Mutations in the Estrogen Receptor Ligand Binding Domain Discriminate between Hormone-dependent Transactivation and Transrepression*

The estrogen receptor (ER) suppresses transcriptional activity of the RelA subunit of nuclear factor-κB in a hormone-dependent manner by a mechanism involving both the receptor DNA binding domain and ligand binding domain (LBD). In this study we examine the role of the ER LBD in mediating ligand-dependent RelA transrepression. Both ERα and ERβ inhibit RelA in response to 17β-estradiol but not in the presence of antihormones. We have identified residues within the ERα LBD that are responsible for receptor dimerization and show that dimerization is necessary for transactivation and transrepression. Moreover we have generated mutant receptors that have lost their ability to inhibit RelA but retain their capacity to stimulate transcription and conversely mutants that are transcriptionally defective but capable of antagonizing RelA. Overexpression of p160 and cAMP-response element-binding protein-binding protein/p300 co-activators failed to relieve repression of RelA, which is consistent with the demonstration that RelA inhibition can occur independently of these co-activators. These findings suggest it is unlikely that sequestration of these cofactors required for ER transcriptional activation can account for hormone-dependent antagonism of RelA. The identification of ER mutants that discriminate between transactivation and transrepression implies that distinct surfaces within the LBD are involved in mediating these two receptor functions.

The pleiotropic effects of estrogens are mediated by estrogen receptors (ERs), 1 which function as hormone-activated transcription factors regulating the expression of a variety of estrogen-responsive genes. In common with other members of the nuclear receptor (NR) superfamily, ER contains three functional domains, an N-terminal region with a hormone-independent activation function (AF1), a conserved central DNA binding domain (DBD), and a C-terminal ligand binding domain (LBD), which is responsible for high affinity ligand binding, dimerization, and hormone-dependent activation (AF2) (1)(2)(3). There are two forms of ER, ER␣, and ER␤, which share a high degree of homology in their DBD and LBD but contain divergent N-terminal domains (4,5). In response to hormone binding, ER homodimers contact estrogen response elements (ERE) located within the regulatory sequences of target genes, resulting in the recruitment of co-activator proteins and consequent initiation of gene transcription (6,7). Elucidation of the crystal structure of ER␣ and ER␤ LBD has revealed that they share a common modular structure with the LBD of related NRs, which can be divided into 12 discrete helices (see Refs. 8 -11, and the references therein). Comparison of these unliganded and liganded crystal structures suggests that upon ligand binding the LBD undergoes a conformational change whereby the C-terminal helix (H12 in ER) is realigned over the ligand binding pocket and in ER is packed against H3, H5/6, and H11 (8 -11). The activity of ER␣ AF2 is dependent on the integrity of a hydrophobic interaction surface generated by conserved amino acids in H3, H4/5, and H12 (12)(13)(14).
In addition to stimulating transcription by directly binding cognate response elements within the promoters of responsive genes, NRs are able to regulate gene expression in the absence of DNA binding by modulating the activity of other transcription factors. A well documented example of cross-talk between transcription factors is the mutual antagonism between AP1 and NRs including glucocorticoid receptor (GR) (15)(16)(17), retinoic acid receptor, and retinoic X receptor (18 -20) and thyroid hormone receptor (21). Functional synergism has been documented between ER and AP1 (22) and ER and SP1 (23), and inhibition of GATA-1 by ER is reported to contribute to the suppressive effect of estrogen on erythroid differentiation (24). Estrogen plays an essential role in bone homeostasis by down-regulating interleukin-6 production in osteoblasts and marrow stromal cells (25). Analysis of the interleukin-6 proximal promoter revealed that the estrogen-dependent reduction in interleukin-6 transcriptional activity occurred in the absence of ER DNA binding (26,27) and was mediated through interleukin-6 transcriptional activators NF-B and to a lesser extent CCAAT enhancerbinding protein (27,28).
NF-B is a key inducible regulator of multiple genes involved in proliferation and immune functions. In its latent state, NF-B is sequestered in the cytoplasm through association with inhibitory IB proteins. Activation of a variety of signaling pathways leads to the phosphorylation and degradation of IB, resulting in the release of NF-B and subsequent translocation to the nucleus where it binds to DNA recognition elements and stimulates gene transcription (29,30). Proteins * This research was supported by grants from the International Association for Cancer Research (to J. E. V.) and the Netherlands Organization for Scientific Research (NWO) (to E. K.). 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  comprising the NF-B family share a conserved N-terminal Rel homology domain, which contains the DNA binding domain, dimerization interface, and nuclear localization signals that are masked when this region interacts with IB. Transcriptionally active NF-B is composed of homo-or heterodimers of various family members, most typically heterodimers of RelA (p65), which contains two C-terminal transcription activation domains, and NF-B1 (p50) (29,30). In addition to ER-mediated repression of NF-B (27,28), hormone-dependent inhibition of RelA transcriptional activity has been described for GR (31)(32)(33), progesterone receptor (34), androgen receptor (AR) (35), and the mineralocortocoid receptor (36). In common with AP1/NR crosstalk, NF-B and hormone receptors are mutually antagonistic as RelA is able to reciprocally inhibit receptor activity (27,28,(33)(34)(35). However, in contrast to AP1, RelA transrepression is restricted to the steroid hormone receptor subgroup of NRs, with thyroid hormone receptor and retinoic acid receptor exhibiting no demonstrable effect on RelA transactivation (32).
A number of models have been proposed to account for the mechanism of steroid hormone receptor-mediated NF-B transrepression including (a) up-regulation of IB expression resulting in sequestration of NF-B in the cytoplasm (33,37), (b) formation of a receptor⅐NF-B complex leading either to inhibition of NF-B DNA binding (33,38,39) or to transcriptional interference of DNA-bound NF-B (28,32,34), or (c) competition for a limiting common transcriptional mediator (40 -42). Functional studies in transiently transfected cells using receptor deletion mutants have indicated that both the DBD and LBD are required for ligand-dependent inhibition of RelA (28,29,32,34,38). In particular the integrity of key conserved amino acids within the C-terminal zinc finger of steroid hormone receptor DBDs is essential for GR transrepression of RelA (36). In this study we have examined the role of specific residues within the ER␣ LBD in mediating ligand-dependent inhibition of RelA activity. Based on the recently determined structural models of NR LBDs, we have identified residues that are essential for receptor dimerization and demonstrate that dimerization is required for both receptor transcriptional activity and antagonism. Using point mutagenesis of the LBD we have identified transcriptionally inactive ER␣ mutants that retain their ability to suppress RelA and conversely mutants that activate transcription but have lost their capacity to inhibit RelA. Discrimination between the transactivation and transrepression properties of the receptor suggests that different surfaces in the LBD are involved in mediating these two ligand-dependent functions and moreover that the p160/CBP cofactors required for ER-induced gene transcription are unlikely to be involved in functional antagonism of NF-B.
Transient Transfection Experiments-HeLa and COS-1 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum (Life Technologies, Inc.). For transient transfection assays, cells were plated into 24-well plates in phenol red-free Dulbecco's modified Eagle's medium containing 5% dextran-charcoal-stripped fetal calf serum. Cells were transfected using a modified calcium phosphate co-precipitation method (49) with either 1 g of ICAM-tk-Luc reporter plasmid, 150 ng of PDMLacZ internal control plasmid, 10 ng of pSG5-RelA, and 250 ng of estrogen receptor expression plasmids or with 1 g of either 2XERE-pS2-CAT or 2XERE-pS2-pGL3 reporter, 150 ng of PDMLacZ, and 10 ng of estrogen receptor expression plasmids. Where appropriate, empty expression vectors were added to equalize the amount of DNA present in each well, and where indicated 100 ng of co-activators SRC1, TIF2, CBP, and p300 were cotransfected with reporter and expression plasmids. After 16 h the cells were washed, fresh medium containing ethanol vehicle, 10 Ϫ8 M 17␤-estradiol (E2), 10 Ϫ7 M 4-hydroxytamoxifen (OHT), or 10 Ϫ7 M ICI 182,780 (ICI) was added as shown, and the cells were incubated for a further 24 h. Subsequently cells were washed in phosphate-buffered saline and harvested in lysis buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.65% Nonidet P-40), and the extracts were assayed for luciferase (12) and CAT (50) activities. Reporter activities were then standardized against ␤-galactosidase activity of the internal control vector PDMLacZ measured by the Galacto-Light chemiluminescent assay (Tropix). For Western analysis, wild-type and mutant receptors were overexpressed in COS-1 cells, whole cell extracts were prepared, and protein content was determined as described previously (13).
Gel Retardation Assay-In vitro translated ER␣ wild-type and mutant receptors were synthesized using the TNT-coupled in vitro translation system (Promega). Proteins were incubated with a 32 P-labeled double-stranded oligonucleotide probe containing a consensus ERE from the vitellogenin A2 gene promoter in the presence of preimmune serum or ER␣-specific MP16 antibody as described previously (51). Receptor-DNA complexes were resolved from unbound DNA on nondenaturing 7% polyacrylamide gels and were visualized by autoradiography.
Western Blotting-Whole cell extracts were prepared from COS-1 cells transfected with wild-type or mutant ER␣ expression vectors. Samples containing 5 g of total protein were separated by SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes by electroblotting. The membranes were blocked in phosphate-buffered saline containing 3% nonfat milk powder, washed in phosphate-buffered saline with 0.05% Tween 20, and incubated for 1 h with either the MP16 polyclonal antibody raised against the 130 -142 region of mouse ER␣ (51) or the H222 monoclonal antibody raised in rat against human ER␣. Following washing, the membranes were incubated for 30 min with horseradish peroxidase-coupled anti-rabbit or anti-rat immunoglobulins (DAKO), respectively, the membranes were washed, and bound antibodies were visualized with the ECL western detection system (Amersham Pharmacia Biotech).
GST Pull-down Assays-GST and GST-SRC1-(570 -780) fusion proteins were expressed in bacteria and bound to glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) using the method previously described (52). Proteins were synthesized in vitro from wild-type and mutant ER␣ cDNA templates using the TNT-coupled in vitro translation system (Promega) in the presence of [ 35 S]methionine. The 35 Slabeled proteins were incubated with GST or GST-SRC1 fusion proteins coupled to beads for a minimum of 3 h in NETN buffer (100 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 8.0, 0.5% Nonidet P-40) containing protease inhibitors in the presence or absence of 10 Ϫ6 M E2. Samples were washed in NETN, and bound proteins were separated by SDSpolyacrylamide gel electrophoresis. Gels were stained with Coomassie Blue to confirm equivalent amounts of GST proteins were present followed by detection of labeled proteins by fluorography.

RESULTS
Estrogen Receptor-mediated Transrepression of RelA Is Abolished by Antihormones-The ability of ER to modulate RelA activity in the presence of various ligands was determined in HeLa cells cotransfected with a RelA expression plasmid and ICAM-tk-Luc reporter containing three NF-B-response elements in the presence or absence of ERs. Cells were subsequently treated with either no hormone (NH), E2, OHT, or ICI, and the effect on RelA transactivation is presented in Fig. 1. Consistent with previous findings, expression of ER␣ in the presence of E2 resulted in a 5-fold decrease in RelA activity. The receptor also inhibited RelA activity in the absence of ligand, a phenomenon that has been previously noted with ER, progesterone receptor, and AR (28,34,35). To address whether this effect could be attributed to residual steroids present in the medium, we made use of the ER␣ mutant G525R, which cannot bind E2 but retains responsiveness to OHT (53). Cotransfection of RelA and G525R resulted in a hormone-independent decrease in RelA activity relative to RelA alone, although this reduction was less than with wild-type ER␣ in the absence of hormone. As expected, the addition of E2 had no further effect on G525R-mediated RelA transrepression. This finding suggests that low levels of estrogens in the culture medium may, in part, account for the observed ligand-independent inhibition; however, the unoccupied receptor is capable of functionally interfering with RelA transcriptional activity. We next examined the effect of hormone on the modulation of RelA activity by ER␤ (Fig. 1). In common with ER␣, expression of ER␤ resulted in a decrease in RelA activity, which was further reduced in the presence of E2.
Antihormones abrogate the ability of the ER to stimulate transcription of ERE-containing reporter genes in many cell types. It was therefore of interest to determine whether these agents also antagonize the inhibitory effect of ER on RelA transcription. Treatment with OHT and ICI had no effect on RelA activity alone but completely relieved transrepression of RelA mediated by ER␣ and ER␤ (Fig. 1). In addition, exposure to OHT abolished the inhibitory effect of the G525R mutant on RelA. The above findings demonstrate that both ER␣ and ER␤ mediate similar ligand-dependent actions on RelA transcriptional activity.
Dimerization Is Necessary for Both the Transactivating and Transrepressing Functions of the ER-We have previously characterized a region within ER␣, which is responsible for receptor dimerization (51). Subsequent elucidation of the crystal structure of the E2-bound LBD of ER␣ enabled us to identify candidate residues within the helix 11 dimerization domain that may be directly involved in forming contacts between ER␣ monomers (8,9). The surface of this domain can be divided into two regions comprising a group of hydrophobic residues at the top of helix 11 (Leu 508 , Leu 512 , Leu 515 ) surrounding Ala 509 and a cluster of mainly charged residues toward the C terminus of helix 11 (Ser 516 , Arg 519 , His 520 , Asn 523 ). To investigate the role of these two regions in dimer formation and their consequent effect on ER transcriptional activation and functional antagonism of RelA, we replaced each of the amino acids with alanine. The hydrophobic leucines were additionally altered to glutamic acid residues (L508E/L512E/L515E), and Ala 509 was replaced with a charged glutamic acid residue (A509E).
All mutant receptors bound ligand with high affinity (data not shown). The DNA binding capacity of in vitro translated wild-type and mutant receptors to a 32 P-labeled ERE was determined by gel shift assay in the presence or absence of E2 and MP16, an antibody specific for ER␣. This antibody is capable of restoring dimerization and DNA binding to dimerization defective mutant receptors ( Fig. 2A) (51). Substitutions L508A/ L512A/L515A and A509E at the N terminus of helix 11 reduced the DNA binding activity of the mutant receptors, whereas replacement of the leucines with glutamic acid residues, L508E/L512E/L515E, completely abolished the capacity of the receptor to bind to an ERE (left panel). The addition of MP16, however, restored the DNA binding capability of all three mutant receptors indicating that their DNA binding domains were intact. Mutations to the C-terminal end of helix 11 were less deleterious to DNA binding as mutants S516A/R519A, H520A/ N523A, and S516A/R519A/H520A/N523A bound DNA in the presence and absence of hormone and were supershifted by MP16 to a comparable extent as wild-type ER␣ (right panel). We conclude that N-terminal Leu 508 /Leu 512 /Leu 515 and Ala 509 residues in the dimer interface identified by the ER␣ crystal structure (8,9) are crucial for dimerization so that their replacement results in reduced DNA binding activity.
We next compared the ability of the mutant receptors to activate transcription from an ERE reporter gene and to modulate RelA transcriptional activity (Fig. 2B). HeLa cells were transiently transfected with wild-type ER␣ or mutant receptors together with a 2XERE-pS2-Luc reporter gene, and the resulting normalized ER transcriptional activity in the presence and absence of E2 is illustrated in the left panel. HeLa cells were then transiently transfected with the RelA expression plasmid, and the wild-type ER␣ or mutant receptors in the presence and absence of E2 and the effect on NF-B-mediated transactivation relative to RelA alone (100%) are shown in the right panel. Western blot analysis demonstrated that the various receptors were expressed at equivalent levels (Fig. 2C).
As anticipated ER␣-activated transcription of the ERE reporter gene in a ligand-dependent manner (Fig. 2B, left panel) and caused an ϳ5-fold reduction in RelA activity in response to hormone (Fig. 2B, right panel). In line with their impaired DNA binding capacities, receptors L508A/L512A/L515A and A509E exhibited attenuated ER transcriptional activity, yet retained their ability to antagonize RelA in an E2-dependent fashion (right panel). In contrast substitution of leucines Leu 508 /Leu 512 /Leu 515 with glutamic acid residues rendered the dimerization defective mutant receptor both transcriptionally  receptors (27, 28, 32, 34). To examine the contribution of discrete residues required for ligand-dependent transactivation within the LBD to hormone-dependent RelA transrepression, we used full-length receptors with point mutations in the LBD, which retain their capacity to bind DNA and ligand. The mutant receptors selected for this study contain alterations of residues within helicies 3, 4, and 12. In the presence of ligand these residues form part of a surface for co-activator interactions and are therefore required for receptor transcriptional activation (12)(13)(14). These mutants have been described elsewhere (see "Experimental Procedures") with the exception of F371A, where the phenylalanine residue at position 371 in helix 4 was replaced by alanine. 2 Western blot analysis with an ER␣-specific antibody H222 revealed that the receptors were expressed at similar levels (Fig. 3A). The ability of the wild-type ER␣ and various LBD mutants to stimulate transcription of a 2XERE-pS2-CAT reporter and to regulate RelA-mediated transactivation in transiently transfected HeLa cells in the presence and absence of E2 is shown in Fig. 3B.
In the presence of E2, ER␣ stimulated transcription from the 2XERE-pS2-CAT reporter by ϳ25-fold (Fig. 3B, left panel) and reduced RelA activity by about 90% (right panel). Interestingly, substitution of the tyrosine residue at position 541 with alanine, which resulted in a constitutively active receptor ( (44), left panel), did not lead to ligand-independent inhibition of RelA, but rather to a hormone-responsive phenotype similar to wild-type ER␣ (right panel). The remaining mutations had a deleterious effect on receptor transcriptional activity (left panel). The transcriptionally defective helix 12 mutants ⌬540 -552 and L543A/L544A ( (12), left panel) demonstrated virtually no ligand-dependent suppression of RelA ((12), right panel) implying that residues within helix 12 are required for both these receptor functions. The helix 4 mutant F371A and the helix 12 mutant D542N/E546Q/D549N were capable of stimulating transcription from the reporter gene, albeit at approximately half the magnitude of ER␣ ((12), left panel), but they completely lost the ability to inhibit RelA in a hormone-dependent manner (right panel, compare F371A and D542N/E546Q/D549N NH and E2 bars). In contrast the helix 3 mutant K366A and the helix 12 mutant M547A/L548A exhibited significantly impaired transcriptional activation ((12, 13), left panel) but retained their capacity for hormone-dependent transrepression (right panel, compare NH versus E2 for each mutant).
The mutants can therefore be divided into several categories, receptors that have impaired ability to transactivate but can inhibit RelA in a ligand-dependent manner (K366A and to a lesser extent M547A/L548A), receptors that are capable of stimulating transcription but have lost their capacity for liganddependent transrepression (F371A and D542N/E546Q/D549N), a constitutively active receptor that exhibits hormone-dependent repression (Y541A), and receptors that are unable to mediate ligand-dependent transcription or transrepression (⌬540 -552 and L543A/L544A). Together these data demonstrate the importance of the integrity of individual residues within the LBD in conferring ER/RelA antagonism and moreover enable discrimination between the transactivation and transrepression properties of the receptor.
Ligand-dependent Repression of RelA Is Independent of ER Co-activator Binding-A potential mechanism underlying functional interference between ER and RelA may be competition for limiting concentrations of common cofactors. Candidates include the cAMP-response element-binding proteinbinding protein CBP/p300 family, which are essential coactivators of both NF-B and NR-mediated transcription and p160 family members including SRC1 and TIF2 that are addi-2 P. Henttu and M. Parker, unpublished results.

FIG. 2. Transactivation and transrepression require ER␣ dimerization.
A, ER␣ dimers are necessary for DNA binding. In vitro translated wild-type and mutant receptors were analyzed for DNA binding activity in a gel shift assay using a 32 P-labeled ERE in the presence or absence of E2 and an ER␣-specific antibody MP16. Protein⅐DNA complexes were separated on 7% native polyacrylamide gels and detected by autoradiography. B, mutations that disrupt dimerization impair ER-mediated transcription and RelA inhibition. HeLa cells were transiently transfected with (left panel) a 2XERE-pS2-pGL3 luciferase reporter together with wild-type ER␣, mutant receptors, or empty pSG5 expression vector or (right panel) the ICAM-tk-Luc reporter and RelA in combination with wild-type ER␣, mutant receptors, or empty pSG5 expression vector. Control cells (C) were transfected with the relevant reporter and empty expression vectors. Cells were exposed to 10 Ϫ8 M E2 or vehicle (NH) for 24 h prior to harvesting. The reporter gene activities were assayed, and results were expressed as ER activity from a 2XERE-pS2-pGL3 normalized for transfection efficiency using a ␤-galactosidase internal control (left panel) and for ICAM-tk-Luc transfections (right panel) relative to ␤-galactosidase standardized RelA activity in the absence of ligand (100%). Results shown represent the mean Ϯ S.E. of at least three independent experiments assayed in duplicate. C, Western blot analysis of ER␣ wild-type and mutant receptors overexpressed in COS-1 cells and detected with the ER␣-specific polyclonal antibody MP16. tionally required for nuclear receptor transcriptional activity and have been proposed to play a role in NF-B-dependent gene expression (42, 54 -59). To address this hypothesis we overexpressed these co-activators in HeLa cells transiently transfected with RelA and an NF-B-responsive ICAM-tk-Luc reporter gene in the presence or absence of ER␣ (Fig. 4A). Under similar experimental conditions in HeLa cells where p160 family members SRC1 and TIF2 potentiated ER␣-stimulated transcription of an ERE reporter by at least 3-fold (data not shown), overexpression of SRC1, TIF2, p300, and CBP had no effect on RelA alone. As anticipated, cotransfection of ER␣ caused a decrease in RelA activity that was further down-regulated in the presence of E2. Addition of the co-activators as illustrated in Fig.  4A and in titration experiments (data not shown) failed to relieve ER-mediated ligand-dependent RelA transrepression.
We next sought to determine whether there was a correlation between the ability of the ER␣ mutants to interact with coactivators and their capacity to inhibit RelA activity. 35 S-labeled in vitro translated wild-type and mutant receptors were incubated with GST alone and GST fused to the region of SRC1 previously shown to bind the LBD of ER␣ (45). As demonstrated in Fig. 4B, the ER␣ LBD receptor mutants described in Fig. 3 differed in their ability to interact with SRC1 in the presence of E2. Wild-type ER␣ and the Y541A receptor variant bound SRC1 (Fig. 4B) and inhibited RelA activity in response to treatment with E2 (Fig. 3B, right panel). Conversely F371A, L543/544A, and ⌬540 -552 failed to interact with GST-SRC1 (Fig. 4B) and were unable to repress RelA in a ligand-dependent manner (Fig. 3B, right panel). Binding of SRC1 to K366A and M547A/L548A was significantly impaired (Fig. 4B), whereas these mutants were capable of down-regulating RelA activity in the presence of hormone (Fig. 3B, right panel). Furthermore the D542N/E546Q/D549N mutant was capable of weakly associating with SRC1 (Fig. 4B) yet had lost the ability to mediate ligand-dependent RelA transrepression (Fig. 3B,   right panel). These results indicate that inhibition of RelA transcriptional activity by the ER can be dissociated from interaction with p160 co-activators.
We have previously identified specific residues of the ER␣ LBD helices 3, 5, and 12, which are critical in forming an interface for p160 co-activator interactions (13,14,45). The amino acids in helix 11 described in Fig. 2, which differed in their capacities to mediate transactivation and transrepression, do not form part of this co-activator binding surface. To further investigate the role of co-activators in ER/RelA crosstalk we examined the relationship between association of these helix 11 mutant receptors with SRC1 and the ability to inhibit RelA in response to E2. As illustrated in Fig. 4C, all in vitro translated mutants were able to bind GST-SRC1 in the presence of E2. Transrepression studies with these mutants (Fig.  2B, right panel) demonstrated that the dimerization defective mutant L508E/L512E/L515E was incapable of suppressing RelA in a hormone-dependent manner, suggesting that binding to SRC1 is not sufficient to allow inhibition of RelA by ER␣. These data support the above finding that the ability to functionally interfere with RelA transcriptional activity can be independent of association with co-activators and furthermore is consistent with the inability to relieve RelA transrepression under conditions where co-activator concentrations are not limiting. Taken together, these results suggest that sequestration of p160 or CBP/p300 co-activators is unlikely to be the principle cause of ER-mediated antagonism of RelA transcriptional activity.
ER-mediated Transrepression of RelA Is Independent of the RelA LXXLL Motif-Recently we and others have identified an ␣-helical leucine motif, LXXLL (where L is leucine and X is any amino acid), present in NR co-activator proteins that is both necessary and sufficient for interaction with NR (60 -62). Substitution of the leucine doublet with a pair of alanines in motifs within co-activators abolished their interaction with NRs (61). The Rel homology domain at the N terminus of RelA has been shown to directly interact with GR in vitro and in addition the C-terminal transactivation domain is necessary for cross-talk with steroid hormone receptors (47). The presence of a leucine motif at position 450 -454 in the C-terminal transactivation domain of RelA led us to investigate whether interference with RelA transactivation may be mediated by direct RelA/ER contact via this sequence. The pair of leucines within the RelA LXXLL motif was replaced with alanines, and the ability of ER␣ to repress the mutated RelA was assessed in transiently transfected HeLa cells. Although the motif is located within a leucine-rich region of the transactivation domain, which is important for RelA transactivation (47,63), mutation of this sequence did not have any effect on RelA transcriptional activity (Fig. 5). Cotransfection with ER␣ repressed the activity of the mutated RelA to a similar extent to wild-type RelA, indicating that this motif is not necessary for conferring transcriptional repression by the ER. DISCUSSION Steroid hormones have been shown to inhibit the transcriptional activity of NF-B by a mechanism that involves both the DBD and LBD of their cognate receptors (28,29,32,34,38).
Several reports have characterized features of the receptor DBD that are necessary to confer antagonism (36,64,65), but little is know about the role of specific regions within the LBD. In our study of transrepression of NF-B by the ER we have examined the contribution of residues implicated in dimerization and required for ligand-dependent transcription mediated by AF2. Because the crystal structure of ER␣ indicates that the dimer interface is comprised primarily of residues in helix 11 (8,9), we generated specific mutations in the N-and C-terminal portions of this helix. We found that replacement of hydrophobic amino acids toward the N terminus of helix 11, Leu 508 / Leu 512 /Leu 515 , with alanine or glutamic acid residues impaired or abolished, respectively, the DNA binding and ligand-dependent transcriptional activity of the receptor. We conclude that reduced DNA binding activity is caused by impaired receptor dimerization. This is supported by our observation that DNA binding activity could be restored in vitro by addition of the antibody MP16, which we have previously demonstrated can restore DNA binding of dimerization defective mutant receptors (51). The substitutions, which do not affect the integrity of the DBD, do not appear to alter the overall structure of the FIG. 4. Functional antagonism of RelA does not correlate with mutant receptor/co-activator interactions. A, overexpression of co-activators fails to rescue ER-mediated RelA repression. HeLa cells were cotransfected with the ICAM-tk-Luc reporter and RelA and with co-activators SRC1, TIF2, p300, or CBP either in the presence or absence of ER␣. Control cells (C) were transfected with the ICAM-tk-Luc reporter and empty expression vectors. Cells were treated without (NH) or with 10 Ϫ8 M E2 for 24 h prior to harvesting. Results are expressed as described in Fig. 1. B, ER␣ LBD mutants differ in their ability to interact with SRC1. In vitro translated [ 35 S]methionine-labeled wild-type and mutant receptors were incubated with GST or GST-SRC1-(570 -780) protein fusions in the absence (Ϫ) or presence of 10 Ϫ6 M E2. Bound complexes were separated by SDS-polyacrylamide gel electrophoresis, and labeled proteins were visualized by fluorography. 1/10 input represents 10% of the total amount of 35 S-labeled receptor used in each reaction. C, ER␣ dimerization defective mutants are able to associate with SRC1. GST pull-down assay with in vitro translated [ 35 S]methionine-labeled wild-type and helix 11 mutant receptors incubated with GST or GST-SRC1-(570 -780) protein fusions in the absence (Ϫ) or presence of 10 Ϫ6 M E2 as described above.
LBD as these mutants retain their ligand binding affinity and the ability to interact with SRC1 (Fig. 4C). Moreover, given that the mutations are predicted to be on the dimer interface of the LBD, they are unlikely to affect other functions. Therefore in agreement with the prediction from the crystal structure of the receptor (8,9), we conclude that residues in the N-terminal portion of helix 11 are essential for ER␣ dimerization and high affinity DNA binding.
The inability of the L508E/L512E/L515E mutant to inhibit RelA activity suggests that receptor dimerization is essential not only for transcriptional activation but also for transrepression. The observation that L508A/L512A/L515A and A509E retain the capacity to antagonize RelA suggests that the extent to which these mutants were able to form dimers was sufficient to mediate transrepression. This finding is consistent with the comparative level of the DNA binding and transactivation activities of these mutants (Fig. 2, A and B) and the amount of wild-type receptor necessary to suppress RelA being greater than that required for transactivation. 3 Our finding that receptor dimerization is necessary for both the transactivating and transrepressing functions of ER␣ seems to differ from the requirements of other steroid hormone receptors (36, 64 -67). However in contrast to our mutants, GR dimerization and AR DBD mutants were generated by introducing substitutions into the DBD thereby reducing association between the monomers on DNA. Thus the LBDs of these mutant receptors remain intact. Our data indicate that dimerization via residues in the N terminus of helix 11 in the ER␣ LBD is required for both transcriptional activation and transrepression properties of the ER␣ receptor.
Ligand-dependent transcription by ER␣ mediated by AF2 has been shown to depend on residues in helices 3, 4/5, and 12 (12)(13)(14). In this study we demonstrate that specific residues within these helicies are essential not only for transactivation but also for hormone-dependent inhibition of RelA (Fig. 3B). Moreover the correct alignment of helix 12 is important for both processes, because transrepression by ER␣ and ER␤ occurs in the presence of estrogen but not tamoxifen, which results in its misalignment (Fig. 1) (8 -10). However we have generated ER␣ LBD mutants, which enable discrimination between the stimulatory and inhibitory properties of the receptor. We have found that some of the residues that are necessary for transactivation (Lys 366 , Met 547 , and Leu 548 ) are not required for transrepression, and conversely certain residues that are required for ligand-dependent inhibition of RelA (Phe 371 , Asp 542 , Glu 546 , Asp 549 ) are less critical for ER transcriptional activity, whereas others are essential for both functions (Leu 543 , Leu 544 ). These findings indicate that the two properties of the receptor are mediated by both common and distinct surfaces on the LBD. Furthermore we have demonstrated that certain LBD mutants (K366A and M547A/L548A), which exhibit impaired transcriptional activity (Fig. 3B) and were unable to interact with SRC1 (Fig. 4B), retained their ability to inhibit RelA (Fig. 3B). Conversely the dimerization defective mutant L508E/L512E/L515E, which was able to bind SRC1 (Fig. 4C), had lost the capacity to repress RelA (Fig. 2B). These observations suggest that the recruitment of p160/CBP families of co-activators is not essential to mediate ER␣ transrepression, which is consistent with the observation that overexpression of these co-activators failed to relieve the inhibitory effects of ER␣ on RelA transcriptional activity (Fig. 4A). Taken together our data indicate that competition and sequestration of the p160/CBP co-activators cannot account for the estrogendependent transrepression of RelA.
In agreement with our findings, overexpression of SRC1 has also been reported to have no effect on modulating ER-dependent down-regulation of NF-B in either HeLa or HEK-293 cells (68). Similar conclusions were drawn for AP1 transrepression by the thyroid hormone receptor, because the integrity of AF2 in the thyroid hormone receptor did not alter the ability of the receptor to inhibit AP1 (69). However other studies have demonstrated that overexpression of CBP in transiently transfected cells does rescue hormone-dependent down-regulation of AP1 by retinoic acid receptor and GR (54) and overcomes reciprocal GR/ RelA (42), AR/RelA, and AR/AP1 cross-talk (41). Thus the importance of co-activators in mediating functional antagonism seems to vary depending on the receptor and target transcription factor. Our results indicate that estrogens inhibit NF-B transcription by a mechanism that does not depend on the integrity of ER␣ AF2 and its ability to interact with p160 co-activators.
We and others have demonstrated that ␣-helical LXXLL motifs present within certain co-activators are sufficient to mediate direct protein-protein contacts with the LBDs of nuclear receptors (60 -62). The C-terminal transactivation domain of RelA contains a consensus LXXLL motif, which is located in a region that is required for efficient transactivation (47,63), for the reciprocal suppression of GR activation (47), and within a domain that is necessary for the recruitment of CBP/p300 (70). Based on these data, it was conceivable that RelA contributed to the E2-dependent inhibitory effects of the receptor by interacting either directly or indirectly with ER␣ via this motif. However, substitution of the leucines with alanines did not alter down-regulation of RelA activity by the ER␣, indicating that the RelA LXXLL motif is not involved in ER ligand-dependent transrepression (Fig. 5). Increasing evidence suggests the mechanism of transrepression of NF-B in vivo may involve a combination of factors, the importance of which may vary between different cell types and steroid hormone receptors. For instance, up-regulation of IB expression appears to contribute to progesterone receptor antagonism of NF-B in T47D cells (64,71) and to GR-mediated inhibition of NF-B activity in HeLa cells (33), T-lymphocytes (37), and pulmonary epithelial A549 cells (71) but not in fibroblast (72) or endothelial cells (72, 73) and does not play a role in ER-or AR-mediated NF-B transrepression (28, 64). In conclusion we have shown a requirement for ER␣ dimerization and identified specific residues within the ER␣ LBD, which are required for ligand-dependent inhibition of RelA. Furthermore we have described ER␣ LBD mutations that discriminate between the stimulatory and inhibitory functions of the receptor and demonstrated that RelA transrepression can occur independently of interaction with co-activators. It is possible that RelA may bind cofactors that also associate with the ER in a ligand-dependent manner; however, our data suggest that it is unlikely that sequestration of co-activators that are required for ER activation can account for hormone-dependent antagonism of NF-B. Estrogen receptor-dependent repression of NF-B-induced genes is important in maintaining bone density (25,26) and in mediating the protective effects of estrogen in the cardiovascular system (74). The demonstration that hormone-dependent transactivation and transrepression by the ER are separable functions enables future identification of therapeutic agents, which selectively modulate these processes.