| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Steroid Hormones Section, National Institute of Diabetes and Digestive and Kidney Diseases/Laboratory of Molecular and Cellular Biology, National Institutes of Health, Bethesda, Maryland 20892
Address all correspondence and requests for reprints to: Dr. S. Stoney Simons, Jr., Building 8, Room B2A-07, National Institute of Diabetes and Digestive and Kidney Disease/Laboratory of Molecular and Cellular Biology, National Institutes of Health, Bethesda, Maryland 20892. E-mail: steroids{at}helix.nih.gov.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Coactivators and corepressors are cofactors that have attracted considerable attention because they increase or decrease the total activity of most steroid-receptor complexes. Coactivators, such as the p160 family of coactivators steroid receptor coactivator 1 (10), transcriptional intermediary factor 2 (TIF2)/GR-interacting protein 1 (GRIP1) (11, 12), and amplified in breast cancer-1 (AIB1) (13) were originally defined as factors that increase the total amount of induced gene product with saturating, or pharmacological, concentrations of hormone (14, 15, 16). Conversely, corepressors such as nuclear receptor corepressor (NCoR) (17) and silencing mediator of retinoid and thyroid hormone receptor (SMRT) (18) are factors that are characterized by their ability to decrease the total amount of gene product. The prevailing model is that the nature of the ligand binding to receptors acts as a molecular switch, with agonist steroids causing both the dissociation of corepressors from ligand-free or antagonist-bound receptors and the association of coactivators (19, 20, 21).
Corepressors were initially discovered on the basis of their ability to bind to ligand-free nuclear receptors, such as the thyroid receptor (17, 18). The conclusion that corepressors bind only to the nuclear receptors was modified when it was found that corepressors interact with antagonist-bound androgen (AR) (22, 23), estrogen (ER) (24, 25), GR (26), and progesterone (27, 28) receptors and often to the ligand-free steroid receptors (reviewed in Ref.29). The sites of interaction in steroid and nuclear receptors for both corepressors and coactivators have been identified to be in the ligand-binding domain (LBD); in fact, the two sites appear to overlap (30, 31, 32, 33, 34).
Interestingly, corepressors also affect several biological properties of steroid receptor-agonist complexes such as the total amount of induced activity and the position of the dose-response curve (and the steroid concentration required for half-maximal induction, or EC50) (28, 29, 35, 36, 37). This suggests that whereas corepressors may not bind to agonist complexes of the nuclear receptors, there is at least a functional interaction with agonist complexes of the classical steroid receptors. The interactions of corepressors with steroid receptors are of additional importance because they may provide a mechanism for differentiating between the activated complexes of various steroid receptors (androgen, glucocorticoid, mineralocorticoid, and progestin) in a cell-specific manner. Whereas each receptor displays preferential binding of the naturally occurring ligands, each activated receptor-steroid complex is able to bind to the same hormone response element (HRE). Thus the specificity of ligand binding is lost due to the commonality of potential HREs. Nevertheless, overexpressed NCoR and SMRT produce diametrically opposite effects on both the dose-response curve of agonists and the partial agonist activity of antagonists bound to PRs vs. GRs for induction of the same gene in the same cells (37). Thus, corepressor interactions with DNA-bound steroid receptor complexes can restore some of the specificity that appears to be lost upon binding to common HREs. Furthermore, the effects of added corepressor on GR complexes are different in CV-1 vs. 1470.2 cells (36, 37). These results suggest that the observed effects of corepressors on GR transcriptional properties can be modified by tissue-specific factors.
Finally, the corepressor SMRT can antagonize the ability of the coactivator TIF2 to modulate the position of the dose-response curve of GR-agonist complexes and the amount of partial agonist activity of GR-antagonist complexes (36). These findings led us to propose an equilibrium model in which coactivators and corepressors each interact with both agonist- and antagonist-bound steroid receptors in a manner in which the intracellular ratio of coactivators to corepressors, as opposed to the absolute concentrations, influences the position of the agonist dose-response curve and the partial agonist activity of antisteroids (29, 35, 36, 37, 38). This model has been supported by the recent findings that coactivators both increase the partial agonist activity of antagonists and physically interact with antagonist-bound GRs in mammalian two-hybrid and glutathione-S-transferase (GST)-pull-down assays (39).
The purpose of this study, therefore, was to further test our proposed model and to determine whether the above biological responses of GRs to corepressors can be accounted for by direct binding interactions between GRs and corepressors. In particular, if there is an interaction, we wanted to know the effect of ligand (agonist vs. antagonist) on this interaction, whether coactivators and corepressors interact with GR in an equilibrium fashion, whether changes in the endogenous levels of corepressors modulate GR properties, and whether cell-specific differences in CV-1 vs. 1470.2 cells modify any of the biochemical properties of corepressor binding to GRs.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
. The action of corepressors in steroid receptor-regulated gene induction is thought to involve histone deacetylation (3, 4, 5). The data of Fig. 1 show that trichostatin A (TSA), which is an inhibitor of histone deacetylases (40), is unable to reverse the effects of added SMRT in CV-1 cells. As expected, SMRT does reduce the total level of transactivation by 2.6 ± 0.8 fold (±SD, n = 3) and the coaddition of 300 nM TSA with SMRT causes a 4.1 ± 0.9 (±range, n = 2) increase in total transactivation (data not shown). These results indicate that whereas the repression of GR-induced total activity by SMRT may involve histone deacetylases, the ability of SMRT to modulate the EC50 and partial agonist activity for GR transactivation is independent of any TSA-sensitive deacetylases.
|
), and the total amount of induced gene product from a luciferase reporter regulated by five repeats of the GAL4 response element was determined. Figure 2B shows a GR-dependent and steroid-dependent increase in the interaction of VP16/GR with GAL/NCoR-RID in CV-1 cells. Whereas the magnitude of the interaction is greatest for GR bound with the antiglucocorticoid RU486, a statistically significant increase over the GAL control is also seen with the agonist dexamethasone (Dex). Because the total level of gene expression is proportional to the strength of the interactions among equally expressed proteins, and GAL and GAL/NCoR-RID are expressed at equal levels (see inset), we can conclude that the significantly increased association of GR complexes with GAL/NCoR-RID vs. GAL is due to both GR-agonist and GR-antagonist complexes combining with NCoR-RID.
|
). Unexpectedly, GR-LBD (VP16/GR486C) is unable to interact with NCoR, regardless of whether GR is bound with agonist or antagonist (Fig. 2B). Western blots show that nearly equal amounts of each chimera are expressed in COS-7 cells (Fig. 2C). Furthermore, cell-free extracts of VP16/GR, VP16/GR486C, and VP16/GR361C that are overexpressed in COS-7 cells each contain approximately equal amounts of functional receptor as shown by similar amounts of specifically bound [3H]Dex (data not shown). Therefore, the inability to detect any interactions of VP16/GR486C with NCoR is not due to low levels of functional VP16/GR486C.
Interestingly, there is no statistically significant association of the amino-terminal half of GR (GRN523) with GAL/NCoR-RID vs. GAL (Fig. 2B). The relative inactivity of NCoR with both the amino- and carboxyl-terminal halves of GR suggests that no one domain is sufficient, and that regions in each half of GR are required for binding to NCoR.
Comparable results are seen for GR association with SMRT in the two-hybrid assay in CV-1 cells (Fig. 2D). Statistically significant SMRT- and steroid-dependent responses are seen with both agonist- and antagonist-bound GRs. As with NCoR, no SMRT-dependent interactions occur with either the amino- or carboxyl-terminal fragments of GR regardless of the steroid present. The major difference in GR interactions with SMRT vs. NCoR is that the steroid-dependent increases are about 20-fold less for SMRT (Fig. 2, panel B vs. panel D).
NCoR Interactions with Agonist-Bound GRs
To obtain additional support for our conclusion that corepressors interact with agonist-bound GRs, we asked whether a mutant NCoR that does not bind to receptors would display decreased responses with GRs in our two-hybrid assay. NCoR-RIDm12 is a GAL/NCoR-RID construct containing mutations in each of the CoRNR box motifs of the two RIDs that are required for NCoR binding to nuclear/steroid receptors (20). A mammalian two-hybrid assay with GAL/NCoR-RID constructs with and without the m12 mutations clearly shows that interactions of both Dex- and RU486-bound GRs require the presence of the wild-type CoRNR box motifs (Fig. 3A). Western blots (Fig. 3A, inset) establish that the inactivity of the NCoR-RIDm12 construct is not due to reduced levels of expressed protein and thus permits the conclusion that the increased levels of gene expression for both agonist- and antagonist-bound VP16/GR with NCoR-RID vs. NCoR-RIDm12 are due to a stronger association of the ligand-bound chimera with NCoR-RID.
|
) supports the conclusion that NCoR interacts with wild-type GR bound by both agonist and antagonist steroids in the context of intact cells. Thus, the above interactions with VP16/GR are not an artifact of the fused VP16 activation domain. The fact that an even stronger interaction of GAL/NCoR-RID is seen with the antagonist RU486 suggests that some conformations of RU486-bound GRs maintain transcriptionally competent structures and that other parameters are responsible for the usually negligible gene induction of RU486-bound GRs from a typical GRE-directed gene (36, 41). The much lower activity of wild-type GR with GAL/SMRT-RID in Fig. 3B is not unexpected considering the 
20-fold lower activity of VP16/GR with SMRT vs. NCoR in CV-1 cells (Fig. 2, panel B vs. panel D).
Finally, we asked whether complexes between GR and corepressors could be formed in intact cells. Cytosols were prepared from Cos-7 cells that had been transiently transfected with GR and either GAL or GAL/NCoR-RID and then treated with Dex. Anti-GAL4 antibody was used to precipitate the GAL or GAL/NCoR-RID. The precipitates were analyzed by SDS-PAGE, and the existence of coimmunoprecipitated GR was determined by Western blotting with anti-GR antibody. Figure 3C shows that steroid-free and -bound GRs are specifically coimmunoprecipitated in a manner that requires the presence of NCoR-RID. We therefore conclude that Dex-bound GRs in addition to ligand-free and antagonist-bound GRs form complexes with corepressors in intact cells.
NCoR Binding to GR
The association of two proteins in mammalian two-hybrid and coimmunoprecipitation (co-IP) assays does not require direct interactions. They may be mediated by an adapter protein. In vitro pull-down assays, where one protein has been overexpressed in bacteria and selectively binds the other protein that has been in vitro translated in reticulocyte lysates, is stronger, but still not unequivocal, evidence for the direct binding. We therefore used a pull-down assay to examine the ability of overexpressed GST/NCoR-RID to immobilize in vitro translated GR in the presence of EtOH, Dex, and RU486 (Fig. 4). Full-length GR displays good binding to GST/NCoR-RID, which depends upon the presence of NCoR and is steroid independent. In view of the dramatically decreased interactions of steroid-bound GR407C and GR486C with NCoR in the mammalian two-hybrid assays (Fig. 2B), the binding of truncated GRs was also examined. Unexpectedly, GR407C exhibits binding to GST/NCoR-RID that is almost indistinguishable from the full-length GR. However, as seen in the two-hybrid assays, significantly reduced or no binding to NCoR is seen for the steroid-bound GR LBD (GR486C) or the amino-terminal 523 amino acids (GRN523), respectively. These results support our earlier conclusion that strong NCoR binding to steroid-bound GRs requires more than one domain of GR.
|
) can disrupt the association of GAL/NCoR-RID with VP16/GR. The observation in Fig. 5B that TIF2 reduces the response of VP16/GR plus GAL/NCoR-RID when GR is bound with either Dex or RU486 indicates that TIF2 competitively inhibits NCoR interactions with both agonist- and antagonist-bound GRs. These results support an equilibrium interaction of coactivators and corepressors with GRs.
|
) (42), displays no competitive activity. These data argue that the TIF2 RIDs themselves are not sufficient to compete for NCoR association with GR-agonist or -antagonist complexes and that GR regions other than the coactivator binding pocket of helices 3, 4, 5, and 12 (44) are important for corepressor binding. Preliminary experiments revealed that the ability of TIF2 to competitively inhibit RU486-bound GR interactions with GAL/SMRT-RID is similar to those depicted in Fig. 5B for GAL/NCoR-RID in that TIF2 and TIF2.0 are the most potent competitors (data not shown).
Studies with the nuclear receptors (30, 31, 32, 33) and ERs (34) indicate that corepressor binding overlaps much of the same receptor surface that is contacted by coactivators. To further examine whether this overlap is less extensive for coactivator and corepressor binding to GRs in our system, we tested the ability of TIF2 fragments to inhibit VP16/GR complex binding to GAL/TIF2.4, which contains the three RIDs of TIF2 (Fig. 5A) (42), in our mammalian two-hybrid assay with CV-1 cells. As we previously reported (39), the association of RU486-bound GRs with TIF2.4 is much less than that with the full-length TIF2. Nevertheless, TIF2 significantly diminishes the response of TIF2.4 with both Dex- and dexamethasone 21-mesylate (Dex-Mes)-bound GRs whereas TIF2.0 and TIF2m123 are inactive (Fig. 5C). Thus, a major difference exists in that TIF2.0, comprising the amino-terminal 627 amino acids of TIF2, competitively inhibits the response of GR association with NCoR (Fig. 5B) but not TIF2 (Fig. 5C). We therefore conclude that corepressor binding to GRs involves more than the TIF2 binding site in the LBD of GRs.
The capacity of TIF2 to inhibit GR interactions with the corepressor NCoR in CV-1 cells is demonstrated in Fig. 5B. In Fig. 5D, we examined the converse: the ability of corepressor to prevent GR binding to TIF2. With both full-length NCoR and SMRT, a concentration-dependent increase in the ability of corepressor to prevent the association of TIF2 with agonist- and antagonist- (41) bound GRs is observed. This is precisely the result that would be expected if the association of coactivators and corepressors to GRs complexes is dictated by equilibrium binding reactions. The much higher amount of NCoR required here than for GAL/NCoR-RID in Fig. 5B is due to the much lower level of expression of the full-length corepressor (Fig. 5E; note that 8 times less GAL/NCoR-RID plasmid is used here). This inhibitory activity of corepressors is not due to a nonspecific effect on transactivation in general. As shown in Fig. 5F, 150 ng of SMRT do not reduce the ability of a GAL/VP16 chimera to induce luciferase expression under the same conditions where SMRT represses the interaction of GAL/TIF2.4 and VP16/GR complexed with either agonist or antagonist.
Competitive Interaction of Coactivators and Corepressors with GRs in Intact Cells
To support the conclusion that coactivators and corepressors competitively inhibit the binding of each other to GR complexes, we turned to whole-cell co-IP assays. Specifically, we asked whether the overexpression of corepressors in cells could reduce the ability of transfected coactivators to bind to cotransfected GRs, as seen by a decrease in the amount of GR that can be coimmunoprecipitated with anticoactivator antibody due to its presence in GR-coactivator complexes. The data of Fig. 6A show that, under conditions where equal amounts of hemagglutinin (HA)/GRIP1 are immunoprecipitated (top panel of "Co-IP" section), approximately equal amounts of GR are associated with GRIP1 regardless of whether GR is ligand free or bound by either the agonist Dex or the antiglucocorticoids Dex-Mes and RU486 (lanes 3–6). The presence of GRs in the immunoprecipitated material is not seen when HA is cotransfected instead of HA/GRIP1 (lanes 2 vs. 3), showing that the co-IP of GR complexes with HA/GRIP1 depends upon the presence of overexpressed GRIP1.
|
; bottom panel of "Co-IP" section, lanes 2 and 4 vs. 1), just as seen in Fig. 6A. When GAL/NCoR-RID is cotransfected, however, there is a decrease in the amount of coimmunoprecipitated GR for receptors bound with both the agonist Dex (lanes 5 vs. 4) and the antagonist RU486 (lanes 3 vs. 2). These data argue that, under whole-cell conditions, corepressors and coactivators compete with each other for the binding to GR complexes of both agonist and antagonist steroids.
Modulation of GR Properties by Varying the Intracellular Ratios of Coactivators to Corepressors
As a final test of whether GR transcription properties reflect the equilibrium binding reaction of different ratios of coactivators to corepressors, we sought to modify this ratio not by overexpressing either cofactor but by adding thyroid hormone receptor ß (TRß) ± ligand to selectively sequester one or another of the available cofactors in intact cells. Ligand-free TRß preferentially binds NCoR over SMRT (45, 46) whereas ligand-bound TRß avidly complexes coactivators (44, 47, 48). Recently, it has been reported that, in CV-1 cells, the p160 coactivators bind to corepressors (49) (data not shown) and that the coactivator ACTR (amplified in breast cancer 1, AIB1) binds to ligand-free TRß only in the presence of corepressor (NCoR) (49). Thus, the presence of excess unliganded TRß should cause a decrease in both NCoR and coactivators. We have reported that NCoR is without effect on GR in CV-1 cells (36). Therefore, overexpressed TRß, by selectively reducing the concentration of NCoR and coactivators, should produce a right shift of the dose-response curve to higher EC50 values and a decrease in the partial agonist activity due to the dominant effect of coactivator concentration (Fig. 7A). The addition of thyroid hormone (T3) leads to the dissociation of corepressors and the binding of coactivators, which should also cause a right shift (and increase in the EC50) and decrease in the partial agonist activity. This scenario further predicts that if corepressor is limiting, the addition of more NCoR will eventually cause ligand-free TRß to complex the same amount of coactivators as T3-bound TRß, thereby giving the same increase in EC50 and decrease in partial agonist activity.
|
), and partial agonist activity of Dex-Mes (Fig. 7C), was determined. There is a slight but statistically significant increase in the Dex EC50 and decrease in Dex-Mes partial agonist activity in the absence of T3, both of which are augmented by adding T3. We believe that the minimal effect of unliganded TRß is due to low levels of endogenous corepressors, as discussed above. These results are also consistent with the report that the lack of repression by GAL/retinoic acid receptor in CV-1 cells is due to a lack of SMRT (50) even though appreciable levels of SMRT and NCoR mRNAs are present in CV-1 cells (51). We therefore examined the consequences of increasing the concentration of NCoR by cotransfecting NCoR plasmid. In the absence of added TRß, there is no effect, just as we have reported earlier (36). Now, however, the presence of unliganded TRß produces a much more dramatic increase in the EC50 and decrease in Dex-Mes partial agonist activity. This is the predicted response as the larger amounts of NCoR available to bind to unliganded TRß will sequester more coactivators (49). The effect of adding T3 to cells containing overexpressed NCoR and TRß is small as the net result is to release NCoR, which is relatively inactive (Fig. 7A). Overexpression of the corepressor SMRT ± TRß produces similar responses. However, SMRT by itself is much more effective than NCoR in modifying the EC50 and partial agonist activity of GR complexes in CV-1 cells (36). The presence of unliganded TRß accentuates the response by removing coactivators. The addition of T3 has no further effect because no additional coactivators are depleted (Fig. 7A). The larger responses of TRß with added SMRT vs. NCoR are probably due to the fact that equilibrium conditions allow some SMRT binding to GRs to give additional repressed effects. In summary, the modulation of the EC50 of GR-agonist complexes and the partial agonist activity of GR-antagonist complexes that is achieved by using exogenous TRß ± T3 to vary the ratios of endogenous coactivators to corepressors offers strong support for our model in which the equilibrium binding of both coactivators and corepressors to GRs can perturb the transcriptional properties of all GR complexes.
Interactions of GR with Corepressors in 1470.2 Cells
We previously reported that the effects of corepressors on the EC50 and partial agonist activity for GR-induced gene expression are different in monkey kidney CV-1 cells and 1470.2 cells, which is a line of mouse mammary adenocarcinoma cells (37). We therefore asked whether the interactions of corepressors with GRs are any different in 1470.2 cells vs. CV-1 cells. As in CV-1 cells (Fig. 2B), a strong interaction of RU486-bound GRs and a much weaker, but still statistically significant, response for Dex-bound GRs is seen with GAL/NCoR-RID that depends upon the presence of GRs (Fig. 8A). Interestingly, the antagonists Dex-Mes and dexamethasone oxetanone (Dex-Ox) afford even less gene activation than the agonist Dex, although all responses are still significant relative to that without GR. Preliminary results indicate that the interactions of GR bound with the antiglucocorticoid progesterone are almost identical to those with Dex-Ox (data not shown). The inability of the GR-LBD in GAL/GR486C to interact with GAL/NCoR-RID in 1470.2 cells is similar to that observed in CV-1 cells (Fig. 2B). However, the amount of steroid-inducible responses is much less in 1470.2 than CV-1 cells, and the decreased responses in CV-1 cells upon deletion of the amino-terminal 360 amino acids to give GR361C are not observed in 1470.2 cells.
|
). Although the magnitude of the interaction is greatest for GR bound with the antiglucocorticoid RU486, statistically significant increases over the EtOH control are also seen with other antagonists (Dex-Mes and Dex-Ox). A significant association of SMRT and GR bound with the agonist Dex that is greater than those with the antagonists Dex-Mes and Dex-Ox is also observed. GR-LBD (VP16/GR486C) is unable to interact with SMRT, regardless of whether GR is bound with agonist or antagonists.
The interaction of SMRT with GR in 1470.2 cells, as judged from the level of total gene activation, is much lower than that with GAL/NCoR-RID (Fig. 8, panel B vs. panel A). Thus, as in CV-1 cells (see Fig. 2, D vs. B), the strength of the interactions of SMRT-RID with GR appear to be much less than that of NCoR-RID. Interestingly, deletion of the N-terminal 360 amino acids of GR significantly diminishes the association of GR with NCoR- or SMRT-RID in CV-1 cells but not in 1470.2 cells (Fig. 2, B and D, vs. Fig. 8, A and B).
Competition of GR-Corepressor Interactions by Added Cofactors in 1470.2 Cells
Because only minor differences for the interactions between corepressors and GRs in 1470.2 vs. CV-1 cells were observed, we examined the ability of coactivators and corepressors to inhibit GR binding to corepressors in 1470.2 cells. Added full-length NCoR and SMRT reduce the total level of induced reporter resulting from the interactions of GAL/NCoR-RID with RU486-bound VP16/GR (data not shown), consistent with the interaction of RU486-bound GRs with corepressors being reversible. Full-length TIF2 inhibits the interactions of both agonist- and antagonist-bound GRs with NCoR-RID in 1470.2 cells (Fig. 9), just as is observed in CV-1 cells (Fig. 5B). In both cell lines, a fragment of TIF2 (TIF2.5) containing the three RIDs that are required for binding to GR (42, 43), i.e. TIF2.5 (42), is inactive whereas the full-length TIF2 with mutated RIDs (TIF2m123) still competes (Figs. 9 and 5B). Thus, in both cell lines, the TIF2 RIDs themselves are not sufficient to compete for NCoR association with GR-agonist or -antagonist complexes, which indicates that GR regions other than the coactivator binding pocket of helices 3, 4, 5, and 12 (44) are important for corepressor binding. A major difference between CV-1 and 1470.2 cells is, however, the ability of amino-terminal fragment (TIF2.0) to compete for GR-NCoR interactions. TIF2.0 is a good competitor of the association of RU486-bound GR with NCoR in CV-1 cells but not in 1470.2 cells (Fig. 5B vs. Fig. 9). These results suggest that cell-specific factors may influence the association of TIF2 and NCoR with GRs.
|
) show not only that CV-1 cells contain more NCoR than do 1470.2 cells, but also that the low amount of transiently transfected NCoR-Flag (cf. Fig. 5E) causes a larger percentage increase in the total amount of NCoR in 1470.2 cells than in CV-1 cells. Thus, the greater activity of NCoR in 1470.2 cells than in CV-1 cells may result from the greater fold increase of NCoR in transiently transfected 1470.2 cells and/or from the higher levels of endogenous NCoR in CV-1 cells being sufficient for a maximal effect on GR properties.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
The ability of ectopic NCoR to alter GR properties in 1470.2, but not CV-1, cells (36, 37) appears to be due to the lower levels of endogenous NCoR in 1470.2 cells, which are more significantly increased by transfected NCoR. Very few differences were observed in the interactions between GR complexes and corepressors in CV-1 vs. 1470.2 cells. However, we cannot eliminate the role of other cell-specific factors that could alter the responses to corepressors in the two cell lines. Cell-specific factors may also account for the fact that TIF2 and TIF2m123 compete for GR-NCoR associations in both cell lines whereas the amino-terminal fragment of TIF2 (TIF2.0) is an effective inhibitor only in CV-1 cells (Figs. 5B vs. Fig. 9). Additional studies are required to test this hypothesis.
Whereas early studies indicated that corepressor binding was restricted to the nuclear receptors, subsequent investigations revealed that corepressors also bind to antagonist complexes of AR (22, 23, 52, 53), ER (24, 25), GR (26), MR (Wang, Q., W. F. Richter, S. L. Anzick, P. S. Meltzer, and S. S. Simons, Jr., manuscript submitted), and PR (27, 28) and to agonist-bound complexes of AR (22, 23, 55, 56) and GR (Ref.26 and reviewed in Ref.29). The ability of corepressors to associate with agonist-bound receptors has not been commonly observed (19, 20, 21), and the magnitude in our experiments is modest. However, the statistically significant increased activities in the mammalian two-hybrid (Figs. 2, B and D, and 8, A and B) and hybrid-and-a-half (Fig. 3B) assays for both agonist- and antagonist-bound GRs, combined with the elimination of this steroid-induced activity when the CoRNR motifs of NCoR are mutated (Fig. 3A), and the association of GR-hormone complexes with NCoR in both pull-down (Fig. 4) and co-IP (Fig. 3C) assays provide strong evidence that both agonist- and antagonist-bound GRs bind to corepressors, albeit not with the same affinity. Collectively, these data support the hypothesis that corepressors can modify the dose-response curve of glucocorticoids and the partial agonist activity of antiglucocorticoids by directly interacting with GR-steroid complexes.
The similar binding of Dex- vs. RU486-bound GRs to NCoR in both the pull-down (Fig. 4) and co-immunoprecipitation (Fig. 3C) assays contrasts with the much larger differences between the biologically productive associations of Dex- vs. RU486-bound GRs with NCoR in the two-hybrid assays (Fig. 2B). This suggests that the interactions seen in soluble GR-NCoR complexes may be modified by other factors in DNA-bound transcriptional complexes. Alternatively, the transcriptional activity of RU486-bound GR complexes that are tethered to DNA, as in the two-hybrid assay, may be unexpectedly high, which is consistent with the strong transactivation of the wild-type GR-RU486 complexes in the hybrid-and-a-half assay (Fig. 3B). Further experiments are needed to resolve this issue.
The ability of ligand-free GR to bind NCoR (Fig. 4) is not unanticipated in view of the fact that corepressors were discovered on the basis of their ability to bind ligand-free nuclear receptors (17, 18). Furthermore, the ligand independence of NCoR binding that we see with GR appears to be a common phenomenon among steroid receptors as it has been reported for AR (22, 56), GR (26), PR (25, 28), and ER (57). The ligand dependency for corepressor activity for GR is most easily explained by the fact that the majority of ligand-free GR is cytoplasmic and does not become nuclear, or bind to GREs to regulate gene expression, until bound by steroid. Therefore, even if a complex is formed in the cytoplasm of intact cells between corepressors and ligand-free GRs, its presence would not be detected in our whole-cell biological activity assays. However, other processes and factors may also be involved.
A major objective of the present study was to test our hypothesis that corepressor binding to GRs does not reflect a switch in the properties of agonist vs. antagonist GR complexes but rather is part of a whole-cell equilibrium binding reaction to GRs that is influenced by the ratio of coactivators to corepressors (29, 35, 36, 37, 38, 39). We now provide several lines of evidence supporting an equilibrium reaction for GR binding to coactivators vs. corepressors. First, the coactivator TIF2 is able to block the productive association of GR with NCoR in both 1470.2 (Fig. 9) and CV-1 cells (Fig. 5B), indicating that this is a general response in intact cells. Second, corepressors reduce the interactions of GR with coactivators in two-hybrid assays (Fig. 5D). Third, the co-immunoprecipitation of both agonist- and antagonist-bound GRs with coactivators in whole cells (Fig. 6A) is reversed by added corepressors (Fig. 6B). It is especially relevant that corepressors reduce the binding of coactivators with GRs regardless of whether the bound steroid is an agonist or an antagonist. Finally, the EC50 of agonists and the partial agonist activity of antisteroids can be modified simply by adding a factor (TRß ± T3) that selectively sequesters the existing cofactors, thereby altering the intracellular ratio of coactivators to corepressors (Fig. 7). Therefore, we conclude that corepressors and coactivators associate with both agonist and antagonist complexes of GR in an equilibrium binding reaction, as previously proposed (29, 35, 36, 37, 38), so that some molecules of any GR-steroid complex are bound by corepressors whereas others are bound by coactivators. With coactivators and corepressors having opposite effects on the values for the EC50 of GR-agonist complexes and the partial agonist activity of GR-antagonist complexes (29, 35, 36, 38, 39), equilibrium interactions of coactivators and corepressors with GRs would permit almost any fraction of full gene activation during development, differentiation, homeostasis, and endocrine therapies. This flexibility will be further increased when the activities of other modulatory factors are included (reviewed in Ref.29).
The molecular details of the equilibrium binding of coactivators and corepressors to GRs are not known. We suggest that each ligand gives rise to an interconverting population of molecular conformations, as has been proposed from studies with ER complexes (58). However, instead of individual tertiary structures of each GR-steroid complex binding only to coactivators or to corepressors, we speculate that many conformers bind both coactivators and corepressors, albeit with different affinities. Corepressors may have a larger interaction surface than coactivators (59), which could lead to a higher binding affinity to GR as has been previously proposed (36). As in all equilibrium interactions, though, it is the product of the affinity and the protein concentrations that determines the net amount of receptor-cofactor complex. Thus, higher concentrations of coactivator could yield more GR-coactivator complexes even if the affinity of coactivators for GRs is less than that of corepressors.
The competitive equilibrium binding of coactivators and corepressors is most easily explained if they both bind to the same surface of GR. Currently, there is abundant evidence that the binding site of coactivators and corepressors overlap in the LBD of nuclear receptors (30, 31, 32, 33) and ERs (34). However, the situation with GRs appears to be more complex as the C-terminal half of GRs (amino acids 486–795, which includes the LBD), when bound with Dex or RU486, shows negligible interaction with NCoR or SMRT in whole-cell bioassays (Fig. 2, B and D, and Fig. 8, A and B) and very little binding in pull-down assays (Fig. 4). There is also no corepressor binding to the isolated amino-terminal half of GRs (amino acids 1–523; Fig. 2, B and D, and Fig. 4). These results argue that both the amino and carboxyl terminal halves of GR are necessary for corepressor binding, not only to Dex-bound complexes but also to RU486-bound complexes. This same conclusion was reached by Schulz et al. (26) when they looked at the ability of truncated GRs bound to a murine mammary tumor virus promoter to interact with a VP16/NCoR chimera. When our data for GR, or VP16/GR constructs, are combined with the results of Schulz et al., we can conclude that the corepressors NCoR and SMRT interact with both ends of GR in a manner that does not require the DNA binding of GR or any DNA-induced conformational changes in GR. This importance of two GR regions for corepressor binding also offers an attractive explanation for how segments of TIF2 that are unable to prevent GR interactions with TIF2 (Fig. 5C) nonetheless can inhibit GR association with corepressors (Fig. 5B and Fig. 9). This conclusion is further supported by our previous observation that the biological consequences of exogenous NCoR and SMRT depend upon the composition of the amino and carboxyl termini of both GR and PR (37). In this respect, there may be fundamental differences in the binding of corepressors to GR and PR vs. the nuclear receptors. ARs may present yet another interaction scheme as corepressors interact with only the amino-terminal sequences in the absence of steroid and in the presence of the antiandrogen CPA (22).
Some coactivators are known to bind to both the amino and carboxyl termini of GRs (60, 61), as seen here for corepressors. However, the molecular details of corepressor binding to GRs appears to be significantly different from that of coactivators. In intact CV-1 cells, TIF2 constructs lacking functional RIDs (e.g. TIF2m123 and TIF2.0) are still effective competitors of NCoR interactions with GRs whereas species with wild-type RID sequences are required to block the association of GRs with TIF2 (cf. Fig. 5, panel B vs. panel C). Thus, TIF2 sequences other than the RIDs are able to reduce corepressor but not coactivator binding to GR. Further experiments are required to unravel the mechanism by which this competition occurs, but it suggests that there are major differences in the contacts of coactivators and corepressors with GRs.
The corepressor SMRT modulates the EC50 of agonist complexes, and the partial agonist activity of antagonist complexes, of GRs in CV-1 cells in a manner that is not altered by the histone deacetylase inhibitor TSA (Fig. 1). Therefore, the acetylation/dea-cetylation of histones does not appear to be relevant for these modulatory activities of SMRT. The fact that TSA increases the levels of total transactivation in the absence of added SMRT or NCoR is of interest and suggests that endogenous deacetylases do affect other properties of GR-agonist complexes. It is also noteworthy that TSA increases the total transactivation in the presence of SMRT by 4-fold but has no effect on the position of the dose-response curve or the amount of partial agonist activity. This result reinforces our previous findings that the ability of cofactors to modulate the position of the dose-response curve of agonists, and the partial agonist activity of antagonists, is unrelated to their capacity to change the total levels of transactivation (36, 38, 39, 62, 63, 64).
In summary, GR-agonist and -antagonist complexes each interact with corepressors and coactivators in an equilibrium reaction. This relaxed specificity of GRs contrasts with the nuclear receptors, which bind corepressors only when ligand free or complexed with antagonists and interact with coactivators when bound with agonist ligands. Part of these differences between GRs and nuclear receptors may be due to GR domains other than the LBD being required for maximal binding under two-hybrid and pull-down conditions. Also, regions of coactivators other than the LxxLL sequences of the RIDs are able to compete for corepressor binding to GRs. Finally, the apparently more extensive association of GR- vs. nuclear receptor-agonist complexes with corepressors may actually reflect a reduced strength in corepressor association of GR-antagonists complexes. For example, about 40% of unliganded TR is reported to bind to corepressors (65) whereas less than 10% of GR binds to corepressors under our conditions. The expanded and overlapping conditions for GR binding to coactivators and corepressors support the model that several GR transcriptional properties are influenced by the intracellular ratio of coactivators to corepressors. This ability of GR properties to assume a continuum of values, as opposed to the discrete values that might be predicted by mechanisms involving "switches," has major mechanistic consequences regarding the differential control of gene transcription by GRs. Further research is required to elucidate the molecular details of these controls and to determine whether other steroid receptors are similarly affected. The report that the p160 coactivators are able to shift the dose-response curve of vitamin D3 receptors to the left (66) suggests that this phenomenon may also extend to the nuclear receptors.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Chemicals, Buffers, and Plasmids
[3H]dexamethasone (Dex, 91 Ci/mmol) was obtained from NEN Life Science Products (Boston, MA). Nonradioactive Dex and T3 are from Sigma Chemical Co. (St. Louis, MO). Dex-Ox (67) and Dex-Mes (68) were prepared as described. Restriction enzymes and digestions were performed according to the manufacturer’s specifications (New England Biolabs, Beverly, MA).
The Renilla null luciferase reporter was purchased from Promega Corp. (Madison, WI) and Renilla-TS reporter was a gift from Drs. Nasreldin M. Ibrahim and Otto Fröhlich (Department of Physiology) and Dr. S. Russ Price (Department of Medicine) at Emory University School of Medicine (Atlanta, GA). GREtkLUC has been previously described (41) and FR-LUC ([UAS]5tkLUC) is from Stratagene (La Jolla, CA). The cDNA plasmid of GR (pSVLGR) was from Keith Yamamoto (University of California at San Francisco, San Francisco, CA). VP16/GR was described previously (69). TIF2, TIF2.0, TIF2.5, and TIF2m123 were from Hinrich Gronemeyer (Institut de Génétique et de Biologie Moléculaire et Cellulaire, Strasbourg, France). NCoR (Michael Rosenfeld, University of California, San Diego, CA), s-SMRT (70) (Ron Evans, Salk Institute, La Jolla, CA), GAL/ and VP16/NCoR-RID (amino acids 1944–2453), GAL/ and VP16/SMRT-RID (amino acids 982-1495) (71), and GST/NCoR-RID (72) were received as gifts from Mitch Lazar (University of Pennsylvania School of Medicine, Philadelphia, PA). TRß was supplied by Paul Yen (National Institute of Diabetes and Digestive and Kidney Diseases). GAL/VP16 (pm3-VP16AD) was purchased from CLONTECH Laboratories, Inc. (Palo Alto, CA).
Construction of Plasmids
The hSA/pBluescript SK– construct was obtained from the I.M.A.G.E. Consortium (Lawrence Livermore National Laboratory, Clone ID 83491). hSA/pSG5 was prepared by subcloning the EcoRI/KpnI digest of hSA/pBluescript SK– directionally into the same restriction sites of pFLAG-CMV2 (Sigma). The hSA was then excised with, and inserted directionally into, the restriction sites in pSG5 for EcoRI and BamHI. hSA/pCMX was constructed by subcloning hSA from the EcoRI/BamHI restriction sites of hSA/pSG5 directionally into the same restriction sites within pCMX. GAL/GR361C was cloned by the PCR amplification of the pSVL-GR plasmid with primers (5'-primer: 5'-ATGGATCCTTTCTCAGCAGCA-GGATCA; 3'-primer: 5'-TATCTAGAGTCATTTTTGATGAAA-CAG) containing BamHI and XbaI linker sequences, respectively. After 30 cycles of amplification, the PCR products were gel purified, cut with BamHI and XbaI, and subsequently cloned into the BamHI/XbaI-cut pM vector (CLONTECH). VP16/GR361C and /GR407C chimeras were made by cutting Gal/GR361C and /GR407C (39) with BamHI and XbaI and inserting the GR fragment from each digest into the corresponding sites of the VP16 plasmid. VP16/GR486C was constructed by digesting pSVLGR with SfaNI, filling in with Klenow, and recutting with XbaI. Similarly VP16 was digested with AccI, filled in with Klenow, and recut with XbaI. The SfaNI/XbaI fragment of pSVLGR was then ligated to the AccI/XbaI fragment of VP16. VP16/GRN523 was prepared by ligating the 1.6-kb fragment from EcoRI/PstI digestion of VP16/GR to the 3.3-kb EcoRI/PstI fragment from VP16.
The following procedures were used to obtain the various GR constructs for in vitro transcription/translation. For pBAL-GRN523, three fragments were ligated: the 0.5-kb fragment from BamHI/SalI digestion of pSVLGR, the 1.2-kb fragment obtained by SalI/EcoRI digestion of the PCR-amplified region of pSVLGR using the forward primer 5'-AATAGGTCGACCAGCGTTCCAG-3' and reverse primer 5'-TAGAATTCGTCATGCAGTGGCTTGCTGAATC-3', and the 3.0-kb fragment from BamHI/EcoRI digestion of pBAL-GR. For pBAL-GR361C, the 1.3-kb fragment from BamHI digestion of the PCR-amplified region of pSVLGR using the forward primer 5'-ATGGATCCAATGTCCCTTTCTCAGCAGCAGG-3' and reverse primer 5'-TATCTAGAGTCATTTTTGATGAAACAG-3' was ligated with the 3.0-kb fragment from BamHI/SmaI digestion of pBAL-GR. pBAL-GR407C and pBAL-GR486C were prepared in the same manner, with the same reverse primer as for pBAL-GR361C, but using the forward primer 5'-ATGGATCCAATGCGGTCAGTGTTTTCTAATG-3' for amino acids 407–795 and the forward primer 5'-ATGG-ATCCAATGATTGATAAAATTCGAAGG-3' for amino acids 486–795.
Cell Culture and Transfection
Monolayer cultures of COS-7, CV-1, and 1470.2 cells were grown as described previously (35, 63). Cells are transfected for 18 h using Lipofectamine (Life Technologies, Inc., Gaithersburg, MD) or FuGene (Roche Diagnostics, Indianapolis, IN) as recommended by the supplier. For each well of a 24-well plate, we use 100 ng of reporter (FR-LUC or GREtkLUC) and 5 or 10 ng Renilla-TS plus various combinations of other expression vectors. Equal molar amounts of expression vectors lacking GR or cofactors (i.e. hSA/pCMX, hSA/pSG5, VP16, or Gal) are included to keep the molar amount of each vector constant, with the total transfected DNA brought to 300 ng/well with pBSK+ unless otherwise indicated. The cells are then treated for 24–40 h with 0.1% ethanol ± steroids in media containing 10% FBS and harvested in 1x Passive Lysis Buffer (150 µl/well; Promega). Cell lysates (50 µl) are used to assay for luciferase activity using the Dual-Luciferase Assay System from Promega according to the supplier. The data are then normalized for the cotransfected Renilla activity.
Bacterial Expression of Proteins
pGEX series of plasmids are transformed into Escherichia coli (BL21[DE3]; Stratagene) according to the manufacturer’s procedure. A single colony is picked and inoculated into 3 ml LB broth with 100 mM Ampicillin. After overnight culture, 1 ml of bacterial culture is diluted into 50 ml of LB broth containing 100 mM Ampicillin, shaken at 37 C for 2 h, adjusted to 0.5 mM isopropyl-ß-D-thiogalactopyranoside, and shaken at 25 C for another 3 h. The overexpressed proteins are isolated after cell lysis with lysozyme and detergent as described by Zamir et al. (72).
In Vitro Transcription and Translation Assays
pRBAL-GR (1 µl), 40 µl Promega TNT (SP6) Quick Coupled Transcription/Translation System master mix, and 2 µl [35S]methionine (Amersham Biosciences, Piscataway, NJ) are br
0.028 ***, P < 0.005. C, Coactivator TIF2 inhibition of GAL/TIF2.4 association with VP16/GR. Triplicate wells of CV-1 cells were transiently transfected with 2.5 ng GAL/TIF2.4, 10 ng VP16/GR chimera, 100 ng FRLUC, and 5 ng Renilla TS plus the molar equivalent of 75 ng of TIF2 for the indicated TIF2 plasmids. The cells were then incubated with steroid, harvested, and analyzed, and the average values of three independent experiments were plotted as described for Fig. 2B
