Equilibrium Binding of Estrogen Receptor with DNA Using Fluorescence Anisotropy*

Interaction of estrogen receptor (ER) with DNA sequences known as estrogen response elements (ERE) is required for estrogen regulation of the expression of target genes. To characterize the affinity and specificity of ER interaction with ERE sequencesin vitro under equilibrium conditions, fluorescence anisotropy assays were performed using recombinant, purified ER and a fluorescein-labeled 35-base pair oligonucleotide bearing an idealized palindromic ERE. In buffer containing 100 mm KCl, the baculovirus-expressed, purified human ER bound with similar affinity to the consensus ERE and a mutant ERE with a single base pair change per half-site. Above 225 mm KCl, ER exhibited discrimination between the consensus and mutated ERE targets. Between 225 and 275 mm KCl, binding to the consensus ERE was independent of salt concentration and occurred with an equilibrium dissociation constant (K d ) of 1.8 ± 0.6 nm, whereas binding to the mutant ERE was not detected at ER concentrations below 100 nm under the same conditions. At 300 mm KCl, the K d for the consensus ERE increased approximately 25-fold, suggesting complex salt concentration dependence. Both estrogen-occupied and unoccupied ER bound to the consensus ERE sequence with similar affinity, indicating that estrogen affects ER activity at a step other than DNA binding. Unlike the full-length ER, the recombinant DNA binding domain of ER did not discriminate between the consensus and mutated ERE sequences even at buffer salt concentrations greater than 200 mm NaCl, suggesting that ER sequences outside the DNA binding domain may be important in promoting specific binding.

As a member of the superfamily of steroid hormone receptors, the estrogen receptor (ER) 1 is a 67-kDa nuclear protein that regulates transcription for genes involved in cellular differentiation, development of the female reproductive system, and homeostasis (1)(2)(3). ER binds to specific DNA sequences designated estrogen response elements (ERE), which are present in promoters of target genes. The consensus ERE is a 13-base pair (bp) palindromic sequence, consisting of inverted repeats of the half-sites 5Ј-GGTCA-3Ј separated by a 3-bp spacer, found in the 5Ј-flanking region of the Xenopus and chicken vitellogenin A2 genes (4 -6). ER consists of conserved structural and functional domains ( Fig. 1) for DNA-binding in region C, nuclear localization in region D, and hormone-binding in region E (7,8). Two transcriptional activation function domains, designated TAF-1 and TAF-2, are found in regions A/B and E, respectively (8 -11). The physiological ligand for ER is 17␤-estradiol (E 2 ), which diffuses into target cells and binds to ER. Binding of E 2 to ER induces conformational changes in the receptor (12)(13)(14) and dissociation of receptor-associated proteins including hsp90 and p59 (15,16), leading to the assembly of an active transcriptional complex responsible for the positive and negative regulation of target genes.
One of the principal requirements for E 2 -induced regulation of target genes is the specific association of ER with the ERE. The ability of unpurified ER from rat uterine cytosolic extracts to discriminate the consensus ERE from other DNA sequences was examined by using an avidin-biotin complexed with DNA assay to quantify the binding interaction (17). These assays indicated that ER recognized the vitellogenin ERE (vitERE) with an affinity that was three orders of magnitude greater than that for a mutated form of the ERE (mutERE) bearing 1 base change per half-site in a salt-dependent manner. This unpurified ER also bound to the vitERE with an affinity that was five orders of magnitude greater than that for plasmid DNA. The highest level of discrimination was found with binding buffer containing physiological levels of salt (17). The ability of purified ER to discriminate between ERE sequences in the absence of other proteins at various salt concentrations has not been examined in such detail under equilibrium conditions.
The amino acid sequences in the DNA binding domain (DBD) of ER that are necessary for specific interaction of the ER with ERE have been located in the N-terminal zinc finger (18,19) and verified by the crystal structure of the DBD bound to the ERE (20). Although a great deal of evidence has been accrued for the ER DBD, there are several functional differences noted between the DBD fragment and the full-length receptor. The DBD binds to the ERE with lower affinity than the full-length ER (21). The DNA bending induced by ER DBD binding was found to be 34°, rather than the 56°bending detected for the full-length receptor (22,23). Studies using a construct consisting of the DBD from region C plus sequences from region D indicated that this larger protein fragment bound to imperfect * These studies were supported, in part, by the College of Agricultural and Life Sciences, University of Wisconsin, by National Institutes of Health Grants HD08192 and HD07259 (to J. G.), HD31299 and HD0728 (to A. M. N.), and a Whitaker Foundation grant (to C. A. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ERE sequences more stably than the DBD alone (19). Thus, these results suggest that protein sequences other than those present in the DBD may be important for high affinity binding of the ER to the ERE.
The role of E 2 in the ability of ER to bind to DNA has been controversial. Previous models of E 2 action suggested that hormone binding induced interaction of ER with DNA. Several studies using gel shift analysis and quantitative DNA binding assays have indicated that E 2 -occupied and unoccupied ER from rat uterine extracts bind to ERE sequences with similar affinity (24,25). Ligand-independent binding of human ER to ERE sequences has also been reported in transiently transfected cells (26) and in whole cells containing endogenous ER (27). However, it has also been shown that human ER expressed from a recombinant baculovirus and produced by in vitro transcription/translation requires E 2 for DNA binding (28).
In these studies, we used fluorescence anisotropy to examine the specificity and affinity of the ER-ERE interaction under equilibrium conditions. The ability of the full-length purified ER and ER DBD to discriminate between the consensus ERE and the mutERE was compared under various salt conditions. Since ER has been reported to interact with the ERE as either a dimer (20,29,30) or a monomer (25), the data were fitted using either a simple binding model of a single, structural ER unit (monomer or dimer) or a cooperative model of DNA-induced ER dimerization. The role of E 2 in ER binding to the ERE was also examined using this equilibrium technique.

EXPERIMENTAL PROCEDURES
Recombinant ER Proteins-The baculovirus-expressed, purified human ER was kindly provided by PanVera Corporation in storage buffer containing 50 mM Tris, pH 7.5, 10% glycerol, 2 mM dithiothreitol (DTT), 1 mM EDTA, 0.5 M KCl, 1 mM sodium vanadate, and 0.02% sodium azide. The purity of the human ER was estimated as 80%. Mouse ER was prepared from recombinant baculovirus-infected SF21 cells using established procedures and purified using a DNA affinity column. 2 The concentration of active ER, based on estrogen-binding ability, was determined by a modification of the hydroxylapatite receptor assay (24). The DNA binding ability of each ER preparation was compared by gel shift analysis (see below). The DBD of the Xenopus ER was prepared as described previously (21).
Oligonucleotides and Fluorescein Labeling-A 35-bp oligonucleotide containing the vitERE and its flanking sequences was synthesized and reverse-phase purified (Research Genetics, Inc.). The sense strand was labeled with fluorescein-maleimide at a thio-reactive ATP generated at the 5Ј end, using the FluoroAmp™ T4 kinase green oligonucleotide labeling system (Promega Corp.). The fluorescein-labeled sense strand in TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0) was separated from free fluorescein on a 20-cm Sephadex G-25 desalting column (31). The absorbance spectrum from 250 -600 nm and the fluorescence anisotropy (see below) of each column fraction was measured. The concentration (C) of the oligonucleotide was calculated from its absorbance at 260 nm (A 260 ), which was corrected for the absorbance of fluorescein at 260 nm by using the following equation where 65,000 is the fluorescein extinction coefficient. Column fractions of oligonucleotides corresponding to a high degree of fluorescein substitution and an anisotropy value for fluorescein-bound oligonucleotide, in comparison to values for free fluorescein, were frozen at Ϫ20°C for further use. The resulting fluorescein-labeled sense strand was mixed with unlabeled antisense strand, heated to 95°C for 2 min, and slowly cooled to room temperature for 2-4 h to anneal. The resulting 35-bp annealed oligonucleotide, designated F-vitERE, is shown in Sequence 1.
The 6-bp inverted repeat half-sites separated by the 3-bp spacer are underlined. An unlabeled version of vitERE for competition assays was annealed as described above.
A 35-bp oligonucleotide containing a mutated vitERE (F-mutERE) was synthesized, labeled with fluorescein-maleimide using the same method as for F-vitERE, and annealed as described previously. F-mutERE is shown in Sequence 2.
The sequences in each of the two half-sites that deviate from F-vitERE are marked with a double underline. Unlabeled mutERE oligonucleotide was also annealed as described previously for use as a competitor.
Fluorescence Anisotropy Assays-The fluorescein-labeled oligonucleotide was diluted to the appropriate concentration in fluorescence anisotropy buffer (10 mM Tris, pH 7.5, 100 M EDTA, 10% glycerol, 1 mM DTT, 50 g/ml soybean trypsin inhibitor, and 100 to 300 mM KCl or NaCl). Ligand-bound ER was prepared by incubating undiluted purified ER with 5-10-fold molar excess of E 2 (Sigma) for at least 1 h at 4°C. The ER was then diluted to the appropriate starting concentration in binding buffer containing fluorescein-labeled oligonucleotide in the first sample tube with salt and oligonucleotide concentrations adjusted to correct for any dilution from the large volume ER addition. The ER was serially diluted from the starting tube into the samples tubes containing fluorescein-labeled oligonucleotide, and the samples were incubated for at least 1 h on ice to allow protein-DNA complex formation. Longer incubations did not increase the anisotropy values observed for the ER⅐ERE complex, indicating that the reactions were at equilibrium. After prewarming samples to 21°C for 2 to 3 min, fluorescence anisotropy and total intensity were measured for each dilution using an ISS KOALA fluorometer (ISS, Inc.) in photon counting mode. Samples were excited in L-format by 488-nm light from a Omnichrome 543-AP CW argon ion laser defocused through an Oriel multifiber optic bundle. The parallel and perpendicular emission components were measured through a Corion 530 high pass filter. Each sample emission measurement was corrected by subtraction of the background buffer emission measurement. Anisotropy of fluorescence was typically calculated from three to four emission measurements, and the average value was used. Standard deviation for the anisotropy values was 0.001 or less.
Data Analysis-Binding curve analysis was conducted using the numerically based BIOEQS program (32)(33)(34). Given the concentration of the protein and DNA, the final stoichiometries of the interaction, and initial guesses for the ⌬G for formation of the various bound species, BIOEQS calculates the concentration of the various species in the equilibrium binding curve numerically using a constrained optimiza-

FIG. 1. The structural and functional domains of human ER.
Domains A/B and E contain transcriptional activation function sequences, designated TAF-1 and TAF-2. The highly conserved DNA binding domain is located in region C. Domain D is known as the hinge region and also contains the antigenic site for antibody ER712. The ligand binding domain is found in region E. tion algorithm in which the mass balance constraints are incorporated as Lagrange multipliers (32). The program then relates this species concentration vector to the observable parameter (in this case anisotropy) at each point in the titration, and fits for the free energy and plateau values by adjusting these floating parameters using a Marquardt-Levenberg nonlinear least squares algorithm.
Since the final stoichiometries of the ER-ERE interaction are not known exactly, both a simple binding model and cooperative DNAinduced dimerization model were tested for their ability to describe the interaction of ER with the ERE. The simple binding model was used to determine ⌬G for the binding of an ER monomer unit or a preformed ER-homodimer to the DNA. This model includes free ERE, free ER, and ER-bound ERE and can be described by The simple binding model does not discriminate between one ER monomer or two noninteracting monomers of equal affinity binding to the DNA. Since the ER concentration used in the assays was expressed as a monomer, the simple binding model was applied here to describe the binding of a single ER monomer to the DNA. This model could be used to describe the binding of the preformed ER dimer by an appropriate adjustment of the units for the concentration axis. The cooperative dimer model describes the DNA-binding of two ER proteins that interact with each other. This model incorporates free ERE, free ER monomer, ER monomer-bound ERE, amd ER dimerbound ERE and can be described as In this more complex model, BIOEQS solves for ⌬G for the formation of the monomeric ER-DNA species (⌬G 1 ) and the overall ⌬G for the formation of the dimeric ER-DNA species (⌬G 2 ) from the free monomeric ER and unbound DNA. From the anisotropy data, the initial plateau corresponding to the unbound ERE and the end plateau corresponding to ER-bound ERE were also adjustable parameters during the fit. In the dimer fit, the plateaus were initially floated. Once the plateaus were defined, the monomer ER-DNA species was then assigned an anisotropy value that is halfway between the unbound ERE and ER-bound ERE plateaus, and the data were refitted. The ⌬G for formation of the complexes is described by the following equations The concentrations of the various species are indicated in the brackets. R represents the gas constant and T is the temperature. BIOEQS solves for the ⌬G values, given the following mass balance constraints where subscript T represents the total concentration of the species indicated in the brackets. Thus, BIOEQS solves this system of equations numerically. The ⌬G values are adjusted using least-squares analysis. If the difference between ⌬G 2 and ⌬G 1 is significantly different from ⌬G 1 alone, this difference in free energy may be attributed to protein-protein coupling. The free energy associated with the cooperativity (⌬G coop ) of the system can be obtained from From the ⌬G 1 values, the equilibrium dissociation constant (K d ) was calculated by the following relationship Several degrees of error analysis were incorporated into the fitting and interpretation of the data. Each individual anisotropy measurement for a titration was obtained three to five times and the average anisotropy from these replicates is reported. The standard deviation of these measurements, which was Ϯ0.001, was included in the BIOEQS input file along with the DNA concentration, protein concentration, and average anisotropy measurements. The 2 (or variance) was calculated in the data fitting for each experiment, based on the residual anisotropy values, which are the differences between the raw and fitted anisotropy data. Rigorous confidence interval tests at the 67% level were calculated for each fit and reported as the error for ⌬G values (35).

Gel Shift Analysis of ER-ERE Interaction-
The protein-DNA complexes formed between purified baculovirus-expressed human ER and F-vitERE were examined by fluorescence-detected gel shift analysis and compared with the complexes formed between purified baculovirus-expressed mouse ER and F-vit-ERE (Fig. 2). ER was occupied with E 2 prior to incubation with the fluorescein-labeled oligonucleotide. A single human ER complex with F-vitERE was detected (lane 7) and supershifted with antibody ER712 (lane 6). Although the peptide antigen used to produce antibody ER712 originated from the conserved hinge region of the rat ER, this peptide contains a two amino acid variation between the rat and human ER (36). Thus, a fourfold excess of antibody ER712 was used and completely supershifted the human ER complex. The human ER complex with F-vitERE and the supershifted complex aligned with the corresponding mouse ER complexes (lanes 2 and 3), indicating that the approximate sizes of the DNA-bound mouse ER and human ER complexes were similar. Although each reaction contained 2 pmol of ER, the intensity of the mouse ER⅐F-vitERE complex was greater than the human ER⅐F-vitERE complex, suggesting that the DNA binding ability of the mouse ER preparation or stability of the complex was better under these conditions. To examine specificity of binding, the complexes were competed with a 10-fold excess of either unlabeled  10 -14). A 10-fold excess of unlabeled vitERE (lanes 4 and 8) or mutERE (lanes 5 and 9) over the amount of fluorescein-labeled oligonucleotide was added to competition reactions. To supershift ER-containing complexes, antibody ER712 was included in the following amounts: 2-fold excess (lanes 2 and 11) or 4-fold excess (lanes 6 and 13) over the ER concentration. A 4-fold amount of antibody ER712 was used to detect human ER complexes as compared with mouse ER complexes, because antibody ER712 was generated against a rat ER conserved region. The protein-DNA complexes were separated by 4% nondenaturing polyacrylamide gel electrophoresis (acrylamide: bisacrylamide, 19:1) in 10 mM Tris, 0.4 M glycine, and 1 mM EDTA. Immediately following electrophoresis, the wet gel was scanned at 750 V in a Fluorimager SI (Molecular Dynamics). vitERE or mutERE. The mutERE is a palindromic sequence containing a single base pair alteration in each of the half sites. The abundance of the single human ER complex with F-vitERE was not reduced when mutERE was included as competitor (lane 9) but its formation was diminished when vitERE was used as competitor (lane 8), indicating that the purified human ER specifically binds to the vitERE sequence. Likewise, the mouse ER complex with F-vitERE was competed with vitERE (lane 4) but not with mutERE (lane 5). To determine whether the purified ER preparations would bind to F-mutERE, several reactions were prepared in which the purified proteins were incubated with F-mutERE. The human ER unexpectedly bound to F-mutERE (lane 14) and the complex was supershifted with antibody ER712 (lane 13). Similarly, the mouse ER also bound to F-mutERE (lanes 11 and 12). The competition experiments verified that both mouse ER and human ER bind with higher affinity to vitERE than to mutERE. These results also indicated that the purified mouse ER and human ER exhibit non-negligible affinity for the mutERE sequence under these conditions.
Effect of Salt on ER⅐ERE Complex Formation-The affinity of E 2 -bound human ER (ER-E 2 ) for F-vitERE and F-mutERE in solution was next examined using fluorescence anisotropy assays. Fluorescence anisotropy has been used to study both protein-DNA and protein-protein interactions (37)(38)(39)(40) and provides a means to measure molecular interactions under equilibrium conditions without the need to separate free and bound molecules (41). The binding buffer conditions used were virtually identical to those for the gel shift assays, except that soybean trypsin inhibitor was substituted for BSA as carrier protein. At a salt concentration of 100 mM KCl, the anisotropy profile (Fig. 3A) for human ER-E 2 binding to F-vitERE was characterized by an initial plateau for the unbound DNA, followed by a steady anisotropy increase as ER complexed with the fluorescein-labeled DNA, and concluded with a short plateau for saturated binding. At the highest ER-E 2 concentrations, the anisotropy increase following the short plateau probably indicates formation of higher order complexes. The total fluorescence intensity did not change significantly during the titration, indicating that the anisotropy signal was not convoluted with changes in fluorescence lifetimes or contaminated by scattered excitation light, and thus directly represents a molar quantity. The equilibrium binding profile for ER-E 2 interacting with F-mutERE was very similar (data not shown). The data were not fitted because the upper plateau was not well defined. In agreement with the gel shift results, purified human ER-E 2 bound with similar apparent affinity to both F-vitERE and F-mutERE at 100 mM KCl.
The specificity of E 2 -occupied human ER binding to F-vit-ERE compared with F-mutERE was next examined in the fluorescence anisotropy assays by using incremental increases in the buffer salt concentration. In the lac repressor system, increasing salt concentration reduces nonspecific protein-DNA apparent affinity more strongly than specific affinity by competing for interaction with the negatively charged phosphate backbone (42,43). The total anisotropy change from the unbound F-vitERE to the human ER-E 2 complex with F-vitERE was 0.02 anisotropy unit at 200 mM KCl (Fig. 3B), significantly smaller than the 0.06 anisotropy unit observed at 100 mM KCl, suggesting that ER may be aggregating on the ERE at 100 mM KCl. The anisotropy profile for ER-E 2 binding to F-vitERE at 200 mM KCl (Fig. 3B) exhibited a starting plateau, an anisotropy increase, and an ending plateau, indicating that this binding was saturable. The anisotropy increase observed for ER-E 2 binding to F-mutERE at 200 mM KCl (Fig. 4A) was shifted to higher protein concentration as compared with vitERE and did not plateau at the highest ER concentrations used, indicating a lower relative affinity for the mutated sequence. The binding of ER-E 2 to F-vitERE at 200 mM KCl was fitted using the BIO-EQS software, as described under "Experimental Procedures." As shown in Fig. 3B, the data for ER-E 2 binding to F-vitERE at 200 mM KCl fit well to either models for simple binding (full lines) or cooperative dimer binding (dashed lines). The intensity did not change over the titration (data not shown), such that no quantum yield weighting term was needed and no lifetime effects were taken into account. The simple binding model describes the binding of a single ER structural unit. Since the stoichiometry of ER in solution has not been absolutely established, the simple binding model could describe the binding of a preformed ER dimer, a single monomer, or two noninteracting monomers, requiring the appropriate adjustment of the units for concentration. It can be seen by examination of the data in Fig. 3 that the signal-to-noise ratio, although quite good as compared with gel shift assays, is too small to accurately discriminate between the simple and cooperative binding models.
The effect of salt on the specificity of human ER-E 2 binding to F-vitERE and F-mutERE was determined using fluorescence anisotropy assays by increasing the salt concentration by 25 mM increments between 225 and 300 mM KCl. At 225 mM KCl, human ER-E 2 bound to F-vitERE (Fig. 3C) over the investigated concentration range, but no significant degree of binding to F-mutERE was detected under these conditions (Fig. 4B). The F-vitERE anisotropy data were fitted to both simple binding (Fig. 3C, solid lines) and cooperative dimer (Fig. 3C, dashed lines) models of ER-E 2 interaction with F-vitERE. At 225 mM KCl, the K d for the interaction of ER-E 2 with F-vitERE was 2.0 nM. Although the 2 values slightly favored the cooperative model at 225 mM KCl (data not shown), the quality of the data does not provide significant differentiation between the two models. Similar binding patterns for ER-E 2 interaction with F-vitERE were observed at 250 mM KCl (Fig. 3D) or 275 mM KCl (Fig. 3E). Fits to both the simple and cooperative binding models were equally acceptable at these salt concentrations as well, and the affinity (as per the simple binding model) was unaffected by salt concentration over this range (Table I). Binding of ER-E 2 to F-mutERE was not detected at either 250 or 275 mM KCl (data not shown). At 300 mM KCl, the anisotropy profile for human ER-E 2 binding to F-vitERE under equilibrium binding conditions was shifted significantly to higher ER-E 2 concentrations (Fig. 3F). The K d of ER-E 2 for F-vitERE at 300 mM KCl was found to be approximately 54 nM (Table I) but since no plateau in anisotropy values was observed at this high salt concentration, the confidence limits for the free energy values recovered from the fits were quite large. As for the previous titrations, the simple binding and cooperative dimer fits were both equally acceptable, preventing an assignment of a model to the binding. Nonetheless, a loss of at least one order of magnitude in affinity is observed upon increasing the salt concentration from 275 to 300 mM KCl. Binding of ER-E 2 to F-mutERE was not detected at 300 mM KCl, verifying that this interaction was abrogated by the higher salt concentration (data not shown). The salt dependence of ER-E 2 /ERE interactions is thus highly complex, changing from nonspecific to specific modes between 100 and 200 mM salt, followed by no change between 200 and 275 mM salt, and then exhibiting a large decrease in affinity between 275 and 300 mM salt. These titrations were highly reproducible as indicated by the error in the free energy values obtained from the 67% confidence limit tests (Table I). They suggest that salt may affect protein-protein as well as protein-DNA interactions.
Binding of ER to ERE Does Not Require Estradiol-To examine the requirement of E 2 for ER binding to DNA, fluorescence anisotropy assays were used to compare the binding of E 2 -occupied ER and unoccupied ER to F-vitERE. The titrations were carried out at 225 mM KCl, because binding of human ER-E 2 to F-vitERE was specific and of high affinity at this salt concentration (see above). As shown in Table II, the affinity of E 2 -occupied ER or unoccupied ER for the F-vitERE sequence was almost identical.
Effect of Salt on ER DBD Interaction with ERE-To determine if the full-length ER and ER DBD fragment exhibit similar specificity and affinity for the vitERE and mutERE sequences, we next examined the ability of the ER DBD to discriminate between F-vitERE and F-mutERE at two salt concentrations by using anisotropy assays. At 100 mM NaCl, the ER DBD bound to both F-vitERE (Fig. 5A) and F-mutERE (Fig. 5B) with similar affinities. The binding did not appear to

TABLE I
Thermodynamic constants for ER binding to the ERE The data for the titrations at each salt concentration are listed for the simple (s) binding model and the cooperative (c) dimer model. The free energy values (⌬G 1 ) were used to calculate the apparent equilibrium dissociation constant (App K d ), by using the following equation: ⌬G 1 ϭ ϪRT ln K d . In most of the cooperative fits, the ⌬G 1 value was recovered as an upper limit rather than an absolute value, because either the monomer ER-DNA species was never significantly populated due to the strong cooperativity of the ER-ERE interaction or the data were not able to resolve the cooperativity of the system. Therefore, the K d values were not calculated for the cooperative dimer fits. ⌬G values for multiple titrations were averaged and the error reported for ⌬G is obtained from 67% confidence interval tests.   Table I. Titrations were performed in buffer containing 225 mM KCl. The apparent K d values were calculated as described in Table I. Data to compare E 2 -occupied and unoccupied ER binding were obtained from a single experiment. The 2 (or variance) was measured in the data fitting, based on the residual anisotropy values, which are the difference between the raw and fitted anisotropy data. be fully saturated even at 400 nM ER DBD. The change in anisotropy from unbound to ER-bound over the ER DBD concentration range tested was 0.085, suggesting that the ER DBD at this salt concentration binds to the targets as a complex of a higher order than dimer. Interestingly, the total intensity decreased by 25% for ER DBD binding to F-vitERE (Fig. 5A) and by 20% for ER DBD binding to F-mutERE (Fig. 5B). This intensity decrease coincided with the anisotropy increase, suggesting that the binding event was detected by both the anisotropy of the complex and the total intensity emitted from the fluorescein label on the oligonucleotide. Such intensity changes could be caused by fluorescence quenching if ER DBD binding to the oligonucleotide interferes with the fluorescein label.
To promote specific binding of the ER DBD to F-vitERE, fluorescence anisotropy titrations were performed at 250 mM NaCl, a salt concentration at which specific binding of the full-length human ER was observed. Again, ER DBD bound to both F-vitERE and F-mutERE even in binding buffer containing 250 mM NaCl (Fig. 5, C and D). The binding curve was shifted to higher protein concentrations, indicating sensitivity to monovalent cation concentration. The total intensity decrease was also shifted to higher ER DBD concentrations, consistent with our assumption that this intensity phenomenon is also reporting on protein binding. The similarity of the binding profiles for ER DBD with either F-vitERE or F-mutERE indicated that the ER DBD exhibits similar affinities for these different oligonucleotides under these solution conditions. Therefore, unlike the full-length protein, the ER DBD does not differentiate between the vitERE and mutERE even at high salt concentration in the fluorescence anisotropy assays. DISCUSSION We have used fluorescence anisotropy to examine the interaction of the full-length ER and ER DBD with consensus and mutated ERE sequences under equilibrium binding conditions. Our results indicate that the Xenopus ER DBD bound indiscriminately to the fluorescein-labeled vitERE and mutERE sequences with identical salt concentration dependence in these solution conditions, suggesting that sequences outside the DBD may be important for specific binding of the ER DBD. Several previous studies indicated that the ER DBD forms specific dimeric complexes with the consensus ERE and monomeric complexes with imperfect ERE sequences. The 66-amino acid zinc finger core of the DBD has been shown to contain the sequences necessary for homodimerization and ER recognition of the ERE sequence (19,44). The Xenopus ER DBD has been observed to exist as a monomer in solution by column chromatography and fluorescence polarization measurements of the rotational relaxation times (21). Purified Xenopus ER DBD formed a single complex with the consensus ERE over a range of 18 to 3600 nM ER DBD added in gel shift assays (21). Mixing different amounts of the ER DBD and another ER truncation mutant, which also retained DNA binding ability, resulted in the formation of gel-shifted complexes intermediate in size between the DBD⅐DNA complex and the ER truncation mutant-DNA complex, suggesting that ER DBD bound as a dimer to the ERE (45). ER DBD complexed as a monomer with an imperfect ERE containing a double base pair change in one arm of the palindrome, as determined in gel shift assays (45). D region amino acids adjacent to the DBD were found to stabilize the ER fragment binding to imperfect ERE sequences (45). Xenopus ER DBD was shown to recognize the consensus ERE with higher affinity than imperfect ERE sequences bearing one or two nucleotide changes in competition gel shift assays (46). In our fluorescence anisotropy assays, which did not require physical separation of free and DNA-bound protein, the Xenopus ER DBD (amino acids 171-281) recognized the palindromic vitERE and mutERE sequences with similar affinity, perhaps due to detection of lower affinity complexes. These results indicate that sequences or conformation present in the fulllength ER protein may contribute to specificity of binding.
Our observations also indicate that purified, full-length human ER-E 2 binds nonspecifically to vitERE and mutERE in buffer containing 100 mM KCl as demonstrated by both gel shift analysis and fluorescence anisotropy titrations (Figs. 2 and 3). Even in the 10 -30 nM physiological range of ER concentration, ER did not discriminate between the consensus and mutated ERE sequences in the fluorescence anisotropy assays. The large anisotropy change observed from unbound to bound oligonucleotide and the lack of a well defined plateau even at high ER concentrations suggested that the purified ER was aggregating on the DNA under these salt conditions. Recent studies have indicated that baculovirus-expressed mouse ER binds to the consensus ERE in an oligomeric form larger than a dimer in buffer containing 100 mM salt. 3 Further experimentation is needed to decipher the oligomeric state of the purified ER bound to DNA at 100 mM KCl, but it is likely that protein-3 I. Anderson and J. Gorski, unpublished data. protein interactions are occurring at this salt concentration.
In buffer conditions that included salt concentrations equal to or greater than 200 mM KCl, ER exhibited discrimination between the consensus and mutant ERE sequences. From 225 to 275 mM KCl, ER-E 2 bound to the F-vitERE but not to the F-mutERE over the protein concentration range examined. Increasing the salt concentration has been shown to decrease the apparent affinity for protein-DNA complexes (42,43). However, in these studies, the mode of binding shifted from a nonspecific oligomeric mode to a specific mode between 100 and 200 mM KCl, likely involving dimeric complexes, given the changes in the values of the anisotropy of the complexes. Between 200 and 275 mM KCl, little change in affinity was observed. However, a further increase in the salt concentration to 300 mM KCl resulted in a shift in the binding curves to higher protein concentration. A drastic reduction in ER⅐ERE complex formation at salt concentrations above physiological levels was previously observed with rat uterine ER (17). This unusual salt dependence strongly suggests that salt may be affecting protein-DNA and protein-protein interactions differentially over this range.
In the salt concentration range between 225 and 275 mM, where the anisotropy profiles were characterized by well defined plateaus, testing of both simple and cooperative binding models revealed that the data did not provide the precision necessary for discriminating between these two possibilities. If a cooperative binding model had consistently described the data better than a simple model, this would have provided strong evidence for dimer binding. Unfortunately, our data are equally consistent with models of noncooperative binding of two monomers, simple binding of one monomer, simple binding of a preformed dimer, or cooperative binding of two monomers.
In our studies, ER-E 2 was able to discriminate between vit-ERE and mutERE only at salt concentrations greater than 200 mM KCl, which was slightly higher than the optimal salt concentrations found for the unpurified rat uterine ER (17). The equilibrium dissociation constant measured in buffer conditions containing 225 to 275 mM KCl was 1.8 Ϯ 0.6 nM, which was about fivefold greater than the value of 390 Ϯ 40 pM obtained for unpurified, E 2 -occupied, rat uterine ER binding to the ERE (24). Since purified ER bound rather indiscriminately and with a slightly lower affinity as compared with the unpurified rat uterine ER at physiological salt concentrations, the mechanism by which ER differentiates between imperfect ERE sequences may include participation of other nuclear proteins to increase formation of active ER complexes at appropriate target genes. Alternatively, some loss of specific DNA binding activity may result from the ER purification procedure. Aggregation phenomena of the DNA-bound ER at high protein concentrations precludes an independent measure of DNA binding activity. In addition, our results verified that E 2 -occupied and unoccupied ER bound with the same affinity to vitERE (Table  II). These findings are similar to those obtained using ER from rat uterine cytosol (24,25). Thus, E 2 -binding to the ER affects another step in the transcription pathway, other than DNA binding, in support of the current model of estrogen action.
In summary, our results indicate that ER sequences outside of the DBD may play an important role in promoting specific interactions with ERE sequences. Unlike the ER DBD, the full-length human ER specifically recognized the ERE sequence at buffer salt concentrations near physiological salt concentrations. Unoccupied and E 2 -occupied ER bound to the ERE sequence with similar affinity, verifying that ER binds to ERE sequences in the absence of estrogen. Moreover, our results suggest that factors other than the receptor itself may serve to provide binding specificity.