Estrogen response elements alter coactivator recruitment through allosteric modulation of estrogen receptor beta conformation.

Estrogen receptor beta (ERbeta) activates transcription by binding to estrogen response elements (EREs) and coactivator proteins that act as bridging proteins between the receptor and the basal transcription machinery. Although the imperfect vitellogenin B1, pS2, and oxytocin (OT) EREs each differ from the consensus vitellogenin A2 ERE sequence by a single base pair, ERbeta activates transcription of reporter plasmids containing A2, pS2, B1, and OT EREs to different extents. To explain how these differences in transactivation might occur, we have examined the interaction of ERbeta with these EREs and monitored recruitment of the coactivators amplified in breast cancer (AIB1) and transcription intermediary factor 2 (TIF2). Protease sensitivity, antibody interaction, and DNA pull-down assays demonstrated that ERbeta undergoes ERE-dependent changes in conformation resulting in differential recruitment of AIB1 and TIF2 to the DNA-bound receptor. Overexpression of TIF2 or AIB1 in transient transfection assays differentially enhanced ERbeta-mediated transcription of reporter plasmids containing the A2, pS2, B1, and OT EREs. Our studies demonstrate that individual ERE sequences induce changes in conformation of the DNA-bound receptor and influence coactivator recruitment. DNA-induced modulation of receptor conformation may contribute to the ability of ERbeta to differentially activate transcription of genes containing divergent ERE sequences.

Estrogen receptor ␤ (ER␤) activates transcription by binding to estrogen response elements (EREs) and coactivator proteins that act as bridging proteins between the receptor and the basal transcription machinery. Although the imperfect vitellogenin B1, pS2, and oxytocin (OT) EREs each differ from the consensus vitellogenin A2 ERE sequence by a single base pair, ER␤ activates transcription of reporter plasmids containing A2, pS2, B1, and OT EREs to different extents. To explain how these differences in transactivation might occur, we have examined the interaction of ER␤ with these EREs and monitored recruitment of the coactivators amplified in breast cancer (AIB1) and transcription intermediary factor 2 (TIF2). Protease sensitivity, antibody interaction, and DNA pull-down assays demonstrated that ER␤ undergoes ERE-dependent changes in conformation resulting in differential recruitment of AIB1 and TIF2 to the DNA-bound receptor. Overexpression of TIF2 or AIB1 in transient transfection assays differentially enhanced ER␤-mediated transcription of reporter plasmids containing the A2, pS2, B1, and OT EREs. Our studies demonstrate that individual ERE sequences induce changes in conformation of the DNA-bound receptor and influence coactivator recruitment. DNA-induced modulation of receptor conformation may contribute to the ability of ER␤ to differentially activate transcription of genes containing divergent ERE sequences.
Transcription activation requires the coordinated interaction of multiple transacting factors with DNA recognition sites and other regulatory proteins. In response to cellular signals, transcription factors bind to specific DNA sequences residing in target genes and interact with numerous regulatory proteins to form an active transcription complex and initiate changes in gene expression. This multistep process provides a mechanism by which cells expressing different populations of proteins can differentially regulate expression of target genes.
The nuclear receptor superfamily is composed of a large number of transcription factors that bind to hormone response elements and modulate transcription. Estrogen receptors (ERs) 1 ␣ and ␤ are members of this nuclear receptor superfamily (1)(2)(3)(4)(5) and function as ligand-induced transcription factors that modulate expression of estrogen-responsive genes. Upon binding hormone, the ER undergoes a conformational change and binds to estrogen response elements (EREs) residing in target genes to initiate changes in transcription (6). The hormone-induced modulation of receptor conformation has been documented in the ligand-binding domains (LBDs) of 17␤-estradiol-and raloxifene-bound ER␣ (7) and in genistein-and raloxifene-bound ER␤ (8), with the most striking changes in conformation occurring in the positioning of helix 12 of the LBD. In addition to modulating receptor conformation, ligand binding influences the interaction of the ER with coactivator proteins such as steroid receptor coactivator 1 (SRC1) (9), transcription intermediary factor 2 (TIF2/GRIP1) (10 -12), amplified in breast cancer 1 (AIB1/ACTR/RAC3) (13)(14)(15), and CREBbinding protein (CBP/p300) (16,17). Crystal structures of the ER␣ LBD with the nuclear receptor interaction domain from GRIP1 (18) indicate that when the LBD is bound to an antagonist, the position of helix 12 interferes with coactivator binding. Thus, ligand-induced alterations in receptor conformation may alter coactivator recruitment and ultimately influence activation of transcription by ER. In addition to ligand-induced changes in conformation, there is a growing body of evidence to suggest that DNA sequences can modulate protein conformation. This allosteric modulation of protein conformation can dramatically alter gene expression, as has been documented with the POU domain-containing transcription factor Pit-1. Pit-1 serves as a potent activator of transcription when bound to its recognition site in the prolactin gene (19), but represses transcription when bound to its recognition sequence in the growth hormone gene. DNA-induced conformational changes can also have more subtle effects resulting in alteration of the level of transcription. Allosteric modulation of nuclear receptor conformation has been implicated in influencing transcription of a number of hormoneresponsive genes (19 -24).
Both ER␣ and ER␤ bind to EREs and activate transcription, but ER␤ is typically a less potent activator of reporter plasmids containing the vitellogenin A2 ERE compared with ER␣ (25)(26)(27)(28). The basis for this differential activation of transcription by ER␣ and ER␤ is unclear. The decreased affinity of ER␤ for the ERE compared with ER␣ could impair its ability to activate transcription (28). Additionally, studies with ER␣/ER␤ chimeric proteins indicate that the amino-terminal activation function 1 (AF-1) of these receptors vary in amino acid sequence and may influence the ability of these receptors to mediate transcription activation (25).
The goal of this study was to determine whether ER␤ conformation is different when bound to different EREs and, if so, to characterize the effect of conformational changes on receptor-coactivator interactions and transactivation. We find that DNA-dependent changes in receptor conformation directly translate into alterations in epitope availability and that these DNA-induced changes in receptor conformation alter interaction of ER␤ with coregulatory proteins. Thus, ERE-induced changes in ER␤ conformation may ultimately influence the ability of ER␤ to induce transcription activation.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-U2 osteosarcoma (U2-OS) cells were maintained in Eagle's minimal essential medium with 10 g/liter phenol red, 25 g/ml gentamycin, 100 units/ml penicillin, 100 g/ml streptomycin, and 15% heat-inactivated fetal calf serum. The medium was changed to Eagle's minimal essential medium containing phenol red with 5% charcoal dextran-treated (29) calf serum 3 days prior to transfection. Two days prior to transfections, cells were transferred to phenol red-free Eagle's minimal essential medium supplemented with 5% charcoal dextran-treated calf serum (Transfection B medium). Cells (4 ϫ 10 5 ) were plated in each well of a 24-well plate and maintained in Transfection B medium for 18 h. Transfections were carried out using Lipofectin (Life Technologies, Inc.) as described (28) with 500 ng of the human ER␤ expression vector CMV5-hER␤ (25), 400 ng of the ␤-galactosidase vector CMV-␤gal (Promega, Madison, WI), and 7.5 g of the indicated chloramphenicol acetyltransferase (CAT) reporter vector: consEREϩ10-CAT, pS2EREϩ10-CAT, ERE2ϩ10-CAT (30), and OTEREϩ10-CAT (28), which contain a single A2, pS2, B1, or oxytocin (OT) ERE, respectively, 3.6 helical turns upstream of a TATA box. Following incubation with the Lipofectin/DNA, cells were maintained in Transfection B medium containing ethanol vehicle or 10 nM 17␤estradiol (E 2 ) for 24 h. ␤-Galactosidase activity was measured as previously described (31) and used to normalize for differences in transfection efficiency. CAT assays were carried out as described (28), scanned with a Molecular Dynamics PhosphorImager, and analyzed using Im-ageQuant Version 5.0 software (Molecular Dynamics, Inc., Sunnyvale, CA). The coactivator expression vectors pSG5-TIF2 (12) and pcDNA3.1-AIB1 (13), kindly provided by Hinrich Gronemeyer (Institut de Genetique et de Biologie Moleculaire et Cellulaire, Illkirch, France) and Paul Meltzer (Laboratory of Cancer Genetics, Bethesda, MD), respectively, were added as indicated.
Partial Proteolysis and Antibody Interaction Experiments-The vectors B3consERE, B3pS2ERE, B3ERE2, and B3OTERE (21, 30), containing the A2, pS2, B1, and OT EREs, respectively, were digested with BamHI and EcoRI; and the resulting 55-base pair DNA fragments were isolated and 32 P-labeled. The resulting probe (10,000 cpm) was combined with 65 fmol of baculovirus-expressed, FLAG-tagged purified ER␤ (28) in binding reaction buffer (10% glycerol, 0.05 mM ZnCl 2 , 4 mM dithiothreitol, 50 mM KCl, 15 mM Tris, 0.2 mM EDTA, and 50 nM E 2 ) in the presence of 2.5 g of bovine serum albumin and 100 ng of poly(dI-dC). The ER␤-DNA binding reaction was incubated for 8 min at 25°C. Staphylococcus aureus protease V8 (Worthington) or proteinase K (Promega) was added as indicated, and the binding reactions were incubated for 10 min at 25°C. Reactions were loaded onto a nondenaturing acrylamide gel and electrophoresed. For the antibody interaction experiments, 90 fmol of ER␤ was incubated for 10 min at 25°C with a 32 P-labeled probe containing the A2, pS2, B1, or OT ERE, followed by addition of a phosphate-buffered saline control or anti-ER␤ antibody CWK-F12 or UICK-98 (kindly provided by B. S. Katzenellenbogen, University of Illinois, Urbana, IL) (32), anti-FLAG antibody M2 (Sigma), or anti-ER␣ antibody H151 (kindly provided by D. P. Edwards, University of Colorado Health Science Center, Denver, CO). The reactions were incubated for 10 min at 25°C and separated on a nondenaturing acrylamide gel. Protease sensitivity and antibody interaction experiments were repeated at least three times and produced similar results.
Pull-down Assays-DNA pull-down assays were carried out as described previously (21). Briefly, 4 pmol of annealed oligonucleotides containing the A2, pS2, B1, or OT ERE or a nonspecific sequence was immobilized on streptavidin paramagnetic beads (Dynal, Inc., Lake Success, NY). The immobilized DNA was then incubated with 4 pmol of baculovirus-expressed purified ER␤ and 1 M E 2 for 10 min. U2-OS nuclear extracts prepared as described previously for HeLa nuclear extracts (21) were added to the reaction. Following a 4-h incubation at 4°C, the beads were washed, and proteins were eluted in SDS sample buffer. After Western analysis, autoradiograms were optically scanned and quantitated using ImageQuant Version 5.0 software. Since binding of ER␤ varied slightly between the different EREs, the association of the coactivator was normalized to the amount of ERE-bound receptor by dividing the level of coactivator bound by the level of ER␤ bound. Coactivator/ER ratios from four independent experiments were combined. To minimize interexperimental variation, each coactivator/ER ratio was divided by the mean coactivator/ER ratio for that experiment and multiplied by the mean coactivator/ER ratio for all experiments.

Estrogen-dependent Transcription by ER␤ through Four Different EREs-ER␤ induces transcription of reporter plasmids
containing the consensus vitellogenin A2 ERE (25)(26)(27)(28). In this study, we have compared the ability of ER␤ to induce transcription of reporter plasmids containing the A2 ERE (GGT-CANNNTGACC) (33) and ERE sequences that vary from the consensus ERE sequence. We have utilized the imperfect vitellogenin B1 (AGTCANNNTGACC) (34) and oxytocin (GGT-GANNNTGACC) (35) EREs, which differ from the A2 ERE in the 5Ј-half-site, and the pS2 ERE (GGTCANNNTGGCC) (36), which differs from the A2 ERE in the 3Ј-half-site. Transient transfection assays were carried out in U2-OS cells to determine the ability of ER␤ to activate transcription of promoters containing a TATA box and a single A2, pS2, B1, or OT ERE. U2-OS cells were transfected with an ERE-containing CAT reporter plasmid, an ER␤ expression vector, and a CMV-␤gal control plasmid and exposed to ethanol vehicle or E 2 . CAT assays were performed and normalized for transfection efficiency. As shown in Fig. 1, ER␤ was able to significantly induce transcription through all four EREs in the presence of estrogen compared with vehicle controls (p Ͻ 0.01). However, ER␤ increased transcription to the highest degree (7.3-fold) through the A2 ERE, to an intermediate degree through the OT ERE (5.0-fold), and to the lowest degree through the pS2 and B1 EREs (2.4-and 1.9-fold, respectively). These findings indicate that ER␤ increases transcription of reporter plasmids contain-FIG. 1. ER␤ induces transcription of reporter plasmids containing the A2, pS2, B1, and OT EREs to different extents. An ER␤ expression vector, a ␤-galactosidase internal control vector, and a CAT reporter vector containing a single A2, pS2, B1, or OT ERE upstream from a TATA box were transiently transfected into U2-OS cells and exposed to ethanol vehicle (white bars) or 10 nM E 2 (gray bars). CAT activity was compared with the amount of ␤-galactosidase activity (cpm/␤-galactosidase (␤-gal) units) to normalize for differences in transfection efficiency. Experiments were repeated three times in duplicate, and data are expressed as the means Ϯ S.E. Asterisks indicate statistically significant induction in the presence of E 2 as determined by Student's t test (p Ͻ 0.01).
ing EREs with subtle differences in nucleotide sequence to different extents. Interestingly, although similar levels of transcription were previously observed with the A2, pS2, and OT ERE-containing reporter plasmids in Chinese hamster ovary cells, transcription of the B1 ERE-containing reporter plasmid was not increased in the presence of E 2 (28), suggesting that the B1 ERE may be differentially regulated by ER␤ in different cell contexts.
DNA-induced Changes in ER␤ Conformation-Studies with other nuclear receptors have suggested that subtle differences in nucleotide sequence can alter the conformation of DNAbound receptors and thereby influence transcription activation (20 -22, 24, 37, 38). To determine whether ER␤ conformation was altered when bound to different ERE sequences, protease sensitivity assays were carried out. Baculovirus-expressed purified ER␤ was combined with 32 P-labeled DNA fragments containing the A2, pS2, B1, or OT ERE. Increasing concentrations of proteinase K, which cleaves at aliphatic and aromatic residues, were added to the ER␤-DNA binding reaction. The receptor-DNA binding reactions were loaded onto a nondenaturing acrylamide gel and separated. In the absence of protease, ER␤ formed a single prominent receptor⅐DNA complex (Fig. 2). A higher order complex, which contained DNA-bound ER␤ and an Sf9 protein that copurified with the receptor, was also sometimes observed. Digestion of A2 ERE-bound ER␤ produced three slowly migrating complexes (P1-P3), whereas digestion of OT ERE-bound ER␤ resulted in two more rapidly migrating complexes (P4 and P5). Digestion of the pS2 and B1 ERE-bound receptors produced a combination of receptor⅐DNA complexes (P1-P5 and P1, P2, and P4, respectively).
A second protease was also utilized to examine amino acid accessibility and to confirm that differences in conformation existed. S. aureus protease V8, which cleaves at glutamic and aspartic acid residues, was added to the binding reactions, and the resulting complexes were fractionated on a nondenaturing gel (Fig. 3). Digestion of A2 ERE-bound ER␤ produced three slowly migrating complexes (V1-V3), whereas digestion of OT ERE-bound ER␤ resulted in two complexes with faster migration (V4 and V5). Digestion of the pS2 and B1 ERE-bound receptors produced two rapidly migrating complexes (V4 and V5) and two more slowly migrating bands (V1 and V2), with a predominance of V4 and V5 for pS2 ERE-bound ER␤ and a more even distribution of the four receptor⅐DNA complexes when ER␤ was bound to the B1 ERE. The unique patterns of ER␤⅐ERE digestion products indicate that there are differences in the availability of ER␤ cleavage sites when the receptor is bound to the four ERE sequences and suggest that individual EREs induce specific changes in receptor conformation.
Differential Interaction of ER␤-specific Antibodies with Receptor⅐ERE Complexes-To examine whether DNA-induced conformational changes occur in ER␤ using another method, antibody interaction studies were carried out. ER␤ was incubated with DNA fragments containing one of the four EREs and no antibody, a polyclonal antibody generated against the human ER␤ LBD (UICK-98), an ER␤-specific monoclonal antibody that recognizes an epitope at the start of the human ER␤ LBD (amino acids 273-285; CWK-F12), an antibody to the FLAG sequence at the amino terminus of the receptor (M2), or an ER␣-specific antibody (H151). ER␤ formed one major complex with the A2, pS2, B1, or OT ERE in the absence of antibody (Fig. 4, lanes 1-4). As anticipated, addition of the ER␣specific antibody H151 did not affect the ER␤⅐ERE complexes (lanes [17][18][19][20]. When the M2 antibody, which recognizes the amino-terminal FLAG sequence of purified ER␤, was added to the reaction mixture, the ER␤⅐DNA complex was supershifted (lanes 9 -12). Interestingly, the trimeric receptor⅐DNA⅐antibody complex formed with the OT ERE migrated more slowly than the receptor⅐DNA⅐antibody complex formed with the other EREs. This lower mobility OT ERE-bound ER␤⅐DNA⅐antibody complex was also observed when the ER␤-specific polyclonal antibody UICK-98 was incubated with the DNA-bound receptor (lanes [13][14][15][16]. In contrast to the antibody supershifts with M2 and UICK-98, incubation of receptor⅐DNA complexes with the LBD-specific monoclonal antibody CWK-F12 resulted in differential interaction with the ER␤⅐ERE complexes ( lanes  5-8). The interaction of ER␤ with the A2 ERE was unaffected by addition of CWK-F12. Binding of ER␤ to the B1 ERE was significantly diminished, and binding of ER␤ to the pS2 and OT FIG. 2. Proteinase K digestion of A2, pS2, B1, or OT ERE-bound ER␤ produces distinct digestion patterns. Baculovirus-expressed purified ER␤ was incubated with 32 P-labeled DNA fragments containing an A2, pS2, B1, or OT ERE. Increasing concentrations of proteinase K (1, 1.5, 10, and 20 ng) were added to the binding reaction and incubated for 10 min. The products were loaded onto nondenaturing acrylamide gels and separated. Final digestion products are indicated by numbers (P1-P5).
EREs was completely disrupted. These striking differences in receptor⅐DNA complex formation with four distinct ERE sequences, along with the differential interaction of the M2 and UICK-98 antibodies with OT ERE-bound ER␤, support the idea that ER␤ epitopes were positioned differently when the receptor was bound to the A2, pS2, B1, and OT EREs.
ERE Sequence-dependent Recruitment of AIB1 and TIF2 by ER␤-The recruitment of coactivator proteins is thought to be an important step in ER-mediated transcription activation (39,40). It seemed possible from our protease sensitivity and antibody interaction studies that allosteric modulation of the receptor conformation by different ERE sequences might influence the recruitment of coregulatory proteins and subsequently alter transcription activation. To determine whether association of coactivator proteins with ER␤ was ERE sequence-dependent, DNA pull-down experiments were carried out using U2-OS nuclear extracts. As shown in Fig. 5A, these U2-OS nuclear extracts contained substantial levels of TIF2 and AIB1, but did not contain ER␤. For the pull-down experiments, biotinylated DNA fragments containing a nonspecific sequence or the A2, pS2, B1, or OT ERE were adsorbed to streptavidinlinked magnetic beads, and ER␤ was allowed to bind to the DNA. U2-OS nuclear extracts were added; the beads were washed; and the ER␤⅐coactivator complexes were eluted. Recruitment of the coactivator proteins TIF2 (12) and AIB1 (13) to DNA-bound ER␤ was quantitated by Western analysis. When oligonucleotides contained a nonspecific DNA sequence, ER␤ was not retained on the DNA, and neither AIB1 nor TIF2 was recruited (Fig. 5B, NS). However, when the DNA fragments contained the A2, pS2, B1, or OT ERE, ER␤ was bound to the DNA, and AIB1 and TIF2 were recruited to the ERE-bound receptor. Interestingly, although the A2 and OT ERE-bound receptors recruited similar amounts of TIF2, the pS2 and B1 ERE-bound receptors recruited significantly less TIF2 than the A2 ERE-bound receptor (Fig. 5C). In contrast, the pS2, B1, or OT ERE-bound receptor recruited less AIB1 than the A2 EREbound receptor (Fig. 5D). Differences in coactivator recruitment could not be attributed to the lower affinity of ER␤ for the imperfect EREs compared with the consensus sequence since the amount of coactivator recruited to ER␤ was expressed as a ratio of coactivator to ER␤ for each sample. Given the difference in coactivator recruitment to ER␤ on the four discrete ERE sequences, our combined data from protease sensitivity, antibody interaction, and DNA pull-down studies suggest that the conformation of ER␤ is different when the receptor is bound to different DNA sequences and that these changes in conformation alter coactivator recruitment.
ERE-specific Enhancement of Transcription with AIB1 and TIF2-A number of laboratories have demonstrated that TIF2 and AIB1 increase transcription of reporter plasmids contain-ing the A2 ERE (11,12,17). However, the involvement of these proteins in transcription from promoters containing imperfect EREs is less clear. To determine whether TIF2 influences transcription of imperfect ERE-driven promoters, a CAT reporter plasmid containing a TATA box and no ERE or the A2, pS2, B1, or OT ERE was cotransfected into cells with an ER␤ expression plasmid, a ␤-galactosidase internal control vector, and increasing concentrations of TIF2 expression vector. Cells were treated with vehicle control or 10 nM E 2 and assayed for CAT activity. TIF2 expression resulted in a dose-dependent increase in transcription of the reporter plasmid containing an A2, pS2, or OT ERE (Fig. 6). Inclusion of 150 or 500 ng of the TIF2 expression vector resulted in 53 and 100% increases in transcription with the A2 ERE, 65 and 85% increases with the pS2 ERE, and 86 and 112% increases with the OT ERE, respectively, compared with no TIF2 expression vector addition. In contrast, no increases in transcription were observed when the TIF2 expression vector was included with the parental plasmid (Fig. 6, Ϫ) or the reporter plasmid containing the B1 ERE. Thus, TIF2 enhanced transcription through the A2 and OT EREs to a greater extent than through the pS2 ERE, but did not affect transcription when the CAT reporter plasmid contained the B1 ERE. When similar transfection experiments were carried out with a reporter plasmid containing one of the four EREs and an AIB1 expression vector, AIB1 enhanced transcription of the A2 ERE-containing reporter plasmid to the greatest extent and enhanced transcription of the OT EREcontaining reporter plasmid to an intermediate extent, but did not affect transcription of the pS2 and B1 ERE-containing reporter plasmids or the parental plasmid (Fig. 7). Inclusion of 150 or 500 ng of the AIB1 expression vector resulted in 42 and 83% increases in transcription with the A2 ERE and 22 and 35% increases with the OT ERE, respectively, compared with no AIB1 expression vector addition. In the absence of the ER, FIG. 4. ER␤-specific antibodies detect ERE-dependent changes in receptor conformation. Baculovirus-expressed purified ER␤ was incubated with a 32 P-labeled probe containing an A2, pS2, B1, or OT ERE. Antibodies (Ab) were added as indicated. Products were loaded onto nondenaturing acrylamide gels and separated.

FIG. 5. Recruitment of TIF2 and AIB1 to ER␤ is influenced by the ERE sequence.
A, 50 g of U2-OS nuclear extract was separated by SDS-polyacrylamide gel electrophoresis; blotted; and subjected to Western analysis with an ER␤-, AIB1-, or TIF2-specific antibody. B-D, pull-down experiments were carried out with immobilized oligonucleotides containing a nonspecific sequence (NS) or an A2, pS2, B1, or OT ERE. Baculovirus-expressed purified ER␤ and U2-OS nuclear extract were allowed to interact with the DNA. Unbound proteins were washed from the beads. Specifically bound complexes were eluted; separated by SDS-polyacrylamide gel electrophoresis; blotted; and probed for ER␤, TIF2, or AIB1. Association of these proteins with nonspecific sequencecontaining and A2, pS2, B1, and OT ERE-containing oligonucleotides is shown. Data for TIF2 (C) and AIB1 (D) association were quantitated from Western blots using ImageQuant software, and values are expressed as a ratio of TIF2 or AIB1 to ER␤. The amount of coactivator recruited to A2 ERE-bound ER␤ was compared with that recruited to pS2, B1, or OT ERE-bound ER␤ by Student's t test. Asterisks indicate statistically significant differences (p Ͻ 0.05).
overexpression of TIF2 or AIB1 failed to enhance transcription of the ERE-containing reporter plasmids (data not shown). Thus, both TIF2 and AIB1 enhanced the ER␤-and E 2 -dependent activation through the A2 and OT EREs. Only TIF2 enhanced transcription of the pS2 ERE, and neither TIF2 nor AIB1 was capable of augmenting transcription through the B1 ERE. DISCUSSION We have demonstrated that four naturally occurring ERE sequences exhibit different levels of ER␤-dependent transactivation in transient cotransfection assays. Our studies provide evidence that individual EREs induce changes in ER␤ conformation and that these conformational changes alter the ability of the receptor to recruit the coactivator proteins TIF2 and AIB1. The differential recruitment of coactivator proteins to the ERE-bound receptor may in turn influence transcription of estrogen-responsive genes.
Binding of ER␤ to Distinct ERE Sequences Results in Allosteric Modulation of Receptor Conformation-Protease sensitivity studies with two different proteases demonstrated that there were differences in accessibility of proteinase K and protease V8 cleavage sites when ER␤ was bound to four different EREs. These data indicate that binding of ER␤ to different ERE sequences elicits specific changes in ER␤ conformation. Interestingly, similar digestion patterns were produced when each ER␤⅐ERE complex was exposed to different proteases. When receptor⅐DNA complexes were digested with either proteinase K or protease V8, the A2 ERE-bound receptor produced lower mobility complexes; the OT ERE-bound receptor produced higher mobility complexes; and the pS2 and B1 EREbound receptors produced complexes with higher and lower mobilities. This similarity in digestion patterns when the receptor was bound to the same ERE but cleaved with a different protease most likely resulted from cleavage of adjacent proteinase K and protease V8 sites on exposed receptor surfaces. Similar digestion patterns were also observed for each of the EREs after chymotrypsin digestion of DNA-bound ER␤ (28).
The differential interaction of antibodies with the A2, pS2, B1, or OT ERE-bound receptor provided additional evidence that individual EREs induce specific changes in ER␤ conformation. Antibodies to both the amino terminus and LBD detected differences in epitope availability when ER␤ was bound to the four different ERE sequences. Although no single antibody differentiated between all four conformations of ER␤, taken together, the three antibodies utilized in our studies demonstrate that each of the four EREs induces unique changes in receptor conformation. Furthermore, since each antibody recognized more than one ER␤⅐DNA complex, the conformation of the entire receptor protein on each ERE sequence is likely to be a composite consisting of epitopes that are common to and variant from the conformation of ER␤ when bound to the other sequences.
Despite only 58% conservation of overall amino acid sequence between human ER␣ and ER␤ in the LBD (3), the crystal structures of the two ER LBDs are quite similar (8,18). X-ray crystallographic and mutation analyses of both the thyroid hormone receptor and ER␣ LBDs have been used to identify a coactivator interaction site for the LXXLL motif found in a number of coactivator proteins (18,41,42). ER␣ amino acids Leu 354 -Lys 362 of helix 3, Phe 367 -Val 368 of helix 4, Leu 370 from the turn between helices 4 and 5, and Gln 375 -Glu 380 of helix 5 form a shallow nonpolar groove with charged ends that coordinate the LXXLL motif of GRIP1 (18). In the ER␤ crystal structure (8), these key amino acids are similarly positioned, suggesting that the coactivator interaction surface is conserved in ER␣ and ER␤. The CWK-F12 antibody used in our studies recognizes ER␤ amino acids Leu 273 -Arg 285 , which map to the carboxyl terminus of helix 2 and the region between helices 2 and 3 (8,32). This region borders the amino acids in ER␣ and the corresponding amino acids in the thyroid hormone receptor that interact with coactivators in crystal structure studies. Significantly, the interaction of CWK-F12 with ER␤⅐ERE complexes was strikingly different, suggesting that ER␤ conformation in the region bordering the coactivator interaction site was influenced by ERE sequence. CWK-F12 blocked or severely reduced the interaction of the receptor with the pS2, B1, and OT EREs. In contrast, when the receptor was bound to the A2 ERE, the formation of the receptor⅐DNA complex was minimally affected by addition of CWK-F12, suggesting that the antibody interaction site was occluded or placed in a conformation that was not recognized. As Feng et al. (41) point out, the coactivator-binding surface in nuclear receptors is small (300 Å). The size of the interaction surface coupled with allosteric FIG. 6. Overexpression of TIF2 selectively enhances ER␤-mediated transcription through the A2, pS2, and OT EREs. An ER␤ expression vector, a ␤-galactosidase internal control vector, and a CAT reporter vector containing a single A2, pS2, B1, or OT ERE or no ERE (Ϫ) upstream from a TATA box were transiently transfected into U2-OS cells and incubated in the presence of vehicle (white bars) or 10 nM E 2 (gray bars). Increasing concentrations of TIF2 expression plasmid were added as indicated. CAT activity was compared with the amount of ␤-galactosidase activity (cpm/␤-galactosidase (␤-gal) units) to normalize for differences in transfection efficiency. Experiments were repeated three times in duplicate, and data are expressed as the means Ϯ S.E.

FIG. 7.
Overexpression of AIB1 selectively enhances ER␤-mediated transcription through the A2 and OT EREs. An ER␤ expression vector, a ␤-galactosidase internal control vector, and a CAT reporter vector containing a single A2, pS2, B1, or OT ERE or no ERE (Ϫ) upstream from a TATA box were transiently transfected into U2-OS cells and incubated in the presence of vehicle (white bars) or 10 nM E 2 (gray bars). Increasing concentrations of AIB1 expression plasmid were added as indicated. CAT activity was compared with the amount of ␤-galactosidase activity (cpm/␤-galactosidase (␤-gal) units) to normalize for differences in transfection efficiency. Data are derived from three independent transfection experiments and are expressed as the means Ϯ S.E. modulation of a nearby epitope in the LBD of ER␤ when the receptor is bound to the A2, pS2, B1, or OT ERE could alter the ability of the receptor to interact with coregulatory proteins. Differences in coactivator recruitment in the pull-down experiments reported here illustrate the functional consequences of the altered ER␤ conformation.
Allosteric Modulation of ER␤ Conformation Influences Recruitment of Coactivator Proteins and Transcription Activation-Coactivator and corepressor proteins bind to agonist-and antagonist-occupied ERs in vitro and play a critical role in transcription activation and repression (9 -14, 16, 17, 43, 44). Interaction of coregulatory proteins with ERs also occurs in vivo. For example, immunoprecipitation assays showed liganddependent interaction of AIB1 with endogenous ER in MCF-7 breast cancer cells (45). In the DNA pull-down experiments presented here, different levels of the coactivator proteins TIF2 and AIB1 were recruited to the A2, pS2, B1, and OT EREbound receptors. Both the pS2 and B1 ERE-bound receptors recruited significantly less TIF2 and AIB1 than the A2 EREbound receptor. The OT ERE-bound receptor recruited significantly less AIB1, but similar levels of TIF2 compared with the A2 ERE-bound receptor. Interestingly, the ability of EREbound ER␤ to recruit AIB1 and TIF2 was related to the ability of the receptor to activate transcription. The pS2 and B1 EREs, which were associated with significantly lower levels of ER␤bound coactivator proteins, were the least effective transcriptional enhancers. The A2 ERE, which was associated with the highest levels of ER␤-bound coactivator proteins, was the most potent transcriptional enhancer. The OT ERE, which was associated with high levels of ER␤-bound TIF2 (but not AIB1), had an intermediate ability to function as a transcriptional enhancer.
We believe that transcription of estrogen-responsive genes is subject to the cooperative interaction of numerous coregulatory proteins with the ER. We have demonstrated that two well characterized coactivators, TIF2 and AIB1, may play a role in selectively altering transcription of promoters containing discrete ERE sequences. The decreased recruitment of TIF2 and AIB1 to the pS2 and B1 ERE-bound receptors, the modest ability of ER␤ to activate transcription through the pS2 and B1 EREs in transient transfection assays, and the decreased ability of overexpressed TIF2 and AIB1 to augment transcription via the pS2 and B1 ERE-containing reporter plasmids compared with the A2 ERE suggest that TIF2 and AIB1 may be less important in E 2 -induced transcription through the pS2 and B1 EREs than through the A2 ERE. The ability of the OT ERE-bound receptor to recruit TIF2 (but not AIB1) in our in vitro assays, the intermediate ability of the OT ERE to function as a transcriptional enhancer, the potent transcriptional enhancement with overexpression of TIF2, and the moderate transcriptional enhancement with AIB1 suggest that TIF2 may play a more important role in regulating transcription of OT ERE-containing promoters than AIB1. We propose that differential recruitment of TIF2, AIB1, and other coregulatory proteins by the ERE-bound receptor plays an important role in regulating transcription of estrogen-responsive genes.
Modulation of Protein Conformation Provides a Mechanism for Differential Regulation of Gene Expression-A recent study demonstrated that the POU domain-containing transcription factor Pit-1 undergoes a conformational change in response to binding to its recognition site in the growth hormone gene, resulting in recruitment of cofactors that mediate transcriptional repression (19). In contrast, Pit-1 activates transcription when bound to its recognition site in the prolactin gene. Crystal structure analysis of Pit-1 bound to the prolactin or growth hormone gene recognition sequences showed dramatic alter-ation in the domain spacing of the Pit-1 protein. When Pit-1 was bound to its recognition sequence in the growth hormone gene, nuclear corepressor (NCoR) was recruited, and transcription was repressed. Scully et al. (19) propose that Pit-1-mediated repression of the growth hormone gene in certain cell types is mediated by conformational changes induced by the Pit-1-binding site and the resulting recruitment of corepressors.
Experiments carried out with thyroid receptor (TR)/retinoid X receptor (RXR) heterodimers support the idea that DNAinduced changes in receptor conformation can alter association of coactivator proteins. TR/RXR heterodimers are more resistant to trypsin digestion when bound to transcriptionally active thyroid response elements (TREs) than when bound to transcriptionally inactive TREs (22), suggesting that there are also differences in the conformation of the TR/RXR heterodimer when it is bound to different TRE sequences. Furthermore, fragments of steroid receptor coactivator 1 associated with TR/RXR heterodimers in the presence of a transcriptionally active TRE, but failed to associate with TR/RXR heterodimers in the presence of DNA containing a transcriptionally inactive TRE (23).
Combined with our previous studies of ER␣ (21), we have now demonstrated that the conformation of ER␣ and ER␤ is modulated by ERE sequence and that these DNA-induced changes in receptor conformation alter recruitment of coactivator proteins. Interestingly, our earlier studies with purified ER␣ and HeLa nuclear extracts showed that A2, pS2, and OT ERE-bound ER␣ receptors recruited similar amounts of TIF2; that B1 ERE-bound ER␣ recruited significantly less TIF2; and that the level of AIB1 recruitment by A2, pS2, B1, or OT ERE-bound ER␣ did not vary. The decreased ability of ER␤ to recruit TIF2 and AIB1 when bound to these same EREs may help to explain the decreased ability of ER␤ to enhance transcription of reporter plasmids containing these ERE sequences compared with ER␣ (28).
Few studies have examined the role of DNA sequence in modulating recruitment of coregulatory proteins to DNA-bound nuclear receptors. Here we have shown that ER␤ undergoes discrete changes in conformation when bound to EREs with slight variations in nucleotide sequence. These studies document the functional consequences of DNA-induced changes in receptor conformation and highlight the importance of each individual ERE sequence in regulating transcription of estrogen-responsive genes. We propose that ERE sequence can alter the conformation of the DNA-bound receptor and influence recruitment of regulatory proteins. Furthermore, the effect of DNA sequence on receptor conformation and subsequent coactivator recruitment is probably not limited to ERs, but likely plays a role in differential expression of other hormone-responsive genes.