In Vitro Fluorescence Anisotropy Analysis of the Interaction of Full-length SRC1a with Estrogen Receptors α and β Supports an Active Displacement Model for Coregulator Utilization*

Binding of full-length P160 coactivators to hormone response element-steroid receptor complexes has been difficult to investigate in vitro. Here, we report a new application of our recently described fluorescence anisotropy microplate assay to investigate binding and dissociation of full-length steroid receptor coactivator-1a (SRC1a) from full-length estrogen receptor α (ERα) or estrogen receptor β (ERβ) bound to a fluorescein-labeled (fl) estrogen response element (ERE). SRC1a exhibited slightly higher affinity binding to flERE·ERβ than to flERE·ERα. Binding of SRC1a to flERE·ERα and to flERE·ERβ was 17β-estradiol (E2)-dependent and was nearly absent when ICI 182,780, raloxifene, or 4-hydroxytamoxifen were bound to the ERs. SRC1a binds to flERE·E2-ERα and flERE·E2-ERβ complexes with a t½ of 15–20 s. Short LXXLL-containing nuclear receptor (NR) box peptides from P160 coactivators competed much better for SRC1a binding to flERE·E2-ER than an NR box peptide from TRAP220. However, ∼40–250-fold molar excess of the P160 NR box peptides was required to inhibit SRC1a binding by 50%. This suggests that whereas the NR box region is a primary site of interaction between SRC1a and ERE·E2-ER, additional contacts between the coactivator and the ligand-receptor-DNA complex make substantial contributions to overall affinity. Increasing amounts of NR box peptides greatly enhanced the rate of dissociation of SRC1a from preformed flERE·E2-ER complexes. The data support a model in which coactivator exchange is facilitated by active displacement and is not simply the result of passive dissociation and replacement. It also shows that an isolated coactivator exhibits an inherent capacity for rapid exchange.

diverse biological effects by binding to the intracellular proteins, estrogen receptor (ER) 3 ␣ and ER␤ (6). ER and other members of the steroid hormone receptor subfamily of nuclear receptors (NRs) share a common domain structure and overall scheme for transcriptional activation (7,8). In the classical model for ER action, binding of estrogens enables the ER to dimerize and bind to specific canonical DNA sequences termed estrogen response elements (EREs). Transcriptional activation by ER is largely mediated by two interacting activation functions. Activation function 1, in the N-terminal A/B domain is important in ligand-independent transcription and activation function 2 in the C-terminal ligand binding domain is important in ligand-dependent recruitment of coactivators. The bound coactivators help assemble a dynamic, rapidly changing, multiprotein complex that facilitates both chromatin remodeling and formation of an active transcription complex (9 -14). When agonists are bound to ER, several ER helices align to form a hydrophobic cleft that is critical for coactivator binding. The long side chains in selective estrogen receptor modulators, such as 4-hydroxytamoxifen (the active metabolite of tamoxifen) and raloxifene (RAL), contact amino acids in helix 12 of the ER ligand binding domain and change the orientation of helix 12 so that helix 12, rather than coactivators, occupies the hydrophobic cleft (15,16).
The steroid receptor coactivator (SRC) or P160 family of coactivators, consisting of SRC1/NCoA-1, SRC2/GRIP1/TIF2/ NCoA2, and SRC3/ACTR/AIB1/p/CIP/RAC3/TRAM1, are perhaps the best known of the many coactivators implicated in ER-mediated transactivation (17,18). The P160 coactivators interact with the binding cleft in the ER ligand binding domain via highly conserved ␣-helical Leu-X-X-Leu-Leu (LXXLL) motifs, also called NR boxes. Most P160 proteins contain a central nuclear receptor-interacting domain made up of three NR boxes arranged in tandem. An alternatively spliced form of SRC1, called SRC1a, contains a fourth NR box at its C terminus (19). Whereas the NR boxes are thought to play a key role in binding of P160 coactivators to ER, amino acid residues flanking this core motif are also important for ER recognition and * This work was supported by National Institutes of Health Grants DK-071909 binding specificity (20,21). In addition, sequences in the N-terminal region of ER play a role in SRC1 binding (22). In vivo, coactivators can interact with diverse proteins including the cointegrators p300/CREB-binding protein and PCAF (23)(24)(25)(26). The DRIP/TRAP/ARC multiprotein complex represents a second class of coactivator complex implicated in ER-mediated transactivation. This complex interacts with liganded ER AF2 via two LXXLL motifs in its DRIP205/TRAP220 subunit (27).
Chromatin immunoprecipitation results and other data suggest that the components of transcription complexes at EREs undergo rapid exchange and replacement. In a simplified formulation of this sequential model of coregulator exchange, the ER initially recruits an SRC⅐p300/CREB-binding protein⅐PCAF complex whose histone acetyltransferase activity acetylates nearby histones, altering the architecture of the promoter regions and enhancing promoter accessibility. The DRIP⅐TRAP⅐ARC complex then replaces the SRC⅐p300/CREBbinding protein⅐PCAF complex and additional proteins including DNA topoisomerase II, that are linked to the DNA repair machinery, are recruited to the complex prior to steroid receptor activation of transcription (9 -11, 13, 14). However, some aspects of the model are difficult to investigate in intact cells, and critical mechanistic questions remain. Addressing questions such as whether components of the complexes can actively displace each other or only enter the complex after the initial components have dissociated, and whether rapid exchange of coactivators is an intrinsic capacity or requires enzymatic modification of the coactivators, requires the development of suitable in vitro approaches.
Previous in vitro studies of ER⅐P160 coactivator interactions primarily used NR box peptides and NRID-containing protein fragments (28 -36). These studies suggest that LXXLL motifs may act only as docking points and that additional contacts outside of the NRID may be required for more complete binding, emphasizing the need to study these interactions with fulllength proteins. In addition, because binding of ER to the ERE is thought to precede coactivator binding (37), and the ERE also influences ER conformation and the ability of the ER to recruit coactivators (38 -45), it is important that ER-coactivator interaction be examined in a context where ER is pre-bound to its DNA binding site.
Perhaps because of instability, the tripartite complex composed of the ERE, ER, and P160 coactivator is not detected using electrophoretic mobility shift assays (data not shown). In contrast to electrophoretic mobility shift assays, surface plasmon resonance and fluorescence anisotropy/fluorescence polarization assays provide real-time analysis of macromolecular interactions. Cheskis and co-workers (29) used surface plasmon resonance to analyze binding of P160 coactivators to liganded ER. However, this important work did not involve a physically isolated P160 coactivator, or ER bound to EREs, and has not stimulated subsequent studies. Recently, we developed the fluorescence anisotropy/polarization microplate assay (FAMA) to analyze the interactions of steroid receptors with their DNA response elements in ultra low-volume microplates (46). In this assay, after excitation with polarized light, the relatively small fluorescein-labeled consensus ERE (flERE) undergoes rotational diffusion more rapidly than the time required for light emission. Therefore, the position of the flERE at the time of emission is largely randomized, resulting in depolarization of most of the emitted light (Fig. 1A). When the ER binds to the flERE, the larger size of the flERE⅐ER complex causes rotational diffusion to be slower, increasing the likelihood that the flERE⅐ER complex will be in the same plane at the time of emission as it was at the time of excitation. Therefore, the emitted light will be more highly polarized (Fig. 1B). This change is observed as an increase in fluorescence polarization, or the closely related parameter, fluorescence anisotropy. We also demonstrated that binding of anti-ER␣ antibodies to the flERE⅐ER␣ complex resulted in a further increase in anisotropy, suggesting that the utility of the FAMA could be extended to the study of ERE⅐ER⅐coregulator interactions (modeled in Fig.  1C). Results from earlier cuvette-based studies also suggested that the FAMA could be developed for studies of coregulator interactions (47,48).
Here, we show that the FAMA provides a reliable in vitro assay for monitoring the binding of a full-length P160 coactivator, SRC1a, to flERE⅐ER complexes and for monitoring the dissociation of flERE⅐ER⅐SRC1 complexes. We analyzed the effects of estrogenic ligands and selective estrogen receptor modulators on SRC1 binding and the relative ability of several NR box peptides to block binding of full-length SRC1 to flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ complexes. The ability of NR box peptides to stimulate dissociation of full-length SRC1a from preformed flERE⅐E 2 -ER␣⅐SRC1a and flERE⅐E 2 -ER␤⅐SRC1a complexes suggests that rapid coactivator exchange results from active displacement of the coactivator from the receptor-DNA complex.

EXPERIMENTAL PROCEDURES
Proteins and Peptides-Full-length FLAG epitope-tagged human ER␣ (ER␣) and FLAG epitope-tagged human ER␤ were expressed in the presence of 200 nM E 2 , purified in the presence of 20 nM E 2 , and quantitated as described previously (45,49). Full-length FLAG epitope-tagged human SRC1a was purified as described (50). Different preparations of SRC1a were used in the several experiments. Optimal binding of SRC1a to flERE⅐ER complexes was determined from saturation curves done with each preparation of SRC1a (data not shown). Different concentrations of SRC1a are therefore used in some experiments.
Oligonucleotides-A 30-bp oligonucleotide containing the cERE was synthesized with fluorescein (6-FAM) at its 5Ј end using phosphoramidite chemistry and PolyPak II (Glen Research Corp., Sterling, VA) purified by the Biotechnology Center (University of Illinois). The sequence of this fluoresceinlabeled sense strand, with the ERE half-sites underlined, is: 5Ј-fl-CTAGATTACAGGTCACAGTGACCTTACTCA-3Ј. A 260 values were measured to calculate the oligonucleotide concentration. The ϳ60% degree of fluorescein incorporation was determined as described (51). The fluorescein-labeled sense strand was annealed with an equimolar concentration of the unlabeled antisense strand.
Fluorescence Anisotropy Microplate Assays-FAMA was carried out with some modifications of our earlier methods (46). The fluorescein-labeled double-stranded cERE was diluted to 1 nM in the anisotropy buffer (15 mM Tris-HCl, pH 7.5, 150 mM KCl, 1 mM dithiothreitol, 5% glycerol, 0.05% Nonidet P-40, 2 ng of poly(dI-dC), 100 nM E 2 ) and added to the wells of a black 96-well low-volume, high efficiency HE microplate (Molecular Devices Corp., Sunnyvale, CA). A concentration of human ER identified in preliminary studies as sufficient for saturated binding of human ER to the flERE probe (usually 15 nM) was added to the wells in a total reaction volume of 20 l. The reactions were mixed by pipetting and incubated at room temperature in the dark. Anisotropy values were measured until the reactions reached equilibrium (usually within 15 min). SRC1a was diluted to appropriate starting concentrations in a buffer containing 20 mM Tris (pH 7.5), 100 mM NaCl, 0.2 mM EDTA, 15% glycerol, 0.1% Nonidet P-40, and 0.5 mg/ml insulin (or bovine serum albumin). Our ongoing studies showed that the presence of Nonidet P-40 and EDTA had no effect and the Nonidet P-40 could be left out (data not shown). The indicated amounts of SRC1a were then added to wells containing the equilibrated flERE⅐ER binding reaction. The reactions were mixed by pipetting and incubating at room temperature in the dark. Anisotropy values were measured until the reactions reached equilibrium (within 10 min, see Fig. 4). Anisotropy changes resulting from SRC1a binding were calculated by subtracting the anisotropy values measured for the flERE⅐ER reactions from the values measured for the flERE⅐ER⅐SRC1a reactions. For studies of binding affinity, a percent bound value was calculated at each concentration of SRC1a using the simple expression: (anisotropy change/maximum anisotropy change) ϫ 100. After plotting the percent probe bound versus SRC1a concentration, an apparent K d was determined as the SRC1a concentration at which 50% of the flERE⅐ER complex was bound to SRC1a. Anisotropy measurements were made on the Ultra Evolution Multifunctional Microplate Reader (Tecan, Research Triangle Park, NC), or on a more sensitive PHERAStar High-End Microplate Reader (BMG LabTech Inc., NC). Although outcomes were independent of the instrument used, because of different gain settings on the instruments, the magnitude of the anisotropy changes is not identical in the different figures. Unless otherwise indicated, Sigma Plot was used to fit the data and calculate data from the curves.
Ligand Exchange Experiments-For ligand exchange experiments, a concentration of E 2 -ER that produces saturated binding to the flERE was added to the anisotropy buffer (plus 15 g of bovine serum albumin to help stabilize the ERs and 100 nM of the indicated ligand) in the wells of the microplate. Following a 1-h incubation at room temperature, flERE was added to a concentration of 1 nM in a total reaction volume of 20 l. The reactions were mixed by pipetting and incubated at room temperature. An amount of SRC1a required to achieve ϳ90 -100% of maximum binding of the flERE⅐ER complex was added and anisotropy was measured as described above.
On Rate Experiments-For on rate experiments, a concentration of E 2 -ER that produces saturated binding to the flERE was added to the anisotropy buffer containing 1 nM flERE and then incubated for 30 min at room temperature. A near saturating concentration of SRC1a (15 nM) was added to the pre-bound flERE⅐E 2 -ER complex in a total reaction volume of 20 l. Kinetic measurements were carried out at room temperature by taking anisotropy measurements at the indicated time points after SRC1a addition. The anisotropy change when binding reached a plateau at equilibrium was set equal to 100% binding.
Peptide Competition-Peptide competition experiments used ER concentrations that achieved maximal binding of 1 nM flERE in the anisotropy buffer. The indicated amounts of peptide were added to equilibrated flERE⅐E 2 -ER binding reactions in a total reaction volume of 20 or 15 l for the experiments comparing the SRC peptides and the TRAP-220 peptide, respectively, and incubated for 30 min at room temperature in the dark. For competition experiments using peptides derived from SRCs, a concentration of SRC1a that produces ϳ90 -100% of maximal binding of the flERE⅐ER complex was added. For the TRAP-220 peptide, a concentration of SRC1a that produces ϳ70 -80% of maximal binding of the flERE⅐ER complex was added. Anisotropy measurements were made as described above. Data for the competition experiments were graphed by curve fitting with Sigma Plot (Figs. 5 and 7A).
Analysis of SRC1a Exchange-Peptide-mediated dissociation experiments used the ER and SRC1a concentrations described above for the peptide competition experiments. Following equilibration, different peptide concentrations were added. The calculated anisotropy change before addition of peptide was set as the maximum anisotropy change at time 0. Kinetic measurements were carried out at room temperature (ϳ22°C) by taking anisotropy measurements at the indicated time points after peptide addition. The Kaleidagraph program was used to fit data collected from the NR box peptide and the control HMTP2 peptides with a logarithmic trend line for the NR box peptides and a linear trend line for the HMTP2 peptide. The SRC1 NR box 2 peptide exhibited intrinsic fluorescence that could in principle nonspecifically reduce anisotropy values and calculated anisotropy changes. However, inhibition results remained the same whether or not the data were corrected for background fluorescence of the peptide.
In the exchange experiments it was important to show that dilution was sufficient to nearly abolish rebinding by SRC1a that dissociated from the flERE⅐E 2 -ER complexes. In the dilution experiments, 1 nM flERE and 15 nM ER were preincubated for 30 min at room temperature. Two l of pre-bound flERE⅐E 2 -ER complex was diluted 1:10 into anisotropy buffer in a final volume of 20 l. SRC1a was present or absent in the anisotropy buffer used for dilution. The SRC1a concentration used here equals the final SRC1a concentration used in the active displacement experiments described below after dilution. Kinetic measurements were carried out at room temperature by taking anisotropy measurements at the indicated time points after dilution of the flERE⅐E 2 -ER complex.
For active displacement experiments, the indicated concentration of SRC1a was added to preincubated flERE⅐E 2 -ER complex in a total reaction volume of 20 l. After SRC1a binding reached equilibrium, 2 l of flERE⅐E 2 -ER⅐SRC1a complex was diluted into 18 l of anisotropy buffer containing the indicated concentrations of SRC1 NR box 4 or GRIP1 NR box 2 peptide. The anisotropy change at time 0 was set equal to 100% binding. Kinetic measurements were carried out at room temperature by taking anisotropy measurements at the indicated time points after diluting the flERE⅐E 2 -ER⅐SRC1a complex.

RESULTS
FAMA Analysis of the Interaction of Full-length SRC1a with flERE⅐E 2 -ER Complexes-Previously, we used the FAMA to analyze binding of steroid hormone receptors and RNA-binding proteins to their response elements in a small sample volume, microplate format (46,52). To extend the utility of the FAMA, we tested its ability to measure binding of full-length SRC1a to a pre-formed flERE⅐E 2 -ER complex. Theoretically, the larger flERE⅐E 2 -ER⅐SRC1a complex should undergo even slower rotational diffusion than the flERE⅐E 2 -ER complex, resulting in an anisotropy increase above that seen on binding of E 2 -ER to the flERE (Fig.  1). Control experiments showed that incubation with SRC1a did not change the basal anisotropy of the flERE probe (data not shown). After prebinding 1 nM flERE with a saturating concentration of liganded ER␣ or ER␤, SRC1a was added to the indicated concentrations (Fig. 2). Robust binding curves were obtained with apparent K d values of ϳ14 and ϳ6 nM for binding of full-length SRC1a to the flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ complexes, respectively (Fig. 2). These data demonstrated that despite the substantial size of the initial flERE⅐ER complexes, we could reliably monitor the binding of SRC1a to these complexes.
Binding of SRC1a to flERE⅐ER Complexes Is Estrogen-dependent-Studies performed using glutathione S-transferase pulldown experiments, and other techniques, demonstrated that SRC proteins bind to E 2 -ER, but not to antagonist-bound ER (53). To test whether the SRC1a binding measured by the FAMA was ligand-specific, we analyzed the ligand dependence of SRC1a binding to flERE⅐ER complexes. First, we carried out in vitro ligand exchange by incubating E 2 -ER with 100 nM E 2 , the pure antagonist ICI 182,780, or the selective estrogen receptor modulators, 4-hydroxytamoxifen and RAL. Binding of flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ with SRC1a was set equal to 100%. When the E 2 bound in the ligand binding pocket of ER␣ or ER␤ was replaced with ICI, 4-hydroxytamoxifen, or RAL, there was a nearly complete loss of SRC1a binding to the flERE⅐ER␣ and flERE⅐ER␤ complexes (Fig. 3).
SRC1a Binds Rapidly to flERE⅐ER Complexes-Studies in intact cells suggest that initial binding of ER␣ and ER␤ to EREs in DNA is quite rapid (54,55). To determine whether isolated SRC1a also binds rapidly to flERE⅐E 2 -ER complexes, we analyzed the initial binding of SRC1a to preformed flERE⅐E 2 -ER A, after excitation with polarized light, flERE rotates rapidly and the emitted light is largely depolarized. B, binding of E 2 -ER to the flERE results in a much larger, less mobile complex. Therefore, the emitted light is more highly polarized. C, binding of the large ϳ160-kDa SRC1 coactivator to the flERE⅐E 2 -ER complex further slows rotation, resulting in emitted light that is even more highly polarized and an additional increase in fluorescence anisotropy. complexes. Binding of SRC1a to flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ complexes was extremely rapid with half-times for binding of 15-20 s for both the ER␣ and ER␤ complexes (Fig. 4, A and  B). These data indicate that purified SRC1a, in the absence of any further modifications by enzymes, exhibits the capacity for very rapid binding to ERE⅐E 2 -ER complexes.
Role of NR Boxes in SRC1a Binding to ER Complexes-When E 2 is bound to ER, the ER ligand binding domain assumes a conformation resulting in a hydrophobic coactivator binding cleft that interacts with the LXXLL motifs in coactivators. Little is known about the relative importance of the LXXLL motifs and amino acid sequences outside of the LXXLL motifs in the interaction of coactivators with ERs. Because coactivator peptides containing LXXLL motifs are quite small relative to the flERE⅐ER complex, their binding to the flERE⅐ER complex does not increase the anisotropy signal. This difference in the anisotropy signal upon binding of an LXXLL peptide, or the fulllength SRC1a, to flERE⅐ER allowed us to carry out competition experiments to analyze the role of the NR boxes in binding of SRC1a to ERE⅐ER complexes.
Previous studies suggested that the interaction of SRC1a with ER may be mediated primarily through NR boxes 2 and 4 (32,33). We therefore tested the ability of peptides containing either SRC1a NR box 2 or box 4, or SRC2/GRIP1 NR box 2 and their native flanking amino acid sequences, to inhibit binding of full-length SRC1a to flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤. The concentrations of peptide required to inhibit SRC1a binding to flERE⅐E 2 -ER␣ by 50% (the IC 50 ) were ϳ0.9, ϳ1.0, and ϳ3.8 M for the GRIP1 NR box 2, SRC1a NR box 2, and SRC1a NR box 4 peptides, respectively (Fig. 5A). For the flERE⅐E 2 -ER␤⅐SRC1a complex, relative IC 50 values were ϳ2.1, ϳ3.6, and ϳ6.2 M for the SRC1a NR box 2, SRC1a NR box 4, and GRIP1 NR box 2 peptides, respectively (Fig. 5B). Because of intrinsic fluorescence, very high concentrations of the GRIP NR box 2 peptide could not be used. Therefore, even at 9 M, the highest concentration of GRIP NR box 2 peptide, binding of SRC1a to the flERE⅐E 2 -ER␤ complex did not approach zero. This data were therefore less accurate. Because SRC1a was present at a saturating concentration of 25 nM, the LXXLL peptides exhibited an ϳ40 to ϳ250-fold lower affinity for the flERE⅐ER complexes than full-length SRC1a. No inhibition was observed with 10 M of a control peptide (HMTP2) lacking the LXXLL motif (data not shown).
When NR Box Peptides Are Present, Pre-formed ERE⅐ER⅐ SRC1a Complexes Dissociate Rapidly-In the sequential model of coregulator recruitment, coactivators rapidly carry out their functions and are replaced by other coregulators and coregulator complexes (9,12,14). To test aspects of this model using isolated proteins, we determined if the SRC1 NR box peptides, mimicking incoming coactivator complexes, could be used to demonstrate rapid dissociation of pre-formed flERE⅐E 2    concentrations that produced maximum inhibition of SRC1a binding to flERE⅐E 2 -ER in the competition assays shown in Fig.  4. Based on logarithmic trend lines fitted to the data, the times (t1 ⁄ 2 ) required for 50% of the pre-bound SRC1a to dissociate from flERE⅐E 2 -ER␣ were ϳ1 and ϳ3 min for the SRC1a NR box 2 and box 4 peptides, respectively (Fig. 6A). The t1 ⁄ 2 values for NR box 2-and NR box 4-mediated dissociation of flERE⅐E 2 -ER␤⅐SRC1a were each ϳ2 min (Fig. 6B).

TRAP-220 NR Box Peptides Compete with ERE⅐ER⅐SRC1a
Complexes-Chromatin immunoprecipitation studies indicate that SRC1 that dissociates from the ERE⅐ER complex would often be replaced by a different coactivator or coactivator complex, such as the DRIP/TRAP/ARC complex (10,11,13). We therefore tested the ability of a peptide containing the TRAP-220 NR box 2 sequence to dissociate pre-formed flERE⅐E 2 -ER⅐SRC1a. Previous studies suggested that TRAP-220 NR box peptides may bind to ERs with much lower affinity than SRC NR box peptides (56). To maximize our ability to detect TRAP-220 NR box peptide-mediated dissociation of flERE⅐E 2 -ER⅐SRC1a, we used concentrations of full-length SRC1a that resulted in only ϳ70 -80% of maximal binding. Because we had to use lower concentrations of SRC1a to carry out the TRAP 220 NR box competitions than the SRC NR box competitions, the anisotropy changes were smaller. Consequently, the variability in the data from these experiments was somewhat worse than for the studies using the SRC NR box peptides and the precision of the data is somewhat lower.
We first tested the potency and efficacy of TRAP-220 NR box 2 peptide-mediated inhibition in competition experiments. The peptide inhibited binding of SRC1a to both flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ with a relative IC 50 of ϳ40 M (Fig.  7A). In the same experiment, 4 M SRC1 NR box 4 peptide inhibited formation of flERE⅐E 2 -ER␣⅐SRC1a by ϳ30% and 3 M SRC1 NR box 4 peptide inhibited formation of flERE⅐E 2 -ER␤⅐SRC1a by ϳ50%, whereas 98 M HMPT2 did not inhibit formation of the flERE⅐E 2 -ER⅐SRC1a complexes (data not shown).
When a high concentration (140 M) of TRAP-220 NR box 2 peptide was present, we observed rapid dissociation of SRC1a from the flERE⅐E 2 -ER complexes. Based on logarithmic trend lines fitted to the data, the t1 ⁄ 2 values for dissociation of SRC1a from flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ were ϳ2.4 and ϳ2.9 min, respectively (Fig. 7, B and C).
NR Box Peptide Actively Displaces SRC1a from Pre-existing ERE⅐ER⅐SRC1a Complexes-The experiments in Figs. 6 and 7 demonstrated that when an excess of NR box peptide was present, binding of SRC1a to the flERE⅐ER complex was rapidly lost. However, these studies did not distinguish between two possible mechanisms: (i) SRC1a rapidly dissociates from flERE⅐ER  and is then prevented from re-binding because the NR box peptide, which is present in large excess, occupies the coactivator binding cleft or (ii) the NR box peptide actively displaces the bound SRC1a. The most straightforward way to block rebinding of dissociated SRC1a to the complex, without adding NR box peptide, is to dilute the samples. In this method, diluting the reaction after formation of the flERE⅐ER⅐SRC1a complex reduces the potential concentration of free SRC1a in solution to a level so low that very little re-binding will occur (see Fig. 2). To carry out these experiments, we used a sensitive instrument that could carry out anisotropy determinations on diluted 20-l samples containing only 0.1 nM flERE. As a control, we took pre-diluted flERE⅐E 2 -ER␣ and flERE⅐E 2 -ER␤ complexes and added a concentration of free SRC1a equivalent to the highest concentration of SRC1a that could be present after diluting the flERE⅐E 2 -ER⅐SRC1a complexes. Adding this concentration of SRC1 to the pre-diluted produced flERE⅐E 2 -ER complexes produced little or no increase in anisotropy (Fig. 8, A  and B). This shows that there is very little rebinding of dissociated SRC1a at the concentration present in the 10-fold diluted samples.
Although there is more electronic noise in assays carried out using 0.1 nM flERE than in standard assays carried out using 1 nM flERE, the outcomes were clear. In the absence of added NR box peptide, at physiologic salt (150 mM), the half-times for dissociation of SRC1a from the flERE⅐E 2 -ER␣⅐SRC1a and flERE⅐E 2 -ER␤⅐SRC1a complexes were ϳ3 and ϳ4 min, respectively. Addition of increasing amounts of SRC1 NR box 4 peptide, which exhibited similar IC 50 values for the flERE⅐E 2 -ER␣⅐SRC1a and flERE⅐E 2 -ER␤⅐SRC1a complexes in the competition experiments (Fig. 5), resulted in similar concentration-dependent increases in the rate of SRC1a dissociation from flERE⅐E 2 -ER␣⅐SRC1a and flERE⅐E 2 -ER␤⅐SRC1a complexes (Fig. 8, C and D). At 25 M, the NR box 4 peptide t1 ⁄ 2 values for displacement of SRC1a from the flERE⅐E 2 -ER␣⅐SRC1a and flERE⅐E 2 -ER␤⅐SRC1a complexes were ϳ33 and ϳ59 s, respectively (Fig. 8, C and D). GRIP1 NR box 2 peptide displaced SRC1a from flERE⅐E 2 -ER␣ more potently than from flERE⅐E 2 -ER␤ (Fig. 5) and displaced SRC1a from the flERE⅐E 2 -ER␣⅐SRC1a complexes much more rapidly than from flERE⅐E 2 -ER␤⅐SRC1a complexes (Fig. 8, E and F). At 2.5 M, the GRIP1 NR box 2 peptide t1 ⁄ 2 value for displacement of SRC1a from the flERE⅐E 2 -ER␣⅐SRC1a was ϳ12 s, whereas at 3 M GRIP1 NR box 2 peptide displaced SRC1 from the flERE⅐E 2 -ER␤⅐SRC1a complex with a t1 ⁄ 2 of ϳ135 s (Fig. 8, E and F). High NR peptide concentrations increased the rate of dissociation of SRC1a by 5-15-fold compared with dissociation of SRC1a in the absence of NR peptide. This work suggests that the competition data of Fig. 5 is due to displacement of the bound coactivator by the NR box peptides and that active displacement plays an important role in rapid coactivator exchange.

Application of the FAMA to Coactivator Binding and High
Throughput Screening-Ligand-dependent recruitment of coactivators to ER is a vital step in ER-regulated gene expression and has therefore been the focus of many recent studies. It is widely accepted that sequences in the coactivators outside of the LXXLL motifs, and the sequence of the hormone response element to which the receptor is bound are important in coactivator binding (20, 21, 38 -44, 57). However, the complexity of assays involving DNA binding sites, full-length coactivators, and full-length ER has led nearly all researchers to focus on studies that involve LXXLL peptides, or the receptor interact- . The data at t ϭ 0 was set equal to 100%. The data represents the average of at least three experiments. Because lower concentrations of SRC1a were used, the maximum anisotropy values were lower than in the other figures. The actual anisotropy values in A for flERE⅐E 2 -ER ϩ peptide and flERE⅐E 2 -ER ϩ SRC1 were 68 and 79 mA for the ER␣ complexes, respectively, and 61 and 89 mA for the ER␤ complexes, respectively. In B, they were 74 and 86 mA units for the ER␣ complexes: Ϫ and ϩ SRC1a at t ϭ 0, respectively, and in C, they were 80 and 107 mA for the ER␤ complexes Ϫ and ϩ SRC1a at t ϭ 0, respectively. FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5 ing domain of P160 coactivators and ER that is not bound to DNA (28 -36). Here we describe a microplate-based assay for analyzing interaction of full-length coactivators with steroid hormone receptors bound to fluorescein-labeled DNA binding sites. Because these assays are carried out in low volumes (20 l), they conserve limited supplies of full-length coactivator.

Active Displacement Enhances Utilization of Coactivators
The important roles of estrogen bound to ER in diseases ranging from breast cancer to osteoporosis has made coactivator binding an attractive target for high throughput screening to identify modulators of ER-coactivator interaction. Classically, these studies involved identification of ER ligands that bind in the ER ligand binding pocket and result in altered interaction with coactivator LXXLL peptides or the receptor interacting domain, or used cell-based screens. We demonstrate here that in vitro ligand exchange using 4-hydroxytamoxifen, RAL, and ICI 182,780 (Flutexamide) is highly effective and is readily combined with FAMA to quantitate coactivator binding (Fig.  3). Ligand exchange followed by FAMA should allow rapid screening of small molecules that bind in the ERs ligand binding pocket and influence coactivator binding. This system should therefore find broad application in screens to identify novel receptor antagonists and agonists.
Many recent studies have focused not on the identification of new receptor ligands, but on the identification of peptides, peptidomimetics, and small molecules that disrupt receptor-coactivator interactions (20, 58 -68). Since the changes in anisotropy we measure in the FAMA result from changes in the size of the complex (Fig. 1), binding of the relatively large full-length P160 coactivator, SRC1a, to flERE⅐ER complexes elicits a substantial increase in anisotropy, whereas binding of small peptides and the SRC1 RID fragment do not produce a significant increase in anisotropy (data not shown). With its low volume and microplate format, the FAMA is therefore well suited to high throughput screening of small molecule libraries to identify novel inhibitors of coactivator binding. Our demonstration that NR box peptides compete with SRC1a for binding to the flERE⅐ER complex (Figs. 4 -7) provides proof-of-principle for this use of the assay.
Binding of SRC1a to flERE⅐ER Complexes-In our binding studies, full-length SRC1a bound saturated flERE⅐E 2 -ER␤ complexes with an ϳ2-fold higher affinity than the flERE⅐E 2 -ER␣ complexes. Using surface plasmon resonance, Cheskis and coworkers (29) examined the interaction between ERs (not bound to an ERE) and full-length SRC1. They did not observe much difference in affinity between binding of SRC1 to ER␣ and ER␤. Because these studies involved immobilized SRC1 purified directly on the support and were done in the absence of the ERE, the fluorescence anisotropy and surface plasmon resonance results are not directly comparable.
Intracellular studies using fluorescence photobleaching and other techniques to probe the binding of ER to DNA suggest that ER moves around rapidly on DNA (54,55). Chromatin immunoprecipitation and other studies led to the concept of rapid coactivator exchange (9 -11, 13, 14). We evaluated SRC1a binding to ERE⅐ER complexes in a defined system in which only the ERE, purified ER, and the coactivator were present. Our assays do not contain an energy source and there are no known enzyme activities that might covalently modify the baculovirusexpressed, purified SRC1a, or the ERs. SRC1a bound very rapidly to the flERE⅐ER complexes with a t1 ⁄ 2 for binding of 15-20 s. Very rapid binding was also observed by Cheskis and co-workers (29) who used surface plasmon resonance to analyze binding of SRC1 to ER␣ and ER␤ in the absence of the ERE (29). These studies suggest that once the ER binds an estrogen, initial binding of P160 coactivators is extremely rapid. Whether the ER binds the P160 coactivator after binding to an ERE, or before binding to an ERE, may depend on the rate at which E 2 -ER locates and binds to EREs compared with the rate at which it binds P160 coactivators.
Role of LXXLL Motifs in SRC1a Binding-Studies using diverse techniques led to the identification of LXXLL motifs as important in coactivator binding (15, 57, 65, 69 -72). However, the contribution of individual LXXLL motifs to binding of fulllength P160 coactivators is not well defined. To explore this, we evaluated the ability of several NR box peptides to block binding of full-length SRC1a to the flERE⅐E 2 -ER complex. At high concentrations, individual NR box peptides were sufficient to largely abolish binding of SRC1a to the flERE⅐E 2 -ER complexes. These data are consistent with the view that the LXXLL-containing NR boxes are necessary for SRC1a binding. However, an ϳ40 -250-fold molar excess of LXXLL peptide over SRC1a was required to inhibit binding of SRC1a to the flERE⅐E 2 -ER complexes by 50%. The large molar excess of each NR box peptide that was required to inhibit SRC1a binding was especially striking because the NR box peptides were preincubated with the flERE⅐E 2 -ER complexes before adding SRC1a. Our data therefore indicates that individual LXXLL motifs are critical for SRC1a binding, but they are insufficient for high affinity binding of SRC1a to the E 2 -ERE⅐ER complex. Therefore, interaction surfaces on SRC1a and ER distinct from the LXXLL motifs in SRC1a and the LXXLL binding cleft in ER make significant contributions to the high affinity with which full-length SRC1a binds to the ERs.
The peptide competition assays allowed exploration of SRC NR box selectivity for ER subtypes using a full-length SRC1a. Our selectivity data are generally consistent with binding studies that determined selectivity by measuring the K d values of the individual P160 NR box peptides for E 2 -ER␣ and E 2 -ER␤ (30,32,33,35). Thus, the SRC NR box selectivity for ER established using NR box peptides seems to hold true even in the context of full-length coactivator. Interestingly, these differences extend to the rate at which NR box peptides displace SRC1a from ER␣ and ER␤. The GRIP NR box 2 peptide displaces SRC1a from flERE⅐E 2 -ER␣ more than 10 times faster than it displaces SRC1a from flERE⅐E 2 -ER␤ (Fig. 8, E and F). It has been proposed that NR box selectivity helps define ER-coactivator selectivity (19,36,56,69). The different NR box preferences we, and others, have observed for ER␣ and ER␤ may help explain their different transcriptional activities in target cells despite similar ligand and DNA binding affinities.
We also performed competition assays to evaluate the relative potency of the TRAP-220 NR box 2 peptide. Warnmark et al. (56) reported that ER binds SRC2/GRIP1/TIF2 NR box peptides with much higher affinity than TRAP-220 peptides and that TRAP-220 displays a strong preference for binding to ER␤. In contrast, Burakov et al. (73) found that ER binds to an NR box-containing TRAP-220 fragment with only slightly lower affinity than the NR box-containing SRC1 fragment and that TRAP-220 binds to both ER␣ and ER␤ equally well. We found that the TRAP-220 LXXLL peptide exhibited similar potency in blocking SRC1a binding to flERE⅐ER␣ and flERE⅐ER␤. However, the TRAP-220 NR box peptide is a much weaker competitor than the SRC NR box peptides. The SRC1a NR box 4 peptide is Ͼ5-fold more effective than the TRAP-220 peptide in inhibiting SRC1a binding.
LXXLL Peptides Actively Displace SRC1a from ERE⅐ER Complexes-It is widely accepted that the process of regulated transcription initiation by steroid/nuclear receptors involves a complex series of events in which numerous proteins interact with and dissociate from the complex (14). However, studies in intact cells have not readily addressed two important issues. (i) Do coactivators leave the complex by simply dissociating from the complex, after which they are replaced by other coregulators that occupy the now empty coactivator binding site, or are the coactivators actively displaced from the complex by incoming proteins? (ii) The process of coactivator exchange in cells is complex and involves both macromolecular interactions and diverse covalent modifications of the coactivators by several enzymes in the complex. Whether the very rapid replacement of coregulators is due at least in part to intrinsic properties of the coregulators, or only to alterations in protein-protein interactions resulting from covalent modification is unclear. Our studies have begun to address these issues. In the absence of any external agent, the flERE⅐E 2 -ER⅐SRC1a complex does not exhibit the very rapid loss of coregulator seen in intact cells. Studies with two NR box peptides show that the NR box-containing peptides can actively displace full-length SRC1a from the complex and produce rapid exchange. This rapid exchange is consistent with the very rapid dissociation of SRC proteins from ER␣ and ER␤ seen in intracellular systems (54,55). This data suggests that binding of SRC1a to ER is a dynamic process in which the diverse contacts that contribute to high affinity binding are rapidly and repeatedly made and broken. Even though SRC1a is bound to the ERE-ER complex, this process of partial dissociation, or coactivator breathing, frequently exposes the coactivator binding cleft. When no other coactivator or LXXLL peptide with high affinity for this site is present, SRC1a, which is tethered to the ER through other contacts and is therefore in both high local concentration and close proximity to the coactivator binding cleft, rapidly rebinds to that segment of the ER. When the coactivator peptide is present and coactivator breathing exposes the coactivator binding cleft, the LXXLL peptide is able to enter the coactivator binding cleft. This prevents rebinding by the corresponding LXXLL-containing segment of the SRC1a and therefore favors further dissociation of SRC1a from the ER. As the concentration of LXXLL peptide increases, there is an increasing probability that it will occupy the binding cleft rather than allowing the dissociated segment of SRC1a to rebind.
Covalent modification of coregulators doubtless plays a key role in altering their protein-protein interactions and association with multiprotein complexes on DNA (12,14). The high affinity of SRC1a for ERE⅐ER complexes we (Fig. 2) and others (29) report raises the possibility that covalent modification of the SRC1a is absolutely required for rapid exchange. However, our simple in vitro system contains only DNA, purified ER, and SRC1a, is devoid of known enzymatic activities, and lacks an energy source or added coenzymes. Our demonstration of rapid exchange of SRC1a with NR box peptides indicates that SRC1a exhibits both high affinity binding to ER and an intrinsic capacity for rapid exchange. Covalent modification of SRC1a then expands on and regulates this capacity. Our data, including the observation that nanomolar concentrations of fulllength SRC1a are required to bind flERE⅐ER complexes, whereas micromolar concentrations of LXXLL peptides are required for competition, is consistent with a model in which SRC1a binds to ERE⅐ER complexes through multiple interactions that are individually fairly weak, but taken together result in high affinity binding. By summing up these multiple weak interactions, SRC1a achieves both high affinity interaction with the ERs and retains the capacity to be rapidly displaced by other proteins.