Different residues of the human estrogen receptor are involved in the recognition of structurally diverse estrogens and antiestrogens.

We have previously examined, by alanine scanning mutagenesis, amino acids 515-535 of the estrogen receptor (ER) ligand binding domain to determine which of these residues are important in estradiol binding. Mutation at four sites that potentially lie along one face of an α-helix, Gly521, His524, Leu525, and Met528, all significantly impaired estradiol binding by the ER (Ekena, K., Weis, K. E., Katzenellenbogen, J. A., and Katzenellenbogen, B. S. (1996) J. Biol. Chem. 271, 20053-20059). In this report, we compare the pattern of residues that are important in the recognition of several structurally diverse estrogen agonists and antagonists (the synthetic nonsteroidal agonist hexestrol, an agonist derived from the mold metabolite zearalenone, P1496, and the partial agonist-antagonist trans-hydroxytamoxifen) with those that are predicted to contact estradiol in the receptor-ligand complex. Although there are some similarities in the pattern of residue recognition among all four ligands, each ligand showed distinct differences as well. Interestingly, alanine substitution at only one residue, the leucine at position 525, was found to inhibit binding of all the ligands tested. Another residue, His524, was found to be important in the recognition of three different agonists but not trans-hydroxytamoxifen (the only ligand lacking a second hydroxyl group). The recognition of estradiol and another agonist, P1496, was impaired by the G521A mutation, whereas ligand-induced activity by the two compounds that lack B- and C-rings, hexestrol and trans-hydroxytamoxifen, was unaffected. Our findings demonstrate that these ligands fit into the ER ligand binding pocket differently and that each contacts a distinct set of amino acids. The smaller ligands (estradiol and hexestrol) have a narrower footprint of interacting residues than the larger ligands (P1496 and trans-hydroxytamoxifen). This pattern of interaction is most consistent with the amino acids within this region being in contact with the portion of these ligands that corresponds to the D-ring end of estradiol. The interplay between the shape of an ER ligand and the residues that support its binding to ER may potentially underlie the selective actions of different ER ligands in various cell and promoter contexts.

Estrogens strongly influence the growth, differentiation, and functioning of the reproductive system, including the mammary gland and the uterus (1,2). The effect of estrogen on the mammary gland is of particular interest because of its link to breast cancer; the proliferation and metastatic activity of nearly 40% of breast cancers are stimulated by estrogens (3).
Estrogens exert their effects by acting through a ligandactivated transcription factor, the estrogen receptor (ER). 1 A member of a superfamily of nuclear receptors, the estrogen receptor contains a highly conserved DNA-binding domain, domain C, and a highly conserved ligand-binding domain, domain E. In addition to its ligand binding activity, the E domain also possesses dimerization activity and a hormone-dependent activation function (AF-2). A hormone-independent activation function (AF-1) is located within the receptor's A/B domain. In the presence of ligand, the ER bound to an estrogen response element (ERE) can either activate or suppress transcription of a downstream target gene in a cell-and promoter-specific manner (4 -11).
Estrogen antagonists such as tamoxifen have been successfully used in the treatment of ER-positive breast cancers (3,12,13). These antagonists compete with estradiol for binding to the receptor but fail to elicit transcriptional activity. Unfortunately, tamoxifen treatment has several drawbacks. First, tamoxifen is not a complete antagonist and has agonistic effects in certain cell and promoter contexts. Second, breast cancer cells often become resistant to this treatment (14). Finally, despite the need to inhibit ER activity in breast cancer cells, the retention of some ER activity is important in a number of other biological processes, including bone maintenance, the cardiovascular system, and liver metabolism (2). Thus, it becomes important to understand how the ER recognizes and interprets ligands of different structure, that is, which amino acids are involved in their binding, and thereby how these ligands might result in different activities in various cell types. Such information could then be used to design new estrogen agonists and antagonists with target cell-and response-selective biological effects.
We and others have identified regions of the ER involved in recognition of ligand (15)(16)(17)(18). We have found amino acids near Cys 530 to be important in estradiol binding, whereas residues C-terminal to position 535 have been shown to contain AF-2 activity (19 -21), and residues 507-514 may contain dimerization activity (20,22). Recently, we used alanine scanning mutagenesis to identify those amino acids in the 515-535 region of the ER ligand binding domain important in estradiol binding (23). Mutation of the amino acids Gly 521 , His 524 , Leu 525 , and Met 528 to alanine affected the ability of the ER to bind estradiol and had a parallel effect on estrogen-dependent transcription without affecting the overall activity of the receptor. These results, taken together with structural predictions based on crystal structures of other ligand-bound nuclear receptors, indicated that these four amino acids of the ER may lie along one face of an ␣-helix and are presumed to be in contact with the bound estradiol.
In this report, we extend those studies to include four other estrogen agonists and antagonists with varied structures: the synthetic nonsteroidal agonist meso-hexestrol; P1496, an agonist derived from the mold metabolite zearalenone; the partial agonist-antagonist trans-hydroxytamoxifen; and the pure antagonist ICI. Interestingly, we observe that different amino acids in the 515-535 region of the ER are important in the recognition of these different ligands. Only one amino acid (Leu 525 ) was found to be important in binding all of the ligands. By comparing the amino acids important in binding each of the ligands and comparing the ER amino acid sequence with other nuclear receptors and published crystal structures, we suggest which portion of these ligands are in contact with this region of the ER ligand binding pocket.
Cell Culture and Transfections-All transfections were performed using the ER-negative human breast cancer cell line MDA-MB-231. Cells were maintained as described previously (10) and were transfected using the calcium phosphate method (27). Cells were incubated in 5% CO 2 for 40 -48 h prior to transfection with 0.1 g of ER expression plasmid, 2.0 or 5.0 g of (ERE) 2 -pS2-CAT reporter plasmid, 0.8 g of pCMV␤ internal control ␤-galactosidase plasmid, and pTZ19R to 15 g of total DNA/100-mm plate. After incubation of cells with DNA for 4 h, cells were glycerol shocked for 2.5 min with 20% glycerol in growth medium and washed with Hanks' balanced salt solution for 2.5 min. Cells were then treated with ligand in growth medium, harvested after 24 h, and lysed by cycles of freezing on dry ice and thawing at 37°C. ER activity was determined by CAT assay of whole cell lysates and was normalized to ␤-galactosidase activity as described previously (28).

Ligand-induced Responsiveness of Alanine-substituted Human Estrogen Receptor Mutants to Estrogen Agonists and Antagonists-
The mutant receptors used in this study, which consist of individual alanine substitutions from amino acid 515 to amino acid 535 of the ER ligand binding domain, were described by us (23) in a study in which their ability to bind estradiol and activate transcription in response to estradiol was analyzed. The responsiveness of these mutant receptors to two different estrogen agonists, hexestrol and P1496, the partial agonist/antagonist, trans-hydroxytamoxifen (TOT), and a pure antagonist, ICI, was investigated in this study. The structures of these ligands are shown in Fig. 1. The ability of these ligands, all of which have high affinity for the ER (29 -31), to induce or antagonize transcriptional activity of the mutant receptors was monitored in the ER-negative breast cancer cell line MDA-MB-231 using an estrogen-responsive reporter gene construct, (ERE) 2 -pS2-CAT, in transient cotransfection experiments. As expected for wild type ER, both hexestrol and P1496 fully induced transactivation to near estradiol-induced levels ( Fig. 2A). TOT induced activity to only 10 -15% that of estradiol, and it antagonized estradiol-induced activity down to its own level of agonist activity (ϳ10 -15%; Fig. 2). ICI, as expected, behaved as a pure antagonist (Fig. 2B).
To assess the importance of the individual amino acids from positions 515-535 in the recognition of these receptor ligands, the mutant ERs were initially screened at a single concentration of each ligand that resulted in near maximal stimulation of wild type (WT) ER. Mutants whose activity differed significantly from wild type at this concentration were then subsequently tested over a range of ligand concentrations to determine whether the mutant receptors were dose-shifted in their response to ligand, suggesting a decreased affinity for ligand, or if they were defective in the ability to achieve maximal transcriptional activity. For all of these receptors, activity was low in the absence of ligand, as reported previously (23).
Hexestrol-The structure of the synthetic nonsteroidal estrogen hexestrol ( Fig. 1) differs from estradiol in that it is symmetrical and lacks a formal B-or C-ring. The mutant ERs were initially screened at 1 ϫ 10 Ϫ10 M hexestrol, which gave 75% of maximal activity with WT ER. As shown in Fig. 3A, alanine substitutions at only two positions, His 524 and Leu 525 , resulted in large reductions in hexestrol-induced transactivation (Ͻ5% of WT activity). The R515A, M528A, and P535A receptors exhibited modest reductions in activity (60 -80% of WT). Dose response curves for the WT, H524A, L525A, and P535A receptors using 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ6 M hexestrol, are shown in Fig.  3B. The most impaired mutant, H524A, was dose-shifted 1000fold, whereas L525A and P535A were shifted 100ϫ and 10ϫ, respectively. The R515A and M528A receptors were also doseshifted slightly, requiring slightly higher levels of hexestrol to reach wild type maximal activity (data not shown). Though His 524 and Leu 525 appear to be important in both estradiol and hexestrol binding by ER (Table I), alanine substitution at position 524 had a much greater effect on hexestrol-induced activation. Interestingly, the G521A mutation, which was found to significantly affect estradiol binding (Table I), had very little effect on hexestrol-induced transactivation (Fig. 3C).
P1496 -P1496, a derivative of the estrogenic mold metabolite zearalenone, has a 14-membered ring resorcylic acid lactone structure that is significantly larger than either estradiol or hexestrol (Fig. 1). Nevertheless, it acts as a potent inducer of ER activity, stimulating ER transactivation to nearly the same level as estradiol (Fig. 2). Like hexestrol, 1 ϫ 10 Ϫ10 M P1496 gave near maximal (80%) stimulation of ER and was therefore used to screen the 21 alanine-substituted ER mutants. Al-though mutation to alanine at only two positions markedly impaired hexestrol-induced activity (Fig. 3A), half of the ER mutants were substantially impaired for P1496-induced activity (Fig. 4A). Using higher concentrations of ligand demonstrated that all of these mutant receptors could reach wild type maximal transactivation at higher concentrations of ligand (Fig. 4, B and C, and data not shown). Like estradiol induction, the positions where mutation to alanine most greatly affected P1496-induced transactivation were Gly 521 , His 524 , Leu 525 , and Met 528 . However, in addition, mutation at residues 515, 522, 526, and 531-535 also had a significant effect on P1496-induced activity. Fig. 4 (B and C) shows the P1496 dose response curves for the R515A, G521A, M522A, H524A, L525A, and M528A receptors. Of particular note is the observation that the M528A mutation affects receptor response to P1496 more than this mutation affects response to any of the other tested li- Transfected cells were then treated with hexestrol for 24 h before preparation of extracts. CAT activities were normalized to ␤-galactosidase activity and are expressed relative to wild type ER activity (100 -200-fold stimulation in different experiments), which is set at 100%. A, bars represent the level of hexestrolinduced activity for all the mutant receptors at 1 ϫ 10 Ϫ10 M ligand. Wild type ER activity at this concentration is set to 100%. B, dose response curves using 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ6 M hexestrol, showing that the H524A, L525A, and P535A receptors are dose-shifted 10 -1000-fold relative to wild type ER in their response to hexestrol. C, the hexestrol dose response curve for the G521A receptor is nearly indistinguishable from that of wild type ER. All values represent the mean and standard deviation from two or more experiments. For some values, error bars are too small to be seen. gands; 100-fold shift for P1496 versus Ͻ10-fold shift for all other ligands.
trans-Hydroxytamoxifen-TOT is an estrogen antagonist with partial agonistic activity in certain cell types, including MDA-MB-231 cells (Fig. 2). To investigate the ability of the mutant ERs to interpret TOT as an agonist, transient transfections were performed as with hexestrol and P1496, except that more reporter plasmid, 5 g rather than 2 g, was used in order to enhance the agonistic effects. The results of screening the mutant receptors with 1 ϫ 10 Ϫ9 M TOT are shown in Fig. 5A and show that like P1496, about half of the mutant receptors were impaired for activity. However, unlike the P1496-induced response, TOT dose response analysis revealed two different classes of mutants. As observed before, many of the mutants that were impaired for activity at the initial concentration of TOT were able to reach wild type levels of activity in the presence of higher concentrations of ligand ( Fig. 5B and data not shown). However, another class of mutants, which consisted of K520A, M522A, N532A, and P535A (and indicated with light shading in Fig. 5A), all exhibited reduced maximal activity and did not appear to be dose-shifted ( Fig. 5C and data not shown). Rather, the decreases in activity observed at 1 ϫ 10 Ϫ9 M TOT correlated with their reduced maximal activities. Western blot analysis of these receptors demonstrated that they were present at levels at least as great as that of the wild type protein (data not shown), indicating that their reduced activity was not attributable to reduced levels of these receptors.
Antagonists-The antagonistic affects of two compounds, TOT and ICI, were examined to determine whether any of the alanine-substituted receptors failed to recognize these ER ligands as antagonists. For these experiments, cells were treated with both estradiol and a 10-fold excess of either TOT or ICI. For most of the mutant receptors, 1 ϫ 10 Ϫ9 M estradiol with 1 ϫ 10 Ϫ8 M antagonist was used. However, for G521A, H524A, and L525A, where 1 ϫ 10 Ϫ9 M estradiol does not induce high levels of transactivation (23), 1 ϫ 10 Ϫ8 M (for G521A and H524A) and 1 ϫ 10 Ϫ7 M (for L525A) estradiol were used with a correspond-ing 10-fold higher concentration of the antagonist. In all cases, both TOT and ICI were effective antagonists, suppressing estradiol-induced transactivation by greater than 85% (data not shown). DISCUSSION We have probed the interaction that various ligands have with a portion of the ligand binding domain of the human estrogen receptor (ER) by alanine scanning mutagenesis, studying in addition to estradiol the behavior of three high affinity ligands: two nonsteroidal agonists, the simple, symmetrical synthetic ligand meso-hexestrol, and P1496, a derivative of the fungal produced estrogen zearalenone, as well as the partial antagonist, trans-hydroxytamoxifen, the potent metabolite of tamoxifen. As in our earlier study with estradiol, we found that the substitution of certain residues in this ER region to alanine caused a significant to marked reduction in the potency of these ligands in inducing transcriptional activity. In addition, the principal finding from this study is that each of the ligands displayed a distinct footprint of interacting resi- a Open circles represent mutant receptors that have decreased maximal activity.  Fig. 3 except that the compound P1496 was used instead of hexestrol. A, bars represent the level of P1496-induced activity for all the mutant receptors at 1 ϫ 10 Ϫ10 M ligand. Wild type ER activity at this concentration is set to 100%. B and C, dose response curves using 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ6 M P1496, for the wild type, R515A, M522A, L525A, G521A, H524A, and M528A receptors. The R515A, M522A, and L525A receptors are dose-shifted 10 -20-fold relative to wild type ER in their response to P1496. The M528, G521A, and H524A receptors are shifted 50-, 400-, and 1000-fold, respectively. All values represent the means and standard deviations of two or more experiments. For some values, error bars are too small to be seen. dues in the alanine scanning mutagenesis, indicating that the structural differences in these ligands result in different patterns of interaction with amino acids of the ER (Table I).
We have confirmed that the alterations in ligand-induced transcriptional activity of the mutant receptors correspond to a dose shift (i.e. change in potency) in the transactivation response. Previously, with estradiol, we established that the decrease in transcription activation potency correlated well with a decreased binding affinity of estradiol to the mutant receptors (23). Thus, these interacting residues are considered to be ones that are important in determining the binding affinity of these ligands.
The effects of mutational change at certain sites were of particular interest. First, the L525A mutation was the only mutation that strongly affected the transcriptional potency of all four ligands (Table I). However, P1496, the bulkiest of the ligands, was least affected, possibly because it may best be able to compensate for the reduced bulk in ER that results from the Leu to Ala mutation. Second, the H524A mutation had a very strong effect on hexestrol and P1496-induced ER activity and a major effect on estradiol but no effect on TOT. It is of note that TOT is the only ligand that lacks a second hydroxyl group with which this histidine may be interacting. Third, the G521A substitution strongly affected estradiol and P1496 transactivation yet had only a modest effect on TOT and no effect on hexestrol. As discussed later, the smaller size of certain portions of the hexestrol and TOT structures, compared with estradiol and P1496, might enable them to tolerate the increased residue size resulting from the G521A substitution. Fourth, the K520A, M522A, N532A, and P535A substitutions had an effect on TOT induced transactivation but did so by reducing the maximal transcriptional activity of the mutant ER-TOT complex rather than by shifting the dose response curve. This decreased transactivation potential suggests that these residues do not involve the binding interaction between TOT and ER but rather affect the manner in which the liganded complex interacts with other components important for activating transcription. It is interesting that certain specific contacts between TOT and ER appear to affect only its balance of agonist and antagonist activity (such as residues 520, 522, and 535), whereas others affect only its binding affinity (such as residues 515, 516, and 525).
From our previous study on the effect of alanine scanning mutagenesis on the binding of estradiol and ligand-induced transactivation of the ER (23), we showed that the residues most affected by alanine substitution, Gly 521 , His 524 , Leu 525 , and Met 528 , could be displayed on three adjacent turns on one face of an ␣-helix. By sequence comparison with the human retinoic acid receptor-␥ (RAR) and the rat thyroid hormone receptor-␣1 (TR), these sites lie on the inner face of helix-11 and correspond to residues in the RAR-retinoic acid (RA) and TR-thyroid hormone (T 3 ) structures that are in close contact with the bound ligands (32,33). In the case of RAR-RA, these residues are in contact with the apolar end of the ligand, the ␤-ionone portion, and in the case of TR-T 3 , these residues contact the distal phenolic ring, with His 381 in TR (corresponding to Gly 521 in ER), making a hydrogen bond to the distal phenolic hydroxyl group.
In this study, we find that the pattern of residues in the 515-535 region of ER that affect ligand-induced transcription differs depending on the structure of the activating ligand. A comparison of the structures of these ligands (Fig. 1) together with their interacting residues (Table I) allows one to formulate a model for the basic orientation of these ligands within the binding cavity of the ER (Fig. 6). By analogy with the TR-T 3 and RAR-RA structures, this model shows the 515-535 region of ER as a helical region (corresponding to a portion of helix-10 and all of helix-11) and the beginning of the loop region between helix-11 and -12. The residues that affect the transcriptional potency of each of the four ligands are indicated schematically with circles. The helix and loop of ER and the structures of the ligands are presented on the same dimensional scale to illustrate the differential interaction that these four ligands have with the ER.
The four activating ligands, estradiol, hexestrol, P1496, and TOT, all have one common structural feature, a phenol, yet they show different patterns of residues whose substitution with alanine affects their transcriptional potency. These differences suggest that the ligands are not making contact with the 515-535 portion of the ER through their structurally common A-ring like feature. The rather narrow footprint of residues that affect hexestrol recognition (principally two residues, His 524 and Leu 525 , on one turn of the helix) suggests that  Fig. 3 except that the compound TOT was used instead of hexestrol. A, bars represent the level of TOT-induced activity for all the mutant receptors at 1 ϫ 10 Ϫ10 M ligand. Wild type ER activity at this concentration is set to 100%. Lightly shaded bars represent mutant receptors that exhibited reduced maximal activity in the presence of higher concentrations of TOT. All other receptors reached wild type levels of activity with higher levels of TOT. B, dose response curves using 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ6 M TOT, showing that the R515A, L525A, and G521A receptors are dose-shifted 5-50-fold relative to wild type ER in their response to hexestrol. The H524A receptor is not dose-shifted relative to wild type ER. C, dose response curves using 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ6 M TOT, showing that the K520A, M522A, and N532A receptors are not dose-shifted but possess a reduced ability to transactivate. All values represent the mean and standard deviation of two or more experiments. For some values, error bars are too small to be seen. contact is being made with a portion of the hexestrol structure that is narrower than the corresponding region in estradiol (whose potency is affected principally by three residues, Gly 521 , His 524 , and Leu 525 , on two helical turns). In contrast, the much wider footprint of residues that affect P1496 binding (four or more residues on at least three turns) suggests that their interaction is with a region of the ligand that is larger than the corresponding region in estradiol. Thus, if the phenol rings of these three ligands are oriented similarly and away from the 515-535 region, then it is the D-ring end of estradiol and the corresponding narrower region of hexestrol and the wider region of P1496 that are likely to be in contact with this region of the ER.
The pattern of residues that affect the transcriptional potency of TOT is rather unique. The interaction of TOT with the middle of the helical region is limited. In particular, its lack of interaction with His 524 suggests that in lacking a second hydroxyl group, TOT is indifferent to the hydrogen bonding characteristics of the residue at this site. On the other hand, TOT is the only ligand whose transcriptional potency is affected by residues near the N terminus of the examined region (Arg 515 and His 516 ), suggesting that its side chain may extend to be in contact with this region of the receptor.
Altogether, the comparison of ligand structure with the pattern of residues where mutation affects their transcriptional potency suggests that selected residues in the 515-535 region of the receptor are in contact with a portion of these ligands that corresponds to the D-ring end of estradiol. Thus, a rough comparison can be made between the orientation of these estrogens in ER and the RA and T 3 ligands in RAR and TR, respectively. It is the D-ring end of estradiol that corresponds to the apolar end of RA and the phenol of T 3 that is in contact with the helix-11 region of these nuclear receptors. The A-ring or phenolic portion of the estrogen ligands, which corresponds to the polar end of the RA and T 3 ligands, is directed away from helix-11.
It is well recognized now that the ER and related nuclear receptors show transcriptional activity that is modulated by the nature of the particular gene promoter and the cellular background, consistent with ligand-receptor interaction with FIG. 6. A schematic representation of the interaction of four ligands, estradiol (A), hexestrol (B), P1496 (C), and trans-hydroxytamoxifen (D) with the 515-535 region of the ER. By analogy with the TR-T 3 and RAR-RA crystal structures, this region of ER is displayed as an ␣-helix representing the C-terminal end of helix-10 (residues 515-520) and helix-11 (residues 521-531) and the beginning of a loop between helices-11 and -12 (residues 532-535). The four ligands, shown on the same dimensional scale as the ER sequence, are oriented with their phenolic function away from the helix. In each case, those residues whose mutation to alanine have a major effect on the transcriptional potency of the ligand are indicated with a circle or series of circles about the residue position: one circle corresponds to two dots in Table I, two circles corresponds to three dots, and three circles corresponds to four dots. (Those residues in ER whose mutation to alanine affects the transcriptional effectiveness but not the potency of TOT, namely 520, 522, 532, and 535, are not displayed.) cell-and promoter-specific factors and transcriptional coactivators (5,34). The interplay between the shape of an ER ligand and the residues that support its binding to ER, as studied here, may potentially underlie the selective actions of different ER ligands in various target cells and promoter contexts.