Analysis of Estrogen Receptor Interaction with a Repressor of Estrogen Receptor Activity (REA) and the Regulation of Estrogen Receptor Transcriptional Activity by REA

Cell Culture and Transfection. Transfections were done in ER-negative human breast cancer MDA-MB-231 cells, human endometrial cancer (HEC-1) cells and Chinese Hamster Ovary (CHO) cells. Cells were plated for transfection in 24-well plates and incubated for 24 h at 37 (cid:176) C with 5% CO 2 . Transfections were performed with an estrogen-responsive reporter plasmid, ER expression vector, and internal b -galactosidase reporter plasmid as described (23,27) except that all transfections utilized Lipofectin reagent (Life Technologies, Inc.) (28) instead of calcium phosphate. The reporter plasmids have been described (23,27,29). The cells were harvested 24 h after ligand treatment and lysed with three cycles of freezing on dry ice and thawing at 37°C. ER transactivation ability was determined CAT or Luc cell lysates. CAT or Luc assays were normalized to b -galactosidase activity from the cotransfected internal reference pCMV b plasmid.


INTRODUCTION
Nuclear hormone receptors comprise a large superfamily of transcription factors whose activity is under hormonal control (1)(2)(3). These receptors are characterized by a central DNA binding domain, which interacts with specific hormone response elements located near the target gene promoter, and by two distinct activation function (AF) domains that contribute to the transcriptional activity of these receptors. The first activation function, AF-1, is located within the amino-terminal portion of the receptor, whereas the second, hormone-dependent AF-2 is located in the carboxyl-terminal half of the molecule, overlapping the ligand binding domain.
AF-1 and AF-2 function in a synergistic manner and are required for full transcriptional activity in most cell contexts (4,5).
In addition to having their activities regulated by specific hormones, the activity of these nuclear receptors is also dramatically impacted by their interaction with coregulator proteins that function either to enhance (coactivators) or repress (corepressors) transcriptional activity (6)(7)(8)(9).
These coregulatory proteins are believed to be interposed between the receptor and the basal transcriptional complex. This tripartite action of nuclear hormone receptors, involving the receptor, its ligands, and its coregulator proteins (10), allows for the precise regulation of the biological effects of these hormones on gene expression.
In the case of the estrogen receptor (ER), which mediates the biological effects of estrogens in a variety of target tissues and in breast cancer, most of its identified coregulators have been coactivators including, most prominently, SRC-1 (11), and related p160 family members (12), as well as CBP (13), and others that act on several families of these nuclear receptors. Most of the coactivators contain a conserved motif, LxxLL (where L is leucine, x is any amino acid), termed the nuclear receptor (NR) box, which has been demonstrated to be necessary and sufficient for mediating the binding of these coactivators to the liganded receptor (14)(15)(16)(17)(18)(19). However, increasing evidence, in part from phage peptide library display, indicates that other peptide motifs are no doubt also involved in ER interaction with its protein partners (20)(21)(22). This protein is recruited to the hormone-occupied ER and selectively represses transcriptional activity of the ER, but not that of other steroid and nonsteroid nuclear receptors (23). In this study, to decipher the mechanisms of REA interaction with the ER, we have characterized the protein-protein interacting surfaces between REA and ER, and elucidated the role of the NR box motif of REA in the repressive activity of REA. We also document that there is competitive binding of REA and coactivators, such as SRC-1, to ER and show that this competitive binding is lost in REA mutants defective in repressing ER, implying that such competition represents at least part of the mechanism underlying REA suppression of ER activity.
Plasmids. pBSK-ERα S554fs was constructed by releasing the ERα S554fs from the pCMV5 vector (24) by BamHI digest. The cDNA insert was subcloned into the BamHI site of pBSK. pBSK-ERα (1-530) was constructed by releasing the ERα (1-530) from the pCMV5 vector by BamHI digest. The cDNA insert was subcloned into BamHI pBSK. The pBSK vector for the wild-type human ERα (wild type-ERα) and the dominant negative ERα L540Q have been described (23). The pET15b ERα (304−554) has also been described before (25). pCMV5-antisense REA was constructed by releasing the REA from the pBSK vector by EcoRI/XbaI digest. The cDNA insert was blunted and subcloned into the SmaI site of pCMV5.
The antisense orientation of REA was verified with appropriate restriction enzymes.
pBSK-ERα (AB) and pBSK-ERα (ABC) mutants were generated by PCR using the fulllength pCMV-ERα plasmid as template. Reactions were performed using VENT DNA polymerase from New England Biolabs (Beverly, MA) according to the manufacturer's recommendations.
The following PCR primers were used (f = forward primer, r = reverse primer): The forward primer contained an ATG and a Kozac sequence and each reverse primer contained a stop codon. PCR fragments were purified, digested with EcoRI and XbaI, and cloned into the EcoRI/XbaI sites of pBSK to generate pBSK-ERα (AB) and pBSK-ERα (ABC). The nucleotide sequence of these mutants was confirmed using the Taq DyeDeoxy and BigDye system (PE Applied Biosystems, Foster City, CA).
All the truncations of REA used for GST pull-down assays were generated by PCR using the full-length REA plasmid, pCMV-REA (23)

) cells and Chinese Hamster
Ovary (CHO) cells. Cells were plated for transfection in 24-well plates and incubated for 24 h at 37°C with 5% CO 2 . Transfections were performed with an estrogen-responsive reporter plasmid, ER expression vector, and internal β-galactosidase reporter plasmid as described (23,27) except that all transfections utilized Lipofectin reagent (Life Technologies, Inc.) (28) instead of calcium phosphate. The reporter plasmids have been described (23,27,29). The cells were harvested 24 h after ligand treatment and lysed with three cycles of freezing on dry ice and thawing at 37°C. ER transactivation ability was determined by CAT or Luc activity of the whole cell lysates.
CAT or Luc assays were normalized to β-galactosidase activity from the cotransfected internal reference pCMVβ plasmid. GST-REA interaction with [ 35 S] ER (wild type and mutants) was performed as described (23).

In vitro
GST-ER interaction with full length REA and REA mutants was done using GST-ERα was visualized with x-ray film using a chemiluminescence system (Pierce system, Rockford, IL), after washing the filters again under the same condition as for the primary wash.

REA repression of ER transcriptional activity at consensus and non-consensus estrogen
responsive DNA gene sites.
Nuclear hormone receptors regulate gene expression by interacting directly with DNA response elements, and indirectly via tethering to other transcription factors that bind to DNA (10,(29)(30)(31). We have previously observed REA-mediated repression of ER transcriptional activation in several cell types with reporter gene constructs containing consensus estrogen response elements (23). To examine if REA could also suppress ER activity at gene sites containing non-consensus estrogen response elements (such as in the lactoferrin gene, or complement C3 promoter; (27)), as well as at gene sites in which ER does not directly bind to the DNA but acts at the gene site via tethering through other transcription factors, as in the TGF-β3 promoter (29), experiments were carried out with these gene constructs. As shown in Fig. 1, REA effectively reduced transcriptional activity of the ER at these three different target gene sites, and it did so in a dose-dependent manner. Thus, REA acts generally to suppress ER transcriptional activity at diverse gene sites under ER regulation.

Enhancement of the magnitude of ER transcriptional activity by neutralization of endogenous REA activity.
To demonstrate whether the endogenous REA protein in cells acts as a repressor of ER activity, we used an antisense REA-encoding plasmid to reduce the endogenous level of REA  Fig. 2A and 2B), implying that endogenous REA normally dampens the stimulatory response to estradiol.

Defining the REA interaction domains on the ER.
To map the REA interaction domains on the ER, a GST-fusion protein containing fulllength REA was tested for its ability to interact in vitro with [ 35 S]-labeled wild type ER and with several ER mutants (Fig. 3). As previously reported (23)  GST interaction assays were used to examine the regions of REA interacting with the ER.
None of the REAs bound to GST alone, and full-length REA interacted well with GST-ER in the presence of estradiol (E 2 ), as shown in Fig. 4 REA contains one LxxLL domain, called the nuclear receptor or NR box, at amino acids 23-27 (LKLLL), a signature sequence known to be important in the interaction of many coregulators with nuclear receptors (16). Although the interaction studies described above with amino-terminal truncated REA proteins indicate that this NR box is not required for REA interaction with the ER, we wished to investigate whether this NR box played a role in the ability of REA to repress transcriptional activity of the ER. To investigate the functional significance of the REA leucine-rich NR box motif, we introduced point mutations in the LxxLL motif of REA so as to create single alanine substitutions for each of the leucines in the NR box (Fig. 5). The ability of each LxxLL mutant to suppress transcriptional activity of the ER was tested in reporter gene transfection assays (Fig. 5A). Each mutant REA was found to be less effective than wild type REA in suppression of ER transcriptional activity. The greatest loss in repressive ability was observed when alanine was substituted in the middle of the LxxLL domain (mutant LKLAL). In this case, the percent repression was reduced 4-fold. These data indicate that the integrity of the LxxLL domain is necessary for optimal repression by REA, although this motif (as shown in Fig. 4) is not required for REA binding to ER. Fig. 5B shows that all of the LxxLL To determine if the differences in the effectiveness of transcriptional repression by the three LxxLL mutants might be explained by a decrease of the ability of the REA mutants to bind ER, we tested their abilities to bind to ER (Fig. 6) and to compete with wild type REA for binding to ER (Fig. 7). The wild type and the mutant REA cDNAs were in vitro transcribed and translated, and then assayed for ER interaction in GST pull-down assays. As seen in Fig. 6, the LxxLL mutants were able to bind to ER in the presence of estradiol, and wild type REA and each mutant showed a similar magnitude of interaction with ER.
To examine the binding of these mutants to ER more quantitatively, we compared the displacement of radiolabeled wild type REA from ER with increasing concentrations of radioinert wild type or LxxLL mutant REA. Fig. 7 indicates that the two LxxLL REA mutants (LKLLA and LKLAL) found to be most compromised in repressive ability (Fig. 5A) were as effective as the wild type REA in blocking binding of the wild type REA to the ER, implying that mutation of the LxxLL motif does not influence the affinity of binding of REA to the ER.

Comparison of the effects of wild type and LKLAL mutant REA on SRC-1 binding to ER
and on repression of SRC-1-mediated enhancement of ER transcriptional activity. As shown in Fig. 8 To examine the underlying mechanisms responsible for the differences in the repressive activities of the wild type and mutant REA proteins, we evaluated the potential competition between wild type or LKLAL mutant REA, and SRC-1 for binding to the ER. We performed GST pull-down assays with GST-ER and radiolabeled REA in the absence or presence of increasing amounts of radiolabeled SRC-1 to determine if these two coregulators mutually competed for occupancy of the ER. As shown in Fig. 9, These findings indicate that wild type REA and SRC-1 mutually compete for binding to the hormone-occupied ER, but that mutational alteration of the LxxLL motif of REA results in an independence between SRC-1 binding and REA binding, such that their interaction with ER is no longer mutually competitive. This may, at least in part, account for the marked difference in the abilities of the wild type and mutant REAs to repress ER, as further discussed below (see Discussion).

DISCUSSION
The transcriptional activity of nuclear hormone receptors is now well documented to be regulated by the character and balance of coregulators present in target cells (6)(7)(8)34 (35)(36)(37)(38). REA differs from other coregulators in that it is estrogen receptor (ER)-selective among the nuclear hormone receptors and it is recruited to the ER by both estrogen and antiestrogen, whereas the p160 family of coregulators are recruited to receptors in the presence of agonist ligand only (8,11,12,16,39,40). By contrast, the corepressors NCoR and SMRT, which interact most notably with the nonsteroid class of nuclear hormone receptors such as the thyroid receptor, retinoic acid receptor and vitamin D receptor, bind to these receptors in their unliganded state and are shed from receptor upon hormone binding (41)(42)(43). The current studies document that REA represses the actions of the ER when it is acting at a variety of DNA responsive elements, including nonconsensus estrogen response elements, as in the C3 complement and lactoferrin genes (27,44), as well as when tethered via other proteins to the DNA site, as with the TGF-β3 gene (29). This broad range of repressive action at a variety of different promoter-enhancer sites, is in addition to its ability to repress ER activity when acting at a consensus estrogen response element (23).
Interestingly, although REA interacts directly with the ER and contains one NR box, our studies reveal that this motif is not required for ER interaction. Recent studies employing phage peptide library display (20)(21)(22) provide evidence that receptors are likely to interact with sequences on proteins that are quite different from those of the signature LxxLL motif. Also, Lee et al. (46) identified TRIP-1 (thyroid-hormone-receptor interacting protein), which does not contain a consensus LxxLL NR box motif, yet was able to interact in a ligand-dependent manner with the thyroid hormone receptor and the retinoid X receptor. This provides a precedent for the existence of binding motifs, other than NR boxes, that can mediate interactions between nuclear receptors and their coregulators.  conformation that does not recruit REA, and is therefore essential for REA recruitment being a ligand-regulated interaction.
As we show in this study (Figure 9), REA and SRC-1 compete for binding to ER in the presence of estrogen. In this regard, the relative interactions of REA and SRC-1 with dominant negative ERs is of interest. Dominant negative ERs are transcriptionally inactive ERs that are able to dimerize with the wild type ER and block wild type ER activity (24,32,33). Consistent Interestingly, the region of the ER involved in interaction with REA does not contain helix 12 (Fig. 10), a region of the ER considered to be of considerable importance in the interaction with several coactivators. The fact that the NR box (LxxLL) motif of REA is not important in REA interaction with ER, is consistent with our hypothesis that the interaction of Delage-Mourroux et al.  15 REA with the ER is mediated in a manner very different than that of most coactivators. It is also consistent with our observations that REA binds to the antiestrogen-occupied ER, because the NR box motif would not be expected to be involved in the interaction with the antagonistoccupied receptor complex. It has been documented that when estrogen binds to the ER, the NR box motif of the p160 coactivators, such as in SRC-1 or GRIP-1, binds to a groove between helices 3, 4, 5 and 12 of the receptor. When the antiestrogen trans-4-hydroxytamoxifen binds to ER, helix 12 shifts to fill this groove and blocks interaction with the LxxLL motif of the p160 coactivators (17,45,48). Thus, our findings that REA interaction with ER is localized to domain E of the ER lacking helix 12 (ending at amino acid 530) and that the LxxLL motif of REA is not involved in ER interaction, would be consistent with our observations that antagonist as well as agonist occupied ER binds to REA.
The integrity of the LxxLL domain was found to be necessary for optimal repressive function of REA, although the LxxLL motif is not involved in the binding of REA to ER. It is of interest that the LxxLL motif in REA (amino acids [23][24][25][26][27] maps to the first of two regions (amino acids 19-49 and 150-174 (23)) required for repression (Fig. 10). That substitution of the leucines in the LxxLL motif in REA with alanines reduces the repressive activity of REA, but not interaction with the ER, implies that this motif is involved in another protein-protein interaction associated with the repression process. Interestingly, REA, which enhances the inhibitory effectiveness of dominant negative ERs and of antiestrogens (23), also moderates the activity of estrogens, because REA is recruited to ER even in the presence of estradiol. Our finding that administration of an antisense REAencoding plasmid resulted in an increase in the magnitude of ER transactivation, implies that endogenous REA normally dampens the stimulatory response to estradiol. The observations described in this paper assist in defining the biochemical basis of the selective interactions between REA and ER, and they provide insights into the molecular mechanisms by which REA acts to modulate activity of this steroid hormone receptor.     100% activity is set at that observed with estradiol in the absence of any added coregulator (SRC-1, REA or LKLAL REA mutant; second bar from left).