Mutations Targeted to a Predicted Helix in the Extreme Carboxyl-terminal Region of the Human Estrogen Receptor-α Alter Its Response to Estradiol and 4-Hydroxytamoxifen*

The human estrogen receptor-α, a member of the nuclear receptor superfamily, is a ligand-regulated transcriptional modulator. Because comparatively little is known about the extreme carboxyl-terminal region of the estrogen receptor (F domain), we used secondary structure prediction to design mutations that delete the F domain (S554stop), disrupt a possible turn (G556L/G557L), and alter a predicted helix (S559A/E562A, Q565P), and we evaluated the effects of these mutations on hormone binding and transcription activation in response to estradiol and the mixed agonist/antagonist 4-hydroxytamoxifen. Mutations that deleted the F domain (S554stop) or targeted the predicted helix (S559A/E562A, Q565P) greatly reduced or eliminated the agonist activity of 4-hydroxytamoxifen. Deleting the F domain increased the affinity of the receptor for estradiol and decreased the antagonist activity of 4-hydroxytamoxifen. The Q565P mutant exhibited a non-cooperative hormone-binding mechanism, as well as an impaired response to estradiol and increased antagonist activity of 4-hydroxytamoxifen. Our results show that mutations in the F domain alter not only the response to estradiol, the affinity for hormone, and the interaction between receptor subunits but can uncouple the agonist and antagonist activities of 4-hydroxytamoxifen. These results suggest that the F domain modulates the activity of the estrogen receptor-α by multiple mechanisms.

The members of the nuclear receptor superfamily act as transcriptional regulatory factors and exhibit a multidomain structure characterized as domains A-E/F (1,2). The F domain, which is present only in certain members of this large superfamily, is located at the extreme COOH terminus of the receptor distal to the larger ligand binding domain (LBD 1 or domain E) (Fig. 1). Among the nuclear receptors for which this region is present, substantial variability exists in both the length of the F domain, from 19 to more than 80 amino acids long, and its sequence. For example, the F domains in the ␣and ␤-forms of the human estrogen receptor (hER) exhibit no significant sequence homology and are suggested to be in part responsible for the differences in the biological activity of these forms (3). Although structural information is available for the DNA-binding domains and the ligand-binding domains of the hER␣, no structural information is yet available for the F domain.
Residues in this domain are not required for ligand binding or transcriptional activation and have no independent activity attributed to them (4). Indeed, in some cases, deletion of the F domain is reported to enhance receptor activity (3)(4)(5). Not only is the ⌬F ER␣ mutant an effective transactivator in response to E 2 , it is also similar to the wt protein in its ability to induce distortions in DNA and directed bend angles (6), as well as its half-life (7). However, Wrenn and Katzenellenbogen (8) have shown that although the S554 frame-shift ER, which contains 35 codons not present in the F domain of the wt ER␣, is a potent mediator of E 2 -stimulated transactivation in yeast, its activity in Chinese hamster ovary cells is markedly impaired despite the demonstration of an almost normal E 2 binding affinity in both cell types. Ince et al. (9) showed that when the same mutant is coexpressed with the wt receptor, transactivation is suppressed. Thus, although deleting the F domain has little effect on the response of the hER␣ to E 2 , certain mutations of the F domain can impair responses to E 2 .
In contrast to the minimal effect of deleting the F domain of the hER␣ on its response to E 2 , deleting this domain eliminates the ability of tamoxifen to act as an agonist (10). Furthermore, by using a yeast fusion protein assay, Nichols et al. (11) have shown that the F domain is involved in the antagonist activity of tamoxifen as well. Thus, the presence of the F domain is a key determinant of the ability of the hER␣ to respond to tamoxifen.
Because the carboxyl terminus of the E 2 -bound hER␣ LBD is located at the dimerization interface between ER monomers (12,13), it is likely that residues in the F domain influence dimerization (5). In addition, the ER is an allosteric protein that binds E 2 with high positive cooperativity, which indicates that information is transferred efficiently between subunits of the homodimer (14). Because of the position of the F domain relative to the dimerization interface, it is likely to be involved in subunit-subunit interactions and the transfer of ligand binding information between subunits. Finally, because helix 12 is reoriented in the 4-hydroxytamoxifen-(4-OHT-) and raloxifene-bound ER LBDs, it is likely that the F domain will also become reoriented and through this reorientation play a role in the response of the receptor to these ligands (12,13).
The poor conservation in sequence and length combined with the relative lack of structural and functional information make the F domain one of the least well understood segments in the multidomain structure of nuclear receptors. Our aim is to understand better the contribution made by residues in the F domain of hER␣ to the changes in conformation and transcriptional activation induced by the binding of different ligands.
We have examined the F domain of hER␣ by performing secondary structure prediction analyses and by constructing mutants designed to perturb selected predicted elements of structure. We have tested the effects of these mutations on hormone binding and on the response to E 2 , 4-OHT, and the combination of E 2 and 4-OHT in transient transfection assays. We have also used energy minimization to construct a model for the F domain. Our results show that specific mutations in the F domain alter the response to E 2 , the agonist and antagonist activities of 4-OHT, the affinity of the ER for E 2 , and the subunit-subunit interactions of the ER. They also show that the effects of mutations on these activities are separable.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-All cell culture reagents were purchased from Invitrogen with the exception of [ 14 C]chloramphenicol (PerkinElmer Life Sciences), acetyl coenzyme A lithium salt (Amersham Biosciences), 17␤-estradiol (E 2 ), and 4-hydroxytamoxifen (4-OHT) which were obtained from Sigma. HeLa (human cervical carcinoma) cells, obtained from the ATCC (Manassas, VA), were routinely maintained in a 5% CO 2 incubator using DMEM/F-12 media (Dulbecco's modified Eagle's F-12) supplemented with 10% fetal bovine serum (FBS) (Invitrogen) plus 0.5% gentamicin antibiotic (Sigma), and were subcultured weekly. Sf9 insect cells, Bac-N-Blue transfection kit, and baculovirus protein expression kit were purchased from Invitrogen. Ovalbumin was purchased from Sigma. 17␤-[6,7-3 H]Estradiol was purchased from PerkinElmer Life Sciences. The experiments studying the S554stop mutant used the highly purified E isomer of 4-hydroxytamoxifen; in the other experiments, this isomer was no longer available, so a mixture of Z and E isomers (50:50) was used.
Plasmids (DNA Constructs)-The pEREBLCAT reporter (4) contains an estrogen-response element (ERE) derived from the vitellogenin A2 promoter upstream of sequences from the herpes simplex viral thymidine kinase gene linked to the coding region of the bacterial gene for chloramphenicol acetyltransferase (CAT). The pSG5-HEGO expression vector containing the wild-type (wt) hER␣ was the generous gift of Drs. Pierre Chambon and Hinrich Gronemeyer (15). Mutants of hER␣ were constructed as described below.
Site-directed Mutagenesis of the hERa-Oligonucleotide site-directed mutagenesis was performed following procedures suggested by the manufacturer (Promega). A 786-bp fragment containing the cDNA corresponding to the carboxyl terminus of the wt ER, including the junction between domains E and F, was subcloned from HEG0-pSG5 into the mutagenesis vector pALTER1 (Promega) by digesting with EcoRI and HindIII endonucleases. Three oligonucleotide primers in which one or two amino acids were altered as designed were obtained from Operon (Alameda, CA). The mutations and the corresponding oligonucleotides are as follows: 1) Q565P, 5Ј-CAAGTGGCTTGGGTCCGTCTC-3Ј; 2) S559A/E562A, 5Ј-TTGGTCCGTCGCCTCCACGGCTGCCCCTCC-3Ј; and 3) G556L/G557L, 5Ј-CTCCTCCACGGATGCCAATAAACGGCTAG-TGGGCGC-3Ј. The F domain deletion mutant (S554stop) was constructed by altering residue Ser-554 of the F domain to a stop codon, using the oligonucleotide 5Ј-GGATGCCCCTCCACGTCAAGTGGGCG-CATGTAG-3Ј. The oligonucleotide primers were annealed to the hER-pALTER1 single-stranded DNA along with two antibiotic resistancealtering primers (the tetracycline knockout primer and the ampicillin repair primer, provided by Promega) in the present of T4 DNA ligase, T4 DNA polymerase, and dNTPs to synthesize the mutated DNA strand. Mutant colonies were selected from LB plates containing 125 mg/ml ampicillin. All mutants were confirmed by DNA sequencing. The cDNA fragments containing the appropriate mutations were cut and ligated back into the HEG0-pSG5 plasmid to generate the full-length Q565P, SS559A/E562A, G556L/G557L, and S554stop mutant hERs in the pSG5 expression plasmid.
Transient Cotransfections and CAT Assays-For transient transfection experiments, HeLa cells were plated at 2 ϫ 10 5 cells/60-mm dish in DMEM/F-12 growth media supplemented with 5% dextran-coated charcoal-stripped FBS and were allowed to recover for 24 h. Transfection experiments were carried out using the Superfect reagent (Qiagen, Valencia, CA), according to the manufacturer's instructions. Each 60-mm plate received 2.0 g of the CAT reporter plasmid (pERE-BLCAT) along with 0.5 g of an expression plasmid for wt or mutant hER␣. The vitellogenin ERE-containing reporter was chosen for the consistently high level of responsiveness to E 2 stimulation that it confers to transiently transfected HeLa cells. The transfected cells were incubated overnight and then refed with fresh media (DMEM/F-12 plus 5% dextran-coated charcoal-stripped FBS) that had been supplemented with either E 2 or 4-OHT at the indicated concentrations or with ethanol vehicle (0.01%). In experiments designed to test the ability of 4-OHT to antagonize E 2 -induced stimulation, cotreatment assays were performed using the indicated concentration of E 2 in the presence of a single concentration of 4-OHT, 10 Ϫ7 M. The transfected cells were exposed to hormone-or vehicle-supplemented media for 24 h before harvesting. In each experiment, the abilities of both the mutant and the wt ER to activate the reporter plasmid were measured in parallel, in the same assay, using duplicate cell culture dishes for each experimental condition. The samples were normalized for protein concentration (16) and assayed for CAT activity as described previously (17,18).
Data quantitation was performed using AMBIS Imaging Acquisition Software (San Diego, CA). The percentage conversion of chloramphenicol to its acetylated products was determined for each sample and is expressed as the fold activity over control or as the percent of the wild-type ER activity induced by corresponding ligand concentrations. Data are the means of 2-7 independent experiments performed in duplicate.
Expression of the ER in Sf9 Cells-The cDNA encoding the wt and mutant ERs was excised from their respective pSG5 vectors by digestion with EcoRI and MslI and ligated into the pBlueBacHis2B vector using standard molecular biological techniques to generate transfer plasmids for generation of recombinant baculovirus. The correct insertions were confirmed by DNA sequencing. The transfer plasmids containing the wild-type and mutant ERs were cotransfected with the linear viral DNA AcMNPV by cationic liposome-mediated transfection using procedures suggested by the manufacturer (Invitrogen). Identification of recombinant virus and generation of a high titer virus stock were carried out using procedures suggested by the manufacturer (Invitrogen). The recombinant wt and mutant ER proteins were expressed by infecting Sf9 cells (10 6 cells/ml) with the corresponding high titer virus stocks for 48 h. After expression, the cells were washed with phosphate-buffered saline and pelleted in 50-ml aliquots. The pellets were stored at Ϫ80°C.
Equilibrium Binding of Estradiol to the wt and Mutant ER-Equilibrium binding experiments were carried out using procedures similar to those described previously (14). Cell extracts were prepared as described above. Aliquots (200 l) of the cell extracts containing wild-type or mutant ERs were incubated with increasing concentrations of [ 3 H]estradiol from 0.5 to 40 nM for 2-4 h at 25°C to reach equilibrium. The non-specific binding was determined in a parallel set of incubations containing a 200-fold molar excess of unlabeled estradiol. Free and bound steroid were separated by dextran/charcoal assay. Preliminary experiments were conducted to determine the time necessary for each mutant to achieve equilibrium. The affinity of the receptor for estradiol and the Hill coefficient were obtained by non-linear regression analysis using the programs Lotus 1-2-3 and GraphPad Prizm to fit the binding data to the Hill equation (20). The data were also graphed according to the method of Scatchard (21). Experiments in which receptor inactivation was greater than 10% were discarded; for the wt, S554stop, and Q565P receptors, inactivation levels were generally 5% or less. Although the baculovirus-expressed G556L/G557L and S559A/E562A ER␣ mutants also bound [ 3 H]estradiol in vitro, the elevated inactivation exhibited by these mutants prohibited accurate measurements of their binding properties (data not shown).
Molecular Modeling-Molecular modeling was carried out using Biopolymer and Discover within InsightII 98 (Molecular Simulations, Inc.) using a Silicon Graphics O2 work station and the consistent valence force field. The F domain, residues 551-595 of the hER␣ (624 atoms), was constructed based on the secondary structure predictions shown in Fig. 1 (GOR IV); hydrogens were added (pH 7.0), and the model was subjected to 100 rounds of steepest descent minimization with charges, but without Morse potentials and cross-terms, followed by an additional 2500 rounds of conjugate gradient minimization with charges, Morse potentials, and cross-terms.

Design of Mutations in the F Domain of hERa-
In the absence of structural information, we used secondary structure prediction (GOR IV, 22) to target areas of the F domain for mutagenesis ( Fig. 1). The major predicted elements of structure include an ␣-helix (residues 559 -570, Fig. 1, shown in boldface) and a ␤-strand (residues 580 -585, Fig. 1, underlined); in addition, the two glycines at 556 and 557, just upstream of the predicted helix, may be a region of high flexibility or a turn, and two residues in the extreme carboxyl-terminal region are also predicted to be ␤-strand (Fig. 1). The remainder of the domain is predicted to be random coil. Similar predictions were obtained using the methods of Chou-Fasman, Delange-Roux, and Levitt (Refs. 23-25 and results not shown).
Molecular modeling based on the secondary structure predictions of residues 551-595 of the hER␣ was also carried out. In the minimized model, the amino-terminal region contains a helical region, whereas the residues carboxyl-terminal to this form an extended region ϳ60 -70 Å long (Fig. 2). The Gly-556 and Gly-557 occupy a turn between the end of the LBD and the predicted helix (Fig. 2). To test whether the initial assignment of secondary structure to residues could influence the minimized model, a region that is not predicted to be ␣-helical was modeled beginning with a helical secondary structure and subjected to energy minimization. This peptide did not remain helical but unfolded during the early rounds of minimization to a more extended structure (not shown). These results provide additional support for the existence of an ␣Ϫhelix in the aminoterminal region in the F domain, as well as a more extended region that is potentially able to interact with other regions of the receptor and/or other proteins.
For the experimental studies, we targeted residues within and near the predicted ␣-helix for mutation, as well as constructing a receptor lacking the entire F domain (Fig. 1). The S554stop mutant was designed to eliminate the entire F domain. The G556L/G557L mutant was designed to disrupt the potential turn or to reduce the local flexibility of the region located on the amino-terminal side of the predicted helix. The S559A/E562A mutant was designed to alter the hydrogen bonding at the beginning of the putative ␣-helix, whereas the Q565P mutant was intended to disrupt the predicted ␣-helix.
Basal Activity of F Domain Mutants of hER␣-We found that mutations to the F domain of ER␣ have widely differing effects on the levels of basal reporter activity in transiently transfected HeLa cells. The basal activity of each of the ER␣ mutants is expressed as a percentage of the basal CAT activity (arbitrarily set at 100%) induced by wt ER␣ transfectants in ethanol-treated control cells, measured in parallel assays. The Q565P and ⌬F ER mutants exhibited basal activity approximately 2-3-fold greater that of the wt ER (Table I). Basal activity of the S559A/E562A mutant was similar to that of the wt ER. By contrast, the G556L/G557L mutant ER␣ exhibited a reduced level of basal activity, ϳ53% of the wt ER (Table I).
Western immunoblotting showed that all mutants were expressed at levels similar to that of the wt protein (not shown).
Estradiol-stimulated Transcription by the F Domain Mutants-We measured the abilities of wt and mutant hER␣ to activate transcription from the ERE-driven reporter in the presence of increasing concentrations of E 2 , expressed relative to the level of activity in the absence of E 2 . In these experiments, E 2 increased reporter gene activity levels mediated by the wt ER in a dose-dependent manner; the maximum increase was 9.7-fold in response to 10 Ϫ7 M. At 1 M E 2 , a decrease in the level of transactivation was observed, so that the overall doseresponse curve was bell-shaped. All of the F domain mutants of ER␣ were able to increase transcription in response to stimulation with E 2 ; the extent of induction, however, varied among the ER transfectants (Fig. 3A).
The S554stop or ⌬F hER␣ truncation mutant, in which the entire F domain is missing, exhibited a robust response to stimulation with E 2 . The E 2 dose-response curve produced by the ⌬F ER␣ transfectant is qualitatively and quantitatively similar to that of the wt receptor, with the exception that it retained activity at the highest E 2 concentration used (1 M), whereas the wt receptor exhibited decreased activity at this concentration (Fig. 3A). These results show that although deleting the F domain produces a receptor mutant that exhibits increased basal activity, the ⌬F ER␣ retains a wt-like sensitivity to stimulation by E 2 . Our results also show that the presence of the F domain is dispensable for E 2 -stimulated gene expression by hER␣. However, experiments that measure the effects of single and double point mutations of residues in the F FIG. 1. The domains of the human ER␣, the sequence and predicted secondary structure of the F domain, and the mutants used in the present study. The domains of the full-length hER␣ (residues 1-595) are shown. The darkly shaded box indicates the F domain, residues 553-595. The amino acid sequence of the carboxylterminal region of the protein (residues 551-595), including the F domain, of hER␣ is shown below. Residues predicted to be an ␣-helix are in bold; residues predicted to be ␤-strand are underlined. The remaining residues are predicted to be random coil. The ER mutants used in the present study are listed, and the altered residues are indicated by arrows.
FIG . 2. Energy-minimized model of residues 551-595 of the hER␣. Minimization was carried out using Biopolymer and Discover as described under "Experimental Procedures." The backbone ribbon and side chains are shown in gray. The residues mutated in the current study are in black. The side chain of Arg-555 has been omitted for clarity. The dashed lines indicate the approximate positions of the predicted helix and a region of ␤-strand. Two additional residues near the carboxyl terminus of the protein are predicted to be ␤-strand. The remainder of the domain, including the region between the predicted helix and ␤-strand and the region following the ␤-strand, is predicted to be random coil.
domain of ER␣ on E 2 -stimulated expression from the same CAT reporter provide additional insight into the role of this segment.
Transcription activation by the S559A/E562A mutant exhibited an interesting dose response. At the middle concentrations of E 2 (1.0 -10 nM E 2 ), the activity of the mutant was similar to that of the wt protein; however, at the highest concentrations (0.1 and 1 M), the mutant exhibited nearly three times the activity of the wt protein (Fig. 3A). Thus, this double mutation led to magnified E 2 -stimulated transactivation at the higher ligand concentrations.
The double point mutant G556L/G557L hER␣ was also tested for its responsiveness to stimulation by E 2 . Estradiol (1 nM) was a potent agonist of the G556L/G557L mutant hER␣driven transactivation response, stimulating a 16-fold increase in the level of reporter activity, roughly twice the activity of the wt protein at this same concentration (Fig. 3A). Transcriptional activation mediated by the G556L/G557L hER␣ at other E 2 concentrations was similar to that of the wt protein. It is worth noting that the G556L/G557L hER␣ is the only F domain mutant that, like the wt hER␣, exhibits a bell-shaped doseresponse curve. It exhibited maximal activity at low/intermediate concentrations of E 2 and decreasing activity at higher E 2 concentrations.
An entirely different profile is observed for the Q565P mutant. Transient cotransfection assays used to evaluate the impact of this single amino acid change on transcriptional activity revealed a reduction (relative to that of wt hER␣) in the response of this mutant to stimulation at each concentration of E 2 used up to 10 Ϫ6 M (Fig. 3A). Moreover, the maximum response of the Q565P ER␣ mutant to E 2 is ϳ45% lower than that of the wt ER␣ at the same concentration (10 Ϫ7 M E 2 ) (Fig. 3A). The decreased response to E 2 and the increased constitutive activity exhibited by this mutant underscore the importance of Gln-565 in mediating optimal E 2 -regulated responsiveness of the ER␣.
F Domain Mutations Diminish the Capacity of hER␣ to Mediate Partial Agonism by 4-OHT-To examine the influence of F domain residues on the agonist activities of the selective estrogen receptor modulator 4-OHT, we tested the dose-dependent effects of 4-OHT on the activity of the same ERE-CAT reporter induced by wt and F domain mutants of ER␣ in transient cotransfection assays. As has been described previously (10) using other reporters incorporating a vitellogenin-derived promoter, our results show 4-OHT to be a weak agonist of the wt ER␣-driven transactivation response. CAT expression levels in wt ER␣-transfected cells increased by 1.7-1.9-fold in response to administration of 10 Ϫ10 to 10 Ϫ7 M 4-OHT and by 2.7-fold in response to 1 M 4-OHT (Fig. 3B).
We, like others, found that deleting the F domain eliminated the ability of 4-OHT to stimulate transcription from a vitellogenin-derived promoter ( Ref. 10; Fig. 3B). The S559A/E562A mutant has a substantially reduced response to 4-OHT; it is only able to stimulate transcription at the M concentration of 4-OHT (Fig. 3B). The Q565P mutant did not respond to 4-OHT as an agonist at any of the concentrations tested. Moreover, at the highest concentration used, 10 Ϫ6 M 4-OHT, repression of the basal transcriptional activity by the Q565P mutant was observed (Fig. 3B). Thus, mutations to residues within the putative ␣-helix of the F domain produce hER␣ mutants that have either a substantially reduced or complete loss of their ability to mediate 4-OHT agonism.
By contrast, in the experiments using the G556L/G557L mutant, the weak agonistic activity associated with 4-OHT is maintained over much of the concentration curve. Reporter activity levels initiated by the G556L/G557L mutant ER␣ in response to 10 Ϫ10 to 10 Ϫ7 M 4-OHT are similar to those of the wt. However, at 1 M 4-OHT, the agonistic activity is lost (Fig.  3B). Thus, unlike the substantial loss of 4-OHT agonism observed in the deletion mutant or the mutants having changes to residues in the predicted ␣-helix, this mutant, which targeted residues outside of the predicted ␣-helix, had little effect on the ability of 4-OHT to exert mild agonism.
In summary, these results show that the integrity of the F domain of hER␣, and of the predicted ␣-helix within that domain, is required for 4-OHT to act as a weak agonist at an ERE-driven reporter. Moreover, a mutation that increased the response to E 2 , S559A/E562A, also substantially decreased the ability of 4-OHT to function as a partial agonist of ER␣-driven transcriptional activation. This shows that the effects of these mutations on E 2 -stimulated and 4-OHT-stimulated transcription activation are separable.

4-OHT-mediated Repression of the E 2 -stimulated Transactivation Response Is Differentially Affected by Mutations in the F
Domain of hERa-To understand better the role played by F domain residues in mediating the anti-estrogenic effects of 4-OHT, we performed coadministration assays with 4-OHT and E 2 . For these experiments, HeLa cells were transiently cotransfected with the pEREBLCAT reporter and expression vectors for the wt hER␣ or the F domain mutants as before, incubated either with ethanol vehicle alone (controls) or with increasing concentrations of E 2 (10 Ϫ10 to 10 Ϫ6 M) in the presence of a single concentration of 4-OHT (10 Ϫ7 M) for 24 h, and assayed for ERE-CAT reporter activity. By using this approach, we were able to measure the ability of E 2 to overcome the inhibition by 4-OHT and assess the contribution of specific residues in the F domain of hER␣ to this process. The 10 Ϫ7 M concentration of 4-OHT was chosen for its ability to repress wt transcriptional activation stimulated by administration of up to 100 nM E 2 (data not shown). In addition, the S554stop, S559A/E562A, and Q565P mutants were transcriptionally inactive in response to 10 Ϫ7 M 4-OHT alone; the G556L/G557L mutant and the wt exhibited some activity (Fig. 3B). Note that no concentration of 4-OHT alone lacked agonist activity for all proteins tested (Fig. 3B).
For the wt protein and all mutants, increasing E 2 concentrations in the presence of 4-OHT led to at least some degree of increased transcription (Fig. 3C). There are notable differences, however, in the responses of the individual proteins. Coadministration of 4-OHT and E 2 to the S554stop and S559A/ E562A mutants led to a dose-dependent increase in reporter activity levels, greatly exceeding those produced by the wt hER␣ at the highest ligand concentrations (Fig. 3C). A more complex response is exhibited by the G556L/G557L mutant ER␣. At E 2 concentrations from 10 Ϫ10 to 10 Ϫ7 M, transcription increases in a manner similar to that of the wt protein (Fig.  3C). However, at 10 Ϫ6 M E 2 , transcription was substantially reduced compared with the wt protein (Fig. 3C). Finally, although transactivation by the Q565P ER␣ mutant was somewhat stimulated by increasing concentrations of E 2 , it never reached the level produced by the wt protein (Fig. 3C). We also compared the activity of each protein in the presence of 4-OHT and E 2 with the activity in the presence of E 2 alone to determine the degree to which the presence of 4-OHT inhibits (or in some cases stimulates) E 2 -driven transcription (Table II). The addition of 4-OHT led to significant reductions in the reporter activity levels induced by cotreatment of wt ER␣ transfectants with 10 Ϫ10 to 10 Ϫ7 E 2 but was slightly stimulatory in the presence of 10 Ϫ6 M E 2 (Table II). Similarly, 4-OHTmediated antagonism was ameliorated by increasing concentrations of E 2 in the S554stop, G556L/G557L, and S559A/ E562A mutants (Table II). Indeed, the presence of 4-OHT with 10 Ϫ6 M E 2 was slightly stimulatory to not only the wt but the S554stop and G556L/G557L mutants as well (Table II). By contrast, for the S559A/E562A mutant, even though 10 Ϫ6 M E 2 stimulated transcription greatly in the presence of 4-OHT, transcription was still reduced 24% compared with E 2 alone (Table II), indicating that the repressive effect of 4-OHT on the activity of this mutant was not completely overcome. Finally, transcription by the Q565P mutant was strongly inhibited by 4-OHT at all concentrations of E 2 used; even at 10 Ϫ6 M E 2 , transcription was still inhibited by 65% (Table II).
By using the data in Table II, we then calculated the concentration of E 2 that would overcome 50% of the inhibition produced by 10 Ϫ7 M 4-OHT (Table III). For example, the presence of 4-OHT reduced transcription activation by the wt protein by 66% at an E 2 concentration of 10 Ϫ7 M but stimulated transcription by 22% at 1 M E 2 (Fig. 2, A and C; Table II); we calculate 50% of the inhibition would be overcome at an E 2 concentration of 0.16 M (Table III). The S554stop mutant required less E 2 to overcome 4-OHT inhibition than the wt protein, 0.008 M (Table III). The G556L/G557L mutant exhibited similar sensitivity to E 2 as the wt protein (Table III). The S559A/E562A mutant required slightly more E 2 , 0.32 M, to overcome 50% of the inhibition by 4-OHT (Table III). The Q565P mutant was the least responsive to E 2 , requiring 320 M E 2 to overcome 50% of the inhibition by 4-OHT (Table III). Clearly, the ability of E 2 to overcome inhibition by 4-OHT can be increased (S554stop) or decreased (Q565P) by mutations in the F domain.
Effect of F Domain Mutations on E 2 Binding by the ERa-We next investigated the effects of mutations in the F domain on the affinity of the ER for estradiol, as well as the cooperativity of the interaction. The cooperativity of binding is measured by the Hill coefficient, n H (20). A Hill coefficient near 1 is characteristic of a non-cooperative binding mechanism, that is either the protein is a monomer or binding of ligand to one subunit does not influence the binding of ligand to another subunit (26). A Hill coefficient greater than 1 indicates not only that there is more than one subunit but that binding of ligand to one subunit favors the binding of ligand to the other subunit(s) (26). Because the F domain has been implicated in receptor dimerization and could influence interactions between monomers of the receptor homodimer, we were especially interested in the effect of mutations on the cooperativity of binding. The S554stop (⌬F) mutant ER␣ has an increased affinity for E 2 relative to that of the wt receptor, 0.05 Ϯ 0.007 nM compared with 0.64 Ϯ 0.41 nM (Table IV; Fig. 4). Deleting the F domain of ER␣ had no detectable effect on the cooperativity of hormone binding, as the Hill coefficients for the binding of [ 3 H]estradiol to the wt and the S554stop mutant were ϳ1.6, and the Scatchard plots were curved (Table IV; Fig. 4). By contrast, the affinity of the Q565P ER␣ mutant for E 2 was similar to that of the wt protein, 0.23 Ϯ 0.01 nM versus 0.64 Ϯ 0.41 nM, yet the cooperativity of E 2 binding as measured by the Hill coefficient was quite different, 0.94 Ϯ 0.1 versus 1.58 Ϯ 0.18 (Table IV); the Scatchard plot of this mutant was linear (Fig. 4). Thus, this point mutation eliminated the positive cooperative E 2 -binding mechanism of the ER␣ (Fig. 4 and Table IV). These results show that although the subunit-subunit interactions necessary for the positive cooperativity of E 2 binding do not require the presence of the F domain, a point mutation within a predicted ␣-helix in the F domain can interfere with them. DISCUSSION We, like others, have shown that deleting the extreme carboxyl-terminal region of hER␣ (F domain) did not reduce E 2stimulated transcriptional activation, yet eliminated the agonist activity of 4-OHT (3-5, 10) (Fig. 3). Moreover, we have shown that deleting the F domain actually increased the affinity of the hER␣ for E 2 (Fig. 4). We also investigated the effects of mutations targeted to a predicted ␣-helix within the F domain. The Q565P mutation, which was designed to distort the predicted ␣-helix, decreased the response of the hER␣ to E 2 (Fig. 3A) and eliminated the cooperativity of E 2 binding ( Fig. 4; Table IV). Although many mutations have been reported that affect the transactivation function of the hER␣ and its affinity for E 2 , this is the first report of a mutant that exhibits a non-cooperative hormone-binding mechanism. By combining the binding and transactivation data, we assessed whether the S554stop and Q565P mutations alter the preference of the ER for the agonist-bound versus the 4-OHT-bound conformation. The S554stop mutant exhibited an increased affinity for E 2 and required less E 2 to override 4-OHT inhibition than the wt protein. This suggests that the S554stop mutant preferentially adopts the agonist-bound, rather than the 4-OHT-bound, conformation. By contrast, the Q565P mutant exhibited non-coop-  erative E 2 binding, a reduced response to E 2 in transient transfection assays, and required substantially more E 2 than the wt to override the inhibition by 4-OHT. This suggests that the Q565P mutant has a reduced ability to adopt the fully active, agonist-bound conformation and that it preferentially adopts a 4-OHT-bound conformation. The G556L/G557L and S559A/E562A mutants were designed to have more subtle effects on ER function than the S554stop and Q565P mutants and display more complex phenotypes. Gly-556 and Gly-557 at the start of the F domain (Fig.  1) occupy a predicted turn/coil region that could act as a flexible linker between the ER LBD and the rest of the F domain. Mutation of the glycines to leucines was predicted to disrupt the potential turn and/or reduce the local flexibility of this region. Because the mutated glycines are part of a consensus sequence for arginine methylases, RGG (27,28), we also cannot exclude the possibility that this mutation alters post-translational modification of hER␣. In transactivation assays, impaired function of this mutant relative to the wt was observed only at the highest concentration of each ligand, 10 Ϫ6 M (Fig. 3, A-C). Thus, mutating these residues, which lie outside the predicted ␣-helix, had a relatively minor effect on transactivation in response to E 2 and 4-OHT.
The S559A/E562A mutant was designed to test whether hydrogen bonding by Ser-559 and Glu-562 at the start of the predicted ␣-helix is important to ER function. The side chains of these residues could hydrogen-bond with the peptide backbone and stabilize the start of the predicted helix (Fig. 2); mutation to alanine would then destabilize, but not eliminate, helix formation (29 -31). The sequence SVEE is also a casein kinase II consensus sequence (28); phosphorylation of Ser-559 could enhance helical stability (32)(33)(34). In transactivation assays, mutating these residues did not blunt the response to E 2 ; indeed, at the highest concentrations, the response of the S559A/E562A mutant to E 2 was nearly three times the response of the wt protein.
However, this mutant responded to 4-OHT as an agonist only at the highest concentration, 10 Ϫ6 M (Fig. 3B). Also, although E 2 could stimulate transcription in the presence of 4-OHT, inhibition was not entirely overridden (Table II) (3). Thus, this mutant exhibited a strong response to E 2 , impaired agonist activity of 4-OHT, but a slightly increased antagonist activity of 4-OHT. This suggests that the predicted helical region is necessary for the agonist activity of 4-OHT and that mutation of this region enhances the antagonist activity of 4-OHT.
The F domain has been proposed to inhibit receptor dimerization (5). The effects of the S554stop and Q565P mutations on hormone binding are consistent with altered dimerization and/or subunit-subunit interactions. However, if the role of the F domain were solely to inhibit dimerization, one would predict that if the activity in response to one ligand were increased because of increased dimerization, then the constitutive activity and the activity in response to other ligands would also be increased. Our results show that a mutation (S559A/E562A) can increase the response to E 2 while eliminating the agonist activity of 4-OHT. Thus, the agonist activity of E 2 can be uncoupled from the agonist activity of 4-OHT. In addition, two other mutants have lost the agonist activity of 4-OHT, yet the antagonist activity of 4-OHT has been reduced in one (S554stop) and increased in the other (Q565P) (Table III). Therefore, mutations targeting the F domain can uncouple the agonist and antagonist activities of 4-OHT as well. It would appear that the F domain has other functions in addition to modulating dimerization.
The results of digestion of the ER by proteases are consistent with our modeling studies. The region of the F domain carboxyl-terminal to the predicted ␣-helix is highly sensitive to proteolysis (35). The core of the ER in tryptic digests extends to residue 571, corresponding with the carboxyl-terminal end of the predicted helix, although other enzymes can cut within the predicted ␣-helix in limit digests (35). It is tempting to speculate that the vulnerability of the F domain to proteolytic attack, and the subsequent alterations in the response of the hER␣ to E 2 and 4-OHT, may play a role in breast cancer progression and response to antihormone therapy.
Overall, our results provide strong support for the idea that the F domain of the hER␣ contains an ␣Ϫhelix, "helix 13." They also show that mutations of this region alter not only the response to E 2 , the affinity for hormone, and the interaction between receptor subunits, but they can uncouple the agonist and antagonist activities of 4-OHT as well. These results suggest that the F domain modulates the activity of the estrogen receptor-␣ through multiple mechanisms.