Identification of Tamoxifen-Induced Coregulator Interaction Surfaces within the Ligand-Binding Domain of Estrogen Receptors

Abstract

Tamoxifen is a selective estrogen receptor (ER) modulator that is clinically used as an antagonist to treat estrogen-dependent breast cancers but displays unwanted agonistic effects in other tissues. Previous studies on ERα have delineated a role of the N-terminal activation function AF-1 in mediating the agonistic effects of tamoxifen, while the mechanisms for how ERβ mediates tamoxifen action remain to be elucidated. As peptides can be used to detect distinct receptor conformations and binding surfaces for coactivators and corepressors, we attempted in this study to identify previously unrecognized peptides that interact specifically with ERs in the presence of tamoxifen. We identified two distinct peptides among others that are highly selective for tamoxifen-bound ERα or ERβ. Domain mapping and mutation analysis suggest that these peptides recognize a novel tamoxifen-induced binding surface within the C-terminal ligand-binding domain that is distinct from the agonist-induced AF-2 surface. Peptide expression specifically inhibited transcriptional ER activity in response to tamoxifen, presumably by preventing the binding of endogenous coactivators. Moreover, tamoxifen-responsive and ER subtype-selective coactivators were engineered by replacing the LXXLL motifs in the coactivator TIF2 with either of the two peptides. Finally, our results indicate that related coactivators may act via the novel tamoxifen-induced binding surface, referred to as AF-T, allowing us to propose a revised model of tamoxifen agonism.

Both estrogen receptor (ER) subtypes, ERα and ERβ, are ligand-activated transcription factors. After binding estrogen, the receptors associate with specific estrogen response elements within the promoters of estrogen-regulated genes or affect the activity of other transcription factor complexes such as AP-1 (Jun-Fos). The two ER subtypes share affinity for the same ligands and DNA response elements. Like other members of the nuclear receptor family, the ERs consist of three distinct domains: an N-terminal domain, a DNA-binding domain (DBD) involved in DNA recognition and binding, and a C-terminal ligand-binding domain (LBD) (16). Transcriptional activation by ERα is mediated by two distinct activation functions, the constitutively active AF-1, located in the N terminus of the receptor, and a ligand-activated AF-2 in the C terminus. AF-1 activity is usually weaker than AF-2 activity, but the two activation domains function synergistically in ERα. In contrast, ERβ appears to have no significant AF-1 activity and thus depends entirely on the ligand-dependent AF-2 (33). Ligand binding to ERs induces conformational changes in the receptors (24) that are crucial for transforming ligand signaling into transcriptional responses. Distinct conformations enable the receptor to interact selectively with coregulatory proteins, coactivators and corepressors, which are necessary for regulating gene expression (16).

Structural studies have shown that agonist-bound ERs adopt a conformation in which LBD helices 3, 5, and 12 form a hydrophobic cleft, constituting AF-2, which represents the binding surface for α-helical leucine-rich peptide motifs, known as LXXLL motifs or NR boxes, in coactivators. While helix 12 is positioned over the ligand-binding cavity in the agonist formation, ER antagonists with a bulky side chain that protrudes from the ligand-binding pocket sterically prevent helix 12 from adopting the agonist-induced conformation. Instead, helix 12 binds to the hydrophobic AF-2 cleft with its own intrinsic NR box (1, 28), thereby preventing coactivator binding and presumably promoting the association of distinct coregulators involved in antagonist action. The existence of distinct ER conformations and protein binding surfaces has been clearly demonstrated in a number of recent studies that have identified specific ER binding peptides by phage display (18, 22). They revealed that multiple peptides could be categorized into different classes according to their ligand and ER subtype specificities. Importantly, ER-peptide binding studies not only provide crucial experimental tools for the high-throughput identification and characterization of novel ligands in vitro and in vivo, they also provide the opportunity to gain further insight into the specific determinants of ligand-dependent ER-coregulator interactions and thus into the molecular actions of diverse ER ligands.

Selective estrogen receptor modulators (SERMs) are synthetic ER ligands that display a tissue-selective pharmacology, opposing the action of estrogens in certain tissues while imitating their action in others (10). The receptor activity is dependent on the nature of the bound ligand, the ER subtype, the gene promoter elements, and the cellular levels of coregulatory factors (11). Tamoxifen is a SERM used clinically to treat ER-positive breast cancers. Antiestrogen therapy with tamoxifen is often initially efficient, but eventually most tumors become refractory to the antiestrogen treatment. In addition, tamoxifen has unwanted growth-promoting effects in the uterus, and furthermore, the breast tumors somehow switch from recognizing tamoxifen as an antagonist to seeing it as an agonist, as in uterine tissue. It is likely that variations in the pattern of coregulators might explain some of these tissue-specific and temporal changes in tamoxifen-ER function. There are indications that increased expression of the coactivator SRC-1 in combination with decreased levels of the corepressors N-CoR and SMRT could help to explain the stimulatory effects of tamoxifen (12, 26). Additional candidate coregulators that affect ERα activity in the presence of tamoxifen have been identified (3, 7, 17, 19, 25), yet the precise interaction requirements (e.g., ligand specificity, ER subtype specificity, and interaction domains and motifs) of these proteins need to be characterized. Moreover, in general, most studies on the agonistic activities of SERMs have focused on ERα, and very little is known about the involvement of ERβ.

To obtain insight into the differences between ERα and ERβ in the context of tamoxifen signaling, we attempted in this study to identify and characterize non-LXXLL peptides that specifically recognize tamoxifen-activated ERα or ERβ.

MATERIALS AND METHODS

Phage affinity selections.

Phage affinity selections were performed essentially as described by Paige et al. (22). Briefly, biotinylated oligonucleotides encoding the vitellogenin estrogen response element (ERE) were incubated in streptavidin-coated Immulon 4 96-well plates (Dynatech) to allow binding; 3 pmol of ERα or ERβ protein (Panvera Corp.) was incubated in each well for 1 h in TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% Tween 20). Affinity selections were performed in TBST or TBST containing 1 μM 17β-estradiol or 4-hydroxy-tamoxifen with multiple libraries of phages displaying unique peptide sequences. Essentially, M13 phage particles distributed among 18 libraries displaying a total of greater than 10¹⁰ different random or biased amino acid sequences were added to the wells containing immobilized target protein and incubated for 3 h at 25°C. Unbound phage was washed away, and the bound phage was eluted with pH 2 glycine. The eluted phage was amplified by infecting Escherichia coli cells. The amplified phage was then added to wells containing immobilized estrogen receptor target protein, and the cycle of affinity selection was repeated. Enrichment of phages displaying target-specific peptides was monitored after each round of affinity selection with an anti-M13 antibody conjugated to horseradish peroxidase in an enzyme-linked immunosorbent assay-type assay. Pools of phages enriched for target-specific peptides were plated for individual plaques. The plaques were picked, the phages were amplified, and the phages were tested for target-specific binding versus nonspecific binding to various control proteins. DNA was prepared from target-specific phage, the DNA sequence of the peptide-encoding region was determined, and the peptide sequence was deduced.

Chemicals.

17β-Estradiol and raloxifene were purchased from Sigma (St. Louis, Mo.); ICI182,780, methylpiperidinopyrazole (MPP), and 4-hydroxy-tamoxifen were gifts from KaroBio AB (Huddinge, Sweden).

Constructs.

Human ERα and ERβ cDNAs cloned into the VP16 expression vector (Clontech) were used as templates for mutagenesis. Site-directed mutations were introduced with the QuikChange XL site-directed mutagenesis kit (Stratagene) according to the manufacturer's protocol. Base changes are in boldface. ERαE380A was made by introduction of the amino acid change with primers 5′-CAGGTCCACCTTCTAGCATGTGCCTGGCTAGAG-3′ (sense) and 5′-CTCTAGCCAGGCACATGCTAGAAGGTGGACCTG-3′ (antisense). ERαD351Y was made with 5′-GACCAACCTGGCATACAGGGAGCTGGTTC-3′ (sense) and 5′-GAACCAGCTCCCTGTATGCCAGGTTGGTC-3′ (antisense). ERαE542A was made with the 5′-CTATGACCTGCTGCTGGCGATGCTGGACGCCC-3′ (sense) and 5′-GGGCGTCCAGCATCGCCAGCAGCAGGTCATAG-3′ (antisense) primers. ERβE448A was made with the 5′-GACCTGCTGCTGGCGATGCTGAATGC-3′ (sense) and 5′-GCATTCAGCATCGCCAGCAGCAGGTC-3′ (antisense) primers.

A stop codon was introduced before helix 12 or the F domain with the following primers: for ERαΔH12, 5′-GTGCAAGAACGTGGTGTAGCTCTATGACCTGCTG-3′ (sense) and 5′-CAGCAGGTCATAGAGCTACACCACGTTCTTGCAC-3′ (antisense); for ERβΔH12, 5′-GCAAAAATGTGGTCTAAGTGTATGACCTGC-3′ (sense) and 5′-GCAGGTCATACACTTAGACCACATTTTTGC-3′ (antisense); for ERαΔF, 5′CCCACCGCCTATAAGCGCCCACTAG-3′ (sense) and 5′-CTAGTGGGCGCTTATAGGCGGTGGG-3′ (antisense); for ERβΔF, 5′-GCCCACGTGCTTTAAGGGTGCAAGTCCTC (sense) and 5′-GAGGACTTGCACCCTTAAAGCACGTGGGC-3′ (antisense).

Mouse TIF2 cDNA cloned into the pSG5 expression vector (Stratagene, La Jolla, Calif.) was used as the template for mutagenesis. All mutations were introduced with the QuikChange XL site-directed mutagenesis kit (Stratagene). The Tif2-NR construct was produced by introduction of Asp718 and SpeI sites with primers 5′-CGCATGGTACCTCGCTCAAGGAGAAGCATAAGATTTTGCACAGACTCTTACAGGACACTAGTAGTTCCCCTGTGG-3 ] (sense) and 5′-CCACAGGGGAACTAGTGTCCTGTAAGAGTCTGTGCAAAATCTTATGCTTCTCCTTGAGCGAGGTACCATGCG-3′ (antisense). An amino acid change disrupting the LXXLL interaction motif in NR box 1 was made with primers 5′-CAAAGGGCAGACCAAAGCCCTGCAGCTGCTG-3′ (sense) and 5′-CAGCAGCTGCAGGGCTTTGGTCTGCCCTTTG-3′ (antisense), which introduces the change L123A.

Tif2-TA1 was constructed by ligation of the annealed TA1 oligonucleotides 5′-GTACCTCGAGGACGCTTCAGCTGGATTGGGGTACTCTGTATTGGTCTAGAA-3′ (sense) and 5′-CTAGTTCTAGACCAATACAGAGTACCCCAATCCAGCTGAAGCGTCCTCGAG-3′ (antisense) into the Asp718 and SpeI sites of Tif2-NR. Tif2-TB3 was made in the same way with TB3 oligonucleotides 5′-GTACCAGTAGTGTAGCAAGTAGAGAGTGGTGGGTAAGAGAATTATCTAGAA-3′ (sense) and 5′-CTAGTTCTAGATAATTCTCTTACCCACCACTCTC TACTTGCTACACTACTG-3′ (antisense).

Tif2-wt was produced by introduction of Asp718 and SpeI sites with the same primers as above (see Tif2-NR construct) and introduction of NheI and EcoRV sites flanking NR box 1 with primers 5′-GACACAGCCGGCTGCTAGCCAGCAAAGGGCAG (sense), 5′-CTGCCCTTTGCTGGCTAGCAGCCGGCTGTGTC (antisense) and primers 5′-CAAGTCCGACCAGATATCGCCTTCACCCTTG (sense) and 5′-CAAGGGTGAAGGCGATATCTGGTCGGACTTG (antisense). An amino acid change disrupting the LXXLL interaction motif in NR box 3 was made with primers 5′-GAGAATGCACTAGCGCGCTATTTGCTC (sense) and 5′-GAGCAAATAGCGCGCTAGTGCATTCTC (antisense), which changes the first L to an A in NR box 3. Tif2-2XTA1 was constructed by ligation of the annealed TA1 oligonucleotides for box 2 (see above) and the annealed TA1 oligonucleotides 5′-CTAGCCTCGAGGACGCTTCAGCTGGATTGGGGTACTCTGTATTGGTCTAGAGACCAGA (sense) and 5′-ATCTGGTCTCTAGACCAATACAGAGTACCCCAATCCAGCTGAAGCGTCCTCGAGG (antisense) into the introduced NheI and EcoRV sites. Tif2-2XTB3 was constructed in the same way with the above-mentioned oligonucleotides for NR box 2 and 5′-CTAGCCAGTAGTGTAGCAAGTAGAGAGTGGTGGGTAAGAGAATTATCTAGATCCGACCAGAT (sense) and 5′-ATCTGGTCGGATCTAGATAATTCTCTTACCCACCACTCTCTACTTGCTACACTACTGG (antisense) in NR box 1.

Cell culture and transient transfections.

HuH7 (human liver) cells were maintained in high-glucose Dulbecco's modified Eagle's medium (DMEM; Gibco, Invitrogen Corp., catalog no. 41966-029) supplemented with 10% fetal bovine serum and 2 mM l-glutamine in a humidified atmosphere of 5% CO₂ in air. The cells were split twice a week. Cos-7 cells were grown in DMEM supplemented with 10% fetal bovine serum and 2 mM l-glutamine and split twice a week. For transient transfection, cells were seeded into 24-well plates 24 h before transfection in phenol red-free medium supplemented with 10% dextran charcoal-stripped fetal bovine serum and 2mM l-glutamine. Cells were transfected with Lipofectamine 2000 as instructed by the manufacturer (Invitrogen Corp.). After transfection, cells were treated with ligands for 16 h before luciferase and β-galactosidase activity was assayed.

Analysis of intracellular localization by confocal microscopy.

HuH7-cells were plated on glass coverslips in six-well cell culture plates and grown in DMEM without phenol red (Gibco) supplemented with 10% dextran charcoal-stripped fetal bovine serum and l-glutamine for 12 h. The cells were transfected for 6 h with 0.2 μg of a plasmid green fluorescent protein (GFP)-tagged ER and 0.5 μg of peptide plasmid with Lipofectamine 2000 (Invitrogen). After transfection, the cells were treated with ligand for 16 h before being fixed with 3% paraformaldehyde in 5% sucrose-phosphate-buffered saline (PBS) for 20 min at room temperature. For indirect immunofluorescence, fixed cells were rinsed three times with PBS before being permeabilized with PBS-Tween (0.1%) three times for 5 min each at room temperature. The cells were blocked with 5% goat serum (Jackson Immuno Research Laboratories, Inc., West Grove, Pa.) in PBS-Tween (1 h at room temperature) before being incubated with monoclonal mouse anti-Gal4 DBD antibody (Santa Cruz Biotechnology, Inc.) for 1 h (diluted 1:200 in PBS-Tween). After removal of the Gal4 DBD antibody by washing three times for 5 min each with PBS-Tween, cells were treated with Lissamin-Rhodamin-conjugated goat anti-mouse immunoglobulin G (heavy and light chains) (Jackson Immuno Research Laboratories, Inc.) for 1 h at room temperature. Cells were washed five times for 5 min each with PBS-Tween before being fixed to slides with the antiphotobleaching agent Fluorosave (Calbiochem, La Jolla, Calif.). Subcellular localization was determined with a Leica laser scanning confocal microscope (Leica Corp., Deerfield, Ill.).

RESULTS

Identification of binding peptides specific for tamoxifen-ERα or tamoxifen-ERβ.

In vitro affinity selection screenings of phage-expressed peptides that bind ERα or ERβ in the absence or presence of estradiol or tamoxifen were performed. In addition to multiple LXXLL peptides, many of which have been characterized previously and are not the subject of this study, a panel of 89 distinct non-LXXLL peptides were isolated. In order to select for high-affinity interactions with ERs under in vivo conditions, we used a mammalian cell-based two-hybrid assay. Peptide-coding cDNA inserts from the 89 phage clones were transferred into a mammalian expression vector so that the peptides were expressed in fusion with the Gal4 DBD. Interactions between Gal4 DBD-peptide fusion proteins and VP16 activation domain-tagged full-length ERs were assayed by determining activity from a Gal4-responsive luciferase reporter in the human hepatoma cell line HuH7 (Fig. 1A).

FIG. 1.

Selection of ER-binding peptides. (A) In vitro affinity selection of phage-displayed peptide libraries with full-length wild-type (wt) ERα or ERβ in the presence of 17β-estradiol (E2) or 4-hydroxy-tamoxifen (OHT) was followed by in vivo identification of high-affinity binding peptides with the mammalian two-hybrid method. L, ligand; R, receptor. (B) Analysis of ER-peptide interactions in mammalian cells. HuH7 cells were transiently transfected with expression vectors for VP16 activation domain (AD)-tagged ERs and Gal4 DBD-tagged peptides together with a Gal4-responsive luciferase reporter and a β-galactosidase reporter as an internal control. Transfections were performed in duplicate in 24-well formats; error bars represent the standard deviation. Duplicate transfections contained 100 ng of VP16-ER, 200 ng of 5×Gal4-TATA-luc, 100 ng of Gal4-peptide fusion construct, and 40 ng of β-galactosidase control plasmid. Cells were treated with ligands (1 μM) as indicated for 16 h before luciferase and β-galactosidase activities were assayed. Luciferase activity was normalized against β-galactosidase activity.

Although many of the in vitro-selected peptides did not display significant binding in the mammalian system, 12 of these 89 peptides interacted well, albeit with different affinities, with ERα and/or ERβ (Fig. 1B). Furthermore, the four peptides that bound in the presence of estradiol but not tamoxifen also displayed ER subtype specificity, i.e., peptides EA1 and EA2 appeared to be ERα specific, and peptides EB1 and EB2 appeared to be ERβ specific. Notably, the identification of peptides that displayed high ER-subtype binding affinity and estradiol selectivity indicates that subtype-specific binding surfaces are exposed in agonist-bound ERs outside the LXXLL binding surface.

Of the eight peptides that specifically interacted with tamoxifen-bound receptors, one peptide (TA1) interacted exclusively with ERα, and three peptides (TB1, TB2, and TB3) appeared to prefer interaction with ERβ, even though some weak interaction could be seen with ERα as well. The remaining four tamoxifen peptides (TAB1 to TAB4) interacted with both receptor subtypes, although here preferences could also be seen. To ensure appropriate functionality of the VP16-ERs under our experimental conditions, we included the LXXLL peptide EAB1 (originally denoted α/β I; see Table 1), which has been reported to interact well with both ERα and ERβ in the presence of estradiol (18). We also analyzed the interactions of ERs with the receptor interaction domain containing three distinct LXXLL motifs of the naturally existing p160 coactivators AIB1/SRC3 and TIF2 (data not shown). Finally, neither ERα nor ERβ interacted with the Gal4 DBD alone in the absence or in the presence of any of the ligands tested (data not shown).

TABLE 1.

Sequences of 12 peptides interacting with the full-length receptor in vivo^a

Peptide	Sequence	Isolation target/selection
EAB1	SSNHQSSRLIELLSR	ERα/E2
EA1	SRLEYWLKWEPGPSR	ERα/E2
EA2	SSKGVLWRMLAEPVSR	ERα/E2
EB1	SSLQAGXWLMHYLRGGDSR	ERβ/E2
EB2	SSTSWLHHYLMGTSR	ERβ/no hormone
TAB1*	SSSMMREFFERELSR	ERβ/no hormone
TAB2*	SSTTMFDFFYERLSR	ERβ/tam
TAB3	SSGPFYVGGMLWPADCLSR	ERβ/tam
TAB4	SSVFTIMDGKVALSR	ERβ/tam
TA1	SRTLQLDWGTLYWSR	ERα/tam
TB1	SSIPPRSWWLSQLSR	ERβ/no hormone
TB2	SSQEEWLLPWRLASR	ERβ/tam
TB3*	SSVASREWWVRELSR	ERβ/tam

^aSequences of high-affinity ER-binding peptides were identified by phage display followed by mammalian two-hybrid analysis (see Fig. 1). For comparison, a previously characterized LXXLL peptide, EAB1, was included (18). *, peptides containing a consensus sequence, (S/M)X(D/E)(W/F)(W/F)XXXL, possibly involved in interactions with tamoxifen-bound ERs (18). E2, 17β-estradiol; tam, tamoxifen.

Table 1 summarizes the results of our peptide selection, shows the individual peptide sequences, classifies them according to binding selectivity in the mammalian two-hybrid system, and provides information regarding their origin, i.e., which ligand and which ER subtype was used for the in vitro screenings by phage display. Evidently, the 12 non-LXXLL high-affinity ER binding peptides identified can be subdivided into distinct classes depending on whether they recognize estradiol-dependent (EA, EB, and EAB) or tamoxifen-dependent (TA, TB, and TAB) conformations and whether they bind to one or both of the ER subtypes. Moreover, the binding features of all 12 selected peptides are compatible between the in vivo and the in vitro systems. Most important for our study were the tamoxifen-specific and ER subtype-selective peptides. TA1, specific for tamoxifen-bound ERα in vivo, was isolated in the in vitro screenings with tamoxifen-bound ERα, and TB3, specific for tamoxifen-bound ERβ in vivo, was isolated in the in vitro screenings with tamoxifen-bound ERβ.

ER domain requirement for peptide binding in the presence of tamoxifen.

Previous work has clearly established that LXXLL peptides and cofactors utilizing these motifs bind exclusively to the LBD in the presence of agonistic ligands such as estradiol in the case of the ERs (2, 28). In contrast, peptide binding surfaces for ERs in the presence of antagonists as well as ER domain requirements for non-LXXLL peptides and proteins are largely unknown. Therefore, we attempted to determine which ER domains are required for binding of the 12 selected peptides. We made various VP16-ER constructs expressing different domains or variants with deleted ERα or ERβ domains (Fig. 2), checked them for relative protein expression by Western blotting, and analyzed them for peptide binding in the mammalian two-hybrid assay (Fig. 3). The results of these experiments can be summarized as follows.

FIG. 2.

Schematic representation of ER constructs used in the mammalian two-hybrid analysis. Domain nomenclature: A/B, modulatory N-terminal domain containing AF-1 in ERα; C, DNA-binding domain; D, hinge domain; E, ligand-binding domain, including AF-2 helix 12 (H12); F, modulatory C-terminal domain. ERβ amino acid residues are indicated in parentheses.

FIG. 3.

Analysis of peptide binding to ER domains. Mammalian two-hybrid experiments with VP-16-tagged ER domains (see Fig. 2) were performed as described for Fig. 1.

(i) EA1 and EA2, specific for estradiol-bound ERα, interacted with both ERα DEF and CDEF. The interaction surface or surfaces for these peptides were destroyed when helix 12 was deleted. Thus, interaction surfaces for non-LXXLL peptides that recognize estradiol-bound ERα are most likely exposed within the LBD. (ii) In contrast, binding surfaces for EB1 and EB2, specific for estradiol-bound ERβ, were apparently not exposed within ERβ DEF or CDEF, while the LXXLL control peptide EAB1 interacted, as expected (Fig. 3A and 3B). (iii) TA1 but none of the other peptides specific for tamoxifen-bound ERα interacted with both ERα DEF and CDEF. Similar to the interactions of the estradiol-dependent peptides with ERα, the TA1 interaction was disrupted when helix 12 was removed (Fig. 3C). Importantly, this indicates that tamoxifen-bound ERα exposes a separate interaction surface for non-LXXLL peptides within the LBD. (iv) TB3 but none of the other peptides specific for tamoxifen-bound ERβ interacted with ERβ DEF, although this interaction was relatively weak. Most intriguingly, removal of helix 12 in ERβ enhanced the interaction with TB3 (Fig. 3C), which is in sharp contrast to the situation with ERα and TA1. For reasons unknown, no interaction was observed with ERβ CDEF, which was expressed, as judged from Western blot analysis (data not shown); also, the EAB1 interaction was weak with this construct. As the corresponding ERα construct was functional, unrecognized differences in the domain organization may exist between the two ERs.

(v) The four peptides interacting with both ERα and ERβ in the presence of tamoxifen (TAB1 to TAB4) did not interact with the DEF and CDEF constructs but only with the full-length receptors. (vi) Deletion of helix 12 caused enhanced interaction with TAB1 to TAB4 in the presence of tamoxifen and surprisingly also in the presence of estradiol. Interestingly, the ERβ-specific peptides TB1 and TB3 interacted with ERα ΔH12 (Fig. 3C). (vii) In contrast, deletion of the F domain alone resulted in generally weaker interactions but no change in the specificity pattern of the peptides. Therefore, a small region encompassing helix 12 (and surrounding residues) was identified as a critical determinant for tamoxifen and ER subtype specificity. (viii) None of the peptides interacted with the ER N termini alone (AB constructs), although original peptide selections have been performed with wild-type ERs. This indicates that the ER N terminus, including AF-1, does not encompass a separate binding surface in the presence of tamoxifen, which is interesting in light of the possible involvement of AF-1 in mediating the partial agonist activity of tamoxifen, at least in the case of ERα.

(ix) Amino acid E542 in ERα has been shown to be involved in the formation of the hydrophobic groove in which LXXLL motifs in coactivators bind. As expected, when this amino acid residue was mutated to an alanine, the interaction between the LXXLL-containing peptide EAB1 and ERα was abolished, in contrast to interactions with TA1 and the other tamoxifen peptides that were not affected by this mutation (Fig. 4B), showing that these interactions occur outside the hydrophobic pocket. When the corresponding amino acid in ERβ was mutated (E448), we observed that binding of EAB1 was affected, as expected, and that interaction with the estradiol-dependent non-LXXLL peptides EB1 and EB2 was also abolished, suggesting that these agonist interactions are dependent on amino acid E448 within helix 12 (Fig. 4C). As in ERα, the tamoxifen-peptide interactions were not affected by this mutation. A mutation in ERα in which amino acid E380 within LBD helix 5 was changed to an alanine (E380A) abolished the interaction with peptides TAB1 to TAB4, while interaction with TA1 and the LXXLL peptide EAB1, surprisingly, was sustained (Fig. 4D). Another naturally occurring mutation, D351Y in ERα, previously reported to allow ERα to interpret tamoxifen as an agonist (13), showed a slightly changed peptide binding pattern compared to the wild type (Fig. 4E). However, this mutation did not affect the antagonist selectivity of the tamoxifen peptides, as no binding was observed in the presence of estradiol. Similarly, binding of the agonist peptides was still estradiol dependent, although affinity changes were observed compared with the wild-type ER. Therefore, while our data support the view that the D351Y mutation introduces minor conformational changes in ERα, the largely unchanged ligand selectivity of our peptides does not explain how tamoxifen would be interpreted as an agonist.

FIG. 4.

Analysis of peptide binding to ER mutants. (A) Schematic representation of amino acid substitutions in LBD helices 3, 5, and 12 of ERα and ERβ. Mutated residues are in bold and underlined. (B, C, D, and E) Mammalian two-hybrid experiments were performed as described for Fig. 1.

We conclude that, for a majority of the peptides interacting with tamoxifen-bound ERs, N-terminal regions within the AB (AF-1) domain seem to be required but not sufficient for interaction, which points to the existence of a more complex interaction surface for those peptides involving both N- and C-terminal ER domains. Most fascinating is the result that the DEF region appears to be sufficient for binding of the tamoxifen- and subtype-specific peptides TA1 and TB3 and that helix 12 is differently involved in the interactions of these peptides with ERα and ERβ. This demonstrates for the first time, to our knowledge, that antagonist-bound LBDs of both ERα and ERβ indeed expose a functional peptide and coregulator binding surface distinct from AF-2. Subsequently, we focused on the further characterization of these two peptides, TA1 and TB3.

Ligand selectivity of the tamoxifen peptides.

Tamoxifen represents one of the most studied and clinically used SERMs, and related compounds may share many but not all features with tamoxifen. Therefore, we next sought to determine whether the tamoxifen-induced peptide binding surfaces are exposed even with other antagonists and compared tamoxifen with various compounds in the mammalian two-hybrid assay (Fig. 5). Raloxifene is a SERM used clinically to prevent osteoporosis, with the same beneficial antiestrogenic effects as tamoxifen in breast tissue but without the agonistic effects of tamoxifen in uterine tissue (9). MPP is an ERα-specific antagonist (30), and ICI 182,780 represents a pure antagonist. The result of this comparative analysis was that all these compounds were generally inefficient in promoting high-affinity peptide binding comparable to that with tamoxifen. However, TA1 was weakly responsive to raloxifene and ICI 182,780, while TB3 displayed strict tamoxifen specificity. This indicates that distinct tamoxifen-induced conformations are recognized by the peptides and that unique receptor binding surfaces are exposed depending on the nature of the antagonist.

FIG. 5.

Analysis of ER-peptide binding in the presence of diverse SERMs and antagonists. (A) Structures of the ER antagonists 4-hydroxy-tamoxifen (OHT), raloxifene (RAL), ICI 182,780, and MPP. (B) Mammalian two-hybrid experiments were performed in triplicate in a 96-well format with 100 ng of VP16-ER, 200 ng of 5×Gal4-TATA-luc, and 100 ng of Gal4 peptide. Error bars represent standard deviation.

Receptor subtype- and tamoxifen-specific peptide interactions visualized in vivo by colocalization analysis.

In order to further confirm the ER subtype- and ligand-specific interactions between tamoxifen-ERα and TA1 and tamoxifen-ERβ and TB3, confocal microscopy was used to monitor the localization of GFP-tagged ERs and Gal4-tagged peptides when individually expressed and when coexpressed in mammalian cells (Fig. 6). First we observed that GFP-ERs and Gal4-TA1 or TB3 were evenly distributed in the nucleus in the absence of ligand when expressed individually. In contrast, in the presence of 10 nM estradiol (Fig. 6A II and C II) or 100 nM tamoxifen (Fig. 6A III and C III), both GFP-ERα and GFP-ERβ were redistributed into subnuclear structures, forming a dotted pattern. The Gal4 peptides remained evenly distributed in the nucleus, indicating that the ligands had no effect on peptide distribution. When expressed simultaneously, GFP-ERs and Gal4 peptides did not change the pattern in the absence of ligand (Fig. 6B I and D I). However, the Gal4 peptides were redistributed to the dotted GFP-ER pattern in the presence of tamoxifen but not estradiol when coexpressed, indicating tamoxifen-induced colocalization within the nucleus. Moreover, when the ERβ-specific peptide TB3 and ERα were coexpressed or the ERα-specific peptide TA1 and ERβ were coexpressed in cells, no colocalization was observed (data not shown). In summary, these observations support our previous conclusion that TA1 interacts exclusively with tamoxifen-ERα and that TB3 prefers interaction with tamoxifen-ERβ.

FIG. 6.

Analysis of intracellular colocalization of ERα and ERβ with subtype-specific peptides TA1 and TB3. GFP-tagged ERs and Gal4-tagged peptides were expressed individually or together in HuH7 cells. Peptides were visualized by indirect immunofluorescence in a confocal microscope. (A) Intracellular localization of individually expressed ERα or TA1 without hormone (I), with estradiol (II), and with tamoxifen (III). (B) Intracellular colocalization of simultaneously expressed ERα and TA1. (C) Intracellular localization of individually expressed ERβ or TB3 without hormone (I), with estradiol (II), and with tamoxifen (III). (D) Intracellular colocalization of simultaneously expressed ERβ and TB3.

Inhibition of tamoxifen-dependent ER activity.

We next wanted to assess whether the specific peptide interactions also had a selective effect on transcriptional regulation mediated by tamoxifen-activated ERα or ERβ. We used reporter constructs and monitored the activity in Cos-7 cells after cotransfection of the receptor and increasing amounts of Gal4 peptide. To measure tamoxifen-induced transcriptional activity mediated by ERα, Cos-7 cells were transfected with the estrogen-responsive C3-luc reporter together with an expression vector for ERα and increasing amounts of plasmids expressing Gal4-peptide fusion proteins (Fig. 7A). The tamoxifen-induced transcriptional activity mediated by ERβ was measured in the same way except that the AP1-responsive collagenase reporter construct was used instead of the C3-luc reporter (Fig. 7B). We observed that ERα-selective peptide TA1 effectively inhibited the tamoxifen-induced transcriptional activity mediated by ERα but not ERβ and that the ERβ-selective peptide TB3 effectively inhibited tamoxifen activity mediated by ERβ but not ERα. Estradiol-dependent activation measured on the C3-luc reporter or an ERE-luc reporter in the case of ERα or an ERE-luc reporter in the case of ERβ could not be inhibited by either of these two peptides (data not shown). This shows that ER subtype-specific inhibition of tamoxifen-dependent transcriptional activation is possible and indicates that the peptides might block putative binding surfaces for endogenous coactivators mediating the agonist activity of tamoxifen.

FIG. 7.

Inhibition of tamoxifen-dependent transcriptional activity by ER subtype-specific peptides TA1 and TB3. (A) Cos-7 cells were transfected with the estrogen-responsive complement 3 (C3-luc) reporter gene, expression vectors for ERα, and β-galactosidase control plasmid along with expression vector for the Gal4-TA1 or Gal4-TB3 fusion in increasing amounts, as indicated. After transfection, cells were induced with 100 nM tamoxifen (OHT) for 16 h before luciferase and β-galactosidase activities were analyzed. The control represents the transcriptional activity of 100 nM tamoxifen in the presence of the Gal4 DBD alone and was set at 100%. Experiments were performed in triplicate, with error bars representing the standard deviation. (B) Cos-7 cells were transfected as in panel A except that the expression vector for ERβ and the AP-1-responsive collagenase reporter construct (pCOL-luc) were used. Triplicate transfections contained 300 ng of reporter gene construct, 300 ng of expression vector for ERα or ERβ, 60 ng of pRSV-β-gal, and either 30, 150, 300, 450, or 600 ng of Gal4-peptide fusion construct. Empty Gal4-DBD vector was used to compensate, so that 600 ng of Gal4 DBD was transfected in each replicate. Induction of the reporters with 100 nM tamoxifen is shown above each graph. NH, no hormone.

Engineering of ER subtype- and tamoxifen-specific coactivators.

To investigate whether the specific interaction surfaces for TA1 and TB3 could represent docking sites for novel coactivators mediating tamoxifen activation, we attempted to modify the well-characterized p160 coactivator TIF2 (also called GRIP1 and SRC2), known to be important for estradiol-dependent coactivation (5, 31). TIF2 contains three NR boxes known to be involved in NR interactions; boxes 1 and 2 have been implicated in high-affinity ER binding (32). Therefore, we exchanged one or two of the LXXLL motifs in the NR boxes with the ERα-specific peptide TA1 or the ERβ-specific peptide TB3 (Fig. 8). When it was cotransfected with GAL4-ERα or -ERβ along with a luciferase reporter into Cos-7 cells, we found that coexpression of the modified TIF2 but not of wild-type TIF2 could increase ER activity in the presence of tamoxifen, while estradiol-dependent coactivation was abolished. The double box exchange had only a small additive effect in the case of ERβ and no effect on ERα. This might indicate that the location and spacing of the peptides in Tif2 are not optimized for the function of the predicted tamoxifen surfaces. We conclude that it is indeed possible to generate tamoxifen-specific ER coactivators and provide a model for the function of related, naturally existing coactivators.

FIG. 8.

Enhancement of tamoxifen-dependent transcriptional activity by ER subtype-specific coactivators. (A) The modification of the coactivator TIF2 by replacement NR box 2 or NR boxes 1 and 2 with tamoxifen-specific peptides is shown schematically. (B) Cos-7 cells were cotransfected with expression vectors for Gal4-tagged ERα DEF or ERβ DEF and wild-type or modified TIF2 together with the Gal4-responsive reporter 5×UAS-TATA-luc. Experiments were performed in triplicate, with error bars representing the standard deviation. Each replicate contained 30 ng of Gal4-ER DEF construct, 150 ng of reporter construct, and 150 ng of TIF2 expression plasmid. After transfection, the cells were treated with 10 nM 17β-estradiol (E2), 100 nM tamoxifen (OHT), or dimethyl sulfoxide for 16 h before luciferase and β-galactosidase activities were analyzed. Transcriptional activity in the presence of empty pSG5 vector was used as a control, and data are presented as induction over the corresponding control activity.

DISCUSSION

In the present study, we have identified previously unrecognized ER binding peptides that do not contain a conserved LXXLL motif and that are specific for tamoxifen-bound ERα or ERβ. The unique characteristics of the selected peptides were supported by analysis in a number of different experimental systems in vitro and in vivo. By mammalian two-hybrid analysis of 89 in vitro-selected ER-binding peptides, we could identify eight peptides that interacted specifically with tamoxifen-bound receptors in vivo. Since the two ER subtypes are closely related proteins, it is not unexpected that they also expose identical interaction surfaces, and indeed we found four peptides that interacted with both ERα and ERβ. More importantly, some of the tamoxifen-interacting peptides also appeared to be ER subtype specific. These included the TA1 peptide, which represents a novel type of binding motif, for ERα and three distinct peptides named TB1, TB2, and TB3 for ERβ. ER subtype and tamoxifen specificity could be confirmed and visualized in vivo by intracellular colocalization analysis with GFP-tagged ERs and nuclear-targeted peptide fusion proteins.

Three of the peptides (TAB1, TAB2, and TB3) contained a consensus sequence previously isolated in a related study in another ER-interacting peptide, denoted α/β V, that was shown to interact with both ERα and ERβ in the presence of tamoxifen (18). The demonstration that TB3 specifically colocalizes in vivo with ERβ but not with ERα in the presence of tamoxifen not only confirms the ER subtype and tamoxifen specificity of TB3 but is also interesting because TB3 contains the above-mentioned consensus sequence predicted to bind both ERα and ERβ. We conclude that the peptide sequence TB3 shows a more subtype-specific interaction than the previously characterized α/β V peptide.

Subtype specificity is presumably achieved by sequence alterations adjacent to the consensus motif, since it was previously shown that altering flanking sequences adjacent to the leucine core of LXXLL peptides changes nuclear receptor, including ER subtype, specificity (2). Alternatively, specificity may be achieved by minor alterations within the consensus sequence. Future mutagenesis studies will have to clarify whether or not functionally relevant consensus sequences for antagonist peptides and naturally existing coregulators exist. When searching databases for coregulators or proteins containing the peptide sequences found in this study, we learned that while no protein or predicted open reading frame that matched the entire peptide sequence was identified, partial homologies were found in several proteins with unknown function (data not shown). Cloning of some of these candidate cofactors and extended mutagenesis studies are necessary to elucidate which amino acids in the peptide sequence are essential for the functional peptide-receptor interactions. These experiments are ongoing and will be included in a future study.

We analyzed the receptor domains required for interaction and demonstrated that the DEF region, including the LBD of ERα and ΕRβ, was sufficient for TA1 and TB3 binding, respectively. In contrast, other tamoxifen-dependent peptides isolated in our study required a full-length receptor for efficient interaction. Similar observations have been made when analyzing domains required for corepressor NR box-derived peptide binding in which a complex interaction surface including both the N and C termini of the receptor was proposed (6). We selected the unique TA1 and TB3 peptides for further characterization because they both recognized novel tamoxifen-induced surfaces within the LBD, distinct from the surface necessary for binding of N-CoR, SMRT, and related proteins (6). Interestingly, TA1 required LBD helix 12 for efficient binding, while TB3 and other tamoxifen peptides interacted even better and also lost the subtype and ligand specificity when helix 12 was removed. Mutation analysis revealed that by mutating amino acid residue E542 in ERα or the corresponding E448 in ERβ, the LXXLL pocket was destroyed, while interaction with TA1 and TB3 was unaffected. These results suggest a different role of helix 12 in ERα versus ERβ LBD interaction surfaces for TA1 and TB3, distinct from the charged clamp surface formed in the agonist-LBD conformation. In the case of TA1, the surface may include parts of ERα helix 12, while in the case of TB3, the surface may exclude helix 12. Removal of helix 12 will therefore prevent interactions between TA1 and tamoxifen-bound ERα, while interactions between TB3 and tamoxifen-bound ERβ are enhanced. No such change in specificity was observed when only the F domain was removed.

An earlier study has shown that mutation of ERα E380 gives a mutant receptor with a higher sensitivity to estradiol than the wild-type ER and less sensitivity to antiestrogens for suppression of transcriptional activity (23). A more recent study suggested that E380 is important for discriminating in the recruitment of corepressors, the A domain, or helix 12 binding to the receptor LBD in the functional interplay between receptor domains and corepressors (15). Consistent with this idea, when amino acid residue E380 in ERα was mutated to an alanine, the interactions with estradiol peptides, including the LXXLL peptide EAB1, were retained, while interactions with the tamoxifen peptides TAB1 to TAB4 but not TA1 were lost. Taken together, these results indicate that this residue is important for the formation of antagonist surfaces. Continued analysis of additional ER mutations will allow further delineation of amino acid residues critical for binding of the tamoxifen-specific peptides identified here. Moreover, it is currently unclear whether these novel epitopes are absolutely ER specific or shared with other nuclear receptor family members. However, neither TA1 nor TB3 (nor any of the other non-LXXLL peptides identified in this study) was isolated in phage displays with other nuclear receptor targets or interacted with glucocorticoid receptor or estrogen receptor-related receptor β in the presence or absence of antagonists when analyzed in the mammalian two-hybrid assay (data not shown).

Earlier studies demonstrated that LXXLL peptides can be used to inhibit estrogen-mediated activation in cells when coexpressed with either ERα or ERβ (14, 18). The mechanism behind this inhibition is that coactivator recruitment to AF-2 is prevented because the binding sites on the receptors are occupied by the LXXLL peptides. Importantly, we show here that TA1 and TB3 could inhibit ER-dependent reporter gene activation in response to tamoxifen ER subtype specifically, which indicates that the exposed surfaces are indeed important for tamoxifen agonism. However, even though the precise mechanism of ER action at AP-1 sites is unknown, tamoxifen agonism through ERβ was measured from an AP-1-luciferase reporter. ERβ in the presence of tamoxifen robustly activates AP-1-dependent transcription. Different pathways for this AP-1 activation have been suggested (33), but regardless of whether activation is AF-2 dependent or not, TB3 effectively inhibits the activation, while TA1 does not. Likewise, we showed that TA1 and not TB3 potently inhibits tamoxifen-ERα-mediated activation from an estrogen-responsive complement 3-luciferase reporter containing elements cooperating with EREs, reconstituting tamoxifen agonism mediated from direct binding of ERα to a DNA response element.

Most of the published studies related to antagonist action have focused on mechanisms related to ERα, due to a potent AF-1 domain which has been suggested to be the major determinant of the partial agonist activity of tamoxifen. Furthermore, all cancer cell lines (MCF7, T47D, and Ishikawa) used in those studies express ERα at different levels. Therefore, with the remarkable exception of one initial study (21), the impact of antagonists on the actions of the second subtype ERβ, the mechanisms behind this, the cofactors involved, and the biological implications still need to be addressed. It is also interesting that the peptide binding surfaces exposed with tamoxifen are not present when the receptors are bound to other antagonists, including raloxifene. These qualitative differences cannot be explained from the existing ER LBD structures but emphasize the importance of pharmacological and biological differences with clinical relevance between various antagonists (8, 9).

Different coregulatory proteins have been implicated in tamoxifen action. With regard to coactivators, there are indications that overexpression of the p160 coactivator SRC-1 could be one reason for partial tamoxifen agonism of ERα in certain cell types (26). With regard to corepressors, it has been suggested that low levels of corepressors such as N-CoR or SMRT may relieve the antagonistic effects of tamoxifen and raloxifene (12), and indirect evidence has shown that these corepressors can interact with ERs and the progesterone receptor in the presence of antagonists, even though the physiological relevance of these interactions is debatable (4, 27, 29). Indeed, a recent study has clearly indicated that mutated but not natural corepressor NR box peptide motifs can bind to antagonist ERs, suggesting that corepressors other than N-CoR or SMRT may be involved in antagonist signaling (6).

Recent two-hybrid screens with antagonist-bound steroid receptors have identified additional coregulators implicated in ER antagonist action. Among the corepressors reported to affect tamoxifen action (3, 19, 20, 25), only one has been suggested to show ER and tamoxifen specificity (17). This corepressor, denoted RTA, was shown to interact with the N-terminal domain of ERs. Regarding coactivators besides the p160 coactivators, one additional coactivator affecting tamoxifen activity has been reported, but without any ER or tamoxifen specificity (7). Since previous studies have focused almost exclusively on the ERα subtype, it is currently unclear whether these identified coregulators (i.e., corepressors) are also relevant for ERβ. Even more importantly, coactivators that are specific for antagonist-bound ERs, particularly for ERβ, and that function independently of AF-1, for example, by binding to the tamoxifen-dependent AF-T surface within the LBD, have not yet been described. Therefore, our demonstration that TIF2 can be converted from an agonist (17β-estradiol) to an antagonist (tamoxifen) coactivator by replacing the ER-binding peptide motifs (e.g., NR box 2 in TA1 or TB3) should stimulate the search for naturally existing TIF2-TA1- or TIF2-TB3-like coactivators. After consideration of the results of this study, we propose a revised model of tamoxifen agonism that applies to both ERs (Fig. 9).

FIG. 9.

Revised model of tamoxifen agonism. Current models focus on the critical role of AF-1 in mediating the transcriptional activity of ERα. They also propose a novel tamoxifen-dependent corepressor binding surface (presumably within the LBD), referred to as repression function RF-T. The transcriptional outcome may depend on cell type-specific relative levels of ERα versus ERβ and AF-1 coactivators versus RF-T corepressors. We extend that model by proposing the existence of a tamoxifen-dependent coactivator binding surface within the LBD, referred to as AF-T. The transcriptional outcome may depend on cell type-specific relative levels of ERα versus ERβ and AF-1/AF-T coactivators versus RF-T corepressors.