Unique Protein Determinants of the Subtype-selective Ligand Responses of the Estrogen Receptors (ER a and ER b ) at AP-1 Sites*

The two subtypes of human estrogen receptor, a (hER a ) and b (hER b ), regulate transcription at an AP-1 response element differently in response to estradiol and the anti-estrogens tamoxifen and raloxifene. To bet-ter understand the protein determinants of these differences, chimeric and deletional mutants of the N-termi-nal domain and the F region of ER a and ER b were made and tested in transient transfection assays at the classical estrogen response element (ERE) site as well as at an AP-1 site. Although the same regions on each receptor subtype appeared to be primarily responsible for estradiol activation at an ERE and in HeLa cells, major differences between ER a and ER b mutants were seen in the estrogen and anti-estrogen responses at an AP-1 site. This differential ligand response maps to the N-terminal domain and the F region. These results suggest that different estrogenic and anti-estrogenic ligands use different mechanisms of activation and inhibition at the AP-1 site. In contrast to previous studies, this work also shows that many of subtype-specific responses are not transferred to the other subtype by swapping the N-terminal domain of the receptor. This implies that there are other unique surfaces presented by each subtype outside of the N-terminal domain, and these surfaces can play a role in subtype-selective signaling. Together, these data suggest a complex interface between ligand, response element, and receptor that underlies ligand activation in estrogen signaling pathways. The estrogen receptors and

The estrogen receptors (ER␣ and ER␤) 1 are members of a large family of nuclear receptors that activate or repress the transcription of hormone-regulated genes upon binding to a ligand (1). One feature of the nuclear receptor family is that a receptor can both activate and repress different sets of genes in response to the same ligand, but the mechanisms behind these differential effects are still not well understood. Estrogen receptor is unusual among the nuclear receptors, because its differential regulatory effects manifest themselves as tissue-specific responses to a given ligand (2). For example, tamoxifen functions as an anti-estrogen in breast tissue, but acts as an estrogen in the uterus and bone. Controlling these tissue-specific effects is the ultimate goal in the design and study of selective estrogen receptor modulators for the treatment of diseases such as breast cancer and osteoporosis (3).
One explanation for the different effects is that a ligand may elicit different responses when the receptor acts through different effector sites (4). The estrogen receptor regulates transcription through binding to estrogen response elements (EREs) in the upstream promoter regions of target genes as well as through interactions with a growing number of "nonclassical" response sites (5,6). These nonclassical sites do not necessarily require DNA-protein interactions between the receptor and the promoter element, but instead regulate transcription through protein-protein interactions between the receptor and other transcription factors such as AP-1 and Sp-1. By receptor interactions with different response elements, the same ligand could then cause both activation and repression of different sets of genes.
The estrogen receptor exists in two different forms, ER␣ and ER␤ (7). It has since been shown that the response to both estrogens and anti-estrogens at an AP-1 site depends on the subtype of the receptor (8); estradiol elicits transcriptional activation with ER␣, but transcriptional repression with ER␤. The two ER subtypes also respond differently to the selective estrogen receptor modulator raloxifene at an AP-1 site; ER␤ shows much stronger activation in response to raloxifene than ER␣. Subtype-selective activities have also been seen at other classical and nonclassical estrogen response elements (6, 9 -14). In an effort to understand the mechanism of estrogen receptor action at the AP-1 site and nuclear receptor action at nonclassical response elements in general, this study describes efforts to identify regions in ER␣ and ER␤ that can account for the differential ligand activation at an AP-1 site. We report that different polypeptide regions on ER␣ and ER␤ in both the N-and C-terminal regions are important for the differential ligand response.

EXPERIMENTAL PROCEDURES
Plasmids-The construction of the expression vectors for both hER␣ (HE0) and hER␤ as well as the AP-1-regulated luciferase construct (Coll73-luc) have been described previously (5,8). The ERE-driven luciferase reporter gene consists of two repeats of the upstream region of the vitellogenin ERE promoter from Ϫ331 to Ϫ289, followed by region Ϫ109 to ϩ45 of the thymidine kinase upstream region and the luciferase gene. The chimeric ER constructs were made with overlap extension PCR and inserted into SG-5 expression vectors. While a point mutant of ER␣ that has a lower hormone-independent response was used in these studies (HE0), constructs were also made using wild-type ER␣ (HEG0), and no significant difference in the activation profiles was seen (data not shown). The N-terminal chimera ␤␣␣, replacing amino acids 1-178 of the ER point mutant HE0 with amino acids 1-96 of hER␤, was made by amplifying a fragment of corresponding to amino acids 179 -476 of HE0 using primers with the sequences 5Ј-AGTCAGTGAGCGAG-GAAGCG (this primer will hereby be known as primer RW1) and 5Ј-CTATGCTTCAGGATATCATTATGGAGTC. The fragment was then digested and inserted into the EcoRV and BamHI sites in the pSG5-hER␤ expression vector.
Chimera ␣␤␤, replacing amino acids 1-96 of hER␤ with amino acids 1-178 of HE0, was made by PCR overlap extension. A SacII site in the N-terminal domain of HE0 was eliminated by first amplifying one fragment with primers with the sequence 5Ј-GATCCCGCGGATGAC-CATGACCCTCCACACC (primer RW2) and 5Ј-CGCGTTGGCGGCG-GCCGCCGCGTTGAACTCGTAG and the other fragment with primers with the sequence 5Ј-CTACGAGTTCAACGCGGCGGCCGCCGCCAAC-GCG and 5Ј-GATCGATATCCTGAAGCATAGTCATTGCACAC (primer RW3). The overlap extension was then performed using primers RW2 and RW3 and then digesting and inserting the resulting fragment into the SacII and EcoRV sites of the pSG5-hERb expression vector. The ␣␤␤ chimera deletion mutant ␣(⌬129 -178)␤␤, which deletes amino acids 130 -177 in the N-terminal domain of ␣␤␤, was made using the same procedure but using the plasmid ⌬129 -178 (15) as the PCR template. The ␣␤␤ chimera deletion mutants ␣(⌬109)␤␤ and ␣(⌬117)␤␤, which deletes amino acids 2-108 and amino acids 2-116, respectively, were made by using primer RW2 and RW3 with the templates n109 and n117, respectively (15), and inserting the fragments into the SacII and EcoRV sites in the pSG5 hER␤ expression vector.
To make the N-terminal deletion mutant ⌬␤␤, which deletes amino acids 2-96 of hER␤, a PCR fragment of ER␤ was generated using the primer RW1 and a primer with the sequence 5Ј-AGGGATCCGCGGAT-GTGCGCTGTCTGCAGCG. The PCR product was then digested and inserted into the SacII and BamHI sites of the pSG5-hER␤ construct. Five other N-terminal ER␤ deletion mutants, ⌬9␤␤, ⌬17␤␤, ⌬21␤␤, ⌬28␤␤, and ⌬53␤␤, were made in the same fashion with primer RW1 and the primer containing the appropriate truncation.
The F region chimera ␤F␣, swapping amino acids 451-477 of hER␤ with amino acids 553-595 of HE0, was made by PCR overlap extension. One fragment was amplified from HE0 using the RW1 primer and a primer with the sequence 5Ј-AGCCGTGGAGGGGCAT. The other fragment was amplified from ER␤ using primers with the sequence 5Ј-AAGAGCTGCCAGGCCTGCCG (primer RW4) and 5Ј-ATGCCCCTC-CACGGCTCTTGCACCCGCGAAG. The overlap extension was then performed using primers RW1 and RW4 and then digesting and inserting the resulting fragment into the SacI and BamHI sites of the pSG5-hER␤ expression vector.
The F region chimera ␣F␤, swapping amino acids 553-595 of HE0 for amino acids 451-477 of hER␤, was made by first inserting a silent mutation to insert a SpeI restriction site into HE0 via Quikchange mutagenesis (Stratagene, La Jolla, CA) using primers with the sequence 5Ј-GCGCCCACTAGTCGTGGAG and 5Ј-CTCCACGACTAGT-GGGCGC. A PCR fragment was then amplified from ER␤ using the RW1 primer and a primer with the sequence 5Ј-GAACTAGTTCCAT-CACGGGGTC. The fragment was then digested and inserted into the BamHI site and the newly generated SpeI site of HE0.
The F domain mutant ␣⌬F, deleting amino acids 553-595 of HE0, was made by PCR amplification of HE0 using primer RW4 and a primer with the sequence 5Ј-ATGGGATCCTCAAGTGGGCGCATGTAGGC. The fragment was digested and inserted into the HindIII and BamHI sites of the HE0 expression vector. The other deletion mutant, ␤⌬F, deleting amino acids 451-477 of hER␤, was made by PCR amplification of ER␤ with primer RW4 and a primer with the sequence 5Ј-ATGG-GATCCTCACTTGCACCCGCGAAG. The fragment was then digested and inserted into the SacI and BamHI sites of the pSG5-hER␤ expression vector.
Tissue Culture, Transfection, and Luciferase Assays-HeLa cells were grown in 0.1 m filtered Dulbecco's modified Eagle's medium supplemented with 4.5 g/liter glucose, 0.876 g/liter glutamine, 100 mg/liter streptomycin sulfate, 100 units/ml penicillin G and 10% newborn calf serum. Cells were grown to a density of not more that 5 ϫ 10 4 cells/cm 2 . For transient transfection assays, cells were suspended in 0.5 ml of electroporation buffer in 0.4-cm gap electroporation cuvettes at ϳ1.5 ϫ 10 6 cells/cuvette with 5 g of the reporter plasmid and the optimal amount of the receptor expression vector. The optimal receptor plasmid concentration for maximal ligand activation was determined for each mutant and was found to be 5 g of plasmid per transfection except for the ␣␤␤ and ␤␣␣ chimera, which required only 1 g expression plasmid per transfection. The electroporation buffer consisted of 0.2 m filtered PBS, 0.1% glucose, and 0.001% Biobrene detergent. Cells were transfected by electroporation at a potential of 0.25 kV and a capacitance of 960 microfarads. The transfected cells were immedi-ately resuspended in growth medium supplemented as described above with the exception that the newborn calf serum had been treated with charcoal. Cells were plated into six-well dishes at 2 ml/well at a density ϳ1 ϫ 10 5 cells/well. After 2 h of incubation at 37°C, hormones were added in 2 l of ethanol (8).
After 48 h of incubation at 37°C, the cells were lysed by first removing the medium from the wells, washing with PBS, and then adding 0.2 ml of lysis buffer consisting of 100 mM potassium phosphate (pH 7.5), 0.2% Triton X-100, and 1 mM dithiothreitol. The plates were frozen at Ϫ80°C, thawed, and scraped with a rubber policeman to loosen the cell fragments. The lysate was centrifuged for 5 min, and 0.1 ml of the supernatant was combined with 0.3 ml of the luciferase assay buffer consisting of 25 mM glycylglycine, 15 mM MgSO 4 , 4 mM EGTA, 15 mM potassium phosphate (pH 7.8) with the addition to a final concentration of 1 mM dithiothreitol, 2 mM ATP, and 0.2 mM luciferin. Luminescence was measured for 10 s with a Monolight 3010 luminometer (Analytical Luminescence Laboratory, San Diego, CA). Each hormone dose was performed in triplicate, and the relative error was determined by calculating the S.E. of the mean. Each construct was tested at least five times, and no significant differences in the results were observed between experiments.
Ligand Binding Assay and Data Normalization-Transfection efficiency and receptor expression levels were tested using a whole cell ligand binding assay (16). After each transfection with the reporter and expression plasmids described above, a portion of the transfected cell suspension was plated into four wells of a 24-well plate at a density of 10 5 cells/well. The cells were grown as described above for ϳ12 h, then the medium was removed, the cells were washed once with PBS and then treated with 200 l of medium minus the newborn calf serum and containing 20 nM [2,4,6,7,16,17-3 H]estradiol. To two of the four wells, diethylstilbestrol was added to a concentration of 10 M. The cells were then incubated at 37°C for approximately 1 h. The medium was removed, the cells were washed three times with 0.5 ml of ice-cold PBS and then extracted twice with ethanol. The ethanol extractions were than diluted in scintillation fluid and counted for activity. Specific binding to the receptor was calculated by subtracting nonspecific binding (measured from the cells treated with diethylstilbestrol) from total binding (measured from the cells not treated with diethylstilbestrol). This assay has been repeated with each construct at least five times, and the specific binding has differed less than 15% between experiments.
A normalization factor for each transfection was then determined by dividing the amount of specific ligand binding for each transfect by the specific ligand binding of a transfection standard performed with every set of transfections using 5 g of HE0 expression plasmid. The relative light unit values and errors determined from the luminescence experiments performed on each transfection were then divided by the normalization factor to give the final, normalized luminescence data.

RESULTS
The N-terminal Domain Was Not Required at ERE-tk Promoter in HeLa Cells-Deletional and chimeric mutants of ER␣ and ER␤ were made to determine the regions of the ER responsible for the different ligand effects at an AP-1 site. Polypeptide regions that had the highest sequence variability between the two subtypes were chosen. The estrogen receptor, like all nuclear receptors, has three structural domains, an N-terminal domain, a DNA binding domain (DBD) and a ligand binding domain (LBD). Of these three domains, only the N-terminal domain shows significant differences between ER␣ and ER␤ in both sequence identity and length. There have been reports of longer forms of ER␤ that extend the N terminus of the receptor (17). These longer forms of ER␤ also possess very low sequence similarity with ER␣ in the N-terminal domain and show no difference in activity at the AP-1 site (15).
Chimeras and deletions of the N-terminal domain of both subtypes were constructed (Fig. 1a) and tested in a transient transfection reporter gene assay with a reporter construct consisting of a luciferase gene driven by a classical ERE (5). To account for potential differences in the expression levels of each chimeric receptor, a whole cell estradiol binding assay was also performed with every transfection (16). The estradiol binding data were used to normalize the luciferase reporter activity. This normalization corrects for differences in transfection effi-ciency and also serves as a control to assay ligand binding activity of the various mutant receptors used in the study. All of the chimeric and deletion mutants showed less than a 2-fold variation in the amount of specific tritiated estradiol binding except for the ER␤ construct containing the N-terminal domain of ER␣ (␣␤␤), which showed 8-fold increased binding activity (Table I). All the mutant receptors also showed maximal ligand activation in the reporter gene assay at the same receptor expression plasmid concentration except for the ␣␤␤ and ␤␣␣ chimeric receptors, which had an optimal concentration 5-fold lower than the others.
In HeLa cells, swapping or deleting the N-terminal domain causes some changes in the magnitude of activation, but all of the mutants were still capable of activating transcription in response to estradiol at an ERE-tk promoter in HeLa cells (Fig.  2). Truncations in amino acids 21-96 of the N-terminal domain of ER␤ or in different regions in the N-terminal domain of the ␣␤␤ chimera also had no effect on estradiol activation at an ERE-driven promoter (data not shown). As expected, little or no activation is seen with tamoxifen or raloxifene at any of the full-length or mutant receptors. The lack of tamoxifen activation at this ERE in HeLa cells is consistent with other reports (5, 18 -20).
The C-terminal Tail Was Not Required at an ERE Site in HeLa Cells-The other region of the receptor that shows great differences between ER␣ and ER␤ is the C-terminal tail (also known as the F region). The F regions of ER␣ and ER␤ share relatively low sequence identity (23%) compared with the rest of the LBD and DBD. In addition, ER␣ also has a longer F region than ER␤ (42 residues for ER␣ versus 26 for ER␤). Chimeras and deletion mutants were also made for the two subtypes (Fig. 1b) and tested in the same transient transfection assays described above. Although there are some changes in the overall magnitude of the activation (Fig. 3), each of the mutants behaved like full-length ER␣ and ER␤ in their response to ligands: estradiol activated transcription, whereas the anti-estrogens raloxifene and tamoxifen did not. A higher level of hormone-independent activation was also seen with the F deletion mutants, but the underlying cause for this increase is still unknown.
Differences between ER␣ and ER␤ at an AP-1 Site-When the chimera and deletion mutants were tested at an AP-1 site, significant differences in ligand activation were seen. Deleting the N-terminal domain of ER␣ (⌬␣␣) or replacing it with the N-terminal domain of ER␤ (␤␣␣) resulted in receptor that showed no activation by estradiol, a weak activation by tamox-ifen, and a much stronger activation response to raloxifene than that of full-length ER␣ (Fig. 4). On the other hand, deletion of the N-terminal domain in ER␤ (⌬␤␤) abolished all ligand activation at an AP-1 site. Deleting the first 53, 28, or 21 amino acids in the N-terminal domain of ER␤ abolished all ligand activation at the AP-1 site as well (Fig. 5). The chimera of ER␤ with the N-terminal domain of ER␣ (␣␤␤) showed transcriptional activation in response to only tamoxifen at a level 3-fold higher than the hormone-independent activation but showed no activation by raloxifene compared with the 10-fold activation seen with full-length ER␤. Deleting either the first 109 or 117 amino acids of ␣␤␤ eliminated any tamoxifen activation but resulted in a 2.5-fold activation by raloxifene (Fig. 6). An internal deletion between amino acids 129 and 178

. Transient transfection assay with N-terminal deletion and chimera mutant receptors and the vitellogenin A2 ERE-tk driven luciferase reporter gene.
All ligand doses were 1 M. Relative light units were normalized for each construct using estradiol binding capacity as described under "Experimental Procedures."

FIG. 3. Transient transfection assay with C-terminal tail deletion and chimera mutant receptors and the vitellogenin A2
ERE-tk-driven luciferase reporter gene. All ligand doses were 1 M. Relative light units were normalized for each construct using estradiol binding capacity as described under "Experimental Procedures." in the N-terminal domain of ␣␤␤ also showed no activation by tamoxifen and a slight activation by raloxifene.
The F region chimeras and deletion mutants were also tested at an AP-1 site. As was seen with the N-terminal domain mutants, ER␣ and ER␤ appear to function differently (Fig. 7). Deleting the F region of ER␤ (␤⌬F) or replacing it with the F region of ER␣ (␤F␣) does not affect the ligand response of the receptor. However, removing the F region of ER␣ (␣⌬F) or replacing the F region with that from ER␤ (␣F␤) results in receptors that show little or no activity in response to estradiol at an AP-1 site but still allows a significant tamoxifen activation. The deletion mutant (␣⌬F) also shows significant activation by raloxifene in contrast to the full-length ER␣ receptor. DISCUSSION It has been previously shown that ER␣ and ER␤ have different ligand activation properties at an AP-1 site (8). Both estrogens and anti-estrogens stimulate transcriptional activation with ER␣ whereas with ER␤, anti-estrogens promote transcriptional activation, while estrogens promote repression. The aim of this study was to define the elements of each ER that are responsible for the differential responses at AP-1. We focused on the N-terminal (A/B) domain and the C-terminal F region of ER␣ and ER␤, since these regions are the most dissimilar in terms of length and sequence between the two ERs.
Of the three structural domains, the function of the N-terminal domain is the least understood, but is generally believed to play a role in transactivation and repression (19 -26), The activation region known as activation function 1 (AF-1) resides in the N-terminal domain of ER. The AF-1 region has been shown to contribute to ligand-independent activation and to synergistic enhancements of ligand-dependent activation with activation function 2 (AF-2), located in the ligand binding domain. Both the constitutive and synergistic effects attributed to the AF-1 region are highly dependent on cell context (20,23,26). While ER␣ possesses this AF-1 region between amino acid 41 and amino acids 120 -150 depending on the cell type, the equivalent AF-1 region in ER␤ is either extremely weak or absent entirely (27,28).
Comparisons of the role of the N-terminal domain of the two ER subtypes on subtype-selective responses have shown that the N-terminal domains of both ER␣ and ER␤ are important for signaling depending on the response element and cell type. Tamoxifen activates transcription in an ER␣-selective manner at certain ERE sites in a number of cell types (though not in HeLa cells, which was also confirmed here) (14,20,(22)(23)(24)26). The importance of the N-terminal domain in this response has been demonstrated by the elimination of the tamoxifen activation through deletion of the N-terminal domain. The tamoxifen activation could also be transferred to ER␤ by making a chimera of ER␤ that contained the N-terminal domain of ER␣ (11). The ER␣ selective response to estradiol at a Sp-1 site was also found to be sensitive to deletion of the N-terminal domain of ER␣ and the activation could be transferable to ER␤ by swapping the N-terminal domain (6,10). An ER␤-selective response to anti-estrogens was also seen with the RAR␣-1 promoter and this response could be significantly reduced by deleting the N-terminal domain of ER␤. In addition, the antiestrogen activation of the RAR␣ 1 promoter could also be transferred to ER␣ by swapping the N-terminal domain (10).
Subtype-selective responses are also seen with the AP-1 promoter, but the domain requirements for these selective responses are more complicated than with any subtype-selective response reported so far. Estradiol causes transcriptional acti- vation at an AP-1 site with ER␣ and transcriptional repression with ER␤. This estradiol activation by ER␣ requires the presence of the N-terminal domain as demonstrated by the absence of activation in the N-terminal deletion mutant (⌬␣␣). This would correlate with a previous report that suggests the estradiol activation at AP-1 is dependent on AF-1 (15). The chimeras which swapped the N-terminal domains of ER␣ and ER␤ (␣␤␤ and ␤␣␣) showed no activation to estradiol. The loss of estradiol activation with the ␤␣␣ construct could be rationalized by the absence of a strong AF-1 in the new chimeric receptor, but the absence of estradiol activation with ␣␤␤ was not expected. The ␣␤␤ chimera does contain the AF-1 region from ER␣, yet still does not activate transcription in the presence of estradiol. This suggests that there is a fundamental difference between ER␣ and ER␤ ligand activation at an AP-1 site that involves protein determinants other than or in addition to the N-terminal domain.
There is also a subtype-selective response to raloxifene at an AP-1 site; ER␤ is activated much more strongly than ER␣ in response to raloxifene. Previous work has suggested that the AF-1 in the N-terminal domain of ER␣ suppresses the raloxifene activation at the AP-1 site and that the raloxifene activity requires only the DNA binding domain and ligand binding domain (15). Deletion of the N-terminal domain of ER␣ converts a weak raloxifene activation into a very strong activation, the ER␣ chimera with the N-terminal domain of ER␤ (␤␣␣) also shows strong activation to raloxifene and the ER␤ chimera with the N-terminal domain of ER␣ (␣␤␤) has a significantly lowered raloxifene activation. However, raloxifene activation at an AP-1 site by ER␤ is eliminated when the N-terminal domain is deleted (⌬␤␤). This indicates that raloxifene activation of ER␤ at AP-1 requires a unique activation function located in the N-terminal domain.
In an attempt to identify the specific residues in the Nterminal domain responsible for the raloxifene activation with ER␤ and raloxifene suppression with ER␣, deletion mutants of ER␤ and ␣␤␤ were constructed. Deletion of the first 21, 25 or 53 amino acids of the N-terminal domain of ER␤ had no significant effect on estradiol activation at an ERE site, but it resulted in abolishment of activation by all ligands at an AP-1 driven promoter. This indicates that an activation region resides in the first 21 amino acids of the N-terminal domain. Attempts to precisely define this activation region were unsuccessful as deletions within the first 21 N-terminal amino acids led to transcriptionally inactive mutants which were also unable to bind estradiol.
Deletion mutants were also used to investigate the suppression of raloxifene activation at the AP-1 site by the N-terminal domain of ER␣. In a previous report, deleting the N-terminal 109 amino acids of ER␣, cutting into the middle of AF-1, caused the receptor to activate transcription in response to raloxifene at an AP-1 site (15). This is also seen here with the ␣(⌬109)␤␤ mutant, suggesting that the AF-1 region of ER␣ can suppress raloxifene activation by ER␤ as well. Deleting between amino acids 129 and 178 in ␣␤␤, which corresponds in ER␣ to a flanking region outside of AF-1 known as iAF-1B (15), showed reduced ligand activation as well, further emphasizing the importance of the AF-1 region in raloxifene repression.
Both ER␣ and ER␤ activate transcription in response to tamoxifen at the AP-1 site in HeLa cells, which is also seen with the regulation of the human quinone reductase gene (12), but the work reported here suggests the two subtypes activate transcription by different mechanisms. Previous work indicated that the tamoxifen activation of ER␣ at an AP-1 site was slightly repressed by AF-1 but required a region in the Nterminal domain outside of the AF-1 region for full activation (15). The relatively small tamoxifen activation with the Nterminal domain deletion mutant of ER␣ (⌬␣␣) seen here is consistent with this hypothesis. Also consistent with this hypothesis is the tamoxifen activation seen with the ER␤ chimera containing the N-terminal domain of ER␣ (␣␤␤). The ␣␤␤ chimera is particularly interesting compared with ER␣ and ER␤ because it is only activated by tamoxifen and not by estradiol or raloxifene. This is the first evidence for a mechanism for tamoxifen activation at an AP-1 site that is different from activation by estradiol. The loss of tamoxifen activation by the N-terminal deletion mutant of ER␤ (⌬␤␤) suggests that a similar activation function in the N-terminal domain of ER␤ exists that is necessary for tamoxifen activation. This activation function in the N-terminal domain of ER␤ is not interchangeable with the N-terminal activation function in ER␣ as evidenced by the lack of tamoxifen activation by the ␤␣␣ chimera. This is further emphasized by the ␣␤␤ deletion mutants. Deleting the first 109 amino acids in ER␣ removes over half the AF-1 region, but the receptor still shows tamoxifen activation at an AP-1 site (15). In contrast, the ␣(⌬109)␤␤ mutant shows no tamoxifen activation at AP-1. This implies that there are mechanistic differences in the response to tamoxifen between ER␣ and ER␤ at AP-1 sites.
The exact function of the F region is not known. Although it has been shown to be unnecessary for transcriptional activation (29,30) or receptor half-life regulation (31), it has been shown to be important in modulating the estrogen and antiestrogen response in some cell types by modulating both AF-1 and AF-2 (32). In addition, experiments with ER fusion proteins suggest that the F region possess different structural orientations depending on whether an agonist or an antagonist binds the receptor (33).
The ER␣-selective estradiol activation depends on the presence of the F region as demonstrated by the diminished estradiol activation in the ␣⌬F deletion mutant relative to the tamoxifen activation. The ER␣ chimera that contained the F region from ER␤ (␣F␤) also showed lowered estradiol activation. Consistent with previous studies on classical EREs (32), the drop in estradiol activity with the ␣⌬F construct can be explained by a decrease in the AF-1 and AF-2 activity of the receptor caused by the deletion of the F-region. However, the loss of activity with ␣F␤ was not expected. The ␣F␤ construct strongly activates transcription in the presence of estradiol at an ERE-tk site, demonstrating that this mutant is functional in terms of ligand binding and transactivation from a classical ERE. It is unclear whether the F-region of ER␣ contains a specific region necessary for the estradiol activation at an AP-1 site or if the AP-1 response is simply more sensitive than an ERE response to F-region attenuation in the AF-1 and AF-2 activity. As was seen with the ␣␤␤ chimera, tamoxifen activation by the ␣⌬F and ␣F␤ mutants is independent of estradiol activation, further emphasizing a difference between the mechanisms of tamoxifen and estradiol activation by ER␣ at the AP-1 site. At the AP-1 site, the F-regions of ER␣ and ER␤ do not appear to have unique roles in ER␤-selective raloxifene activation or tamoxifen activation by either receptor, since deleting the F-region of either ER␣ or ER␤ has no effect on tamoxifen activation.
From these chimera and deletion mutant studies, it is clear that the two subtypes of the estrogen receptor use different mechanisms to respond to different ligands at an AP-1 site. Such a marked signaling difference between two subtypes of a receptor suggests that the receptors may be designed by nature to have different roles in signaling from AP-1 sites. In contrast to other subtype-selective responses (6, 10, 11, 28), differences in the responses of ER␣ and ER␤ can not be explained solely by the presence of an AF-1 region in ER␣ that is absent in ER␤ because swapping N-terminal domains does not entirely swap the ligand response profile. It appears that there is a unique region in the N-terminal domain of ER␤ that is necessary for activation at an AP-1 site by anti-estrogens and a region in another part of the ER␤ protein that is responsible for preventing activation by estrogens at an AP-1 site.
In the case of the F region, it is known that the receptor undergoes a major structural change upon ligand binding that results in the reorganization of helix 12 in the LBD (34 -36). Because the F region is attached to helix 12, it is likely that the F region is susceptible to structural perturbations that could put it in contact with other regions of the receptor. Disturbing this interaction apparently disrupts activation by ER␣ in response to estrogens at both an ERE and an AP-1 site. The chemical extensions of anti-estrogens such as tamoxifen and raloxifene disrupt the reorganization of helix 12 and cause it to interact with a different part of the ligand binding domain (1,35,36). In this alternate conformation, changing the F-region by mutation or deletion apparently does disrupt activation by tamoxifen at an ERE-tk site but does not affect activation by anti-estrogens at an AP-1 site.
Differences in the mechanism of ER␣ and ER␤ activity and differences in the response of ER␣ to estrogens and anti-estrogens have also been suggested by peptide blocking studies using phage display (37,38). Peptides were identified that could specifically block estradiol activation at an ERE site by one ER subtype and not the other. Peptides that could specifically block ER␣ activation by estradiol or by tamoxifen were also reported. The peptides that were specific for blocking either estradiol or tamoxifen activation by ER␣ at an ERE site also blocked activation at an AP-1 site. Interestingly all of those peptides bound to regions in the ligand binding domain of the receptor. Perhaps some of these selective peptides are binding to surfaces on the LBD that communicate with other regions of the receptor, for example the N-terminal domain. Although direct protein-protein interactions between the N-terminal domain and the LBD have not been detected using two hybrid systems, 2 unique surfaces on the LBD of each subtype could interact indirectly through accessory proteins with surfaces on the N-terminal domain that are unique to each subtype. Regardless of mechanisms, this suggests a subtle and complex program of transcriptional regulation by the estrogen receptors.