desired deletions of the N-terminal domains. pSG5-Myc-SMRT(1-1968) was generated by deleting the SMRTα carboxy-terminal domain by using BssHII cleavage and religation. Codon substitutions into the nRID and cRID were created by using a QuikChange mutagenesis kit (Stratagene, La Jolla, CA).
CV-1 cells were maintained in phenol red-free Dulbecco modified Eagle's medium (DMEM) (Invitrogen, Carlsbad, CA) supplemented with 5% heat-inactivated fetal bovine serum (FBS; HyClone, Logan, UT) in a humidified atmosphere of 5% CO2 at 37°C. For transfections, cells were trypsinized and plated at a density of 1.45 × 104 cells per square centimeter in 24-well plates and allowed to attach overnight. Immediately prior to transfection, the cells were rinsed with phosphate-buffered saline and placed into DMEM supplemented with 5% heat-inactivated, hormone-depleted FBS. Transfections were performed by using the Effectene protocol as recommended by the manufacturer (Qiagen, Valencia, CA), using 50 ng of a reporter containing ERα response elements (pMMTV-2xERE-Luc), 10 ng of the appropriate pSG5.1-human ERα expression vector (unless otherwise noted), 10 ng of the appropriate pSG5.1 SMRT expression vector, 5 ng of LacZ vector (used as an internal transfection control), sufficient pUC18 to bring the total DNA to 250 ng, 2 μl of Enhancer reagent, and 3 μl of Effectene reagent per well. The culture medium was replaced 24 h later with fresh medium containing 1 μM 17-β-estradiol (Sigma-Aldrich, St. Louis, MO), 1 μM 4-OH-tamoxifen (Sigma-Aldrich, St. Louis, MO), or an equivalent amount of ethanol carrier. After 24 h of incubation at 37°C, the cells were harvested and lysed, and the luciferase and β-galactosidase activities were determined as previously described (16).
Glutathione S-transferase (GST)-corepressor and GST-nuclear receptor protein fusions were produced in Escherichia coli strain BL-21 cells bearing the corresponding pGEX-recombinant vector by a modification of a procedure described previously (67). The bacteria were lysed by sonication, and the GST fusion proteins were bound to glutathione-agarose beads. 35S-radiolabeled corepressors or nuclear receptors were synthesized in vitro by using a TnT-coupled reticulocyte lysate system (Promega, Madison, WI). Each radiolabeled protein (representing 5 μl of TnT product per reaction) was then incubated with the immobilized GST fusion protein of interest immobilized to 20 μl of agarose beads in a total volume of 200 μl of HEMG buffer (4 mM HEPES [pH 7.8], 100 mM KCl, 0.2 mM EDTA, 5 mM MgCl2, 0.1% NP-40, 10% glycerol, 1.5 mM dithiothreitol). The binding reactions were performed at 4°C in 1.5-ml microcentrifuge tubes placed on a rotating platform to ensure thorough mixing. After a 1-h incubation, the immobilized proteins were washed by repeated rounds of centrifugation and resuspension using 1 ml of ice-cold HEMG buffer per tube each round. Radiolabeled proteins remaining bound to the immobilized GST fusion proteins were eluted with 50 μl of 10 mM glutathione in 50 mM Tris-HCl (pH 7.8). The eluted proteins were resolved by SDS-PAGE and were visualized and quantified by using a Storm PhosphorImager system (GE Healthcare Bio-Sciences Corp., Piscataway, NJ). Coimmunoprecipitation assays were performed as described previously (50).
To determine the role of corepressors in ERα function, we first examined the ability of SMRTα to repress the ERα-dependent transcription of a prototypical target gene in CV1 cells (a cell line lacking endogenous ERs). A pSG5-human ERα expression vector was introduced into these cells by transient transfection together with an ERα-responsive luciferase reporter and either an expression vector for full-length SMRTα or the equivalent empty vector control. No reporter gene regulation was detected in the absence of ectopic ERα expression, whereas a strong, E2-dependent activation of luciferase expression was observed in the presence of the ectopic ERα (Fig. (Fig.1).1). Tamoxifen alone was a weak agonist on ERα in these cells but was a very strong antagonist of E2-mediated activation, consistent with the previous identification of this hormone as a SERM (Fig. (Fig.1)1) (55). The introduction of SMRTα in the absence of ERα had little or no effect on reporter gene expression (Fig. (Fig.1).1). In contrast, the introduction of SMRTα together with ERα inhibited E2-induced gene expression, abolished the weak agonistic actions of tamoxifen when assayed alone, and enhanced the antagonistic actions of tamoxifen when assayed with E2 (Fig. (Fig.1).1). These results indicate that SMRTα counteracts the agonistic activities of ERα ligands in vivo and agree with results from previously reported studies demonstrating that SMRTα plays a role in the downmodulation of ERα transcriptional activity (23, 24, 27, 55-57, 59, 62, 64).
Given the functional interactions between ERα and SMRTα in vivo, we used GST pulldown assays to examine the ability of ERα to interact in vitro with full-length SMRTα and with its various subdomains (Fig. (Fig.2A).2A). A GST fusion of full-length ERα bound to radiolabeled, full-length SMRTα(1-2470) under these conditions (Fig. (Fig.2B).2B). Surprisingly, this binding was preserved when using a SMRTα(1-1968) construct lacking the known cRID (Fig. (Fig.2B).2B). Analysis of a series of additional, nested SMRT N-terminal constructs revealed that the first 289 amino acids of SMRTα were sufficient to bind to ERα under these conditions, whereas the first 268 amino acids of SMRTα were not (Fig. (Fig.2B).2B). A similar interaction was also observed in a reverse GST pulldown protocol using a GST-SMRTα(1-289) fusion and radiolabeled, native ERα (Fig. (Fig.2C)2C) and in coimmunoprecipitation assays of cells (Fig. (Fig.2D).2D). We denote this SMRTα region the N-terminal receptor interaction domain (“nRID”).
To determine where the SMRT nRID contacts ERα, we repeated the GST pulldown with a GST-nRID construct and different ERα domains. ERα has been conceptually divided into domains A to F (Fig. (Fig.2C);2C); the N-terminal “A/B” domain contains a ligand-independent activation function, the “C” domain makes direct contacts with the major groove of the DNA response element, the “D” domain both is a linker and (in certain nuclear receptors) makes contacts with the minor groove of the DNA, the “E” domain both binds hormone and tethers coregulators in a hormone-regulated fashion (including corepressors), and the F domain can modify the transcriptional response of the receptor (7). GST fusions containing the SMRT nRID region interacted strongly with the ERα “A-to-D” and “C/D” domains (Fig. (Fig.2C).2C). In contrast, little or no interaction of the SMRT nRID constructs was observed with the ERα “A/B” domain or with the ERα “E/F” domain, despite the latter being the principal site of SMRT interactions by TRs and RARs (Fig. (Fig.2C).2C). Reciprocal GST pulldown experiments using GST-ERα constructs and radiolabeled SMRT subdomains confirmed these results and demonstrated that the nRID of SMRTα contacts ERα primarily through amino acids in the DNA binding “C” domain of this receptor (Fig. (Fig.33).
Expanding these studies to additional SMRTα constructs further narrowed the site of contact of the SMRTα nRID to the “C” domain of ERα and revealed that constructs containing the previously identified cRID of the corepressor [e.g., SMRTα(1210-2470)] also bound to full-length ERα (Fig. (Fig.33 and and4).4). Surprisingly, given observations on most other nuclear receptors, constructs containing the SMRTα cRID displayed little or no interaction with the ERα “A/B,” “D,” or “E/F” domain but (in parallel with our nRID results) mediated a strong interaction with the ERα “C” domain (Fig. (Fig.33).
SMRT constructs lacking both the nRID and cRID interacted relatively weakly with ERα (Fig. (Fig.3);3); intriguingly, even this residual interaction was also directed primarily toward the ERα “C” and “C/D” domains. Depending on the specific constructs involved, the presence of these internal regions of SMRTα either further stabilized or interfered with the interactions mediated by the nRID and cRID (Fig. (Fig.3).3). We conclude that the interaction of SMRTα with ERα is mediated primarily by nontraditional contacts between CoRNR box-like motifs present in both the nRID and cRID of the former and the DNA binding domain of the latter. This contrasts with the previously elucidated mode of TR or RAR interactions, which involves contacts between CoRNR boxes in the corepressor cRID with the receptor “E/F” hormone binding domain.
Inspection of the SMRTα(1-289) nRID revealed an amino acid sequence, LVQII, that matches the (I/L)XX(I/V)I consensus CoRNR box motifs previously implicated in corepressor binding to other nuclear receptors (Fig. (Fig.4A)4A) (20, 45). The substitution of this nRID sequence with an LVQAA or AAAAA sequence impaired the ability of the SMRTα(1-289) domain to bind to ERα (Fig. (Fig.4B).4B). Notably, the N-terminal domain of N-CoR also contains two CoRNR motifs (Fig. (Fig.4A),4A), and an N-CoR(1-459) construct was also able to bind to ERα in vitro (Fig. (Fig.4C).4C). The SMRTα cRID has two CoRNR boxes (denoted S1 and S2) (Fig. (Fig.4A).4A). A strong interaction was observed between CoRNR box S1 and either full-length ERα or the ER C/D domain; a mutation of CoRNR box S1 abolished this interaction (Fig. (Fig.4D).4D). Neither the S2 CoRNR box in SMRTα nor an S3 CoRNR box found in the alternatively spliced, SMRTγ form of the corepressor was able to bind to ERα under these conditions (Fig. (Fig.4D).4D). The substitution of the dysfunctional S2 CoRNR box (ISEVI) into the nRID sequence inhibited nRID binding to ERα (Fig. (Fig.4B).4B). We conclude that corepressors can be recruited to ERα through CoRNR-like motifs present in at least two distinct regions: the previously reported cRID and this newly recognized nRID. It should be noted, however, that in addition to the CoRNR box motifs themselves, additional corepressor-flanking sequences appear to contribute to RID function. This is particularly true of the nRID: a SMRT(1-268) construct containing the CoRNR box motif but lacking more C-terminal sequences failed to interact with ERα, whereas a SMRT(268-427) construct lacking the CoRNR box but containing more C-terminal flanking sequences retained an ability to interact with ERα (Fig. (Fig.2B2B).
SMRT and NCoR bind to TRs and RARs through interactions between CoRNR box motifs within the corepressors' cRID and a hydrophobic cleft formed by helices 3, 4, and 5 within the receptors' hormone binding domains (20, 33, 45, 74). The binding of hormone agonist by these receptors induces a repositioning of receptor helix 12, which occludes the corepressor docking surface, resulting in the release of SMRT and NCoR (10, 48). In contrast, our experiments indicated that nRID and cRID interact with the DNA binding domain of ERα; consistent with this suggestion, the interaction between full-length SMRTα and ERα in vitro was not significantly influenced by E2 or by tamoxifen, hormones that might be expected to occlude or enhance access to a helix 3, 4, and 5 docking surface on the receptor (Fig. (Fig.5A5A).
Our use of full-length SMRTα and ERα in these assays might have obscured a hormone-dependent subcomponent of the overall interaction. We therefore also examined this question by using individual subdomains of ERα and SMRTα. ERα constructs limited to the receptor “C” domain interacted in a ligand-independent manner with either the nRID or cRID of SMRTα, whereas the “E/F” domain of ERα did not interact with any domain of SMRTα under any condition tested (Fig. (Fig.5B).5B). In equivalent experiments, the interaction of the cRID construct of SMRTα with TRs and with RARs was strongly inhibited by the appropriate hormone agonists for these receptors (data not shown). As a positive control, the ERα “E/F” domain was able to bind to the SRC1 coactivator in response to E2 (but not to tamoxifen), demonstrating that the “E/F” domain was properly folded and able to bind hormone (Fig. (Fig.5C).5C). We conclude that the atypical mode of contact between SMRTα and the DNA binding domain of ERα results in a hormone-independent interaction in vitro.
Given that the major SMRTα interaction site on ERα mapped to the DNA binding “C” domain, we asked if this interaction was altered in the presence of a cognate DNA binding site. We repeated our GST pulldown assays in the presence of estrogen response elements (EREs) found in the pS2 and vitellogenin genes (pS2-ERE and Vit-ERE) and, as a negative control, an irrelevant response element for the transcription factor RBP-Jk, which does not bind to ERα. The interaction between ERα and the isolated SMRTα nRID was inhibited by the presence of either of the two EREs, whereas the RBP-JkE oligonucleotide had no effect (Fig. (Fig.6A).6A). The same ERE-mediated inhibition was observed for the ERα interaction with the isolated SMRTα cRID (Fig. (Fig.6B).6B). As a control, we tested TRα, which interacts with SMRTα primarily through cRID contacts with the receptor “E/F” domain (20, 33, 45, 74); we used a TR binding site (LysF2-TRE) in addition to the pS2-ERE, Vit-ERE, and RBP-JkE oligonucleotides. None of the tested response elements inhibited the interaction between TRα and SMRT cRID (Fig. (Fig.6C),6C), further indicating that the SMRT cRID interaction with ERα is mechanistically distinct from its interactions with other nuclear receptors, such as TRα. Interestingly, when we repeated these experiments with ERα and a full-length SMRTα in place of the isolated nRID or cRID constructs, little or no effect of the cognate EREs was observed (Fig. (Fig.6D).6D). Our inability to detect an ERE-mediated inhibition of the interaction of full-length SMRTα with ERα may reflect an additional stabilization of the ERα interaction due to having both nRID and cRID present simultaneously (perhaps operating together with the internal RIDs mapped in Fig. Fig.2B)2B) or may reflect distinct properties imposed on the ERα interaction by the native corepressor protein (see Discussion).
To extend our studies to an in vivo context, we next explored whether the nRID or cRID was essential for the ability of SMRT to modulate ERα function in cells. We introduced ERα and the ERE-luciferase reporter gene into cells together with full-length SMRTα(1-2470), with SMRTα(310-2470) (lacking the nRID), or with SMRTα(1-1968) (lacking the cRID) and determined the effect on reporter gene expression (Fig. (Fig.7).7). Whereas the full-length corepressor significantly decreased the ERα-dependent activity induced by E2, tamoxifen, or both, the overexpression of either SMRTα(310-2470) or SMRTα(1-1968) not only failed to repress but also (in certain contexts) reversed the usual repressive effects of native SMRTα, suggesting that these constructs could operate as dominant negative inhibitors of endogenous corepressor function (Fig. (Fig.7A).7A). The introduction of SMRTα in the absence of ERα had no effect on reporter expression, and the SMRT deletions were expressed at levels comparable to or higher than that of the full-length corepressor (data not shown). These results indicate that both the nRID and cRID are necessary for SMRTα to antagonize ERα target gene activation in cells. SMRTα(1-2470) constructs containing either an AAAAA or ISEVI substitution in the nRID CoRNR box were also impaired in their ability to repress ERα gene activation although to a more limited extent than that observed for the nRID deletion represented by the SMRTα(310-2470) construct (Fig. (Fig.7B);7B); this finding is consistent with our results demonstrating that these CoRNR box substitutions reduce but do not eliminate ERα binding (Fig. (Fig.4B)4B) and that the nRID is not restricted to the CoRNR box motif alone but encompasses flanking sequences as well (Fig. (Fig.2B2B).
It was possible that the nRID mutations influenced the repression rather than the receptor interaction properties of SMRTα. We therefore repeated the experiment with TRα and a TRE-luciferase reporter gene (Fig. (Fig.7C).7C). The TRE-luciferase reporter displayed a modest level of basal expression in the absence of TRα (defined as 1) (Fig. (Fig.7C).7C). The introduction of TRα led to a significant repression of reporter basal transcriptional activity, which was further pronounced by the cointroduction of either full-length SMRTα(1-2470) or SMRTα(310-2570) bearing an nRID deletion. In contrast, SMRTα(1-1968), lacking the cRID, reversed repression by TRα in a fashion similar to that of its effects on ERα (Fig. (Fig.7C).7C). We conclude that SMRTα-mediated repression by both ERα and TRα requires the corepressor cRID in these cells, whereas the requirement for the corepressor nRID is specific to ERα and likely reflects the ability of the nRID to interact with this receptor in vitro.
We further tested the ability of SMRTα fragments restricted to the nRID or cRID to act as competitive inhibitors of full-length SMRTα function with ERα. Unlike full-length SMRTα, the introduction of a SMRTα(1-427) (representing the nRID) or SMRTα(2050-2470) (representing the cRID) construct had little or no effect on ERα-driven reporter gene expression in the absence of hormone or in response to E2 (Fig. (Fig.7D).7D). However, both of these SMRTα fragments enhanced the agonist effects of tamoxifen and counteracted the inhibitory effects of this SERM on E2 (Fig. (Fig.7D),7D), presumably by competing with an endogenous corepressor in these cells. The dominant negative effects of the abstracted nRID and cRID fragments were also evident when cointroduced together with ectopic full-length SMRTα (Fig. (Fig.7D).7D). Of note, the SMRTα(1-289) fragment partially reversed tamoxifen antagonism in the absence of an ectopic, full-length SMRTα but less efficiently than did the SMRT(1-427) fragment and was unable to interfere when the full-length SMRTα corepressor was cointroduced (data not shown). These weak dominant negative properties of the SMRT(1-289) fragment relative to those of SMRTα(1-427) may be due to a less efficient folding of the shorter fragment. Alternatively, the larger SMRTα(1-427) fragment may possess additional sequences that stabilize its interaction with ERα. In any case, our findings indicate that both the nRID and the cRID of SMRTα contribute to corepressor recruitment by, and function with, ERα in vivo.
Most previous studies of corepressor recruitment by nuclear receptors typically focused on the interaction between the receptor “E/F” hormone binding domain and the corepressor cRID, and the possibility of additional SMRT interaction surfaces was often not examined. To determine whether the SMRTα nRID is an interaction surface specific to ERα, we extended our GST pulldown assays to additional nuclear receptors (Fig. (Fig.8).8). We tested each one in the presence or absence of a cognate ligand agonist. The SMRTα nRID interaction with ERα was clearly the strongest of those of the receptors tested by this method (Fig. (Fig.8).8). However, TRα, TRβ1, PPARα, and FXR also displayed detectable binding to the SMRTα nRID, whereas RARα, RARβ, RARγ, glucocorticoid receptor, LXR, PPARβ, and PPARγ did not associate significantly with the SMRTα nRID under these conditions (Fig. (Fig.8).8). No effect of the agonist on the nRID interaction, positive or negative, was detected for any receptor (Fig. (Fig.8).8). Unlike ERα, which bound to the nRID and cRID approximately equally, TRα bound to the nRID much more weakly than to the cRID; this likely accounts for why TRα-mediated repression was not significantly influenced by the presence or absence of the nRID in our transfection assays. Although not required for TRα-mediated repression in our reporter gene transfections, the nRID corepressor surface may conceivably contribute to TRα, PPARα, or FXR function in other promoter or cellular contexts.
SMRT and NCoR were initially identified based on their ability to interact strongly with TRs and RARs in the absence of hormone (conditions under which these receptors typically repress transcription) (5, 19, 53). Several lines of evidence indicate that NCoR and SMRT contribute to the inhibitory functions of ERα when this receptor is bound to antagonists or to mixed agonists/antagonists/SERMs (22, 35, 59, 64). Tamoxifen, for example, is an ERα antagonist in mammary cells and recruits SMRT and NCoR to ERα target genes in these cells (55, 56). Interestingly, tamoxifen is an ERα agonist in the uterine endometrium, and this tissue specificity appears to reflect, in part, the different ratios of corepressors and coactivators in breast versus uterus (54, 55). Reducing corepressor levels (or increasing coactivator levels) favors an agonist response to tamoxifen (23, 24, 27, 55, 56, 59, 64), whereas (as shown here) increasing corepressor levels favors an antagonist response.
Despite this strong evidence of a functional role for SMRT/NCoR complexes in the modulation of ERα function, the molecular basis behind the physical interaction between these corepressors and ERα has remained poorly understood. The cRID of SMRT and NCoR binds to ERα in vitro; however, this interaction is minimally, if at all, stabilized by tamoxifen or released by E2, as might be predicted by analogy with the previously elucidated interactions of SMRT and NCoR with their docking surfaces in the “E/F” domains of TR and RAR (e.g., see reference 79). In fact, based on structural and biochemical studies, ERα helix 12 may occlude the corepressor docking surface in the ERα E/F domain regardless of hormone status (2, 17, 26, 58).
Here we report that it is the ERα DNA binding “C” domain that recruits SMRTα and does so by contacting at least two domains in the corepressor: the previously identified cRID and a newly recognized nRID. A similar nRID is present in N-CoR. These observations resolve how a corepressor can bind to ERα despite the occluded nature of the docking surface in the ERα hormone binding “E/F” domain and are also consistent with the ligand independence of this interaction in vitro. Surprisingly, given its contact with the ERα “C” domain, the cRID interaction with ERα is mediated, at least in part, by the same SMRT S1 CoRNR motif that is required for the corepressor interaction with the helix 3, 4, and 5 docking surface in the HBDs of other nuclear receptors. In fact, a similar CoRNR-like motif in the SMRT N-terminal domain, operating together with flanking sequences, participates in the ability of the nRID to interact with ERα. These CoRNR motifs are believed to form extended amphipathic α-helices (20, 45) and may create functionally diverse surfaces that can be exploited to dock corepressors to either the hydrophobic cleft within the hormone binding domain of many receptors or an alternative (and not yet fully characterized) docking surface within the DNA binding domains of specific nuclear receptors, such as ERα, as determined here.
Notably, previous studies have shown that the center region of N-CoR can mediate a subcellular redistribution of androgen receptors, glucocorticoid receptors, and ERs and can also suppress transcriptional activation by these receptors in the presence of agonists (73). Although we have focused on the nRID and cRID regions, consistent with data from the previous study, we did detect a residual interaction between ERα and the central region of SMRTα if both N- and C-terminal corepressor domains were deleted. Other studies have shown that the A/B domains of PRs and GRs, although insufficient to interact with corepressor alone, are able to help stabilize the interaction of the E/F domains of these receptors with N-CoR and SMRT (18, 68). We conclude that the steroid receptors recruit the corepressor through a web of multiple interactions, including novel contacts not previously detected by using TRs and RARs.
Elevated levels of expression of full-length SMRTα decreased the agonist activities of E2 and intensified the antagonistic properties of tamoxifen. In our hands, the deletion of either the nRID or cRID impaired the ability of SMRTα to function in this fashion, indicating that both were required for SMRTα to functionally partner with ERα, whereas the nRID was not required for the SMRTα suppression of TRα-mediated gene activation. In fact, the overexpression of certain constructs, such as an isolated nRID, interfered with full-length SMRTα function in a dominant negative fashion, consistent with a competition between this fragment and the native corepressor for recruitment to ERα. Our results from these in vivo transcription experiments therefore parallel the role of the nRID and cRID in corepressor recruitment by ERα in vitro.
Our data favor a model in which the nRID is necessary for effective corepressor recruitment by ERα. However, the nRID is also a site of contact of SMRT with additional components of the corepressor complex (24, 38, 42, 48). By making contact with the SMRT nRID, ERα might displace specific subunits of the corepressor complex that require the nRID sequences for tethering to SMRTα (e.g., see reference 75). ERα might thereby recruit a distinct corepressor complex (with distinct repression properties) from that recruited by nuclear receptors that do not make nRID contacts. The loss of function by the SMRTα nRID deletion in our transcription assays may reflect, in part, a specific requirement for these N-terminally tethered corepressor proteins in ERα-mediated repression but not in TRα-mediated repression. Therefore, we do not exclude the possibility that the nRID sequences contribute not only to stabilizing corepressor recruitment but also to regulating corepressor complex composition and function.
SMRT is expressed by alternative mRNA splicing to generate a family of interrelated corepressor variants; these variants differ in relative abundance in different cell types and display distinct biochemical and biological functions (11, 32). At least one of these splice variants (denoted 44−) retains the nRID but lacks the cRID S1 CoRNR box, potentially rendering this splice variant less effective as an ERα coregulator than splice variants, such as the SMRTα form tested here, that retain the S1 CoRNR box. Another previously reported splice variant, SMRTβ, retains the cRID S1 sequence but deletes the corepressor N terminus, including the nRID sequences identified here (43). Although we have as yet been unable to detect the equivalent of the previously reported SMRTβ isoform in our own studies by using reverse transcription (RT)-PCR, we did observe SMRT-related proteins in immunoblots that are considerably smaller than SMRTα and may, in common with the previously reported SMRTβ sequence, lack the nRID (N. Varlakhanova and M. L. Privalsky, unpublished results). Therefore, whatever their genesis, these different SMRT protein variants are likely to differ in their abilities to interact with ERα and, depending on their relative abundances, may serve to customize ERα transcriptional function to different cell types under different conditions. These differences in SMRT and N-CoR splicing in different cell types may help explain the observation that SMRT can paradoxically enhance agonist-driven transcriptional activation by ERα in certain contexts (46).
Our results demonstrate that SMRTα is recruited to ERα, at least in part, by contacts between the ERα DNA binding domain and the corepressor's nRID and cRID regions and that these nRID and cRID contacts are individually inhibited in vitro by the presence of a cognate ERα DNA binding sequence. In a sense, DNA appears to regulate these nRID and cRID contacts by ERα in a manner loosely analogous to how hormone agonists regulate cRID corepressor contacts by TRs and RARs. Nonetheless, previous studies have shown that SMRT can be recruited to ERα on target promoters in cells (e.g., see references 55 and 56); therefore, ERα can, in at least many contexts, bind to both chromatin and SMRT. It is notable in this regard that unlike the individual nRID and cRID interactions, the interaction of full-length SMRTα with ERα was not inhibited by cognate DNA in our in vitro assay. This may be due to a stabilization of the SMRTα/ERα interaction by the presence of multiple RIDs (perhaps including the internal SMRT interaction domains noted previously), or the nRID and cRID regions may be altered in the native SMRTα protein so as to render them more resistant to challenge by the EREs that we tested here. Alternatively, DNA binding by ERα might serve to regulate corepressor function and not recruitment; DNA binding, by releasing either the corepressor nRID or cRID from the receptor, could permit the access of additional components of the corepressor holocomplex to these surfaces while tethering the receptor through the remaining RID. Under this scenario, the corepressor complex recruited by ERα might change its composition when ERα is recruited to different target genes by different DNA sequences or when ERα is recruited to a promoter by protein-protein contacts instead of protein-DNA contacts. This may operate in addition to DNA-induced allosteric changes that were previously observed for the receptor itself (e.g., see reference 76).
E2 binding induces, whereas tamoxifen prevents, the formation of a coactivator docking surface on ERα (2, 30, 58). However, neither the conformation of ERα plus E2 nor the conformation of ERα plus tamoxifen appears to allow corepressor access to the corepressor docking surface found within the “E/F” domain of nuclear receptors. Our mapping of the primary SMRT binding surface in ERα to the DNA binding domain helps explain how SMRT can be recruited to ERα despite this unavailability of the traditional (hormone binding domain) docking surface. Considered alone, however, this discovery falls short of explaining how antagonists promote, and how agonists counteract, corepressor recruitment to ERα-regulated promoters in vivo. To explain this phenomenon, we suggest that it is the well-established ability of E2 to recruit coactivators to the ERα hormone binding domain that may be the driving force in vivo for corepressor release from the ERα DNA binding domain. Corepressors and coactivators each form very large complexes, and steric hindrance may prohibit both of them from binding simultaneously to the same ERα (e.g., see references 29, 40, 69, and 77). Therefore, the E2 recruitment of the coactivator to the hormone binding domain might favor corepressor release from the DNA binding domain, whereas the tamoxifen-driven release of the coactivator might favor corepressor recruitment, as has been observed in vivo. This competition would not be observed in our in vitro studies, which, due to technical limitations, neither utilized full-length coregulators nor reconstituted the still-larger multiprotein complexes characteristic of living cells. Alternatively, there may be yet other cellular proteins that confer, directly or indirectly, hormone-regulated corepressor recruitment to ERα in vivo.
Although the interaction of the SMRT nRID was strongest with ERα, we did detect weaker interactions between this domain and additional members of the nuclear receptor family, including TRα1, TRβ1, PPARα, and FXR. In addition, what appear to be analogous interactions between the “C” domain of the COUP-TFI orphan nuclear receptor and the NCoR corepressor have been previously reported (78). Unlike ERα, however, most of these other nuclear receptors also make strong E/F domain contacts with the corepressor cRID. The weaker nRID contacts that we detected in our studies may help stabilize the recruitment of SMRT to these receptors under specific circumstances (although under the conditions that we tested here, the nRID was not required for SMRT-mediated repression by TRα), or, as noted for ERα, the additional contacts made by the nRID may regulate the composition of the overall corepressor complex tethered by these receptors rather than SMRT or N-CoR recruitment per se. For example, nRID contacts may play a role in “negative response genes,” which are activated in the absence of a hormone agonist and repressed in its presence. It was previously suggested that SMRT and N-CoR function paradoxically as coactivators on these negative response genes (63, 70), and it is possible that the nRID contacts modify corepressor function to generate the inverted transcriptional response. Considered as a whole, our observations indicate that the traditional model of corepressor/nuclear receptor interaction will need to be broadened to more fully reflect this diversity of contact surfaces and modes of regulation that are observable among the different members of the nuclear receptor family.
We thank Liming Liu for excellent technical support and assistance.
This work was supported by U.S. Public Health Service grant R01DK53528 from the NIDDK, National Institutes of Health. Johnnie B. Hahm was supported, in part, by PHS predoctoral training grant award T32-GM007377 from the National Institute of General Medical Sciences.
Published ahead of print on 11 January 2010.