Differential SERM Effects on Corepressor Binding Dictate ERα Activity in Vivo *

Selective estrogen receptor modulators (SERMs) show differential effects upon ERα activation function 1 (AF-1). Tamoxifen allows strong ERα AF-1 activity, whereas raloxifene allows less and ICI 182,780 (ICI) allows none. Here, we show that blockade of corepressor histone de-acetylase (HDAC) activity reverses the differential inhibitory effect of SERMs upon AF-1 activity in MCF-7 cells. This suggests that differential SERM repression of AF-1 involves HDAC-dependent corepressors. Consistent with this, ICI and raloxifene are more potent than tamoxifen in promoting ERα-dependent sequestration of progesterone receptor-associated corepressors. Moreover, ICI and raloxifene are more efficient than tamoxifen in promoting ERα binding to the corepressor N-CoR in vivo and in vitro. An ERα mutation (537X) that increases N-CoR binding in the presence of all SERMs blocks AF-1 activity. An ERα mutation (L379R) that decreases N-CoR binding increases AF-1 activity in the presence of ICI and raloxifene and reverses the effect of the 537X mutation. The 537X and L379R mutations also alter the ligand preference of ERα action at AP-1 sites and C3 complement, an action that also involves AF-1. Together, our results suggest that differential SERM effects on corepressor binding can explain differences in SERM effects on ERα activity. We propose a model for differential effects of SERMs on N-CoR binding.

by recruiting a large coactivator complex composed of p160 coactivators such as GRIP1 and SRC-1 and the histone acetyltransferases p300/CREB-binding protein and pCAF (4). Like other nuclear receptors, the ERs are comprised of an N-terminal domain (NTD), a central DNA-binding domain (DBD), and a C-terminal ligand-binding domain (LBD). The DBD mediates ERE recognition, and the NTD and LBD contain distinct activation functions (AF-1 and AF-2, respectively) that mediate coactivator recruitment. The ERs also modulate expression of genes with alternate estrogen response elements, such as AP-1 sites and SP-1 sites (5)(6)(7). ER␣ acts at AP-1 sites via protein-protein interactions and recruits p160s in a process that requires ER␣ AF-1 and AF-2 and their cognate binding sites within the p160s (6, 8 -10). Thus, estrogen action at classical EREs and alternate response elements involve similar coactivators.
Selective estrogen receptor modulators (SERMs) are used in treatment and prevention of estrogen-dependent breast cancer (11). The SERMs inhibit ER action by blocking AF-2 activity, and the mechanism of this effect is now understood at the atomic level (2). AF-2 is composed of surface-exposed residues from LBD helices (H) 3, 5, and 12, which form a hydrophobic cleft that binds short nuclear receptor (NR) boxes (consensus Leu-X-X-Leu-Leu or LXXLL) found in each p160 (12,13). Tamoxifen and raloxifene both protrude through the LBD surface and displace H12, which rotates 110°and utilizes a sequence that resembles the NR box (LLEML) to bind the upper part of the cleft (14 -17). The antagonist ICI 182,780 (ICI) also possesses a longer extension that protrudes through the LBD surface and displaces H12, but H12 is not resolved in this structure (18).
While SERMs act as antagonists in breast, they exhibit estrogen-like effects in other tissues. Tamoxifen shows beneficial agonist effects, including prevention of osteoporosis, and harmful agonist effects, including emergence of tamoxifenstimulated breast cancers and increased rates of uterine cancer (19). Raloxifene and GW5638 show agonist effects in bone and other tissues, but lack harmful uterotropic effects (20). ICI shows agonist effects in bone (21), but generally acts as a pure antiestrogen (22). Understanding why SERMs exhibit differential estrogen-like activities will help in identification of new SERMs with favorable profiles of activity.
Other estrogen-like effects of SERMs stem from ER action at genes with alternate response elements (5)(6)(7). ER␣ enhances AP-1 activity in the presence of tamoxifen but shows weaker activity in the presence of raloxifene or other SERMs (5,6,32,33). Moreover, these tamoxifen effects require the ER␣ NTD (6). Thus, SERMs differentially regulate some aspect of ER␣ AF-1 activity at AP-1 sites, just as they do at classical EREs. SERMs also enhance AP-1 activity via a second mechanism that is independent of ER activation functions and up-regulated in ER␣ truncations that lack AF-1 and in ER␤, which naturally lacks AF-1 (6,32,34). We have suggested that these effects may involve sequestration of corepressors that inhibit AP-1 activity (35), possibly of the N-CoR/SMRT family (4). ICI and raloxifene strongly enhance ER␣ action in this AF-independent pathway, whereas tamoxifen does not. Thus, SERMs differentially regulate two mechanisms of ER action at AP-1 sites.
Why do SERMs show differential effects on ER␣ activity? One possible mechanism involves effects upon ER␣ turnover. Tamoxifen increases ER␣ steady state levels, raloxifene and GW7604 reduce ER␣ levels, and ICI reduces ER␣ levels by more than 90% (31,36). These effects correlate well with effects on ER␣ AF-1 activity, but in vivo competition assays reveal that the amount of ER␣ that occupies the ERE is relatively unaffected by SERMs (31,37,38). Moreover, ICI and raloxifene act as ER␣ agonists at genes with AP-1 sites (6). Thus, raloxifene and ICI inhibition of ER␣ activity is not simply a consequence of elimination of functional ERs from the cell. Another possible mechanism involves differences in corepressor recruitment. SERMs promote ER␣ interactions with the corepressors N-CoR and SMRT (39 -42) and N-CoR recruitment to estrogenregulated promoters in vivo (10,43). Several lines of evidence indicate that corepressors inhibit the activity of the SERM-ER␣ complexes (reviewed in Ref. 4). Moreover, tamoxifen is less efficient than raloxifene at recruiting N-CoR to estrogenregulated genes with alternate response elements in Ishikawa uterine cells (10), and this effect may explain why tamoxifen behaves as an agonist in this context when raloxifene does not (10,26,32).
While SERMs differ in their ability to promote N-CoR recruitment at one type of gene, it is not clear why. SERM-ER␣ complexes interact relatively strongly with the corepressor complex in vivo (10,40,43,44) but, at best, only weakly bind to bacterially expressed N-CoR and SMRT in vitro (42,44). It also remains unclear whether SERMs directly influence ER␣ interactions with N-CoR itself, or whether these effects involve N-CoR-associated proteins. Finally, it is unclear whether differences in corepressor recruitment explain differences in SERM agonist activity in contexts besides ER action at genes with alternate response elements in uterine cells. Here, we examine SERM effects upon ER␣ interactions with N-CoR and ask whether these effects underlie differences in SERM agonist activity in a variety of contexts in vivo.
ER␣537X and DBD-LBD-537X were prepared by exchanging a restriction fragment from a 537X mutant LBD into full-length ER␣ and DBD-LBD (46). ER␣L379R, ER␣T371R, and DBD-LBD L379R were prepared by PCR point mutagenesis (Stratagene). GST-SMRT (amino acids 987-1491) was made by PCR amplification of SMRT cDNA (gift of Dr. R. Evans, University of California, San Diego) using oligonucleotides with EcoRI and SalI sites. The resulting fragment was digested and cloned into pGEX-5X-1.
Transfections-Cells were transfected by electroporation (47). Transfections contained 2 g of reporters and, where indicated, 1 g of ER␣ expression vectors and 5 g of coactivator expression vectors or control vectors. Luciferase and ␤-galactosidase activities were measured using luciferase (Promega, Madison, WI) and Galacto-Light assay systems (Tropix, Bedford, MA).
Bacterial Expression and Protein Binding Assays-GST fusions were expressed in Escherichia coli BL21 and purified as before (26). Briefly, cultures were grown to A 600 1.5 at room temperatures (ϳ22°C), and protein production was initiated by addition of isopropyl-1-thio-␤-Dgalactopyranoside to 1 mM. After 4 h, bacterial pellets were obtained, resuspended in 20 mM HEPES, pH 7.9/80 mM, KCl/6 mM, MgCl 2 /1 mM, dithiothreitol/1 mM, ATP/0.2 mM, phenylmethylsulfonyl fluoride, and protease inhibitors and sonicated. Debris was pelleted by centrifugation in an ss34 rotor for 1 h at 12,000 rpm. The supernatant was incubated with glutathione-Sepharose 4B beads (Amersham Biosciences) and washed as previously described. Protein preparations were stored at Ϫ20°C in 20% glycerol.
Interactions between [ 35 S]methionine-labeled ER␣ and bacterially expressed GST fusions were assessed (5). Labeled ERs were produced using coupled in vitro transcription-translation (TNT kit, Promega). Briefly, assays were carried out in a volume of 150 l that contained 137.5 l of ice-cold protein binding buffer (PBB) along with 10 l of GST-bead slurry corresponding to 3 g of fusion protein, 1 l of in vitro translated protein, and 1.5 l of ligand or vehicle. PBB was generally freshly prepared in 24-ml aliquots composed of 20 ml of A-150 (20 mM HEPES, 150 mM KCl, 10 mM MgCl 2 , and 1% glycerol), and 2 ml each of phosphate-buffered saline supplemented, respectively, with 1% Triton X-100 and 1% Nonidet P-40. Phenylmethylsulfonyl fluoride, dithiothreitol, bovine serum albumin, and protease inhibitor mixture (Novagen) were added to 0.1 mM, 1 mM, 2 g/ml, and 1/1000 dilution, respectively. The mix was incubated for 2 h in the cold room with gentle agitation, the beads were pelleted by spinning briefly on a bench top Eppendorf centrifuge and washed four times with PBB containing no bovine serum albumin, and the pellet was dried under vacuum for 20 min. Labeled ER was subjected to SDS-PAGE and autoradiography. The amount of bound protein in GST pulldown assays was quantified by scanning light exposures of gels into a Hewlett-Packard Scanjet ADF. Images were converted to PDF format and quantified using NIH image as previously described (8).
Immunoprecipitations-10-cm dishes of transfected HeLa cells were treated with ligands for 24 h, chilled on ice, washed with ice-cold PBS, incubated in lysis buffer (0.2% Triton X-100, 150 mM NaCl, 50 mM Tris pH 7.8, 1 mM EDTA, 1/1000 dilution Novagen protease inhibitor mixture) for 10 min, scraped, and centrifuged at 4°C in an Eppendorf bench top centrifuge at maximum speed for 15 min. Protein concentration in cell extract was determined by standard methods. The concentration of each sample was then normalized by addition of the appropriate volume of cold lysis buffer to ϳ5 mg of total protein in 1 ml of total volume. Each sample was precleared at 4°C for 2 h with 25 l of protein G plus agarose beads (Santa Cruz antibodies, Santa Cruz, CA) that were prewashed twice with lysis buffer. The beads were then pelleted by brief centrifugation, and the supernatant was mixed with 25 l of fresh prewashed protein G plus agarose beads and 5 g of GAL4 antibody RK5C1 (Santa Cruz antibodies). The mix was shaken overnight in the cold room. The following day, the beads were pelleted, washed three times in lysis buffer, dried, and resuspended in SDS-PAGE loading buffer.
Western Blots-For analysis of immunoprecipitation, protein from a standard volume of beads was used. For whole cell extracts, protein concentration was determined in cellular extract by standard methods, and 20 -30 g of proteins were separated by SDS-PAGE. Proteins were transferred to a prewetted Immuno-Blot polyvinylidene difluoride membrane (Bio-Rad, Hercules CA) overnight at 90 mA/30 V using a standard transfer apparatus. Following transfer, the membrane was incubated at room temperature in 5% non-fat milk in PBS-T (1ϫ PBS, 0.1% Tween 20) for 1 h and washed twice in PBS-T for 10 min. The primary ER␣ antibodies that were used in this study were HC-20 (Santa Cruz antibodies) directed against the ER␣ C terminus, or number 1600024 (Geneka, Montreal, Canada) directed against the ER␣ NTD. The primary N-CoR antibody was N-19 (Santa Cruz antibodies). Primary antibody was diluted 1:2000 in PBS-T and incubated with the membrane for 1 h, followed by PBS-T washes, 1 ϫ 15 min and then 2 ϫ 5 min. The membrane was incubated for 45 min with horseradish peroxidase-conjugated anti-rabbit or anti-goat IgG (Santa Cruz antibodies) diluted 1:2000 in PBS-T, followed by PBS-T washes (1 ϫ 15 min and 4 ϫ 5 min). After the last wash, the membrane was developed with a standard ECL kit (Amersham Biosciences), covered with Saran wrap, and exposed to x-ray film. Westerns were quantified as described for GST pulldown assays.

A HDAC Inhibitor Eliminates Differential SERM AF-1 Inhibition at an ERE-responsive Promoter in MCF-7 Cells-
The nuclear receptor corepressor complex inhibits transcription by recruiting histone deacetylases, or HDACs (4). To begin this study, we determined whether inhibition of HDACs would affect SERM inhibition of ER␣ action. We chose tamoxifen, raloxifene, and ICI as representative SERMs that permit, partially inhibit, and completely inhibit AF-1 activity, respectively. Fig. 1A shows that, as often noted, an ERE-responsive reporter (ERE-TK-LUC) exhibited weak constitutive activity in MCF-7 breast tumor cells and that this activity was enhanced by estrogens and reduced by SERMs, especially by ICI. A similar reporter (TK-LUC), which lacked an ERE, was not affected by ligand. Addition of HDAC inhibitor (trichostatin A, TSA) enhanced basal activity of both reporters. However, TSA also specifically blocked the ability of SERMs to inhibit EREdependent transcription, with the result that fairly equivalent levels of ERE-dependent transcription were obtained in the absence of hormone and in the presence of each ligand. Similar results were also obtained in T47D breast cancer cells and GHFT1-5 pituitary cells (not shown).
To highlight TSA effects upon ER␣ ligand preference, we calculated fold TSA induction in the presence of each ligand (Fig. 1B). 6-fold TSA induction was obtained in the absence of ligand, 22-fold induction was obtained with ICI, which does not allow AF-1 activity, 12-fold with raloxifene, which allows intermediate AF-1 activity, and 7.5-fold with tamoxifen, which allows the most AF-1 activity. These differential TSA effects were not seen at the core TK promoter, and TSA did not alter the overall profile of SERM effects on ER␣ steady state levels (Fig.  1C). Thus, TSA enhances ERE-dependent transcription more potently in the presence of SERMs. This suggests that SERMs increase TSA-sensitive HDAC-dependent inhibition at the ERE-responsive promoter but not at the parental TK core promoter. Moreover, the fact that TSA induction is larger in the presence of ICI and raloxifene relative to tamoxifen shows that the magnitude of this HDAC-dependent effect varies with individual SERMs.
SERMs Show Differential Ability to Sequester Corepressors from Progesterone Receptor-Next, we determined whether SERMs would differentially influence ER␣ interactions with the corepressor complex in vivo. To perform this experiment, we investigated whether SERMs would enhance the ability of ER␣ to potentiate the activity of DNA-bound antiprogestin-PR complexes using the system developed by Bagchi and coworkers (41). This assay measures the ability of ER␣ to sequester unspecified N-CoR containing corepressor complexes and therefore gives an indication of overall ER␣ interactions with shared corepressor complexes rather than a direct indication of ER␣ interactions with particular corepressor complex components.
In agreement with previous results, ER␣ potentiated the activity of a PR-RU486 complex bound to a progestin-responsive reporter gene (TAT 3 -LUC) in the presence of tamoxifen but not estradiol (Fig. 2). However, ER␣ also potentiated PR activity more potently in the presence of raloxifene and most strongly with ICI. We conclude that ICI and raloxifene are more efficient at promoting sequestration of corepressors from the PR-RU486 complex in vivo.
Differential Ability of SERMs to Promote ER␣ Interactions with N-CoR-N-CoR consists of separable N-terminal repression domains that interact with HDACs and other components of the corepressor complex and a C-terminal domain that contains the nuclear receptor-interacting region (4). SERMs pro- mote ER␣ interactions with N-CoR in vivo (10,39,40,42,43). We next determined whether SERMs would promote ER␣ interactions with the N-CoR C-terminal nuclear receptor interacting domain in vivo. Transiently expressed ER␣ coprecipitated with the N-CoR C terminus in the absence of ligand and in presence of tamoxifen and raloxifene (IP-galϩW ER␣, Fig.  3A). A lower, but detectable, level of ER␣ coprecipitated in the presence of ICI. This pattern is similar to that observed between ER␣ and full-length N-CoR in other studies ((40, 42) and also our data not shown). Thus, ER␣ binds to the N-CoR C terminus in vivo.
As expected (31), Western blotting of cell extracts prior to immunoprecipitation confirmed that ICI and raloxifene reduced steady state ER␣ levels and that tamoxifen increased steady state ER␣ levels (W ER␣). Comparison of the intensity of bands corresponding to ER␣ in both blots (Fig. 3B) indicates that tamoxifen generally increased the amount of ER␣ in the N-CoR-enriched fraction by about 2-fold (although there was significant variation between samples), whereas raloxifene increased the amount of ER␣ in the N-CoR-enriched fraction about 3-fold. More surprisingly, ICI increased the amount of ER␣ in the N-CoR-enriched fraction by at least 7-fold relative to the low amount of ICI-liganded ER␣ in transfected cells.
To further investigate the ligand preference of ER␣ interactions with N-CoR, we examined ER␣ interactions with bacterially expressed N-CoR C terminus in vitro. SERMs, but not estradiol, enhanced ER␣ interactions with the N-CoR C terminus (Fig. 4A). A similar (albeit weaker) binding profile was obtained with the SMRT C terminus. Control binding assays confirmed that ER␣ interactions with the NR box region of the coactivator GRIP1 were reduced by SERMs and promoted by estradiol. Quantification of a series of similar pulldown assays (n ϭ 15, Fig. 4B) indicated that individual SERMs exhibited differences in their ability to promote N-CoR binding in vitro. ICI gave the best binding (about 3.5% of input); raloxifene gave intermediate levels (2.2% of input); and tamoxifen gave the least binding (1% of input). Thus, SERMs differ in their ability to promote ER␣ interactions with N-CoR, and the ligand preference of these effects correlates qualitatively with SERM effects on repression of AF-1 activity (see the introduction), HDAC inhibition of an ERE-responsive promoter in MCF-7 cells (Fig. 1B), and sequestration of corepressors from the PR-RU486 complex (Fig. 2).
A Mutation (ER␣537X) That Increases N-CoR Binding also Inhibits AF-1 Activity-While SERMs show intriguing effects on ER␣ interactions with N-CoR, it is also clear that these interactions are relatively weak (see "Discussion"). To further investigate the significance of these interactions, we sought ER␣ mutations that would alter the pattern of ER␣ binding to N-CoR and then investigated the properties of these ER␣ mutants in transfection assays in vivo.
It is known that the nuclear receptor corepressor-binding site lies in the LBD hydrophobic cleft and that corepressor binding is enhanced by truncation of H12 (see for example Ref. 48). We first confirmed that truncation of ER␣ H12 would enhance N-CoR binding (Fig. 5A). An ER␣ mutant that lacked H12 (ER␣537X) showed enhanced N-CoR binding in the absence of ligand and in the presence of tamoxifen and estradiol (see Fig. 5B for quantification). ER␣537X did not show appreciably increased N-CoR binding in the presence of ICI, suggesting that H12 does not influence ER␣ binding to N-CoR in the presence of ICI. Surprisingly, ER␣537X did exhibit increased binding in the presence of raloxifene (8% input). This suggests that raloxifene exerts unspecified effects on core-LBD structure that promote N-CoR binding.
The effect of the H12 truncation on SERM ligand preference was also apparent in the context of the isolated ER␣ LBD (LBD-537X). Nevertheless, the isolated ER␣ LBD failed to bind  N-CoR suggesting that other regions of ER␣ contribute to N-CoR binding. The ER␣ DBD-LBD region did bind N-CoR with a ligand preference that resembled full-length ER␣, but an ER␣ deletion mutant that lacked the hinge domain (ER⌬hinge) and the isolated AB-DBD and AB regions did not bind N-CoR. Thus, ER␣ interactions with N-CoR require the hydrophobic cleft and the DBD-hinge.
Because the 537X mutation allowed N-CoR binding with all SERMs in vitro, we examined the phenotype of ER␣537X in vivo. As expected (24), wild type ER␣ enhanced transcription of an ERE-responsive reporter (ERE-TK-LUC) in the absence of ligand in MDA-MB-453 breast cells, and estradiol gave further stimulation (Fig. 6A). The SERMs tamoxifen and raloxifene allowed some residual activity that stems from AF-1 (24), but ICI allowed none. ER␣G400V, which exhibits reduced constitutive activity (49,50), showed similar activity to ER␣ in the presence of each ligand. In parallel, ER␣537X showed three differences from wild type ER␣. First, ER␣537X showed reduced estrogen response. This is consistent with the requirement for H12 in p160 binding (15). Second, ER␣537X showed reduced constitutive activity. This phenotype is common in ER␣ LBD mutants (24). Third, and most importantly, ER␣537X failed to elicit a tamoxifen response. This lack of response persisted even in the presence of a 10-fold excess of ER␣537X expression vector (not shown). Fig. 6B shows that ER␣537X was expressed at comparable levels to wild type ER␣ in the presence of SERMs. Thus, the lack of tamoxifen response in the presence of ER␣537X is not a consequence of a large reduction in ER␣ levels. Moreover, our analysis of the amount of ER␣ and ER␣537X transfection vectors that are required to elicit responses from the ERE-responsive reporter indicate that ER␣537X has not become superactive at low levels of transfected receptor and is not squelching its own activity under these conditions (data not shown). We conclude that an ER␣ mutation (537X) that enhances corepressor binding also eliminates AF-1 activity.
An ER␣ Mutation That Reduces N-CoR Binding Allows Equal AF-1 Activity with SERMs-The nuclear receptor corepressor-binding surface consists of residues that overlap the hydrophobic cleft that comprises part of the AF-2 surface (48,(51)(52)(53)(54). In particular, a Leu residue at the base of H5 is important for corepressor binding (51). We therefore next determined whether mutation of the equivalent ER␣ residue (Leu-379) would influence SERM effects on corepressor binding. Fig. 7 shows that N-CoR binding was reduced in all conditions in the presence of ER␣L379R but that these effects were especially prominent in the presence of ICI and raloxifene (about 0.5% of ICI-liganded ER␣L379R bound N-CoR relative to 4% of ICIliganded wild type ER␣; about 0.2% of raloxifene-liganded ER␣L379R bound to N-CoR relative to more than 2% of wild type ER␣). ER␣L379R also reduced GRIP1 binding in the presence of estradiol (Ref. 15 and not shown). Mutations in nearby residues (K362A, T371R, V376R, and E542K) did not show large effect effects on the overall ligand preference of N-CoR binding, even though some (K362A, V376R, and E542K) did reduce GRIP1 binding (Ref. 12 and not shown). Thus, Leu-379 is important for SERM-dependent ER␣ interactions with N-CoR in vitro, and these interactions are deficient in the ER␣L379R mutant, but not in other mutations in the AF-2 surface.
We then tested ER␣L379R in vivo. Like other ER␣ AF-2 mutants, ER␣L379R showed reduced estrogen response and constitutive activity (Fig. 8A). More importantly, ER␣L379R showed increased raloxifene and ICI response. Other ER␣ AF-2 mutants (K362A, V376R, and E542K) showed reduced estrogen response and constitutive activity or comparable estrogen response and constitutive activity to wild type ER␣ (T371R), but all retained the same SERM ligand preference as wild type ER␣. Fig. 8B shows that ER␣L379R showed a modest increase in steady state levels in the presence of ICI and decreased expression levels in the presence of raloxifene and tamoxifen. However, modest alterations in ER␣ levels were also observed with other mutants (especially E542K), and these alterations did not translate into alterations in SERM activity in vivo. Thus, ER␣L379R activity is not simply related to ER␣ steady state levels. Truncation of the ER␣L379R NTD (DBD-LBDL379R) abolished all ER␣L379R-dependent SERM agonist effects, confirming that they require AF-1 (not shown). Thus, the L379R mutation increases AF-1 activity at an ERE in the presence of ICI and raloxifene.
We then examined whether the L379R mutation would affect the behavior of the 537X mutation, which, as shown above, increases N-CoR binding and reduces AF-1 activity. Fig. 8C shows that, in contrast to the behavior of ER␣537X, an ER␣L379R/537X double mutant did enhance ERE-dependent transcription in the presence of SERMs, even though it was poorly expressed (see inset). Thus, the L379R mutation reverses the phenotype of ER␣537X.
Finally, ER␣ AF-1 activity is usually low in HeLa cells (23), but can be enhanced by the AF-1 target coactivator proteins, GRIP1 and p300 (6,26). To determine whether alteration of N-CoR binding would affect the AF-1 activity that is mediated by this coactivator complex, we examined effects of overexpression of AF-1 target coactivators upon ER␣L379R. As expected (26), ER␣ and ER␣G400V enhanced transcription with tamoxifen but not with raloxifene or ICI in HeLa cells and did so only when GRIP1 and p300 were supplied (Fig. 9, compare left and right panels). In the same conditions, ER␣L379R enhanced transcription with tamoxifen (albeit not quite as strongly as wild type ER␣) but also enhanced transcription with raloxifene and ICI. By contrast, ER␣K362A enhanced transcription only with tamoxifen. Again, ER␣L379R levels were slightly increased in the presence of ICI, but similar alterations in ER␣ levels were also obtained with ER␣K362A, which did not show altered SERM ligand preference. Thus, the L379R mutation allows increased ICI and raloxifene agonist activity in conditions in which AF-1 activity is mediated by the p160/p300 complex. Together, our results indicate that the L379R mutation permits AF-1 activity in the presence of a range of SERMs. We conclude that the amount of ER␣ binding to N-CoR (whether influenced by SERMs or mutations) correlates with the amount of ER␣ AF-1 activity at simple EREs.
Mutations That Affect N-CoR Binding Have Parallel Effects on ER␣ Action at AP-1 Sites-ER␣ enhances AP-1 activity in the presence of estradiol and the partial agonist tamoxifen (5,32), and these effects require ER␣ activation functions and p160s (6,10). However, ER␣ also enhances AP-1 activity via a second AF-independent mechanism that is up-regulated in ER␣ deletions that lack AF-1 and strongly potentiated by ICI and raloxifene (6,32,34) and may involve corepressors (6,35). Thus, a mutation that reduces N-CoR binding should potentiate ER␣ AF-1 activity at AP-1 sites but should also inhibit the activity of ER␣ truncations that are committed to the AFindependent pathway. Moreover, mutations that enhance N-CoR binding may show increased ability to potentiate ER␣ activity in the AF-independent pathway.
To test these ideas, we examined the effect of the L379R mutation upon ER␣ action at AP-1 sites. Fig. 10A confirms that ER␣ enhanced the activity of an AP-1-responsive promoter (Coll73-LUC) in the absence of ligand and in the presence of tamoxifen, estradiol, and DES in HeLa cells and that ER␣G400V showed a similar profile. The same data show that ER␣L379R exhibited equivalent activity at the AP-1-responsive reporter in the absence of ligand and in the presence of each SERM. The L379R mutation also enhanced estradiol and DES response at the AP-1 site. This is consistent with the suggestion that the corepressor complex restricts estrogen-liganded ER␣ activity in some contexts (44). Fig. 10B confirms that truncation of the ER␣L379R NTD (DBD-LBDL379R) completely abolished all SERM effects. Thus, increased SERM activities that are observed with ER␣L379R require AF-1 just as they do at EREs. The same data confirm that the DBD-LBD enhances AP-1 activity in the presence of ICI and raloxifene but not tamoxifen or estradiol (6,24) and that the L379R mutation abolishes these responses. Thus, the ICI and raloxifene response through the AF-independent pathway requires the N-CoR binding surface.
We then examined the effect of the 537X mutation, which increases and equalizes N-CoR binding in the presence of SERMs. Like ER␣L379R, ER␣537X showed equivalent activity at the AP-1-responsive reporter in the absence of ligand and in the presence of the three SERMs (Fig. 10A). However, unlike ER␣L379R, truncation of the NTD did not abolish these SERM effects (Fig. 10B). Instead, ICI and raloxifene responses were preserved and tamoxifen and estradiol responses were increased. Thus, the 537X mutation equalizes the activity of ligands in the AF-independent pathway. We conclude that mutations that affect N-CoR binding alter the ligand preference of ER␣ action at AP-1 sites through two mechanisms An Exception to Parallels between N-CoR Binding and AF-1 Activity, the C3 Complement-ER␣ shows potent AF-1 activity at the C3 complement promoter (C3) (25,30). Fig. 11 confirms that ER␣ and ER␣G400V enhanced the C3 activity in the presence of estradiol and tamoxifen but not ICI or raloxifene in MDA-MB-453 cells. ER␣537X failed to enhance C3 in the presence of any ligand, just as it did at a simple ERE. Moreover, ER␣L379R showed enhanced ICI and raloxifene responses relative to wild type ER␣, just as it did at a simple ERE and the AP-1 site, and enhanced estradiol and DES response, just as did at the AP-1 site. ICI and raloxifene effects were dependent upon the presence of the NTD (not shown). Thus, a mutation (L379R) that reduces N-CoR binding enhances AF-1 activity at C3 in the presence of ICI and raloxifene.
Unlike its activity at EREs and the AP-1 site, however, the L379R mutation did not preserve any tamoxifen response at C3 in MDA-MB-453 cells (shown here) and other cell types (includ- ing HeLa and HepG2; data not shown). This finding, though not expected, is in line with previous observations that indicate that ER␣ tamoxifen activation of C3 involves an unspecified contribution from the LBD (25,30). It is possible that this contribution is deficient in the L379R mutant. We conclude that N-CoR inhibits ICI and raloxifene activity at C3 but that low levels of N-CoR binding in the presence of tamoxifen do not fully explain why tamoxifen is a potent activator in this context. DISCUSSION Differential Corepressor recruitment underlies differential SERM effects at genes with alternate response elements in uterine cells (10). However, the structural basis for these differential SERM effects on corepressor recruitment is unclear, and it is controversial whether ER␣ shows meaningful direct interactions with N-CoR. It is also unclear whether similar mechanisms underlie differences in SERM activity in other contexts.
In this study, we showed that ICI and raloxifene promote HDAC repression at an ERE-responsive reporter in MCF-7 cells and ER␣-dependent sequestration of PR-associated corepressor complexes, and that they do so more potently than tamoxifen. We also found that SERMs promote ER␣ interactions with the nuclear receptor interaction domain of N-CoR (C terminus) in immunoprecipitations, and estimates of ER␣ levels in the N-CoR-enriched fraction versus the cell extract suggested that the ICI and raloxifene are better than tamoxifen in promoting this interaction. Likewise, ICI and raloxifene are more potent than tamoxifen in promoting ER␣ association with the N-CoR C terminus in GST pulldown assays in vitro. Thus, SERMs show differential effects on ER␣ interactions with N-CoR and do so with a ligand preference that parallels their effects on suppression of AF-1.
While SERMs promote N-CoR binding in vitro, the interac- tions are weak compared with other nuclear receptors. Between 1-3% of input ER␣ was retained on the GST-N-CoR beads, whereas 20 -40% of input TR was retained on the GST-N-CoR beads in parallel (data not shown, see Ref. 45). Moreover, peptides that correspond to isolated N-CoR nuclear receptorinteracting domains 1 and 3 (45) failed to compete for ER␣/N-CoR interactions (not shown). We can infer that the affinity of ER␣ for these isolated peptides is very low (greater than 30 M compared with 1 M affinity of LXXLL peptides for ER␣ AF-2, discussed in Ref. 55). Finally, we were unable to detect ER␣ LBD interactions with N-CoR in mammalian two-hybrid assays (not shown). To evaluate the significance of interactions between ER␣ and N-CoR, we therefore identified ER␣ mutations that influence N-CoR binding in vitro and examined their function in vivo. We found that a mutation (537X) that increases and equalizes N-CoR binding in the presence of SERMs eliminates AF-1 activity at simple EREs and the C3 promoter and equalizes SERM activation of AP-1 sites via the AF-independent mechanism. A mutation (L379R) that reduces and equalizes N-CoR binding with SERMs allows equivalent AF-1 activity at simple EREs and AP-1 sites in the presence of SERMs and increases ICI and raloxifene response at C3. The same mutant also reverses the effect of the 537X mutation and eliminates ER␣ action through the AFindependent pathway at AP-1 sites. Thus, mutations that affect N-CoR binding have parallel effects on ER␣ activity. We therefore propose that differential effects on direct corepressor interactions underlie many instances of differential SERM activities in vivo. This idea is in line with observations that SERMs enhance ER␣ interactions with peptides corresponding to corepressor nuclear receptor-interacting domains in vivo (43,44) and that mutations that influence ER␣ interactions with N-CoR or N-CoR-like peptides alter the ligand preference of ER␣ action in vivo (42,44).
How can we reconcile the paradox that SERMs promote interactions with N-CoR in vivo, yet ER␣ only binds weakly to N-CoR in vitro? We suggest that ER␣ interactions with the N-CoR C terminus correspond to the ligand-dependent component of ER␣/ corepressor interactions, but that ER␣ also contacts other surfaces of the corepressor complex that augment corepressor recruitment. We also recognize that our binding assays may not faithfully recreate conditions that are required for maximal ER␣/ N-CoR interactions (perhaps ER␣ or N-CoR need specific modifications or cofactor interactions) and that our studies do not address the roles of N-CoR versus SMRT versus other repressors and certainly do not exclude the possibility that there are multiple repressors of ER␣ action. It will be interesting to ask whether ER␣ uses a similar surface to recruit other repressors, such as REA and HET/SAFb (56,57).
The diverse effects of SERMs on ER␣ interactions with N-CoR contribute to an emerging picture in which nuclear receptor antagonists exert diverse effects upon corepressor binding. The PPAR antagonist GW6471 promotes corepressor binding (58), whereas the TR antagonist NH-3 promotes corepressor release (59). PR antagonists show differential effects on corepressor binding (60). ER␣ residues that mediate corepressor binding (Leu-379 and others (42,44,48)) lie within the hydrophobic cleft. Estrogens permit docking of H12 against the lower part of the cleft and FIG. 12. Model to explain differential SERM effects upon N-CoR binding. The ER␣ coactivator/corepressor-binding surface is shown in schematic, with H3 shown in white, H5 in dark gray, and H12 in pale gray. N-CoR binding residues are shown as circles. Three possible liganded conformations of the ER␣-LBD are presented, and their effects on N-CoR binding are discussed in the text. I, estradiol promotes formation of the AF-2 hydrophobic cleft (composed of H3, H5, and H12) and blocks N-CoR binding. ii, tamoxifen or raloxifene protrudes through the LBD surface, thereby displacing H12, which folds into the cleft and blocks N-CoR binding. iii, SERMs (especially raloxifene and ICI) also displace H12 but here, H12 occupies an unusual position that exposes the cleft and allows N-CoR to bind.

FIG. 10. Mutations in ER␣ that affect N-CoR binding affect SERM activity at an AP-1-responsive reporter.
A, HeLa transfection in which ER␣ activity was compared with ER␣L379R and ER␣537X at an AP-1-responsive reporter (Coll73-Luc). B, as in A except that the DBD-LBD truncation, missing AF-1, was used.
FIG. 11. The L379R mutation allows some AF-1 activity in the presence of ICI and raloxifene at the complement C3 promoter but decreases tamoxifen activation. Transfections were performed as in Fig. 5A except that the C3 complement promoter was used.
should prevent corepressor binding (Fig. 12, i). By contrast, SERMs displace H12. H12 can fold over the cleft as in published ER-SERM crystal structures (Fig. 12, ii) (14,15,17,18). The results described here suggest that this conformation occludes the H3-H5 region that mediates corepressor binding. SERMs must therefore allow H12 to dock in another position that exposes the H3-H5 region and promotes N-CoR binding (Fig. 12,  iii). Perhaps this configuration resembles the PPAR-SMRT structure (58) or the ER␤-ICI structure (18). We therefore suggest that overall levels on N-CoR binding depend on equilibrium between H12 positions (ii and iii); tamoxifen would favor ii, ICI would favor iii, and raloxifene would permit both. Factors that might influence this equilibrium include SERM effects upon the length of the H11-H12 loop, which affect the ease of H12 docking in the cleft (14,15), contacts between the SERM extension and either ER␣ or N-CoR, or unspecified effects of SERMs on core-LBD structure (see Fig. 5).
To what extent does differential corepressor binding account for SERM activity in vivo? The L379R mutation eliminates tamoxifen response at C3 (Fig. 11) and reduces tamoxifen response at EREs and AP-1 sites in HeLa cells (Figs. 9, 10). Thus, some tamoxifen responses require a contribution from the LBD that is unrelated to corepressor binding. Moreover, while SERM activity is not absolutely related to ER␣ levels over the short time course of our experiments (Figs. 1C, 6B, 8B, and 9B), it is likely that SERM-dependent alterations in ER␣ levels play a more important role over longer times. Nevertheless, differential SERM effects on corepressor binding contribute to differential SERM activity in a variety of contexts. It will be especially important to understand corepressor recruitment to genes with alternate response elements, which are important for ER effects on proliferation (5,10,61,62).