RU486-induced glucocorticoid receptor agonism is controlled by the receptor N terminus and by corepressor binding.

Glucocorticoid-induced gene transcription has been shown to be mediated by coactivators bound to the glucocorticoid receptor (GR). The glucocorticoid antagonist RU486 interferes with the steroid-mediated activation and can also exhibit partial agonist activity, a response in which corepressors have been implicated. Here we have shown that deletion of the N terminus of GR totally abolishes the agonist activity of RU486. Furthermore, we have demonstrated that corepressors bind directly to the RU486-bound GR as determined by glutathione S-transferase pull-down, mammalian two-hybrid assay, and coimmunoprecipitation. Fine mapping of the interaction regions within GR and the corepressor NCoR reveals a complex interaction profile that involves a number of domains in each protein. Notably, the N and the C termini of GR are both involved in corepressor binding. Thus, the N terminus of GR is a major determinant for RU486-dependent NCoR interaction as well as for RU486-mediated agonist activity.

Glucocorticoid-induced gene transcription has been shown to be mediated by coactivators bound to the glucocorticoid receptor (GR). The glucocorticoid antagonist RU486 interferes with the steroid-mediated activation and can also exhibit partial agonist activity, a response in which corepressors have been implicated. Here we have shown that deletion of the N terminus of GR totally abolishes the agonist activity of RU486. Furthermore, we have demonstrated that corepressors bind directly to the RU486-bound GR as determined by glutathione S-transferase pull-down, mammalian two-hybrid assay, and coimmunoprecipitation. Fine mapping of the interaction regions within GR and the corepressor NCoR reveals a complex interaction profile that involves a number of domains in each protein. Notably, the N and the C termini of GR are both involved in corepressor binding. Thus, the N terminus of GR is a major determinant for RU486-dependent NCoR interaction as well as for RU486-mediated agonist activity.
The glucocorticoid receptor (GR) 1 is a ligand-dependent transcription factor of the steroid hormone receptor superfamily (1). On binding to glucocorticoid, the GR translocates to the nucleus and binds to specific palindromic DNA sequences in the promoter region of target genes. These binding sites function as glucocorticoid response elements (GRE), which confer glucocorticoid-dependent transcriptional regulation to the neighboring genes. Prominent examples of GRE-containing promoters are the tyrosine aminotransferase gene and the long terminal repeat in the mouse mammary tumor virus (MMTV) genome (2,3). DNA-bound steroid receptors have been shown to interact with coactivators, such as the CREB-binding protein (CBP) (4) and members of the p160 family (5)(6)(7).
Similar to GR, the thyroid hormone receptor (TR) binds to thyroid hormone response elements (TRE) and modulates gene transcription. Repression by TR has been characterized in de-tail. In contrast to steroid hormone receptors, unliganded TR binds to TRE as well. Depending on the presence or absence of thyroid hormone (T 3 ), TR induces or represses/silences gene transcription in an active manner (for review, see Ref. 8). This switch between repression and induction is controlled not only by the ligand but also by the nature of the DNA-binding element. So-called positive and negative response elements have been described (9). As for steroid hormone receptors, gene activation requires the binding of coactivators, whereas gene repression is mediated by corepressors (for review, see Ref. 10). The search for corepressors conferring transcriptional repression of unliganded receptors led to the identification of a protein named nuclear corepressor (NCoR) and another factor called silencing mediator for retinoic acid and thyroid hormone receptors (SMRT) (11,12). In contrast to other corepressors like Alien (13) and Sharp (14), NCoR and SMRT are related both structurally and functionally. It has been shown that they interact strongly in vitro with unliganded TR and retinoic acid receptor (RAR) (11,12). The exchange of corepressor and coactivator complexes with opposite chromatin-modifying enzyme activity is therefore a possible mechanism for the ligand-induced switch of nuclear receptors between repression and activation (15)(16)(17).
Another aspect of steroid hormone action is the fact that hormone antagonists interfere with steroid-mediated activation but often confer a residual agonist activity to these receptors. By using a mammalian two-hybrid assay, it has been shown that the receptor corepressors NCoR and SMRT preferentially associate with the progesterone receptor occupied by the antagonist RU486 (18 -20). The magnitude of the partial agonist activity of steroid antagonists and the dose-response curve of the agonist were shown to be influenced by the concentration of cofactors (18,19,21,22).
Here, we have assessed whether the RU486-induced effects on GR are mediated by direct binding of NCoR and SMRT corepressors. In addition, we wanted to know which domains of the corepressor and of GR are required for the interaction. Using a mammalian one-hybrid assay, coimmunoprecipitation, and in vitro binding, we have demonstrated a RU486-dependent in vivo interaction involving several domains of the GR. Detailed in vitro binding analyses show a direct interaction and demonstrate the involvement of several corepressor domains and the N terminus of GR.

EXPERIMENTAL PROCEDURES
Plasmids-The expression vectors for GR and GR deletion mutants (23), VP16, VP16⅐cSMRT, and VP16⅐cNCoR (24), were described before. To generate the VP16⅐GR fusion construct, the cSMRT cDNA in VP16⅐cSMRT was replaced by in-frame insertion of GR cDNA obtained by digestion of pCMV-GST⅐hGR-(1-777) (25) with BamHI and SmaI. The pRSV-TR␤ plasmid and the corresponding empty vector and GRE 2 * This work was supported by the Fonds der Chemischen Industrie, the Stiftung W. G. Kerckhoff, Bad Nauheim, and a stipend (to M. S.) from the Hessische Graduiertenfoerderung. 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.
Cell Culture-CV1 cells were grown in Dulbecco's modified Eagle's medium with 10% charcoal-treated fetal calf serum at 37°C/5% CO 2 in 6-well plates. Cells (10 5 cells/well) were cotransfected with 0.88 g of reporter plasmid and 0.18 g of GR expression vector, unless otherwise stated, in 216 l of transfection mixture using the CaPO 4 method as described earlier (32). For overexpression of TR␤, 0.88 g of pRSV-TR␤ or empty vector was used. In modified mammalian two-hybrid experiments, 0.18 g of either VP16⅐cNCoR, VP16⅐cSMRT, or VP16 expression vector was used. The total amount of transfected DNA was adjusted to 5.4 g with pBluescript SKII (Stratagene). For hormonal studies, 10 Ϫ7 M hormone was added 12 h after transfection. Cells were harvested after 24 -72 h and assayed for luciferase activity. All transfection assays shown were performed in duplicate or triplicate and were repeated at least twice. The error bars represent the mean Ϯ S.D.
GST Pull-down Experiments-Expression of GST and GST⅐NCoR proteins was induced in BL21(DE3) bacteria with 0.2 mM IPTG for 3 h at room temperature. Purification of the GST fusion proteins and interaction studies using in vitro translated, [ 35 S]methionine-labeled GR were carried out as described earlier (33). Polyacrylamide gels were stained with Coomassie Brilliant Blue to ensure the presence of comparable amounts of GST fusion proteins in the binding reactions. Bound, labeled proteins were visualized by autoradiography. For fulllength GR and GR-(418 -777), RU486 (10 Ϫ7 M) was included during both the transcription/translation and binding reactions.
Coimmunoprecipitation-100 l of GR-containing extract (derived from baculovirus-infected SF9 cells) was incubated with dexamethasone or RU486 for 5 min at 30°C. Subsequently, 300 l of 293T cell extract and NETN buffer (31) were added to a final volume of 800 l. Immunoprecipitation was performed for 1 h at 4°C with 5 l of GR antibody (G85920, Transduction Laboratories) and 60 l of protein A/G-Sepharose beads (Amersham Biosciences). Bound proteins were washed 3-4 times with NETN. Proteins were solubilized with Laemmli buffer and heated to 100°C for 5 min before loading onto SDS-polyacrylamide gels. Western blotting was carried out using antiserum to the C terminus of NCoR (31).

Binding of Corepressors Limits the Agonistic Activity of the RU486-bound Glucocorticoid
Receptor-To study the partial agonistic activity of the antihormone RU486 on the glucocorticoid receptor, we transfected the MMTV-Luc reporter construct into CV1 cells. Cells were either treated with ethanol, RU486, or dexamethasone ( Fig. 1). Cotransfection of increasing amounts of GR expression vector showed no significant effect on the basal level activity of the reporter gene in the absence of hormone. Addition of an agonist, dexamethasone, resulted in strong reporter gene activation dependent on the amount of GR expression plasmid used. Incubation with the antihormone RU486 also resulted in transactivation dependent on the concentration of GR, although at a much lower level compared with the agonist dexamethasone. Clearly, RU486 induces the MMTV reporter gene to a much lesser extent than the agonist, but partial agonistic activity can be measured. To rule out the possibility that the low magnitude of agonistic activity mediated by RU486 was due to a reduced DNA binding of the RU486-complexed GR, we fused the VP16 transactivation domain to the GR sequence. In this way the nuclear translocation and DNA binding activity can be separated from the transactivation function of the DNA-bound receptor within the nucleus, because after nuclear translocation VP16 constructs transactivate the reporter gene efficiently. The results (Fig. 2) show that in the presence of either dexamethasone or RU486 the VP16⅐GR fusion protein must have been translocated to the nucleus and bound to DNA with similar efficiency. In contrast, wild-type GR again only shows a partial agonistic activity in the presence of RU486. This result rules out the hypothesis that the partial agonistic activity of RU486 can be explained only by reduced DNA binding of an activated receptor. Rather, DNA-bound RU486-GR and dexamethasone-GR may recruit different transcription factor complexes. We therefore tested whether endogenous corepressors are involved in the limited activation by RU486. This was analyzed by overexpression of the thyroid hormone receptor, which interacts with corepressors only in the absence of thyroid hormone (11,12). Addition of hormone results in release of corepressors from TR. The MMTV-Luc reporter was cotransfected with expression vectors coding for GR and TR␤. As seen in Fig. 3, co-expression of TR leads to strong agonistic activity of RU486, concluding that unliganded TR squelches cofactors from RU486-bound GR. This increase in agonistic activity not only depends on the absence of T 3 but also on the presence of TR because cotransfection of the empty expression vector (control) had no effect. Such squelching experiments can be controlled by T 3 (34,35).
In the presence of thyroid hormone, TR expression had no effect on the partial agonistic activity of RU486 (Fig. 3). The agonist activity of RU486 is enhanced by TR, and T 3 relieves this enhancement.
These results strongly suggest that the activation function of RU486-complexed GR is impaired by binding to the same corepressors that bind to the unliganded TR.
RU486-dependent in Vivo Binding of Corepressors to the Glucocorticoid Receptor-Unliganded TR binds to corepressors, including SMRT and NCoR. To test the possible interaction of GR with NCoR or SMRT, we performed a modified mammalian two-hybrid analysis. For this assay, we again used the MMTV-Luc reporter cotransfected with expression plasmids coding for full-length human GR and an expression plasmid coding for either a VP16 fusion with the C-terminal half of SMRT (VP16⅐cSMRT), the C-terminal half of NCoR (VP16⅐cNCoR), or the VP16 activation domain itself. When corepressor binding occurs, the VP16 domain should transactivate the reporter gene. In the presence of RU486 both fusion proteins, VP16⅐cSMRT and VP16⅐cNCoR, result in a strong activation, whereas the VP16 activation domain by itself has no effect (Fig.  4). This strong activation reflects the interaction with GR and is dependent on RU486, because the addition of dexamethasone results in only a marginal activation. Because the MMTV reporter contains several GR binding sites (28), we tested whether this strong RU486-dependent interaction of GR with VP16⅐cSMRT and VP16⅐cNCoR is dependent on the binding element. Use of a duplicated palindromic consensus binding site for GR (GRE 2 ; Ref. 27) in a luciferase reporter construct resulted in a similar RU486-dependent interaction and transactivation as seen with the MMTV-Luc reporter (data not shown), showing that this interaction is not restricted to GR bound to the MMTV promoter.
In vivo complexes of corepressors NCoR or SMRT with the thyroid hormone receptor have been shown, as well as abrogation of corepressor binding by the addition of T 3 in two-hybrid assays (11,12). Here, we analyzed whether the antagonistbound GR is complexed with the corepressor NCoR in cell extracts. We enriched 293T cell extracts with baculovirus-expressed GR, immunoprecipitated the glucocorticoid receptor under various hormone conditions, and looked for coprecipitation of NCoR by Western blotting.
Immunoprecipitation of GR in the presence of RU486 results in coprecipitation of NCoR, showing that NCoR is complexed with RU486-bound GR in cell extracts (Fig. 5). This binding is specific since no NCoR is coprecipitated with the same antibody in the absence of hormone. As expected from the weak inter-action in a mammalian two-hybrid assay (Fig. 4) in the presence of dexamethasone, no complex could be detected.
Association of Glucocorticoid Receptor and NCoR-After having shown that the extent of agonist activity displayed by RU486-bound GR is dependent on the amount of available corepressors in the cell (see above), we wondered whether this RU486-mediated transactivation is confined to the C-terminal half of the receptor as has been shown for glucocorticoid-induced transactivation (36). Therefore we tested RU486-mediated transactivation of the MMTV-Luc reporter by wild-type GR (Fig. 6, GR 1-777), or a C-terminal region of GR (GR  418 -777). Surprisingly, deletion of the N-terminal region in the receptor construct GR-(418 -777) abolishes RU486-mediated After the addition of 293T cell extracts, they were subjected to immunoprecipitation with an antibody specific for GR in the absence or presence of the indicated ligands. Immunoprecipitated material, analyzed by Western blotting using antiserum directed against NCoR input (5%), and a control precipitation with an unrelated antibody are shown.
induction, whereas the dexamethasone induction is retained (not shown). This demonstrates that the N terminus of GR is critical for RU486-mediated agonism. Consequently, we wanted to identify the domains of GR involved in corepressor binding. We utilized a modified mammalian two-hybrid system to delineate the regions of GR required for NCoR interaction. In this setup, the natural DNA binding domain of GR is used for DNA binding, and different GR deletions are tested for interaction with corepressor VP16 fusions. Full-length GR, GR-(1-777), was expressed together with VP16 or VP16⅐cNCoR (Fig.  7A). As seen above, there is a strong, 80-fold RU486-dependent interaction detected. Dexamethasone-induced activation by the VP16⅐cNCoR construct is only 6-fold (Fig. 7D). The N-terminal deletion mutant GR-(418 -777) results in a reduced activation by VP16⅐cNCoR (Fig. 7, B and D). With this deletion, both agonist and antagonist induce a similar weak interaction of the corepressor (about 5-fold, see Fig. 7D), showing that the specific effect of the antihormone is completely abolished (Fig. 7B). GR deletions of the C terminus result in receptor constructs mediating hormone-independent, constitutive gene activation (37). Coexpression of VP16⅐cNCoR (Fig. 7, C and D) confers an interaction-mediated activation of moderate activity for the GR-(1-550) construct in contrast to the other deletions used, which show only a low level interaction. In particular, deletion of the N-terminal 261 amino acids results in a severe reduction in binding, indicating that this region is critical for interaction of NCoR with GR.
From these data it can be concluded that both the N and the C termini regions of the receptor are required for efficient NCoR interaction. A major determinant for antagonist-dependent NCoR interaction is mediated by the N terminus of GR, as is the antagonist-mediated activation.
In order to determine whether NCoR interacts directly with GR, bacterially expressed GST⅐NCoR fusion proteins were incubated with in vitro translated, 35 S radio-labeled full-length GR or deletion constructs (Fig. 8). The longest NCoR construct FIG. 7. N-terminal as well as C-terminal domains of GR are required for efficient interaction with NCoR. The indicated GR constructs were analyzed for binding to VP16⅐NCoR in the modified mammalian two-hybrid system using the MMTV-Luc reporter. CV1 cells were cotransfected with MMTV-Luc, expression vectors, and the indicated proteins. Cells were harvested after 48 h and assayed for luciferase activity. Full-length GR (A) and GR constructs containing an intact ligand binding domain (B) were tested in the absence or presence of the indicated hormones. C, GR constructs in which the ligand binding domain is deleted were tested in the absence of hormone only. Luciferase activity is expressed as Fold activation relative to VP16 in the absence of ligand. D, schematic view of full-length GR and GR deletion mutants used in A-C. The ligand binding domain (LBD) and the DNA binding domain (DBD) are marked. The strength of interaction is represented by the ratio of luciferase activity after transfection of VP16⅐cNCoR relative to the VP16 control under the same hormone conditions. that was used spans amino acids 1679 to 2453, a region containing three nuclear receptor interaction domains (Fig. 8A, N1, N2, N3) that coordinate binding to the TR (38 -40). This GST⅐NCoR protein, bound to glutathione beads, retains fulllength GR as well as the two deletion mutants, showing that NCoR can bind directly to GR (Fig. 8C). Further delineation of the interaction domain within NCoR showed that the very C-terminal constructs, GST⅐NCoR-(2296 -2453) and GST⅐NCoR-(2239 -2453), do not interact with full-length GR or with GR deletions even when the N1 domain is present (Fig.  8C). In contrast, GST⅐NCoR constructs spanning amino acids 1745-2215 and 1954 -2453 clearly show an interaction, although reduced, with full-length GR. Electrophoretic migration of full-length GR is perturbed by the GST⅐NCoR protein 1745-2215, which migrates at the very same position. Individual combinations of all of the GR constructs with all of the GST⅐NCoR constructs show a complex pattern of NCoR as well as of GR domains involved in binding. In particular, the region of NCoR from 2239 -2453, which contains the N1 domain, is not sufficient to bind GR. However, this domain contributes to binding to the GR region 1-488, because deletion of this region leads to a dramatic reduction in binding to GR region 1-488. Additionally, all of the NCoR constructs that bind to full-length GR are able to bind GR-(418 -777).
Therefore, both the receptor N terminus and the C terminus not only contribute to NCoR binding but can also bind individually. Furthermore, three NCoR regions from amino acids 1745-1954, 1954 -2215, and 2239 -2453 contribute to binding to GR. Thus, we have identified several domains of NCoR and GR that are involved in the direct interaction between these two factors. DISCUSSION Steroid hormone receptors are ligand-inducible transcription factors that, after steroid binding, regulate genes harboring steroid response elements. So-called hormone antagonists are highly relevant for endocrine therapies because they prevent steroid receptors from binding to their natural ligands. Antihormones have been grouped according to their effect on the conformational change of the receptor observed after ligand binding. Type I antihormones induce a marginal change in the receptor such that it more resembles the inactive receptor conformation. Hormone antagonists classified into a group of a higher type number more resemble the fully induced receptor after agonist binding. Therefore, the antiglucocorticoid RU486 is often classified as a type II antagonist leading to a mixed agonist/antagonist transcriptional activity (41)(42)(43). Both the estrogen receptor as well as the progesterone receptor bound by a mixed agonist/antagonist have been shown to be associated with the corepressors SMRT or NCoR (18,19,44). The function of the glucocorticoid receptor bound to an agonist has been shown to be affected by overexpression of corepressors. In addition, the agonist activity of mixed antagonists is changed upon corepressor expression (45). Here we have shown by in vitro binding that the glucocorticoid receptor is able to interact directly with the corepressor NCoR. Such an association is confirmed by an in vivo interaction assay, as well as by coimmunoprecipitation. Interestingly, both efficient coimmunoprecipitation and functional in vivo interaction depend on the presence of RU486. In the presence of the agonist dexamethasone, NCoR is not immunoprecipitated with GR. Similarly, functional interaction of GR with either NCoR or SMRT is dependent on the presence of RU486 and is only weakly detectable in the presence of dexamethasone. Upon deletion of the N-terminal half of the glucocorticoid receptor, GR-(418 -777), the strong RU486-specific two-hybrid interaction with VP16⅐cNCoR is lost, but the low level of dexamethasone-induced interaction is maintained. This result suggests that the N-terminal half of GR is required for efficient corepressor interaction. Similarly, deletion of the C-terminal ligand-binding domain diminishes corepressor interaction.
Corepressor domains interacting with nuclear receptors have been characterized, in particular the consensus sequence (I/L)XX(I/V)I (CoRNR box) (38 -40). Detailed analyses of NCoR and SMRT have revealed a difference in these interaction domains between these corepressors. For NCoR, three interaction domains have been characterized that confer receptor specificity (46,47). Two of these domains are necessary for full interactions with nuclear receptors, with a specific decrease in binding of NCoR to TR in case of mutation of the N3 domain (46). However, only the receptor C termini, not the N termini, have been analyzed in detail for corepressor interaction. Our analy- sis of in vitro interaction of NCoR with GR revealed a similar complexity. It is noteworthy that deletion of N1 reduced the binding to the N-terminal half of GR, GR-(1-488), whereas two of these interaction domains (N1 plus N2 or N2 plus N3) are sufficient for binding to full-length GR. Therefore, similar domains of NCoR involved in binding to the GR C terminus are also involved in binding to TR or RAR. The striking result of our experiments is that efficient GR/NCoR interaction depends on the antagonist RU486 and on the N-terminal receptor domain. Since ligand binding and ligand-induced conformational changes are confined to the C terminus, one might envisage cooperativity of both the N and C termini in binding to NCoR.
In addition to corepressor binding, the receptor N terminus mediates agonist activity in the presence of RU486. This region has been shown to contain one of the major transcriptional activation functions (AF1; Ref. 29). RU486-induced transcriptional activity seems to be limited by corepressors, because corepressor squelching results in a strong receptor activation. Therefore, the ratio of simultaneously binding coactivators and corepressors may explain the transcriptional effects of GR by ligands with mixed agonist/antagonist activities (22). The knowledge of these interaction mechanisms will, in the context of antihormone treatment, allow modulation of the mixed agonist/antagonist effects depending on the cell type and the respective cellular corepressor content.