Differential modulation of DNA conformation by estrogen receptors α and β

The human estrogen receptor (ER) induces transcription of estrogen-responsive genes upon binding to estrogen and the estrogen response element (ERE). To determine whether receptor-induced changes in DNA structure are related to transactivation, we compared the abilities of ERα and ERβ to activate transcription and induce distortion and bending in DNA. ERα induced higher levels of transcription than ERβ in the presence of 17β-estradiol. In circular permutation experiments ERα induced greater distortion in DNA fragments containing the consensus ERE sequence than ERβ. Phasing analysis indicated that ERα induced a bend directed toward the major groove of the DNA helix but that ERβ failed to induce a directed DNA bend. Likewise, the ERα DNA binding domain (DBD) and hinge region induced a bend directed toward the major groove of the DNA helix, but the ERβ DBD and hinge region failed to bend ERE-containing DNA fragments. Using receptor chimeras we demonstrated that the ERα DBD C-terminal extension is required for directed DNA bending. Transient transfection assays revealed that appropriately oriented DNA bending enhances receptor-mediated transactivation. The different abilities of ERα and ERβ to induce change in DNA structure could foster or inhibit the interaction of regulatory proteins with the receptor and other transcription factors and help to explain how estrogen-responsive genes are differentially regulated by these two receptors.

The human estrogen receptor (ER) induces transcription of estrogen-responsive genes upon binding to estrogen and the estrogen response element (ERE). To determine whether receptor-induced changes in DNA structure are related to transactivation, we compared the abilities of ER␣ and ER␤ to activate transcription and induce distortion and bending in DNA. ER␣ induced higher levels of transcription than ER␤ in the presence of 17␤-estradiol. In circular permutation experiments ER␣ induced greater distortion in DNA fragments containing the consensus ERE sequence than ER␤. Phasing analysis indicated that ER␣ induced a bend directed toward the major groove of the DNA helix but that ER␤ failed to induce a directed DNA bend. Likewise, the ER␣ DNA binding domain (DBD) and hinge region induced a bend directed toward the major groove of the DNA helix, but the ER␤ DBD and hinge region failed to bend EREcontaining DNA fragments. Using receptor chimeras we demonstrated that the ER␣ DBD C-terminal extension is required for directed DNA bending. Transient transfection assays revealed that appropriately oriented DNA bending enhances receptor-mediated transactivation. The different abilities of ER␣ and ER␤ to induce change in DNA structure could foster or inhibit the interaction of regulatory proteins with the receptor and other transcription factors and help to explain how estrogen-responsive genes are differentially regulated by these two receptors.
Estrogen is a hormone of critical importance in regulating the development, growth, and maintenance of reproductive tissues and also influences cardiovascular, neural, and skeletal cell function (1)(2)(3)(4)(5)(6). The effects of estrogen are mediated through interaction of the estrogen receptor (ER) 1 with estrogen response elements (EREs) residing in target genes. Because the ER-ERE interaction plays such a crucial role in gene expression, there has been and continues to be great interest in understanding how this interaction leads to changes in transcription.
With the recent revelation that two ER subtypes exist, the previously identified ER␣ (7)(8)(9)(10) and the more recently cloned ER␤ (11)(12)(13)(14), our vision of how estrogen brings about its effects in target cells must now be reevaluated and take into consideration the actions of both receptor subtypes. Although the amino acid sequence in the DNA and hormone binding domains of ER␣ and ER␤ is highly conserved, other regions of the two receptors are more varied (15). These differences in amino acid sequence affect the affinities of ER␣ and ER␤ for various ligands and ERE sequences, influence the association of the receptors with coactivator proteins, and thereby alter the abilities of these two receptors to modulate gene expression (13, 14, 16 -25).
A number of thermodynamic and structural studies have demonstrated that specific contacts between protein and DNA are often accompanied by conformational changes in protein, DNA, or both (26 -31). We and others have demonstrated that binding of ER␣ to ERE-containing DNA fragments induces conformational changes in DNA structure suggesting that the ability of ER␣ to induce changes in DNA structure may be related to its ability to activate transcription (32)(33)(34)(35)(36)(37)(38)(39)(40). To determine whether ER␤ was capable of altering DNA conformation and whether ER␣-and ER␤-induced transcription activation might be related to the abilities of these two receptors to induce conformational changes in DNA structure, we have compared the abilities of ER␣ and ER␤ to activate transcription and induce flexibility and directed bending in ERE-containing DNA fragments and determined the effects of DNA bending on transcription activation.

EXPERIMENTAL PROCEDURES
Transient Transfections-Transient cotransfections carried out with Chinese hamster ovary (CHO) cells utilized 3 g of the CAT reporter plasmid, ERE TATA-CAT (41), which contains an ERE in the BglII site of ATC0, 200 ng of the ␤-galactosidase expression vector pCH110 (Amersham Biosciences, Inc.), and 5 ng of the human ER␣ expression vector pCMV5 hER (42) or the human ER␤ expression vector pCMV5 hER␤ (43). Transient transfections carried out in U2 osteosarcoma (U2-OS) cells utilized 25 ng of ER␣ or ER␤ expression vector and 3 g of a CAT reporter plasmid containing an ERE in the SalI site of ATC0 (ATC1, Ref. 44) alone or in combination with an intrinsically bent DNA sequence (ATC1-OP and ATC1-IP, previously described as 3A 6 /ERE3-CAT and 3A 6 /ERE3.4-CAT, Ref. 36). Cells were transfected with Lipofectin (Life Technologies, Inc.) as described (25). CAT activity was quantitated as described (24). ␤-galactosidase activity was determined (45) and used to normalize for transfection efficiency.
Expression and Purification of Wild Type and Truncated ER␣ and ER␤ Proteins-FLAG-tagged ER␣ and ER␤ were expressed and purified as previously described (23,24). Protein purity was assayed by Coomassie blue-stained SDS-PAGE. ER concentration was determined using a hydroxyapatite binding assay (24).
CD␣, which contained the DBD and hinge region of ER␣, was expressed using the expression vector pET-21b(ϩ):flag:hER␣DBD (39) kindly provided by David J. Shapiro (University of Illinois, Urbana, IL). CD␤ was constructed as described above to encode the region of ER␤ (amino acids 134 -260) corresponding to the region included in the ER␣ construct (amino acids 166 -307). 500-ml cultures of transformed BL21DE3plysS cells were induced with 1 mM isopropyl-1-thio-␤-d-galactopyranosideisopropyl-1-thio-␤-D-galactopyranoside. 10 M ZnCl 2 was added to help stabilize the DBD zinc fingers. Induced cultures were incubated for 3 h with shaking at 37°C and then chilled on ice for 5 min and pelleted at 4700 ϫ g for 5 min at 4°C. Cells were lysed with one freeze/thaw cycle. 5 ml of TEGDZ (50 mM Tris, pH 7.5, 1 mM EDTA, pH 8.0, 10% glycerol, 5 mM dithiothreitol, 50 M ZnCl 2 , 25 g/ml aprotinin, 5 g/ml phenylmethylsulfonyl fluoride, 50 g/ml leupeptin, 1 g/ml pepstatin) with 0.5% Triton X-100 was added to the cell lysate and incubated with rotation for 30 min at 4°C. The lysate was centrifuged at 142,000 ϫ g for 30 min at 4°C, and the pellet was discarded. Polyethyleneimine was added dropwise to the supernatant to 0.1%, incubated on ice for 10 min, and centrifuged at 142,000 ϫ g for 30 min at 4°C. The extract was diluted with 2 ml of TEGDZ and incubated with anti-FLAG M2 agarose (A1205, Sigma) with rotation for 4 h at 4°C. The CD-antibody complexes were washed and eluted as described for the full-length receptors (23,24), except that the wash and elution buffers for CD␣ contained 400 mM NaCl.
Protein purity of CD␣, CD␤, CD␣/␤, and CD␤/␣ was assessed on Coomassie blue-stained SDS-PAGE. Protein concentrations were determined using the Bio-Rad protein assay (Bio-Rad, Hercules, CA) with bovine serum albumin as a standard.
Preparation of DNA Probes-Circular permutation experiments were carried out with 427-bp DNA fragments isolated from the DNA bending vector B3ConsERE (41). B3ConsERE was digested with EcoRI, HindIII, EcoRV, NheI, or BamHI to produce DNA fragments containing a consensus ERE near the 3Ј end, at an intermediate 3Ј position, in the middle of the fragment, at an intermediate 5Ј position, or near the 5Ј end, respectively. All DNA fragments were composed of identical nucleotide sequence and differed only in the location of the ERE. The DNA fragments were isolated and 32 P labeled as described (32).
For phasing analysis, gel mobility shift assays were performed with 10,000 cpm of 32 P-labeled DNA fragments containing an ERE and an intrinsic DNA bend in binding buffer with 50 M ZnCl 2 . Full-length receptor assays used 100 fmol of purified baculovirus-expressed ER␣ or ER␤ and included 50 ng of poly(dI/dC), 20 mM KCl, and 50 nM E 2 in the binding reaction buffer. Assays with truncated DBD-hinge proteins used 350 or 600 fmol of purified, bacterially expressed CD␣ or CD␤, respectively, and included 100 ng of poly(dI/dC) and 50 mM KCl (CD␣) or no additional KCl (CD␤) in the binding buffer. Assays with chimeric DBD-CTE proteins utilized 50 or 10 fmol of purified, bacterially expressed CD␣/␤ or CD␤/␣, respectively, and included 100 ng of poly (dI/dC) and 50 mM or 10 mM KCl in the binding buffer. Reactions were incubated for 10 -15 min at room temperature prior to fractionation on low ionic strength polyacrylamide gels (48) at 4°C with buffer recirculation. 2500 cpm each of NheI-or BamHI-digested bending standards were included on each gel.
The migration distances of the ER-DNA complexes, free probe, and DNA bending standards were determined using a Molecular Dynamics PhosphorImager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA).
Calculation of DNA Distortion and Bending Angles-The magnitude of each ER-induced distortion angle was determined from circular permutation experiments using two methods. The magnitude of the distortion angle was determined graphically by plotting the relative mobilities of the DNA bending standards (47) as a function of their known DNA bend angles. Polynomial regression analysis was used to determine the distortion angles induced by ER␣ and ER␤. Distortion angles were also determined using the empirical equation m / e ϭ cos (k␣/2), where m is the relative mobility when the DNA bend is at the middle of the fragment and e is the relative mobility when the bend is at the end of the fragment (47). A k value of 1.130 was determined by comparing the relative mobility of the DNA bending standards with the known bend angles as described (37). Calculations of DNA distortion angles are presented as the mean Ϯ S.E. from four (ER␣) or six (ER␤) independent experiments.
The magnitude of the directed DNA bend was determined from phasing analysis experiments by comparing the relative mobility of each ER-DNA complex with the relative mobility of the DNA bending standards. The magnitude of the directed bend angle (␣ B ) was determined using the equation tan (k␣ B /2) ϭ (A PH /2)/(tan(k␣ C /2)) where ␣ C is the intrinsic bend (49). The k value of 1.154 was determined as described for circular permutation analysis. The relative mobility of each ER-DNA complex was calculated by dividing the mobility of the ER-DNA complex by the mobility of the corresponding uncomplexed DNA fragment. Interexperimental variation was eliminated by dividing the relative mobility of the ER-DNA complex by the average relative mobility to obtain the normalized relative mobility (). The amplitude of phasing, A PH , was calculated by subtracting min from max . The fragment that migrated the farthest was determined using the greatest normalized relative mobility ( max ). Calculations for directed bend angles are presented as the mean Ϯ S.E. from 4 (ER␣), 10 (ER␤), 11 (CD␣), 8 (CD␤), or 6 (CD␣/␤ and CD␤/␣) independent experiments.

ER␣ and ER␤ Activate Transcription to Different Extents-
Transient transfection assays were used to examine the ability of ER␣ and ER␤ to activate transcription of a CAT reporter vector containing a single consensus ERE. CHO cells were transiently transfected with an ER␣ or ER␤ expression vector, a CAT reporter plasmid containing a consensus ERE upstream of a TATA box, and a ␤-galactosidase control vector. Both receptors significantly induced transcription in the presence of 10 nM E 2 when compared with an ethanol vehicle (Fig. 1, p Ͻ 0.0001). ER␣ and ER␤ induced 19-and 11-fold increases in transcription, respectively. Similar results were obtained in U2-OS cells. 2 Thus, ER␣ was a more potent activator of transcription than ER␤.
ER␣ and ER␤ Induce Distortion in ERE-containing DNA Fragments-Previous studies from our laboratory have demonstrated that binding of ER␣ to ERE-containing DNA fragments induces distortion and directed DNA bending (32)(33)(34)(35)(36)(37)(38)(39). These studies and others suggest that the ability of a protein to induce conformational changes in DNA structure may be related to its ability to enhance transcription (50 -52). To determine whether ER␤ induces distortion in ERE-containing DNA fragments and whether the magnitude of distortion might be related to the ability of ER␣ and ER␤ to activate transcription, circular permutation analysis was carried out with ER␣ and ER␤. Circular permutation assays assess the migration of pro-tein-DNA complexes through a native acrylamide gel to determine the magnitude of protein-induced DNA distortion. Because the geometry of a DNA fragment affects its migration through an acrylamide gel, a DNA fragment with a distortion near the end will migrate farther through the gel matrix than a DNA fragment with a distortion in the middle.
427-bp 32 P-labeled DNA fragments containing a consensus ERE near the end, in the middle, or at an intermediate position in the DNA fragment were incubated with purified, baculovirus-expressed ER␣ or ER␤ and fractionated on a nondenaturing polyacrylamide gel. DNA bending standards, which contained 54, 72, or 90°intrinsic DNA bends (53,54) were also included on all gels. As seen in Fig. 2A, distinct differences in the migration of the ER-bound DNA fragments were observed with both ER␣ and ER␤. The migration of the receptor-DNA complexes was dependent on the location of the ERE in the DNA fragment. When the ERE was near the end of the DNA fragment (R, B), the ER-DNA complex migrated farther than when the ERE was present at an intermediate position in the DNA fragment (H, N). When the ERE was at the middle of the DNA fragment (V), the ER-DNA complex migrated even more slowly indicating that both ER␣ and ER␤ induced distortion in DNA structure. The magnitude of distortion was determined from four (ER␣) or six (ER␤) independent experiments using two methods. Graphical analysis, which compares the relative mobilities of the receptor-DNA complexes with the relative mobilities of the DNA bending standards (Fig. 2B), indicated that ER␣ and ER␤ induced distortion angles of 74.6 Ϯ 4.1°and 63.9 Ϯ 0.9°, respectively. Using the empirical equation m / e ϭ cos (k␣/2) (47), similar distortion angles were obtained for ER␣ (76.0 Ϯ 4.3°) and ER␤ (64.5 Ϯ 1.0°). Thus, both ER␣ and ER␤ induce distortion in ERE-containing DNA fragments, but ER␣ induces a larger distortion angle than ER␤ (p Ͻ 0.05 comparing ER␣ to ER␤ by Student's t test).
ER␣, but Not ER␤, Induces a DNA Bend toward the Major Groove of the DNA Helix-Circular permutation analysis can be used to determine the degree of distortion or flexibility induced by protein binding, but it does not provide information about the orientation of a DNA bend. To determine the magnitude and direction of receptor-induced DNA bending, phasing analysis was carried out. This assay utilizes DNA fragments with an ERE 26, 28, 30, 32, 34, or 36 bp upstream of an intrinsic DNA bend so that the orientation of the ER-induced and the intrinsic DNA bends are incrementally varied over one turn of the DNA helix. At some point the ER-induced and intrinsic DNA bends will be in phase, a larger DNA bend will be produced, and the migration of the ER-DNA complex through the gel matrix will be inhibited. Likewise, at some point the ER-induced and intrinsic DNA bends will be out of phase, the DNA fragment will have a smaller overall bend, and the ER-DNA complex will migrate more rapidly through the gel. By monitoring the migration of the ER-DNA complexes, it is possible to determine the magnitude and direction of an ER-induced DNA bend. 32 P-labeled DNA fragments containing an ERE and an intrinsic DNA bend were incubated with purified, baculovirusexpressed ER␣ or ER␤ and separated on a nondenaturing polyacrylamide gel. When the DNA fragment contained an ERE 32 bp upstream of the intrinsic DNA bend, the ER␣-DNA complex migrated more rapidly (Fig. 3). This 32-bp spacing places the ERE and intrinsic DNA bend on the same side of the DNA helix. The rapid migration of the ER␣-DNA complex indicates that the ER␣-induced and intrinsic DNA bends must be out of phase. Since we know that the intrinsic DNA bend is oriented toward the minor groove of the DNA helix (55), the ER␣-induced DNA bend must be directed toward the major groove of the DNA helix. Combined data from four independent experiments indicated that the ER␣-directed DNA bend was 11.0 Ϯ 1.6°. These findings are in good agreement with earlier work examining ER␣-induced DNA bending (37). Surprisingly, when ER␤ was incubated with these same DNA fragments, none of the ER␤-bound DNA fragments consistently migrated farther than the others in 10 independent experiments, demonstrating that ER␤ does not induce a directed bend in EREcontaining DNA fragments. Only ER␣ was capable of inducing a directed bend in these ERE-containing DNA fragments and that bend was directed toward the major groove of the DNA helix. Thus, our results with ER␤ are in stark contrast to our previous studies of 11 different wild type and mutant ER␣ proteins, each of which induced a DNA bend directed toward the major groove of the DNA helix (36,37,56). 2 The DNA Binding Domain and Hinge Region of ER␣, but Not ER␤, Induce a Directed Bend in ERE-containing DNA Fragments-We have previously demonstrated that wild type, truncated, and mutant ER␣s induce a DNA bend toward the major groove of the DNA helix (36 -38, 56) and were quite surprised that full-length ER␤ was incapable of inducing a directed DNA bend. To determine how this difference in ER-induced DNA bending might occur, we carried out phasing analysis with the were incubated with purified, baculovirus-expressed ER␣ or ER␤ and fractionated on a nondenaturing polyacrylamide gel. 32 P-labeled DNA bending standards containing A 6 -tracts with 54, 72, or 90°bends in the middle (upper bands) or end (lower bands) of the DNA fragment were included on the same gel. B, a standard curve was developed by plotting the relative mobilities of the bending standards against their corresponding distortion angles. The distortion angles induced by ER␣ and ER␤ were determined by plotting the relative mobilities of the ER␣-or ER␤-DNA complexes on the standard curve. region of the ER responsible for specific binding, the DBD. Because the DBD alone has a low affinity for the ERE (46,57), the hinge region (domain D) was also included. The DBD and hinge region of ER␣ (CD␣) or ER␤ (CD␤) shown in Fig. 4 were expressed and purified. When the 32 P-labeled DNA phasing probes were combined with purified CD␣, the complex containing DNA fragments with 32 bp between the ERE and intrinsic DNA bend migrated the farthest (Fig. 5A). These findings demonstrate that CD␣, like wild type, truncated, and mutant ER␣s from our previous studies (36 -38, 56), induced a bend directed toward the major groove of the DNA helix. Data from 11 independent experiments demonstrated that the magnitude of the CD␣-directed DNA bend was 6.7 Ϯ 0.6°. In contrast, the migration of the CD␤-DNA complexes did not differ consistently, but followed the same pattern of migration as the free DNA indicating that CD␤ was incapable of inducing a directed DNA bend. These findings demonstrate that CD␣ could induce a bend toward the major groove of the DNA helix in EREcontaining DNA fragments, but CD␤ could not.
The amino acids that comprise the ER␣ and ER␤ DBDs are nearly identical. However, the hinge regions of these receptors vary significantly in amino acid sequence. Thus, we suspected that the hinge region of these two receptors might account for the difference in receptor-induced DNA bending we observed. Previous studies with the Type II and orphan nuclear receptors have defined a region C-terminal to the DBD, the CTE, which has been implicated in stabilizing the DBD-DNA interaction (58 -60). This region has also been implicated in stabilizing ER-DNA binding (46,57). To determine whether the CTE of ER␣ and ER␤ played a role in DNA bending, chimeric proteins were expressed, purified, and used in phasing analysis experiments (Fig. 4). These proteins contained the ER␣ DBD and the ER␤ CTE (CD␣/␤) or the ER␤ DBD and the ER␣ CTE (CD␤/␣). When these proteins were combined with the 32 P-labeled DNA fragments containing an ERE and an intrinsic DNA bend, the migration of complexes containing CD␤/␣ chimera varied when the orientation of the CD␤/␣-induced and intrinsic DNA bends was altered. The complex containing 34 bp between the ERE and intrinsic DNA bend consistently migrated more rapidly than the other CD␤/␣-containing complexes (Fig. 5B). Assuming 10.5 nucleotides per DNA turn, CD␤/␣, like the full-length ER␣ and CD␣, induced a DNA bend directed toward the major groove of the DNA helix. Using data from six independent experiments, we determined that the magnitude of the CD␤/␣induced DNA bend was 5.6 Ϯ 0.7°. In contrast, when CD␣/␤ was utilized, the migration of protein-bound phasing probes mimicked the migration of unbound probes indicating that CD␣/␤ was unable to alter DNA conformation. Thus, the ER␣ CTE plays an important role in directed DNA bending.
DNA Bending Influences ER-mediated Transactivation-Our in vitro binding assays demonstrated that ER␣, CD␣, and CD␤/␣ induced DNA bending directed toward the major groove of the DNA helix, but that ER␤, CD␤, and CD␣/␤ failed to induce a directed DNA bend. To determine whether receptorinduced DNA bending might play a role in E 2 -mediated transactivation, another set of transfection assays was carried out. Reporter plasmids that contained an ERE alone or in combination with an intrinsic DNA bend directed toward the minor groove of the DNA helix were tested for their abilities to enhance ER-mediated transcription activation. When the reporter plasmid contained a single ERE (Fig. 6, ATC1), ER␣ significantly increased CAT activity in the presence of E 2 . The magnitude of the increase was less than observed in Fig. 1, however, because the ATC1 reporter plasmid contained a less potent promoter than ERE TATA-CAT used in those experiments. When the reporter plasmid contained a promoter with the centers of the intrinsic and ER␣-induced DNA bends on the same sides of the DNA helix so that the intrinsic and induced DNA bends were out of phase (ATC1-OP), the level of E 2induced CAT activity was similar to the level observed with ATC1. When the reporter plasmid contained a promoter with the centers of the intrinsic and ER␣-induced DNA bends on opposite sides of the DNA helix so that the intrinsic and induced DNA bends were in phase (ATC1-IP), the level of E 2induced CAT activity was greater than the level observed with ATC1 or ATC1-OP. These data suggest that orienting an intrinsic DNA bend so that it complements an ER␣-induced DNA bend enhances ER␣-mediated transcription.
When the ER␤ expression plasmid was utilized in transfection assays, no increase in CAT activity was observed with ATC1, again reflecting the decreased potency of the ATC1 promoter compared with the ERE TATA-CAT promoter (Fig.  1). When ATC1-OP was utilized, ER␤ was still transcriptionally inert. However, when ATC1-IP was utilized, ER␤ was capable of substantially increasing CAT activity in the presence of E 2 . These combined experiments are consistent with previous studies carried out with ER␣ in CHO cells (36) and highlight the importance of DNA bending and the orientation of that bend in E 2 -mediated transactivation. DISCUSSION We have demonstrated that ER␣ and ER␤ have different abilities to activate transcription and alter DNA structure. Although both ER␣ and ER␤ can distort DNA structure, only ER␣ can induce a directed DNA bend. Taken together with earlier studies, we have now demonstrated that wild type ER␣ and 10 different ER␣ mutants induce directed DNA bending (36,37,56). 2 These 11 ER␣ proteins all contained the CTE, and all induced a bend directed toward the major groove of the DNA helix. Even the smallest ER␣ construct utilized, which included the ER␣ DBD and hinge region (CD␣), was capable of inducing a bend directed toward the major groove of the DNA helix. However, when the ER␣ DBD was combined with the ER␤ CTE (CD␣/␤), the chimeric protein failed to bend ERE-containing DNA fragments. Neither full-length ER␤ nor the DBD and hinge region of ER␤ (CD␤) was capable of inducing a directed DNA bend. However, if the ER␤ DBD was combined with the ER␣ CTE (CD␤/␣), this chimeric protein gained the ability to bend ERE-containing DNA fragments. Thus, replacing the ER␤ CTE with the ER␣ CTE resulted in a gain-of-function and replacing the ER␣ CTE with the ER␤ CTE resulted in a lossof-function. Taken together, these combined experiments highlight the importance of the ER␣ CTE in receptor-induced DNA bending.
The amino acids that comprise the human ER␣ and ER␤ DBDs are 96% conserved and differ by only two amino acids (12,61). However the CTE adjacent to the DBD varies substantially in amino acid sequence. Our experiments demonstrate that the ER␣ CTE plays a critical role in receptor-induced DNA bending. It is unclear at this point whether the entire ER␣ CTE or a specific domain within the ER␣ CTE is required for inducing these structural changes in ERE-containing DNA fragments. Interestingly, the type II thyroid hormone and retinoid X receptors and the orphan receptor RevErb contain ordered structures in the CTE that interact with the minor groove of the DNA helix and stabilize the protein-DNA interaction (58 -60). Of particular interest is the Grip-box in the RevErb CTE, which forms direct contacts with bases in the minor groove. In contrast to these nuclear receptors, the ER␣ CTE is unstructured. However, it is possible that a region in the ER␣ CTE plays a similar role in stabilizing the receptor-DNA interaction and fostering structural changes in ERE-containing DNA fragments.
The ability of a protein to induce conformational changes in DNA structure is not limited to ER␣ and ␤. A number of nuclear receptors induce DNA bending upon binding to their cognate response elements including the retinoid X receptor, thyroid hormone receptor, glucocorticoid, and progesterone receptors (62)(63)(64)(65)(66). Proteins involved in the formation of transcription complexes also induce DNA bending. Crystal structure analysis has revealed that T7 RNA polymerase induces a 63 o bend (67). This polymerase-induced DNA bending has been implicated in lowering the DNA melting point to allow for separation of the DNA strands and initiation of transcription. Fos/Jun heterodimers and Jun/Jun homodimers induce bends oriented in opposite directions (54), which may influence their abilities to recruit and interact with different transcription factors. LEF1-and SRY-induced DNA bending appears to play a role in formation of higher order transcription complexes (68 -70). The C-terminal zinc finger of the GATA-1 transcription factor induces a DNA bend that allows for homodimerization and stable DNA binding (71). Prokaryotic transcription factors, including the Borrelia burgdorferi Hbb protein, which is thought to be important in bacterial DNA replication (72), numerous cytosine-5 methyltransferases (73), and the E. coli catabolite activator protein CAP (74) cause conformational changes in their cognate DNA recognition sequences. Thus, the modification of DNA structure by proteins involved in DNA FIG. 5. Determination of DBD-induced directed DNA bends. A, 32 P-labeled ERE-containing DNA fragments were incubated with purified, bacterially expressed CD␣ or CD␤ and fractionated on a nondenaturing polyacrylamide gel. B, 32 P-labeled ERE-containing DNA fragments were incubated with purified, bacterially expressed CD␣/␤ or CD␤/␣ and fractionated on a nondenaturing polyacrylamide gel.
FIG. 6. Influence of an intrinsic DNA bend on ER␣-and ER␤-mediated transcription. Cells were transiently transfected with a human ER␣ or ER␤ expression vector, a reporter plasmid containing an ERE (ATC1) alone or in combination with an intrinsic DNA bend that is out of phase (ATC1-OP) or in phase (ATC1-IP) with the ER␣-induced DNA bend, and a ␤-galactosidase control vector and treated with ethanol vehicle (open bars) or 10 nM E 2 (shaded bars). Data from three independent experiments were combined and are expressed as the mean Ϯ S.E. Student's t tests were used to determine whether statistical differences between ethanol-and E 2 -treated groups existed. Asterisks (*) indicate significant differences in CAT activity in the ethanol-and E 2 -treated cells (p Ͻ 0.05).
replication, modification, and transcription activation is widely used in both prokaryotic and eukaryotic systems.
Our laboratory and others have documented the enhanced ability of ER␣ to activate transcription of reporter plasmids in transient transfection experiments compared with ER␤ (11,12,20,23,43,75). We have now demonstrated that full-length ER␣, CD␣, and a CD␤/␣ chimera induce directed DNA bending but that full-length ER␤, CD␤, and a CD␣/␤ chimera cannot. The ability of an appropriately aligned DNA bend to enhance E 2 -mediated transcription provides evidence that changes in DNA structure can influence ER␣-mediated transactivation. Furthermore, the fact that an appropriately positioned intrinsic DNA bend is able to help establish ER␤-mediated estrogen responsiveness substantiates the role of DNA bending in ER␤mediated transactivation. It seems possible that the decreased ability of ER␤ to activate transcription may in part be due to its limited ability to induce changes in DNA flexibility and directed bending. ER-induced changes in DNA architecture could provide structural motifs required for recruitment of transcription factors or the scaffolding required for assembly of a basal transcription complex. Alternatively, DNA bending could help stabilize DNA loop formation and foster binding and interaction of multiple DNA-bound transcription factors. Our combined studies suggest that DNA bending influences the ability of ER␣ and ER␤ to activate transcription and that intrinsic and protein-induced changes in DNA structure could play an important role in regulating expression of estrogen-responsive genes in target cells.