Identification of amino acids in the hormone binding domain of the human estrogen receptor important in estrogen binding.

The initial step in the regulation of the transcriptional activity of the estrogen receptor (ER) is the binding of hormone. Previous studies have suggested that the region between amino acids 515 and 535 near the C terminus of the human ER is likely to be important in ligand binding. In order to explicitly define which amino acids in this region are critical for ligand recognition and binding, we have utilized alanine-scanning mutagenesis over the complete 515-535 region. The ability of these 21 mutants to activate transcription in response to the natural estrogen, 17β-estradiol (E2), was evaluated in cell co-transfection assays with estrogen-responsive reporter genes. In addition, their ability to bind E2 was also tested. Mutations at four sites in the 521-528 region had the greatest effects on E2-induced transcription, with L525A reducing responsiveness 250-fold, G521A and H524A 35-fold, and M528A 11-fold. Mutations at other sites had either no effect or a 4-fold or lesser reduction in sensitivity to E2 (M517A, Y526A, N532A, and P535A). Three of the mutants most affected in their transcriptional response, G521A, H524A, and M528A, showed a coordinate reduction in E2 binding affinity. E2 binding by the most affected mutant, L525A, could not be detected. Thus, the altered transcriptional response of these ER mutants appears to derive solely from an alteration in their affinity for the ligand E2. The four sites most affected by alanine substitution, 521, 524, 525, and 528, follow an α-helical periodicity, such that they would be positioned on one face of an α-helix. Furthermore, they correspond precisely to residues in an α-helix shown to be in contact with ligand in the recently described x-ray crystal structures of two other members of the nuclear hormone receptor superfamily, namely the retinoic acid receptor- and thyroid hormone receptor-ligand complexes. Our findings, which broaden observations to the steroid receptor family within the superfamily of nuclear receptors, suggest that this region of the estrogen receptor is in contact with its cognate ligand in a similar fashion.

The initial step in the regulation of the transcriptional activity of the estrogen receptor (ER) is the binding of hormone. Previous studies have suggested that the region between amino acids 515 and 535 near the C terminus of the human ER is likely to be important in ligand binding. In order to explicitly define which amino acids in this region are critical for ligand recognition and binding, we have utilized alanine-scanning mutagenesis over the complete 515-535 region. The ability of these 21 mutants to activate transcription in response to the natural estrogen, 17␤-estradiol (E 2 ), was evaluated in cell co-transfection assays with estrogen-responsive reporter genes. In addition, their ability to bind E 2 was also tested. Mutations at four sites in the 521-528 region had the greatest effects on E 2 -induced transcription, with L525A reducing responsiveness 250-fold, G521A and H524A 35-fold, and M528A 11-fold. Mutations at other sites had either no effect or a 4-fold or lesser reduction in sensitivity to E 2 (M517A, Y526A, N532A, and P535A). Three of the mutants most affected in their transcriptional response, G521A, H524A, and M528A, showed a coordinate reduction in E 2 binding affinity. E 2 binding by the most affected mutant, L525A, could not be detected. Thus, the altered transcriptional response of these ER mutants appears to derive solely from an alteration in their affinity for the ligand E 2 . The four sites most affected by alanine substitution, 521, 524, 525, and 528, follow an ␣-helical periodicity, such that they would be positioned on one face of an ␣-helix. Furthermore, they correspond precisely to residues in an ␣-helix shown to be in contact with ligand in the recently described x-ray crystal structures of two other members of the nuclear hormone receptor superfamily, namely the retinoic acid receptor-and thyroid hormone receptorligand complexes. Our findings, which broaden observations to the steroid receptor family within the superfamily of nuclear receptors, suggest that this region of the estrogen receptor is in contact with its cognate ligand in a similar fashion.
The estrogen receptor (ER) 1 is a ligand-dependent transcrip-tional regulator that belongs to a superfamily of proteins that includes the receptors for steroid hormones as well as those for retinoic acid, thyroid hormone, and vitamin D (1,2). In the absence of ligand, the ER resides within the cell nucleus and is transcriptionally inactive. Binding of the hormone estrogen is believed to cause the ER to undergo a conformational shift into a transcriptionally active form. The ligand-occupied ER, which binds to estrogen response elements (EREs) upstream of a promoter, is then able to interact with other cellular components to either activate or suppress transcription of a target gene in a promoter-and cell-specific manner (2)(3)(4)(5)(6)(7)(8).
The ER is of particular interest because of the important activities that its ligands have on the development and function of tissues of the reproductive system. In addition, estrogens, acting through the ER, also exert crucial actions in other tissues, including the pituitary, hypothalamus, bone, liver, and cardiovascular system (9,10). The beneficial effects of estrogens in bone maintenance, on blood lipid profile, and in the cardiovascular system, account for their importance in hormone replacement in postmenopausal women and underscore the role of ER function in these systems. Estrogen, however, also stimulates the growth of about 40% of breast cancers, specifically those hormone-responsive tumors containing significant levels of the ER protein (11). Thus, it is important to understand in detail how the ER recognizes and binds estrogen and to identify regions of the protein that are in contact with the hormone.
The steroid receptors are composed of six domains based on amino acid similarity and mutational analysis (Refs. 1 and 2 and references therein). The A/B domain at the N termini of these proteins contains a hormone-independent transcriptional activation function (AF-1). DNA binding occurs through a helix-loop-helix motif in the highly conserved C domain. The hinge or D domain is a hypervariable region, while the E/F domain is a conserved ligand-binding domain (LBD). In addition to ligand binding, the E/F domain also possesses dimerization and transcriptional activation function (AF-2) activities. Considerable effort has been directed at elucidating the amino acids important in each function of the E domain.
Previous work has suggested that several residues in the region of amino acids 515-535 of the LBD of the human ER play a critical role in ligand binding. Our studies with affinitylabeling ligands indicated that estradiol binding involves amino acids close to cysteine 530 (12), and a triple mutation that altered the charge of this region, K529Q,K531Q,N532D, resulted in a receptor having reduced affinity for estradiol but unaltered affinity for antiestrogens (13,14). Work in other laboratories has pointed to additional amino acids in this re-gion of the receptor being important for ligand binding. For instance, mutating the glycine at position 525 of the mouse ER (Gly 521 of the human ER) to arginine nearly completely abolished E 2 binding and activity, and deleting residues 521 and 522 (517 and 518 of the human ER) was shown to reduce E 2 binding by 200-fold (15).
Other experiments have demonstrated that regions more C-terminal and N-terminal in the LBD play important roles in hormone-dependent transactivation and receptor dimerization, respectively. A predicted amphipathic ␣-helix spanning residues 534 -548 of the E domain appears to be critical in AF-2 function (16 -18), and amino acids 507-518 of the mouse ER (503-514 of the human ER) have been implicated as being important in the dimerization of the receptor (17,19).
In order to explicitly define which amino acid residues from positions 515-535 of the estrogen receptor are critical for ligand recognition and binding, we have utilized alanine-scanning mutagenesis (20). Alanine substitution is a more conservative mutation than has previously been utilized in the limited analyses in this region, and it has enabled us to assess the individual contributions of each of these 21 amino acids to the transcriptional activity induced by the natural ligand estradiol and the binding of this ligand. With this approach, we identified four amino acids that are of particular importance in ligand binding. Interestingly, their positioning suggests that these four residues lie along one face of an ␣-helix in the ER LBD and are likely to make contact with ligand. These results are interpreted in light of the recently published crystal structures of LBD-ligand complexes from two related receptors of the nuclear receptor superfamily (21)(22)(23).

EXPERIMENTAL PROCEDURES
Plasmids and General Reagents-The plasmids 2ERE-pS2-CAT (24), pCMV5hER (16), and pCMV␤ (Clonetech, Palo Alto, CA) have been previously described. The vector pTZ19R was kindly provided by Dr. Byron Kemper (University of Illinois, Urbana, IL) (25), and pBluescript II SK ϩ was from Stratagene (La Jolla, CA). Plasmids were purified for transfection using either CsCl gradient centrifugation or a plasmid preparation kit according to the manufacturer's instructions (Qiagen, Chatsworth, CA). Restriction enzymes were purchased from Life Technologies, Inc. and New England Biolabs (Beverly, MA). Cell culture media, calf serum, and other reagents for cell culture were purchased from Life Technologies, Inc. and Sigma. For Western analysis, nitrocellulose membrane was obtained from Millipore Corp. (Marlborough, MA), the H226 antibody was kindly provided by Dr. Geoffrey Greene (University of Chicago), and rabbit anti-rat IgG was purchased from Zymed (San Francisco, CA). Radioisotopes for chloramphenicol acetyltransferase (CAT) assays, sequencing, hormone binding assays, and Western blotting were purchased from DuPont NEN and Amersham Corp. All other reagents were purchased from Sigma, Fisher (Pittsburgh, PA), and Amersham.
Oligomer-directed Mutagenesis-The 1.8-kilobase ER-containing BamHI fragment from pCMV5hER was cloned into the BamHI site of pBluescript II SK ϩ . Site-directed mutagenesis was then performed according to Kunkel et al. (26) using the oligonucleotides listed in Table  I, generating the listed ER alanine substitutions. Oligonucleotides were purchased from Life Technologies, Inc. Screening for the desired ER mutations was done by restriction enzyme analysis (Table I). Following mutagenesis, the ER cDNAs were excised from pBluescript II SK ϩ using BamHI and ligated into the BamHI site of the cytomegalovirusdriven expression vector, pCMV5, kindly provided by Dr. David Russell (University of Texas, Dallas, TX) (27). All ER mutations were then confirmed by dideoxy sequence analysis using a Sequenase 2.0 kit (U.S. Biochemical Corp./Amersham).
Cell Culture and Transfections-All transfections were done in ERnegative human breast cancer MDA-MB-231 cells. Cells were maintained and transfected as described previously (7). Cells were plated for transfection at a density of 3 ϫ 10 6 cells/100-mm dish and incubated for 40 -48 h at 37°C with 5% CO 2 . Transfections were performed using 2.0 g of the reporter plasmid 2ERE-pS2-CAT, 0.8 g of the ␤-galactosidase reporter plasmid pCMV␤, 0.1 g of ER expression vector, and pTZ19R carrier plasmid to 15 g of total DNA/100-mm diameter dish. Cells were incubated with calcium phosphate-precipitated DNA for 4 h and then subjected to a 2.5-min glycerol shock, using 20% glycerol in growth medium, followed by a 2.5 min rinse in Hanks' balanced salt solution and ligand treatment in growth medium. Cells were harvested 24 h after ligand treatment and lysed by three cycles of freezing on dry ice and thawing at 37°C. ER transactivation ability was determined by CAT activity of the whole cell lysates and assayed as described previously (28). CAT assays were normalized to ␤-galactosidase activity from the co-transfected pCMV␤ plasmid.
Western Analysis-231 cells were transfected in 100-mm dishes with 10 g of ER expression vector, 0.8 g of pCMV␤, and pTZ19R carrier plasmid to 15 g of total DNA. Cells were then treated with hormone or ethanol vehicle and incubated for 24 h before harvesting in cold HBSS. The cells were centrifuged at 200 ϫ g for 5 min and resuspended in 20 mM Tris (pH 7.4), 0.5 M NaCl, 1.0 mM dithiothreitol, 10% glycerol (v/v), 50 g/ml leupeptin, 50 g/ml aprotinin, 2.5 g/ml pepstatin, and 0.2 mM phenylmethylsulfonyl fluoride. Whole cell extracts were obtained by subjecting cells to three rounds of freezing on dry ice and thawing on wet ice followed by centrifugation at 15,000 ϫ g to remove cell debris. Cell extracts for expression studies were normalized to ␤-galactosidase activity, from the co-transfected pCMV␤ reporter, before loading equivalent ␤-galactosidase units on a 10% SDS-polyacrylamide gel (29). Electrophoresis and Western blotting were done according to standard methods (30). Nitrocellulose blots were probed with the human ERspecific primary antibody H226 at 2.0 g/ml and then incubated with rabbit anti-rat IgG (1 g/ml), and detected with 125 I-conjugated protein A.
Hormone Binding Assays-Binding assays for estradiol and Scatchard analysis were performed as described previously (28). 231 cells were transfected and whole cell extracts were prepared as for Western blot analysis. Cell extracts were then incubated with varying concentrations of [ 3 H]estradiol (1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ7 M [ 3 H]estradiol) in the presence or absence of a 100-fold excess of radioinert estradiol to determine nonspecific and total binding, respectively. Whole cell extracts and ligand were incubated together at 4°C overnight, and unbound estradiol was removed from the samples by treatment with dextrantreated charcoal for 15 min at 4°C. Approximately 1.0 g of total protein was assayed at each concentration of hormone. Equilibrium dissociation constants (K d ) for the wild type and mutant ERs were determined by Scatchard analysis (31).

Screen for Alanine-substituted Mutant Estrogen Receptors
with Altered Transcriptional Activity-Alanine mutants in the ligand binding domain of the human estrogen receptor (hER; amino acids 515-535) were created by oligomer-directed mutagenesis of the ER cDNA (see Table I). The presence of the correct mutations was confirmed by restriction enzyme analysis and sequencing. The ability of each of the 21 mutant ERs to transactivate an ERE-dependent gene was first determined in a simple screening assay ( Fig. 1 and Table II). Human breast cancer MDA-MB-231 cells, which lack endogenous ER, were cotransfected with one of the ER expression vectors together with an (ERE) 2 -pS2-CAT reporter, and CAT activity in cells was monitored after treatment with a single concentration of estradiol (Fig. 1). For this screening we used 1 ϫ 10 Ϫ9 M E 2 and 100 ng of ER expression vector, since wild type ER reached near maximal activity at this concentration of E 2 , and under these conditions the level of activation was independent of the amount of transfected ER DNA over the range of 50 -400 ng (data not shown). CAT activity was very low in all receptors treated with control 0.1% ethanol vehicle and was induced 100 -200-fold over control for wild type ER by the addition of E 2 . Fig. 1 shows the transcriptional activity of each of the mutants. Alanine substitutions N-terminal to position 520 and C-terminal to position 529 showed little to no effect on the ability of the ER to activate transcription, although slight reductions in transactivation were observed for M517A, S518A, N532A, and P535A. As expected from previous work, changing the cysteine at position 530 to alanine had no effect (28). Larger effects on ER activity were observed when alanine was substituted at certain sites in the region spanning amino acids 521-528. Particularly affected were amino acids Gly 521 , His 524 , and Leu 525 , with the latter mutant having almost no activity at 1 ϫ 10 Ϫ9 M E 2 . A lesser but still significant reduction in activity was also observed for M528A.
Western Blot Analysis-By substituting alanine, a relatively conservative amino acid change, for each amino acid across this 21-amino acid region of the hER, we hoped to avoid changes in receptor activity that might be caused by global structural alterations in the protein that might affect its overall stability. In order to demonstrate that the mutant receptors were expressed as full-length stable proteins, Western blot analyses were performed. Cells were transfected with ER expression vector and treated with either ethanol control vehicle or hormone. ER protein was detected using H226 antibody, which recognizes an epitope in the B domain of the ER, well away from the altered region. With the exceptions of K520A and L525A, all of the mutant receptors were expressed at levels at least as great as that of wild type ER, both in the presence and absence of estradiol ( Fig. 2; M517A, S518A, M522A, and N532A were also tested and found to be present at wild type levels (data not shown)).
Surprisingly, despite its decreased transcriptional activity, the G521A ER mutant was found to be present at higher levels than wild type, both in the presence and absence of estradiol (Fig. 2, compare lanes 1 and 2 with lanes 5 and 6). Conversely, both K520A and L525A were present at low levels in the absence of estradiol, although their levels increased somewhat (for K520A) or to wild type levels (for L525A) in the presence of 1 ϫ 10 Ϫ7 M estradiol (Fig. 2, lanes 3 and 4 and lanes 9 and 10). Increases in protein levels of K520A and L525A were also observed with 1 ϫ 10 Ϫ8 M E 2 (data not shown). This result indicates that the L525A protein is more stable in the presence of estradiol, in contrast to wild type ER (Fig. 2).
Estradiol Dose-Response Curves for the Mutant Receptors-Since our goal was to identify residues that are important in estrogen receptor ligand binding and recognition, we measured the transcriptional activity of the mutant receptors over a range of E 2 concentrations, from 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ7 M. These dose-response curves enabled us to determine whether mutants, which exhibited reduced activity at the single E 2 concentration tested in Fig. 1, were dose-shifted in their response to E 2 or were defective in their activation function. For wild type ER, maximal activity was reached at an E 2 concentration between 1 ϫ 10 Ϫ9 M and 1 ϫ 10 Ϫ8 M, and half-maximal activity (EC 50 ) was reached at about 1 ϫ 10 Ϫ10 M E 2 (Fig. 3A). For several of the mutant receptors that exhibited wild type or near wild type activity at 1 ϫ 10 Ϫ9 M E 2 , dose-response curves were not significantly different from wild type ER (K520A) or were only slightly altered (Fig. 3, A and B, and data not shown).
All of the mutations that resulted in significantly reduced activity of receptor at 1 ϫ 10 Ϫ9 M E 2 , G521A, H524A, L525A, and M528A, proved to be dose-shifted in their response to E 2 , requiring elevated levels of E 2 relative to wild type ER in order   to reach half-maximal and maximal activity (Fig. 3, C and D). Of these four mutants, the M528A dose-response was shifted the least, while the L525A ER was the most defective in E 2 response. L525A required 1 ϫ 10 Ϫ7 M or more estradiol to reach wild type activity, and its dose-response curve was shifted to the right more than 250-fold (Fig. 3D). Both G521A and H524A were dose-shifted nearly 35-fold (Fig. 3C). As might be expected for mutations that cause a defect only in affinity for ligand, each mutant receptor achieved wild type or near wild type activity in the presence of sufficiently high concentrations of estradiol. Over the concentration ranges we studied, none of the 21 mutant receptors exhibited higher activity than wild type ER at any E 2 concentration ( Fig. 3

and data not shown).
Binding of Estradiol by the Mutant Receptors-The affinities of the mutant estrogen receptors for estradiol were determined by saturation binding and Scatchard analysis. Cells were transfected with ER expression vector, and cell extracts were incubated with 1 ϫ 10 Ϫ12 M to 1 ϫ 10 Ϫ8 M [ 3 H]estradiol in the presence or absence of excess radioinert estradiol. The K d values are given in Table II and represent the average of three independent experiments. Our observed K d of 0.12 nM for wild type ER is in close agreement with previously published values (32) and with the level of E 2 necessary to achieve half-maximal stimulation in our transient transfection assays (Fig. 3A). The dissociation constants calculated for the mutant receptors (K d values) correlated well with their predicted binding based on the estradiol dose-response curves in the co-transfection assay (EC 50 values), as shown in Fig. 4. M528A, with a K d of 0.45 nM, was the least impaired in hormone binding and in sensitivity to E 2 -induced transactivation, followed by G521A and H524A, which had K d values of 0.78 nM and 1.40 nM, respectively. These values represent 4 -12-fold decreases in hormone binding and 10 -35-fold decreases in sensitivity to E 2 -induced transactivation, compared with wild type ER.
We were unable to obtain a K d value for the L525A mutant for two reasons. First, the protein was unstable in the absence of hormone, limiting the quantity of receptor present in cell extracts used in the hormone binding assays (Fig. 2). Second, L525A bound estradiol extremely poorly. Considering the fact that transactivation response of L525A is dose-shifted by a factor of nearly 250, it is not surprising that we experienced difficulty with this mutant in our hormone binding experiments. Dose-response data suggest that the K d for estradiol binding by this mutant is at least 100-fold greater than for wild type ER (i.e. Ͼ10 nM). DISCUSSION The initial step in the regulation of the transcriptional activity by the estrogen receptor is the binding of ligand. Therefore, determining which residues of the ER are involved in ligand binding will be essential in elucidating the mechanism of receptor activity. Our findings, using alanine-scanning mutagenesis, suggest that amino acids in the region 521-528 of the human ER are intimately involved in the recognition and binding of hormone and also provide evidence that this region of the ER is ␣-helical in its ligand-occupied form.
Specifically, we have identified four amino acid residues, Gly 521 , His 524 , Leu 525 , and Met 528 , that when mutated to alanine, resulted in a significantly reduced sensitivity of the ER to E 2 -stimulated transcription activation. However, despite requiring higher levels of E 2 to elicit a response, each of the mutant receptors reached wild type activity at sufficiently high E 2 concentrations. This suggests that the mutant ERs are impaired in terms of hormone binding rather than AF-2 transactivation function or DNA binding. Substitution of alanine for other amino acids in the region from 515 to 535 of the ER affected transactivation to a lesser degree or not at all.
Directly testing hormone binding by the mutant receptors confirmed that they were impaired for E 2 binding. The calculated K d values correlated well with the shifts observed in the E 2 dose-response transactivation experiments (EC 50 values). This correlation provides strong support that this region plays a direct and crucial role in hormone binding. In contrast, changes in helix 12, the adjacent, more C-terminal portion of a Transactivation of the mutant receptors relative to wild type at 1 ϫ 10 Ϫ9 M E 2 , which is set at 100%. b Effective concentration of E 2 required for the receptor to reach half-maximal activity in the transactivation assay. Values are obtained from E 2 dose-response data and are only given where full dose response assays were done.
c Where no value is given, K d was not determined. d -, Hormone binding assay performed, but specific binding by the L525A receptor protein was too low to obtain an accurate K d value.
the LBD, affect transactivation but not hormone binding by the receptor (16,18).
Other laboratories have previously examined some mutations in this 515-535 region of the ER. For instance, Danielian et al. (15) showed that changing the methionine at position 532 of the mouse ER (Met 528 in hER) to an arginine impaired hormone binding to the same degree as our M528A mutation, increasing its K d value for estradiol 3.7-fold. Interestingly, these two very different amino acid substitutions, small hydrophobic versus large negatively charged, at the position corresponding to hER Met 528 , affected the ER to a similarly modest extent, suggesting that this methionine plays an important, although not critical, role in estradiol binding.
The Gly 521 residue has also been previously identified as potentially important in ligand binding. Changing the corresponding residue in the mouse ER (Gly 525 ) to arginine has been shown to render mER incapable of binding E 2 or of activating transcription even in the presence of 1 ϫ 10 Ϫ6 M E 2 (15,17). For this reason, it has been assumed that this glycine plays a critical role either in the structure of the ligand binding domain or in hormone binding directly. Our more conservative alanine substitution had a much less severe effect on ER activity than the mER G525R mutation; the G521A mutant exhibited only a 50 -60% reduction in activity at 1 ϫ 10 Ϫ9 M E 2 . Furthermore, the affinity of our G521A mutant receptor for E 2 was only 6.5-fold lower than that of wild type ER. Surprisingly, when comparing the levels of ER protein in transfected cells, G521A was found to be considerably more abundant than wild type ER, both in the presence and absence of E 2 . This G521A substitution must therefore alter the structure of the ER in some way that makes the protein more resistant to degradation, possibly by stabilizing an ␣-helix.
Our findings can be interpreted in the context of three recently reported x-ray crystallographic structures of the ligand binding domains of other members of the nuclear receptor superfamily, namely the LBD of the human retinoid X receptor ␣ (hRXR␣) without ligand and the LBD-ligand complexes of the human retinoic acid receptor ␥ (hRAR␥) and rat thyroid hormone receptor-␣ 1 (rTR␣ 1 ) (21)(22)(23). Although these three receptors all have nonsteroidal ligands and have relatively modest amino acid sequence identity and similarity to the hER, it is anticipated that the LBDs of all nuclear hormone receptors will share a similar tertiary structure (33). A sequence alignment encompassing the hER region 515-535 is given in Fig. 5, together with an annotation of secondary structure elements described in the three reports and an identification of the residues in hRAR␥ and rTR␣ 1 that are reported to be in contact with the ligand.
Residues Gly 521 , His 524 , Leu 525 , and Met 528 of the hER, all correspond to positions identified as ligand contact residues in the hRAR␥ structure, Gly 393 , Arg 396 , Ala 397 , and Leu 400 . Positions 521 and 528 of the hER also correspond to ligand contact sites in the rTR␣ 1 structure, residues His 381 and Met 388 . This correlation highlights a basic similarity with which at least a portion of these nuclear receptors interact with their ligands, whether steroidal or not, and suggests that Gly 521 , His 524 , Leu 525 , and maybe Met 528 are in fact ligand contact sites in the hER.
It is interesting to note the location of these presumed ligand FIG. 2. Western immunoblot analysis of wild type and mutant ER expression levels. Extracts were prepared from MDA-MB-231 cells transfected with 10 g of the indicated ER expression plasmid and treated for 24 h with either ethanol vehicle (Ϫestradiol) or 1 ϫ 10 Ϫ7 M E 2 (ϩestradiol). Extracts were prepared and separated by SDS-PAGE as described under "Experimental Procedures." Approximately 100 g of total cell extract was loaded per lane. The 66-kDa ER protein (denoted by an arrow) was detected using the anti-ER H226 antibody.

FIG. 3. Analysis of E 2 -induced transactivation by mutant ERs
relative to wild type receptor. MDA-MB-231 cells were cotransfected with 2ERE-pS2-CAT reporter, pCMV␤ internal control, and the indicated ER expression plasmid. Transfected cells were treated with 1 ϫ 10 Ϫ12 to 1 ϫ 10 Ϫ7 M E 2 for 24 h before preparation of extracts. CAT activities were normalized to ␤-galactosidase activity and are expressed relative to maximal wild type ER activity (greater than 100-fold stimulation), which is set at 100%. A, the E 2 dose response of K520A is nearly identical to that of wild type ER. B, M517A, Y526A, and N532A are slightly shifted to the right in their E 2 dose response. C and D, mutant ERs, G521R, H524A, L525A, and M528A, are 10 -250-fold less sensitive to E 2 -induced stimulation. All values represent the mean and standard deviation from two or more experiments. For some values, error bars are too small to be seen. contact residues in the hER within the context of the LBD secondary and tertiary structure shown in the crystal structures. By analogy to the hRAR␥ and rTR␣ 1 ligand-receptor complexes, the hER residue positions 521, 524, 525, and 528 lie in an ␣-helical region. In the helical face map of the residues in the 515-535 region of the hER (shown in Fig. 6), it is clear that the three residues most affected by mutation, Gly 521 , His 524 , and Leu 525 , are arranged as a compact unit on one face of two successive turns of a putative ␣-helix. Met 528 , which is of somewhat lesser importance in hormone binding and transactivation, is also located in this same facial region, but on the next helical turn. Alanine substitution of certain residues further removed from this facial region, Met 517 , Ser 518 , Tyr 526 , Asn 532 , and Pro 535 , have a modest but detectable effect on E 2 -induced transcriptional activity of the receptor (Fig. 1) and may likely be affecting ligand binding, although to a lesser degree. Interestingly, all of these residues, except one (Tyr 526 ), are on the same face of the ␣-helix as Gly 521 , His 524 , and Leu 525 .
In any case, given the helical periodicity of the sites where mutation to alanine most affects ligand binding, and the homology of this region of the hER with ␣-helical regions in hRAR␥ and rTR␣ 1 , it seems most likely that the 521-528 region of the hER will adopt an ␣-helical secondary structure in the hER-estradiol complex. However, not all of these residues may be in a helix in the absence of ligand. Residues 521-525 of the hER correspond to a loop region between ␣-helices 10 and 11 in the unliganded hRXR␣ structure (21). Thus, a conformational change may occur in the hER 521-525 region upon ligand binding that might involve an increase in the ␣-helical character in the liganded state compared with the unliganded state. A definitive determination of such changes will require direct structural studies on these two states of the ER itself.
If one accepts that the hER Gly 521 , His 524 , Leu 525 , and Met 528 residues are in contact with the ligand in the hERestradiol complex, it is interesting to speculate what portion of the ligand is making contact with these residues. The analogous residues in ␣-helix 11 in the hRAR␥ and rTR␣ 1 ligand complexes are in contact with the ␤-ionone ring of all-transretinoic acid and the phenol portion of the thyroid hormone derivatives, respectively (22,23). The latter interaction, in particular, suggests that the phenolic A-ring of estradiol may be in contact with these residues. It may be that His 524 in the hER is the residue that is in contact with the 3-hydroxyl group of estradiol. Given this interaction, one would presume that the other residues, Gly 521 , Leu 525 , and Met 528 , would be in contact FIG. 4. Correlation between E 2 -induced transcriptional activation and E 2 binding for wild type and mutant ERs. The K d values for E 2 binding and the concentration of E 2 required to induce halfmaximal transcriptional activity (EC 50 ) are displayed in a log-log plot. The linearity of this plot indicates that E 2 binding affinity to the various ERs and ER sensitivity to E 2 -induced transcriptional activation are correlated.
FIG. 5. Amino acid residues required for E 2 binding by the hER correspond to residues of RAR and TR that have been shown to contact ligand. A sequence alignment of hER with hRAR␥, rTR␣1, and hRXR␣ (taken from Ref. 33) is shown along with representations of secondary ␣-helical structure as revealed by x-ray crystallography. Shaded residues in the hER sequence identify amino acids that we have shown are important in E 2 binding by the ER. These residues correspond exactly with ligand contact sites identified in the hRAR␥ and rTR␣1 crystal structures (boxed residues). Note that amino acids 521-525 of the hER correspond to a loop region between helices 10 and 11 of the unliganded hRXR␣ crystal structure but that this region is ␣-helical in the ligand-bound forms of hRAR␥ and rTR␣1. with other regions of estradiol that are within or near the A-ring.
It will be informative to discover whether the changes that affect E 2 binding by the ER also affect the binding of other estrogens or antiestrogens, or whether different amino acids are involved in interactions with ligands of different structures.