Differential recruitment of co-regulatory proteins to the human estrogen receptor 1 in response to xenoestrogens

Abstract

The diverse biological effects of xenoestrogens may be explained by their ability to differentially recruit co-regulatory proteins to the estrogen receptor (ER). We employed high-throughput receptor affinity binding and co-regulatory protein recruitment screening assays based on fluorescence polarization and time resolved florescence resonance energy transfer (TR-FRET), respectively, to assess xenoestrogen-specific binding and co-regulatory protein recruitment to the ER. Then we used a functional proteomic assay based on co-immunoprecipitation of ER-bound proteins to isolate and identify intact co-regulatory proteins recruited to a ligand-activated ER. Through these approaches, we revealed differential binding affinity of bisphenol-A (BPA) and genistein (GEN) to the human ERα (ESR1) and ligand-dependent recruitment of SRC-1 and SRC-3 peptides. Recruitment profiles were variable for each ligand and in some cases were distinct compared to 17β-estradiol (E2). For example, E2 and GEN recruited both SRC-1 and -3 peptides whereas BPA recruited only SRC-1 peptides. Results of the functional proteomic assay showed differential recruitment between ligands where E2 recruited the greatest number of proteins followed by BPA then GEN. A number of proteins share previously identified relationships with ESR1 as determined by STRING analysis. Although there was limited overlap in proteins identified between treatments, all ligands recruited proteins involved in cell growth as determined by subnetwork enrichment analysis (p < 0.05). A comparative, in silico analysis revealed that fewer interactions exist between zebrafish (Danio rerio) esr1 and zebrafish orthologs of proteins identified in our functional proteomic analysis. Taken together these results identify recruitment of known and previously unknown co-regulatory proteins to ESR1 and highlight new methods to assay recruitment of low abundant and intact, endogenous co-regulatory proteins to ESR1 or other nuclear receptors, in both human and aquatic species.

1. Introduction

Estrogen (E2) is a steroid hormone that controls a diverse array of normal biological processes in vertebrates but is also implicated in reproductive and non-reproductive disease (Dahlman-Wright et al., 2006). Classically, E2 exerts its effects through nuclear estrogen receptors (ERs), which are ligand-dependent transcription factors that regulate gene expression through a complex network of protein–protein and protein–DNA interactions (Gruber et al., 2002). Upon ligand induction, the ER undergoes a series of regulatory modulations, including dimerization, nuclear translocation, and recruitment of co-regulatory proteins to form a macromolecular transcriptional complex. This complex consists of co-activators and/or co-repressors that activate or repress downstream gene targets in response to agonists and antagonists, respectively (Gruber et al., 2002).

It has been hypothesized that differential association of the ERs with co-regulatory proteins directs activation at discrete promoter elements, leading to selectivity in downstream gene expression (Shibata et al., 1997; Shang and Brown, 2002). This phenomenon came to be known as the ‘co-activator hypothesis’ which describes how the same ligand can manifest different biological activities in discrete tissues within the same animal (McDonnell, 2005). For example, while TAM acts as an ER antagonist in human breast tissue, it acts as an ER agonist in other tissues (McDonnell et al., 2002). This further led to the realization that structurally diverse ER ligands could induce specific structural changes in the conformation of ERs, which lead to variations in co-regulator protein affinity for the receptor, thus, impacting downstream gene activation or repression. (McDonnell et al., 2002).

To date, over 30 ER co-regulatory proteins have been identified in mammalian systems which interact with ER transcriptional complex machinery in response to E2 and alternate ligands. The most well-characterized ER complex members are those that comprise the p160 family, including steroid receptor co-activators, SRC-1 (NCoA-1), SRC-2 (TIF2, GRIP-1), and SRC-3 (AIB1, ACTR, p/CIP, RAC3, TRAM-1) (Karmakar et al., 2009). SRC-1, has demonstrated direct binding of specific peptides to the ER in response to E2 through yeast two-hybrid and phage ELISA assays (Hall and Korach, 2002; Lee et al., 2002). Interactions between ER and SRCs have also been suggested in activation of the receptor by employing siRNA strategies as loss of SRC-1 or -3 alters E2-driven cellular processes in human astrocytoma cells (González-Arenas et al., 2012). The family of p160 proteins are also highly expressed in a number of tumors and have been proposed to account for resistance to therapeutics such as aromatase inhibitors or TAM (Tikkanen et al., 2000b; Lauritsen et al., 2002; Xu and Li, 2003; Shao et al., 2004; Labhart et al., 2005; Karmakar et al., 2009; McBryan et al., 2012), highlighting their importance in disease development and treatment regimes.

While the interaction of co-accessory proteins with the ER in response to the endogenous ligand E2 has been rigorously examined, modulation by environmentally-relevant chemicals has yet to be explored in-depth. Hormonally active agents compose a diverse class of natural and synthetic compounds (Kuiper et al., 1998a; Sonnenschein and Soto, 1998; Okubo et al., 2004), and can disrupt steroid signaling in organisms through a variety of mechanisms. For example, xenoestrogens alter E2 signaling by modulation of endogenous steroid synthesis and/or direct modulation of the ERs through agonist or antagonistic means (Shanle and Xu, 2010). Many of these xenoestrogens have been causally linked to adverse health outcomes in target organisms when exposed under environmental conditions (Bergman et al., 2012).

Differential recruitment of co-regulatory proteins represents an additional plausible mechanism for endocrine disruption in humans and other species exposed to xenoestrogens. To date, research on the ligand-dependence of co-regulatory protein recruitment to ERs has focused primarily on therapeutic SERMs that have been used in recent years for the treatment of ER-positive breast cancers. These studies indicated that structurally diverse ligands can drive differential recruitment of co-regulatory proteins within a single cellular context. While most of the investigations of ER-co-regulatory recruitment have focused on the well-studied p160 family members (SRCs) in response to E2, only a handful of studies have reported interactions of these proteins with the receptor in response to binding of ER by the xenoestrogens genistein (GEN), diethylstilbestrol, (DES), bisphenol-A (BPA), and nonylphenol (NP). These targeted analyses were performed primarily by employing glutathione S-transferase (GST) pull-down, combinatorial phage display, yeast two hybrid, and surface plasmon resonance assays (Nishikawa et al., 1999; Paige et al., 1999; Kraichely et al., 2000; Parker et al., 2000; Routledge et al., 2000; Wong et al., 2001), although there is growing interest in employing non-targeted methods to evaluate global co-regulatory protein recruitment.

Based on the current gaps in knowledge, the objectives of our study were to investigate xenoestrogen-induced activation of the human estrogen receptor α (hERα, also known as ESR1) using a suite of in vitro, high-throughput real-time receptor binding (fluorescent polarization) assays, to investigate the potential for ligand-specific co-regulatory protein recruitment by SRC-1 and SRC-3 recruitment assays (time resolved fluorescent resonance energy transfer, TR-FRET), and to identify intact co-regulatory proteins bound to endogenous E2- and xenoestrogen-activated ESR1 transcriptional complexes using a co-immunoprecipitation method. Finally, we sought to take a comparative approach to both highlight the utility of our methods in the assessment of differential co-regulatory protein recruitment to nuclear receptors in aquatic species and to compare co-regulatory protein recruitment between human ESR1 and zebrafish (Danio rerio) esr1 in response to E2 binding as ER signaling pathways are moderately conserved across species (Lam et al., 2011).

2. Materials and methods

2.1. Mammalian cell culture

MCF-7 cells (American Type Culture Collection (ATCC), Manassas, VA) were cultured in phenol-red free Eagle's Minimum Essential Medium (MEM, Corning cellgro, 17305-CV) supplemented with 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 1% penicillin–streptomycin (Hyclone, SV30010), 1.5 g/L sodium bicarbonate (Corning cellgro, 25-035-CL), 1% l-glutamine (25-005-CI), and 10% fetal bovine serum (Corning cellgro, 35-010-CV). Human Embryonic Kidney 293 (HEK293) cells (ATCC) were cultured in phenol-red free Dulbecco's Modified Essential Medium (DMEM, Corning cellgro, 17-205-CV) supplemented with 1% l-glutamine, 1% penicillin–streptomycin, and 10% fetal bovine serum.

2.2. Fluorescence polarization ligand binding assay

Fluorescence polarization (FP) was first proposed by Adamczyk et al. (2002)) and offers significant advantages compared to conventional radioligand binding protocols. Unlike radioligand-based assays, the FP assay is performed in a homogeneous format that relies on displacement of a fluorescent-labeled estrogen probe from free-solution ER in order to calculate relative binding affinities of ligands. All ligands (17β-estradiol (E2, Sigma, 2578), 4-OH-tamoxifen (TAM, Sigma, T5648), Genistein (GEN, Sigma, G6649), and bisphenol-A (BPA, supplied by NIEHS)) were prepared as concentrated stocks in DMSO and further diluted into binding buffer (10 mM Tris–HCl pH 7.4 (Sigma, 154,563), 50 mM KCl (Sigma, P5405), 10% glycerol (Sigma, G5516), 0.1 mM DTT (Promega, V3151), 0.02% sodium azide (Sigma, S2002), 1 μg/mL bovine gamma globulin (BGG, Sigma, G5009), 0.2% CHAPS (Sigma, C9426)). The pET-32b vector (Millipore, 69016) containing human estrogen receptor alpha ligand binding domain (ESR1-LBD, a gift of Marc Ruff, L'Institut de génétique et de biologie moléculaire et cellulaire (IGBMC), Strasbourg, France), was expressed in origami cells (Millipore) and purified by HPLC (Agilent 1100). Serial dilutions of ligands were prepared in triplicate by adding 200 μL of the working ligand stocks to the first well of each row in a 96-well plate and transferring 100 μL to each additional well containing buffer. Purified ESR1-LBD was diluted and added to each well to give a final ER concentration of 10 nM. Finally, the fluorescent probe (F-E₁), an estrone conjugate coupled to fluorescein-5-thiosemicarbazide (FTSC), was added to give a final concentration of 0.5 nM. The plate was incubated at room temperature in the dark for 1 h prior to analysis. Fluorescence polarization (FP) was measured on a Biotek Synergy H1 spectrophotometer using an excitation wavelength of 485 nm and emission wavelength of 525 nm. FP was converted to percent inhibition (I_% = (A₀ − A) / (A₀ − A₁₀₀) ∗ 100) where A = absorbance and plotted against concentration of ligand using SigmaPlot 11 (Systat Software, Inc., San Jose, CA). Curves were fit by transforming the x-axis to a logarithmic scale and applying a nonlinear regression curve using a sigmoidal-dose response with variable slope to obtain IC₅₀ values for the corresponding ligands in SigmaPlot 11.

2.3. Time-resolved fluorescence resonance energy transfer (TR-FRET)

To elucidate the mechanism of xenoestrogen action in recruiting co-activators SRC-1 and SRC-3, we utilized a TR-FRET assay. In this assay, compounds were measured for their ability to recruit model peptides derived from distinct adapter regions of the human SRC-1 and SRC-3 proteins to the ER-ligand complex. The peptides employed in the assay each contain an LXXLL motif and occur in the receptor interacting domain (RID) of the co-regulatory protein. One additional LXXLL peptide for SRC-1 is present in the C-terminal region of the protein (Fig. 2).The Lanthascreen estrogen receptor co-activator kit was purchased from Invitrogen Corporation (A15885), and each assay was performed following the manufacturer's instructions. Briefly, all compounds (E2, TAM, BPA, GEN) were serially diluted in DMSO to yield a 12-point dilution series (range of final concentrations 0.01 nM–10.0 μM). Aliquots (10 μL) of each ligand were transferred to 4 wells of a white low-volume 384-well plate (Corning Inc., Corning, NY). The ESR1 Ligand-binding domain tagged with glutathione-S-transferase (GST-ER LBD) was diluted to 4 × in complete buffer and 5 μL added to each well to give a final assay concentration of 7 nM. A premixed solution of 4 × fluorescein peptide and terbium labeled anti-GST antibody were diluted in complete buffer and 5 μL added to each well such that final assay concentrations were 250 nM and 5 nM, respectively. The assay plate was incubated for 1 h at ambient temperature in the dark followed by TR-FRET measurements in a Spectramax M5 dual monochromator spectrophotometer (Molecular Devices) with delay and integration times of 100 μs and 200 μs, respectively. The terbium species was excited at 340 nm with a bandwidth of 30 nm and both terbium and fluorescein emissions were measured at 478 and 520 nm, respectively. The emission ratio was calculated as the raw intensity of acceptor (520 nm) to donor (478 nm) fluorescence with 4 replicates measured per concentration of ligand. The dose–response curves were generated by plotting emission ratios (y axis) against ligand concentration (x axis) using Prism 4 (GraphPad Software, La Jolla, CA). The curves were analyzed by transforming the x-axis to a logarithmic scale and applying a nonlinear regression curve using a sigmoidal-dose response with variable slope using SigmaPlot 11 (Systat Software Inc., San Jose, CA).

Fig. 2.

General SRC protein box schematic representing the seven regions containing LXXLL motifs located within the receptor-interacting and activation domains (RID) (adapted from (Edwards, 2000) and SRC peptides screened for recruitment to the ESR1 using TR-FRET with their respective sequences and domain locations.

2.4. Protein extraction

MCF-7 cells were cultured in phenol-red free Eagle's Minimum Essential Medium (MEM, Cellgro 17-305-CV, Manassas, VA) in T225 cell culture flasks. Exposures were conducted in triplicate for each compound (two flasks of confluent cells per replicate exposure, with an additional two flasks to serve as an IgG-only, non-specific binding control as described below) to 10 nM E2, 1 μM BPA, or 1 μM GEN, as well as to a DMSO carrier-control for a period of 3.5 h. Nuclear extracts were prepared with the Active Motif Nuclear Extract Kit (40,010). Briefly, media were aspirated from dishes and cells were washed with ice-cold PBS/phosphatase inhibitor solution. Cells were removed from each dish by gentle scraping and cells from individual plates for a given treatment replicate were pooled and washed with hypotonic buffer on ice. Lysis was performed with hypotonic lysis buffer by gentle pipetting. Lysates were centrifuged for 30 s at 14,000 × g at 4 °C and supernatant (cytosolic fraction) was discarded. Nuclear extracts were prepared from the pellet using a hypertonic digestion buffer with DNase using gentle vortexing. The extracts were clarified at 14,000 × g at 4 °C for 10 min and processed for immunoprecipitation immediately without freezing.

2.5. Co-immunoprecipitation of intact co-regulatory proteins

We applied a mass-spectrometry compatible immunoprecipitation (IP) strategy for isolating ER transcriptional complexes from nuclear extracts of xenoestrogen-exposed MCF-7 cells. In this method, anti-ESR1 antibodies (Active Motif, 61,035) were covalently crosslinked to protein A/G agarose (Pierce Biotechnology, 20,421) using dimethylpimelimidate (DMP, Pierce Biotechnology, 21,667) prior to immunoprecipitation of the complexes. For each IP reaction, 20–30 μg of antibody (either rabbit anti-ESR1 (Active Motif) or mouse IgG1 (Cell Signaling, 5415) for nonspecific-binding controls) were bound to 30 μL of protein A/G beads overnight in 1 mL PBS (Thermo Scientific, SH30256.01) prior to washing 3 × with 0.2 M sodium borate buffer (pH 9.0). Beads were resuspended in 1 mL of 0.2 M borate buffer containing 20 mM DMP and crosslinked for 40 min at room temperature. The crosslinked antibody beads (IgG and anti-ESR1) were washed 1 × with 0.2 M ethanolamine (pH 8.0, Sigma, E-6133) for 2 h at room temperature to quench residual DMP. Non-crosslinked IgG antibodies were removed by washing with 3 × 1 mL 0.5% v/v acetic acid (Fisher, A38), 150 mM NaCl (Fisher, S271). The beads were then washed 3 × with ice cold PBS prior to use. Nuclear extracts from xenoestrogen exposed cells (0.1–0.2 mg) were added to 25 μL of uncoupled protein A/G resin and incubated for 30 min to “pre-clear” the extracts prior to IP. The precleared extracts were then adjusted to 1.0 mL with nuclear extraction buffer and added to 25 μL of either anti-ESR1- or IgG-coupled resin. The samples were incubated overnight at 4 °C with end-over-end mixing. The resin was then rinsed 3 × with buffer containing 150 mM NaCl, 50 mM Tris (Fisher, BP152), 10 mM EGTA (Sigma, E-4378), and 0.2% NP40 detergent (Sigma, NP40S) and subsequently washed 3 × with 50 mM ammonium bicarbonate (Fisher, BP2413). Finally, immunoprecipitated ESR1 transcriptional complexes were eluted by adding 50 μL of 0.2% Rapigest SF (Waters Corp., 186,001,860) surfactant in 50 mM ammonium bicarbonate to the beads and boiling at 95 °C. The beads were removed by centrifugation and the supernatant was retained for analysis by LC-MS/MS.

2.6. LC-MS/MS

Immunoprecipitated ESR1 complexes (and IgG non-specific binding controls) were digested using Trypsin prior to MS analysis. Supernatants from IP experiments were reduced with 10 mM DTT and alkylated with 200 mM iodoacetamide at room temperature in the dark for 30 min. Sequencing grade trypsin was added to the samples at a loading of ~ 1:50 trypsin:protein (based on original nuclear extract concentration) and samples were digested overnight at 37 °C with shaking. Following digestion, samples were centrifuged at 14,000 x g and the supernatant was adjusted to 1.0% trifluoroacetic acid (TFA) and 2% acetonitrile v/v.

Proteomic analysis of the samples was conducted using a gel-free HPLC-MS/MS approach, with a reversed-phase nanocapillary HPLC separation (C₁₈) followed by data-independent MS/MS (MS^E) analysis with a Synapt G2 (Waters Corp.) High Definition Mass Spectrometer (quadrupole-time-of-flight).

2.7. LC-MS/MS analysis

Mascot was set up to search the SwissProt_2013x database (unknown version, 539,829 entries) also assuming trypsin. IdentityE was searched with a fragment ion mass tolerance of 0.025 Da and a parent ion tolerance of 0.0100 Da. Mascot was searched with a fragment ion mass tolerance of 0.040 Da and a parent ion tolerance of 10.0 PPM. Carbamidomethyl of cysteine was specified in IdentityE and Mascot as a fixed modification. Deamidation of asparagine and glutamine, oxidation of methionine and phosphorylation of serine, threonine and tyrosine were specified in IdentityE as variable modifications. Oxidation of methionine was specified in Mascot as a variable modification. Scaffold (version Scaffold_4.4.3, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 98.0% probability to achieve an FDR less than 1.0%. Peptide Probabilities from IdentityE and Mascot were assigned by the Peptide Prophet algorithm (Keller et al., 2002) with Scaffold delta-mass correction. Peptide Probabilities from IdentityE were assigned by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 87.0% probability to achieve an FDR less than 5.0% and contained at least 1 identified peptide. Protein Probabilities were assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters. Protein complex predictions and assemblies were performed using the online Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) 10 resource (http://string-db.org) and Subnetwork Enrichment Analysis was performed using Pathway Studio 9.0 (Elsevier). For the analysis, 2 entities (proteins) had to be present in a subnetwork for inclusion and results were limited to 200 subnetworks with best p-value (p < 0.05) for enrichment cut-off.

2.8. Cross-species analysis

Gene symbols corresponding to the proteins identified in our analysis were obtained using the retrieve/ID mapping option on the UniProt website (uniprot.org/uploadlists). The gene names were converted to their zebrafish orthologs using The Zebrafish Model Organism Database (ZFIN, zfin.org). The Human Gene Database (GeneCards, genecards.org) was used to convert genes not found in The Zebrafish Model Organism Database. STRING was used to identify known and predicted protein–protein interactions using the zebrafish orthologs and specifying D. rerio as the organism.

3. Results

3.1. Xenoestrogens bind ESR1 with varying affinities

We have examined a suite of putative xenoestrogens to quantitatively assess binding to ESR1 using a fluorescence-based competitive ligand binding assay known as fluorescence polarization (FP). We first performed direct binding studies to assess the interaction of ESR1 and a custom-synthesized, fluorescent-labeled probe compound (Fl-E₁ fluorescent probe) that was prepared by coupling of fluorescein-5-thiosemicarbazide to the 17-keto position of estrone (Fig. 1). Saturation binding curves for this labeled compound (data not shown) indicate a binding affinity of 2.3 nM (K_d) for the ESR1–Fl-E₁ complex, consistent with previous reports and similar to the affinity of the native ligand E2 (Freyberger et al., 2010).

Fig. 1.

(A) Chemical structure of the fluorescent estrogen (F-E1) conjugate used in fluorescence polarization competitive ligand binding studies with the human estrogen receptor α. This synthetic probe was prepared by labeling of estrone at the 17-position with fluorescein thiosemicarbazide. (B–E) Binding curves for E2, TAM, BPA, and GEN, respectively, to ESR1 using a fluorescence polarization assay after 1 h incubation in the dark at room temperature. Fluorescence polarization (FP) was measured on a Biotek Synergy H1 spectrophotometer using an excitation wavelength of 485 nm and emission wavelength of 525 nm. FP was converted to percent inhibition (I_% = (A₀ − A) / (A₀ − A₁₀₀) ∗ 100) where A = absorbance and plotted against concentration of ligand using SigmaPlot 11 (Systat Software, Inc., San Jose, CA). Curves were fit by transforming the x-axis to a logarithmic scale. Concentrations (nM) of each ligand are plotted as log concentration compared to % inhibition (n = 3).

Following validation of the Fl-E₁ probe and FP assay, competitive binding curves for 2 putative xenoestrogens were generated based on their ability to displace the fluorescent probe and bind to purified recombinant ESR1-LBD (ligand binding domain) (Fig. 1). In this study, all compounds exhibited specific binding to the ESR1 as indicated by their IC₅₀ values and relative binding affinities (RBAs). The antagonist TAM displayed a stronger affinity for ESR1 with an IC₅₀ value of 2.8 nM compared to that of E2 (9.5 nM). All other compounds exhibited weak binding relative to the endogenous hormone, E2, (TAM > E2 ≫ GEN > BPA) (Fig. 1).

3.2. Xenoestrogens differentially recruit SRC-1 and SRC-3 peptides to the ER complex

Structurally-diverse xenoestrogens may act via divergent mechanisms to stimulate estrogen-signaling, including differential, ligand-dependent recruitment of co-regulatory proteins to human ESR1. Upon optimization of experimental parameters, we screened the afore-mentioned xenoestrogens for recruitment activity by the ESR1 using these select peptides. Estradiol was used as a positive control, since its ability to activate and induce recruitment of co-regulators to ESR1 is known. E2 yielded a maximum TR-FRET signal relative to all other compounds in its ability to induce ESR1-SRC interactions (Fig. 3). However, E2 differentially recruited specific peptides and based on the EC₅₀ values (Table 2) show the following interaction; SRC-1(3) > SRC-3(3) > SRC-1(2) > SRC-3(2) > SRC-1(1). Estrogen failed to recruit the C-terminal peptide of SRC-1 and SRC-3(3). As anticipated, the antagonist TAM did not recruit any of the co-regulator peptides which is consistent with the lack of binding affinity of this compound to ESR1 determined by the FP assay. BPA failed to induce interaction of the receptor with SRC-3 peptides, but did induce recruitment of two SRC-1 derived peptides (SRC-1(2) and SRC-1(3)).

Fig. 3.

ESR1 ligands differentially recruit SRC-1 and SRC-3 co-regulatory peptides to the human ESR1. Recruitment was measured using time resolved fluorescent resonance energy transfer after 1 h incubation at room temperature. Recruitment curves are presented as the ratio of fluorescent units (520 nm/478 nm) emitted after excitation at 340 nm versus twelve concentrations (nM) of each ligand. Points indicate average fluorescence ratio of each dose (n = 4). Dose–response curves were generated by plotting emission ratios (y axis) against ligand concentration (x axis) using SigmaPlot 11 (Systat Software, Inc., San Jose, CA). Curves were analyzed by transforming the x-axis to a logarithmic scale and applying a nonlinear regression curve using a sigmoidal-dose response with variable slope using SigmaPlot 11.

Table 2.

Calculated EC₅₀ values of each SRC peptide for the ESR1 in response to each ligand in the TR-FRET assays. Dose–response curves were generated by plotting emission ratios (y axis) against ligand concentration (x axis) using SigmaPlot 11 (Systat Software, Inc., San Jose, CA). Curves were analyzed by transforming the x-axis to a logarithmic scale and applying a nonlinear regression curve using a sigmoidal-dose response with variable slope and EC₅₀ values calculated using SigmaPlot 11.

	SRC-1 (nM)				SRC-3 (nM)
Ligand	(1)	(2)	(3)	(4)	(1)	(2)	(3)
17β-Estradiol	8.44	5.08	2.61	–	–	7.19	2.74
4-OH-Tamoxifen	–	–	–	–	–	–	–
Genistein	–	25.0	–	511	17.0	38.8	72.4
Bisphenol-A	–	32.3	3230	–	–	–	–

Perhaps the most interesting result for the compounds studied was that of GEN, which variably recruited SRC-1 and -3 peptides but generated a maximum TR-FRET signal for SRC-3(1). In fact, all three SRC-3 peptides were recruited to the complex by GEN whereas only SRC-3 (2 and 3) were recruited by E2. BPA failed to recruit any of the SRC-3 peptides. Overall, SRC recruitment profiles were highly ligand specific.

3.3. E2 and xenoestrogens differentially recruit co-regulatory proteins to activated ESR1-transcriptional complexes

Proteomic analysis of MCF-7 nuclear complexes immunoprecipitated using anti-ESR1 antibodies after xenoestrogen exposure resulted in a large number of total protein identifications. Across all samples (including nonspecific-binding IgG controls), 913 unique proteins were identified with at least one high-scoring peptide match. Within those data, 657 proteins were found to have been isolated due to non-specific interaction artifacts in the immunoprecipitation procedure. These proteins were identified in at least one of the five immunoprecipitation control preparations in which nuclear extracts were isolated using mouse IgG crosslinked to protein A/G beads and were considered to be non-interactors (or non-specific interactors) with the ESR1. Veracity of the immunoprecipitation procedure and the sensitivity of the proteomics assay were illustrated by the fact that the human ER (ESR1, the target of the IP antibody) was identified in all samples which were immunoprecipitated with anti-ESR1 coupled protein A/G beads. Of the proteins that were not detected in IgG controls, 152 were identified in at least one replicate of DMSO carrier controls, 135 in at least one replicate of the E2-treated cells, 37 in at least one replicate of BPA-treated cells, and 15 in at least one replicate of the GEN-treated cells. As illustrated in Fig. 4, there was considerable overlap in the identity of ESR1 transcriptional complex members present in xenoestrogen-treated MCF-7 cells; however a number of treatment-specific proteins were observed for E2, GEN, and BPA. The list of proteins identified from xenoestrogen-treated proteins in this manner is presented in Table 3.

Fig. 4.

Estrogen receptor α (ESR1) ligands differentially recruit proteins to ESR1. Venn diagram depicting proteins identified in at least one replicate ESR1 transcriptional complex isolated from MCF-7 nuclear extracts after exposure 10 nM 17β-estradiol (E2), 1 μM bisphenol-A (BPA), 1 μM genistein, or DMSO carrier-control (Control) for a period of 3.5 h (n ≥ 3). Only proteins that were not observed in IgG nonspecific binding immunoprecipitation controls are represented.

Table 3.

Proteins identified from ESR1 transcriptional complexes of MCF-7 cells after exposure to 10 nM 17β-estradiol (E2), 1 μM bisphenol-A (BPA), or 1 μM genistein (GEN) for 3.5 h, excluding proteins identified in control group. Proteins are organized by treatment and number of identifications among triplicate samples within a treatment group (IDs).

Ligand	UniProt Accession Number	Gene Symbol	Protein Name	Molecular Weight	IDs
E2	P03372	ESR1	Estrogen receptor	66 kDa	3
	Q9UBS4	DNAJB11	DnaJ homolog subfamily B member 11‡	41 kDa	3
	P62330	ARF6	ADP-ribosylation factor 6‡	20 kDa	2
	Q13185	CBX3	Chromobox protein homolog 3‡	21 kDa	2
	P35222	CTNNB1	Catenin beta-1†	85 kDa	2
	P08574	CYC1	Cytochrome c1, heme protein, mitochondrial‡	35 kDa	2
	Q14315	FLNC	Filamin-C‡	291 kDa	2
	P50148	GNAQ	Guanine nucleotide-binding protein G(q) subunit alpha†	42 kDa	2
	O15049	N4BP3	NEDD4-binding protein 3	60 kDa	2
	O75489	NDUFS3	NADH dehydrogenase [ubiquinone] iron–sulfur protein 3, mitochondrial‡	30 kDa	2
	P61289	PSME3	Proteasome activator complex subunit 3†	30 kDa	2
	Q9Y2W1	THRAP3	Thyroid hormone receptor-associated protein 3‡	109 kDa	2
	P07311	ACYP1	Acylphosphatase-1	11 kDa	1
	P20648	ATP4A	Potassium-transporting ATPase alpha chain 1‡	114 kDa	1
	Q5SY80	C1orf101	Uncharacterized protein C1orf101	110 kDa	1
	Q4AC94	C2CD3	C2 domain-containing protein 3	260 kDa	1
	Q5M9N0	CCDC158	Coiled-coil domain-containing protein 158	127 kDa	1
	Q96A33	CCDC47	Coiled-coil domain-containing protein 47	56 kDa	1
	P19256	CD58	Lymphocyte function-associated antigen 3	28 kDa	1
	Q9UNH5	CDC14A	Dual specificity protein phosphatase CDC14A‡	67 kDa	1
	Q96JB2	COG3	Conserved oligomeric Golgi complex subunit 3	94 kDa	1
	Q96EY1	DNAJA3	DnaJ homolog subfamily A member 3, mitochondrial‡	52 kDa	1
	O75923	DYSF	Dysferlin	237 kDa	1
	Q96C19	EFHD2	EF-hand domain-containing protein D2	27 kDa	1
	Q14156	EFR3A	Protein EFR3 homolog A	93 kDa	1
	P05198	EIF2S1	Eukaryotic translation initiation factor 2 subunit 1‡	36 kDa	1
	Q8N8S7	ENAH	Protein enabled homolog‡	67 kDa	1
	O14908	GIPC1	PDZ domain-containing protein GIPC1	36 kDa	1
	P00367	GLUD1	Glutamate dehydrogenase 1, mitochondrial‡	61 kDa	1
	Q03113	GNA12	Guanine nucleotide-binding protein subunit alpha-12‡	44 kDa	1
	P16520	GNB3	Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-3‡	37 kDa	1
	O75487	GPC4	Glypican-4‡	62 kDa	1
	P16402	HIST1H1D	Histone H1.3‡	22 kDa	1
	Q9H910	HN1L	Hematological and neurological expressed 1-like protein	20 kDa	1
	Q99714	HSD17B10	3-hydroxyacyl-CoA dehydrogenase type-2	27 kDa	1
	P48741	HSPA7	Putative heat shock 70 kDa protein 7‡	40 kDa	1
	P50213	IDH3A	Isocitrate dehydrogenase [NAD] subunit alpha, mitochondrial‡	40 kDa	1
	Q27J81	INF2	Inverted formin-2	136 kDa	1
	Q15046	KARS	Lysine–tRNA ligase‡	68 kDa	1
	Q5VWX1	KHDRBS2	KH domain-containing, RNA-binding, signal transduction-associated protein 2‡	39 kDa	1
	Q7Z3Y7	KRT28	Keratin, type I cytoskeletal 28	51 kDa	1
	Q14525	KRT33B	Keratin, type I cuticular Ha3-II	46 kDa	1
	Q71F78	LCA10	Putative lung carcinoma-associated protein 10	17 kDa	1
	Q03252	LMNB2	Lamin-B2‡	68 kDa	1
	Q32MZ4	LRRFIP1	Leucine-rich repeat flightless-interacting protein 1‡	89 kDa	1
	P61326	MAGOH	Protein mago nashi homolog‡	17 kDa	1
	O60524	NEMF	Nuclear export mediator factor NEMF‡	123 kDa	1
	Q9BSD7	NTPCR	Cancer-related nucleoside-triphosphatase	21 kDa	1
	Q99571	P2RX4	P2X purinoceptor 4	43 kDa	1
	Q86U42	PABPN1	Polyadenylate-binding protein 2‡	33 kDa	1
	O43252	PAPSS1	Bifunctional 3′-phosphoadenosine 5′-phosphosulfate synthase 1	71 kDa	1
	Q15120	PDK3	[Pyruvate dehydrogenase [lipoamide]] kinase isozyme 3, mitochondrial	47 kDa	1
	Q8N0Y7	PGAM4	Probable phosphoglycerate mutase 4	29 kDa	1
	A1L390	PLEKHG3	Pleckstrin homology domain-containing family G member 3‡	134 kDa	1
	Q9UHX1	PUF60	Poly(U)-binding-splicing factor PUF60‡	60 kDa	1
	Q15276	RABEP1	Rab GTPase-binding effector protein 1	99 kDa	1
	Q16576	RBBP7	Histone-binding protein RBBP7†	48 kDa	1
	Q8WZ73	RFFL	E3 ubiquitin-protein ligase rififylin	41 kDa	1
	P04279	SEMG1	Semenogelin-1	52 kDa	1
	Q15637	SF1	Splicing factor 1†	68 kDa	1
	P53007	SLC25A1	Tricarboxylate transport protein, mitochondrial	34 kDa	1
	Q9NTJ3	SMC4	Structural maintenance of chromosomes protein 4‡	147 kDa	1
	Q92485	SMPDL3B	Acid sphingomyelinase-like phosphodiesterase 3b	51 kDa	1
	Q8NB12	SMYD1	SET and MYND domain-containing protein 1	57 kDa	1
	P62306	SNRPF	Small nuclear ribonucleoprotein F‡	10 kDa	1
	Q9BX66	SORBS1	Sorbin and SH3 domain-containing protein 1‡	143 kDa	1
	Q07617	SPAG1	Sperm-associated antigen 1	104 kDa	1
	P11277	SPTB	Spectrin beta chain, erythrocytic‡	246 kDa	1
	Q9Y5M8	SRPRB	Signal recognition particle receptor subunit beta‡	30 kDa	1
	Q8WUY1	THEM6	Protein THEM6	24 kDa	1
	Q6UW68	TMEM205	Transmembrane protein 205	21 kDa	1
	O94826	TOMM70A	Mitochondrial import receptor subunit TOM70	67 kDa	1
	Q9C040	TRIM2	Tripartite motif-containing protein 2	82 kDa	1
	P43897	TSFM	Elongation factor Ts, mitochondrial‡	35 kDa	1
	Q13885	TUBB2A	Tubulin beta-2 A chain†	50 kDa	1
	P08670	VIM	Vimentin‡	54 kDa	1
BPA	P03372	ESR1	Estrogen receptor	66 kDa	3
	P29279	CTGF	Connective tissue growth factor†	38 kDa	1
	P49448	GLUD2	Glutamate dehydrogenase 2, mitochondrial	61 kDa	1
	Q03113	GNA12	Guanine nucleotide-binding protein subunit alpha-12‡	44 kDa	1
	Q58FG0	HSP90AA5P	Putative heat shock protein HSP 90-alpha A5	39 kDa	1
	Q86UX2	ITIH5	Inter-alpha-trypsin inhibitor heavy chain H5	105 kDa	1
	Q9BW19	KIFC1	Kinesin-like protein KIFC1	74 kDa	1
	Q14CN4	KRT72	Keratin, type II cytoskeletal 72	56 kDa	1
	P78386	KRT85	Keratin, type II cuticular Hb5	56 kDa	1
	O43813	LANCL1	LanC-like protein 1	45 kDa	1
	Q13203	MYBPH	Myosin-binding protein H	52 kDa	1
	P0CAP1	MYZAP	Myocardial zonula adherens protein	54 kDa	1
	P54920	NAPA	Alpha-soluble NSF attachment protein‡	33 kDa	1
	P15531	NME1	Nucleoside diphosphate kinase A†	17 kDa	1
	Q04721	NOTCH2	Neurogenic locus notch homolog protein 2	265 kDa	1
	Q7Z3S9	NOTCH2NL	Notch homolog 2 N-terminal-like protein	26 kDa	1
	Q99497	PARK7	Protein DJ-1	20 kDa	1
	P16435	POR	NADPH–cytochrome P450 reductase	77 kDa	1
	P36873	PPP1CC	Serine/threonine-protein phosphatase PP1-gamma catalytic subunit	37 kDa	1
	Q6WKZ4	RAB11FIP1	Rab11 family-interacting protein 1	137 kDa	1
	Q96DA2	RAB39B	Ras-related protein Rab-39B	25 kDa	1
	Q9H930	SP140L	Nuclear body protein SP140-like protein	67 kDa	1
	Q6UW68	TMEM205	Transmembrane protein 205	21 kDa	1
	Q9NZR1	TMOD2	Tropomodulin-2	40 kDa	1
	Q9NY65	TUBA8	Tubulin alpha-8 chain†	50 kDa	1
GEN	P03372	ESR1	Estrogen receptor	66 kDa	3
	P55957	BID	BH3-interacting domain death agonist	22 kDa	1
	P16402	HIST1H1D	Histone H1.3	22 kDa	1
	Q58FF8	HSP90AB2P	Putative heat shock protein HSP 90-beta 2†	44 kDa	1
	Q5VWX1	KHDRBS2	KH domain-containing, RNA-binding, signal transduction-associated protein 2	39 kDa	1
	Q8N1N4	KRT78	Keratin, type II cytoskeletal 78‡	57 kDa	1
	P18615	NELFE	Negative elongation factor E	43 kDa	1
	P60866	RPS20	40S ribosomal protein S20	13 kDa	1
	Q6UW68	TMEM205	Transmembrane protein 205	21 kDa	1
	Q86VY4	TSPYL5	Testis-specific Y-encoded-like protein 5	45 kDa	1

^†after the protein name denotes a primary relationship with ESR1

^‡denotes a secondary relationship as determined by STRING analysis.

3.4. Proteins recruited in response to E2 and xenoestrogen exposure share a number of known and unknown relationships with ESR1

STRING software was used to identify proteins sharing known protein–protein interactions based on the published literature. A number of the identified proteins were determined to share primary relationships with ESR1 (Fig. 5). E2 recruited six proteins sharing known primary interactions with ESR1; Catenin beta-1 (CTNNB1), Guanine nucleotide-binding protein G(q) subunit alpha (GNAQ), Histone-binding protein RBBP7 (RBBP7), TUBB2A protein (TUBB2A), Proteasome activator complex subunit 3 (PSME3), and Splicing factor 1 (SF1). BPA recruited three proteins with known primary interactions: Connective tissue growth factor (CTGF), Nucleoside diphosphate kinase A (NME1), Tubulin alpha-8 chain (TUBA8). Genistein recruited one protein with known primary interactions, Heat shock protein HSP 90-beta (HSP90AB1). The proteins sharing known primary interactions with ESR1 also share relationships with a number of other proteins identified in our analysis as determined by STRING software and may interact with ESR1 via tethering mechanisms. Proteins that do not share known relationships with ESR1 based on the published literature represent potential novel interactors and co-regulatory proteins. Interestingly, a larger percentage of the proteins identified in the BPA and genistein exposed groups did not share any relationship with ESR1 either through direct interactions or via secondary interactions (70% and 60%, respectively) compared to E2 (42%).

Fig. 5.

Proteins recruited to ESR1 in response to ESR1 ligands share both known and unknown interactions. String networks depict interactions among putative ESR1 co-regulatory proteins (shown with gene name) bound to ESR1 after 3.5 h exposure to 10 nM 17β-estradiol (A), 1 μM bisphenol-A (B), or 1 μM genistein (C) in MCF-7 cells based on identification of proteins listed in Table 3. Nodes represent individual proteins, and connecting lines denote known (experimentally-derived or inferred) protein–protein interactions reported in the literature. Figure compiled from http://string-db.org.

3.5. E2 and BPA recruited proteins involved in similar cellular processes while genistein recruited proteins involved in unique cellular processes

Subnetwork enrichment analysis was performed using Pathway Studio to identify cellular processes that were enriched with proteins identified in each exposure group (Fig. 6). All three treatments caused enrichment of proteins involved in cell growth. A high degree of overlap existed in cell processes enriched after exposure to E2 and BPA and included apoptosis, cell adhesion, cell cycle, cell death, cell differentiation, cell growth, and cell migration. E2 specifically caused enrichment of proteins involved in endocytosis and mitosis while BPA specifically caused enrichment of proteins involved in cell survival, and S phase. Both BPA and genistein caused enrichment of oxidative stress proteins and both E2 and genistein caused enrichment of RNA splicing proteins. Cell processes specifically enriched by genistein included mitochondrial damage, DNA damage checkpoint, fatty acid oxidation, gene silencing, lipid transport, oncogenesis, and response to oxidative stress. The cellular process enriched with the largest percentage of proteins recruited to ESR1 in response to E2 and BPA were apoptosis and cell differentiation whereas genistein most recruited proteins involved in cell growth and RNA splicing.

Fig. 6.

Pie charts depicting statistically significant (p < 0.05) enrichment of cell processes as determined by subnetwork enrichment analysis. Each slice represents the percentage of proteins sorted into each biological process related to the total number of proteins identified in MCF-7 cells exposure to either 10 nM 17β-estradiol (A), 1 μM bisphenol-A (B), or 1 μM genistein (C).

3.6. Fewer protein–protein interactions exist between zebrafish (D. rerio) orthologs of esr1 and other proteins identified in our co-immunoprecipitation analysis

The proteins identified in the E2 exposed group were converted to their zebrafish orthologs and an interaction network was constructed based on known and inferred protein–protein interactions using STRING analysis. In all, 67 of the 76 proteins identified in the E2 group were successfully mapped to zebrafish orthologs using ZFIN and GeneCards. Of the 67 proteins, 64 were mapped in the zebrafish STRING database and used to construct the interaction network (Fig. 7). Results of the in silico analysis revealed that ten proteins shared known or inferred relationships with esr1 either directly (primary) or indirectly (secondary). This accounted for 15.6% of mapped proteins while 37 of the 64 proteins successfully mapped to their zebrafish orthologs shared known or inferred relationships with ESR1 in the human STRING database and accounted for 57.8%. These results are detailed in Table 4.

Fig. 7.

Summary of interactions among putative zebrafish esr1 co-regulatory proteins (shown with gene name) in the 17β-estradiol (E2) exposed group. Nodes represent individual proteins, and connecting lines denote known (experimentally-derived or inferred) protein–protein interactions reported in the literature. Figure compiled from http://string-db.org.

Table 4.

Summary of known and inferred protein-protein interactions between zebrafish esr1 and successfully mapped zebrafish orthologs of putative co-regulatory proteins identified in the 17β-estradiol (E2) exposed group in our study as determined by STRING analysis.

Human gene symbol	Zebrafish gene symbol	Protein name	Human interaction	Zebrafish interaction
ESR1	esr1	Estrogen receptor	–	–
ACYP1	acyp1	Acylphosphatase-1
ARF6	arf6a	ADP-ribosylation factor 6	‡	‡
C2CD3	c2cd3	C2 domain-containing protein 3
CBX3	cbx3a	Chromobox protein homolog 3
CCDC47	ccdc47	Coiled-coil domain-containing protein 47
CDC14A	cdc14aa	Dual specificity protein phosphatase CDC14A
COG3	cog3	Conserved oligomeric Golgi complex subunit 3
CTNNB1	ctnnb1	Catenin beta-1	†
CYC1	cyc1	Cytochrome c1, heme protein, mitochondrial	‡
DNAJA3	dnaja3b	DnaJ homolog subfamily A member 3, mitochondrial	‡
DYSF	dysf	Dysferlin
EFHD2	efhd2	EF-hand domain-containing protein D2
EFR3A	efr3a	Protein EFR3 homolog A
EIF2S1	eif2s1a	Eukaryotic translation initiation factor 2 subunit 1	‡	‡
ENAH	enah	Protein enabled homolog	‡	‡
FLNC	flnca	Filamin-C	‡	‡
GIPC1	gipc1	PDZ domain-containing protein GIPC1
GLUD1	glud1a	Glutamate dehydrogenase 1, mitochondrial	‡
GNA12	gna12a	Guanine nucleotide-binding protein subunit alpha-12	‡	†
GNAQ	gnaq	Guanine nucleotide-binding protein G(q) subunit alpha	†	†
GNB3	gnb3a	Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-3	‡	‡
GPC4	gpc4	Glypican-4	‡
HN1L	hn1l	Hematological and neurological expressed 1-like protein
HSD17B10	hsd17b10	3-hydroxyacyl-CoA dehydrogenase type-2
IDH3A	idh3a	Isocitrate dehydrogenase [NAD] subunit alpha, mitochondrial
INF2	inf2	Inverted formin-2
KARS	kars	Lysine–tRNA ligase	‡
KHDRBS2	khdrbs2	KH domain-containing, RNA-binding, signal transduction-associated protein 2	‡
LMNB2	lmnb2	Lamin-B2	‡
LRRFIP1	lrrfip1a	Leucine-rich repeat flightless-interacting protein 1	‡
MAGOH	magoh	Protein mago nashi homolog	‡
N4BP3	n4bp1	NEDD4-binding protein 3
NDUFS3	ndufs3	NADH dehydrogenase [ubiquinone] iron–sulfur protein 3, mitochondrial
NTPCR	ntpcr	Cancer-related nucleoside-triphosphatase
P2RX4	p2rx4a	P2X purinoceptor 4
PABPN1	pabpn1l	Polyadenylate-binding protein 2
PAPSS1	papss1	Bifunctional 3′-phosphoadenosine 5′-phosphosulfate synthase 1
PDK3	pdk3a	[Pyruvate dehydrogenase [lipoamide]] kinase isozyme 3, mitochondrial
PSME3	psme3	Proteasome activator complex subunit 3	†	‡
PUF60	puf60a	Poly(U)-binding-splicing factor PUF60	‡
RFFL	rffl	E3 ubiquitin-protein ligase rififylin
SF1	sf1	Splicing factor 1
SLC25A1	slc25a1a	Tricarboxylate transport protein, mitochondrial
SMC4	smc4	Structural maintenance of chromosomes protein 4
SMPDL3B	smpdl3b	Acid sphingomyelinase-like phosphodiesterase 3b
SMYD1	smyd1a	SET and MYND domain-containing protein 1
SNRPF	snrpf	Small nuclear ribonucleoprotein F	‡
SORBS1	sorbs1	Sorbin and SH3 domain-containing protein 1	‡
SPAG1	spag1a	Sperm-associated antigen 1
SPTB	sptb	Spectrin beta chain, erythrocytic	‡
SRPRB	srprb	Signal recognition particle receptor subunit beta	‡
TMEM205	tmem205	Transmembrane protein 205
TOMM70A	tomm70a	Mitochondrial import receptor subunit TOM70
TRIM2	trim2a	Tripartite motif-containing protein 2
TSFM	tsfm	Elongation factor Ts, mitochondrial	‡
TUBB2A	tubb2	Tubulin beta-2 A chain	†
VIM	vim	Vimentin	‡
ATP4A	atp1a1a.4	Potassium-transporting ATPase alpha chain 1	‡
HIST1H1D	histh1l	Histone H1.3	‡
PGAM4	pgam1a	Probable phosphoglycerate mutase 4
PLEKHG3	plekhg3	Pleckstrin homology domain-containing family G member 3	‡
RBBP7	rbb4l	Histone-binding protein RBBP7	†
THEM6	them6	Protein THEM6

^†Proteins sharing primary relationship with zebrafish esr1 are denoted

^‡proteins sharing secondary relationships with esr1 are denoted

4. Discussion

In the classical model of ER action, the endogenous hormone E2 mediates transactivation in the nucleus. A cascade of signaling events occur such that co-regulatory proteins (co-activators or co-repressors) are recruited to a macromolecular transcriptional complex, the composition and stoichiometry of which is dependent on the conformation adopted by the ER when bound to agonist or antagonist ligands (Paige et al., 1999; O'Malley, 2005). While the influence of native estrogens such as E2 on these transactivation events has been well-studied, the role of xenoestrogens on this system is poorly defined in comparison. Since xenoestrogens have the ability to bind to ER and are thought to cause variable conformational changes in the protein, it is possible that differential recruitment of co-regulators may occur. This xenoestrogen-specific recruitment may have consequences on downstream transcription of gene targets by directing the complex to alternate promoter elements.

To begin to tease apart ligand-specific effects on human ESR1, we performed binding assays to measure binding affinities of various ESR1 ligands for the receptor. In lieu of performing radioligand binding studies, we employed a high-throughput fluorescence polarization (FP) assay to screen the select xenoestrogens for their ability to bind human ESR1. Calculated RBAs, compared to E2 set at 100%, indicated weak binding of most ligands tested except TAM. These values correspond to the range of RBAs (or IC₅₀ values) that have been previously reported based on radioligand, polarization anisotropy and fluorescence-based assays (Table 1) (Bolger et al., 1998; Kuiper et al., 1998a; Kuiper et al., 1998b; Matthews et al., 2000; Nikov et al., 2000; Parker et al., 2000; Nikov et al., 2001; Kuramitz et al., 2002; Ohno et al., 2002; Ohno et al., 2003; Mueller et al., 2004; Olsen et al., 2005; Matsui, 2007; Freyberger et al., 2010; Kwok and Cheung, 2010; McLachlan et al., 2011). In general, our assay produces slightly increased affinities for tested xenoestrogen ligands compared to radioligand assays, a result that is likely due to the fact that FP is a homogeneous assay. This alleviates concern over kinetically-limited binding and hysteretic conditions that may impact results from heterogeneous assays such as the radioligand binding assay. This observation is consistent with RBAs gained from contemporary fluorescent-based techniques, where higher values for BPA, NP and GEN have been reported compared to standard radioligand assays (Ohno et al., 2003). Like other FP assays, ours is fast, convenient, and ideal for high-throughput screening but does have the limitation of not utilizing the entire receptor. It has the additional advantage of using a novel fluorescent ligand (F-E₁) that is easily prepared in high-yield by direct reaction of estrone with commercially-available fluorescein thiosemicarbazide and subsequent purification by reversed-phase HPLC. Overall, our results illustrate the utility of this method for high-fidelity screening of molecular binding to protein receptor targets.

Table 1.

Relative binding affinities calculated from IC₅₀ values of xenoestrogen ligands screened in fluorescence polarization (FP) assay compared to those cited in the literature. Nonlinear regression curves were fit using a sigmoidal-dose response with variable slope to obtain IC₅₀ values for the corresponding ligands using SigmaPlot 11.

Ligand	IC50 (nM)	IC50 95% CI (nM)	RBA (%)	Literature RBA (%)
17β-Estradiol	9.5	8.5–10.6	100	–
4-OH-Tamoxifen	2.8	2.0–3.9	339	4.7–257a–f
Genistein	2120	1830–2470	0.45	0.2–16a,c–f,j,k
Bisphenol-A	3270	2810–3800	0.29	0.006–0.315a,d–i

^aOhno et al. (2003).

^b Nikov et al. (2001).

^cKwok et al. (2010).

^dMatsui et al. (2006).

^e Matthews et al. (2000).

^fKuiper et al. (1998a, 1998b).

^g Olsen et al. (2005).

^h Bolger et al. (1998).

ⁱKuramitz et al. (2002).

^jNikov et al. (2000).

^kMueller et al. (2004).

It is well documented that xenoestrogens generally have low affinity for ERs relative to native and therapeutic ligands; however, transcriptional activation at various response elements does not necessarily correlate with the strength of ER binding (Kuiper et al., 1998a). In addition, RBAs do not differentiate between agonist and antagonist abilities, most clearly demonstrated by TAM and other SERM compounds. This is also the case ligandssuch as GEN, which has a greater RBA to human estrogen receptor β (ESR2) (~87%) but activation potential can be greater for ESR1 in transactivation assays (Kuiper et al., 1998a). That same study showed that despite a greater RBA of certain OH-PCBs for the ERs compared to other xenoestrogens, similar activation of the receptors was not refiected in transactivation assays.

A number of studies have addressed the potential role of differential recruitment of co-regulatory proteins to explain ligand and cell type-specific effects of E2 by focusing on the p160 family of coactivators, specifically SRCs (SRC-1,-2,-3) (Chang et al., 1999a; Mc Ilroy et al., 2006; Suzuki et al., 2007). Fewer studies have focused on the role of these coactivators in ligand-specific transcriptional activation by ESR1, particularly with respect to differential cellular responses to xenoestrogens. As such, using TR-FRET, we demonstrate ligand-specific recruitment profiles of SRC-1 and SRC-3 peptides by E2, TAM, GEN and BPA. TAM did not induce recruitment of any of the tested SRC peptides to the ESR1 complex, indicating it may antagonize receptor activity by inducing a conformational change not suitable for binding of co-activator proteins. These compounds may in fact act as allosteric antagonists by recruiting co-repressor proteins rather than co-activators as would be expected with TAM in certain tissues. None of the ligands tested recruited all peptides nor were the profiles identical to E2. For example, all centrally located SRC-1 peptides were recruited by E2 while GEN and BPA were unable to induce recruitment of all three distinct LXXLL motif-containing peptides. Conversely, GEN weakly induced SRC-1(4) interaction with the ER, but this relation was not induced by E2. These results are not in complete agreement with a previous report where the complete RID of SRC-1 was recruited to ESR1 by GEN and BPA with K_d values of 21 nM and 104 nM, respectively (Suzuki et al., 2007).Perhaps differential sensitivity between the assays may account for this discrepancy. The moderate binding of SRC-1(2) to the ESR1 induced by GEN and BPA suggest that this region of the co-regulator may be positioned for maximal interaction as a result of ligand directed conformational change of the receptor.

Recently there has been increased interest in shifting from a targeted analysis of co-regulatory protein recruitment to a global analysis of the ER transcriptional complex and protein–protein interactions therein. One method requires isolating overexpressed ER transcriptional complexes using an immobilized estrogen response element on a sepharose column (Nalvarte et al., 2010). Another method relies on immunoprecipitation of bait protein and subsequent release of complex members in lysis buffer and identification by mass spectrometry (Ewing et al., 2007). Both methods require the use of recombinant bait proteins which is problematic as the overexpression of bait protein can lead to biases in the results by increasing the identification of false positives (Malovannaya et al., 2010). As such, we developed a functional proteomic method to isolate intact co-regulatory proteins bound to endogenous human ESR1 by co-immunoprecipitation using an anti-ESR1 IgG in MCF-7 cells. We used our method to identify proteins recruited to ESR1 after exposure to E2, GEN, and BPA. The analysis was sensitive enough to identify the ESR1 in each replicate from every treatment showing that it is possible to isolate transcriptional complexes from human cell culture without overexpressing the target protein. This allowed us to circumvent problems associated with abnormally high levels of bait protein (Malovannaya et al., 2010). E2 recruited the greatest number of proteins to ESR1 followed by BPA and then genistein. These results are in contrast to the SRC-1/3 peptide recruitment experiments which indicated that genistein recruited more SRC-1/3 peptides than BPA (Table 2) and highlight ligand specific responses. In support of this notion, little overlap in recruited proteins among the treatment groups occurred (Fig. 4) suggesting that each ligand induces a unique conformational change that favors the recruitment of different co-regulatory proteins.

Interestingly, SRC-3 was not among the identified proteins recruited after 3.5 h of ligand induction even though SRC-3 peptides were shown to be recruited by E2 and GEN in the TR-FRET experiments. This result was surprising as other groups have reported high concentrations of SRC-3 in MCF-7 cells (Thenot et al., 1999) and have successfully co-immunoprecipitated SRC-3 with ESR1 antibodies after 3.5 h of E2 induction (Tikkanen et al., 2000a). The discrepancy could be a result of variable stringencies in co-immunoprecipitation methods or the transient and labile nature of co-regulator interaction. Further, we potentially identified some nonspecific interactors in our analysis that could have been eliminated by adding an ultracentrifugation step after primary antibody incubation (Malovannaya et al., 2010). Nonetheless, the identification of a suite of known and previously unknown co-regulatory proteins in our analysis confirms the robustness of our mass spectrometric identification method.

Next, we sought to identify relationships between proteins recruited to human ESR1 in response to each ligand using STRING 10. This analysis revealed that a number of the identified proteins share known relationships with ESR1 through direct interactions while others share secondary relationships via interactions through other proteins (Fig. 5). For example, we identified Catenin beta-1 (CTNNB1), a protein sharing known primary relationships with ESR1, in E2 exposed cells. CTNNB1 is the downstream regulator of the Wnt pathway and an indiscriminate nuclear receptor coactivator that binds LXXLL motifs and is especially important in androgen receptor activity and cellular proliferation (Mulholland et al., 2005). It has been found to form a complex with ESR1 and facilitate transcription in the brain (Varea et al., 2009). Splicing factor 1 (SF1), also identified in the E2 exposed group, shares a direct relationship with ESR1 and affects gene transactivation and pre-mRNA splicing activities of CTNNB1 complex in colorectal cancer (Shitashige et al., 2007). The identification of CTNNB1 and SF1 in the E2 group suggests that E2 may modulate components of the Wnt signaling pathway through recruitment of Wnt-related co-regulatory proteins to ESR1 in MCF-7 cells.

We also identified potential novel ESR1 co-regulatory proteins such as Filamin-C (FLNC), an actin binding protein that is suggested to function as an androgen receptor coactivator mediating the growth and differentiation of muscle cells (Ting and Chang, 2008). While no evidence for a direct interaction between ESR1 and FLNC was found in the STRING analysis, the identification of FLNC in the E2 exposed group suggests that it might function as a previously unknown estrogen receptor co-regulatory protein and interact with ESR1 through direct or tethering mechanisms thereby coordinating regulation of the cytoskeleton. Identification of the structural protein Vimentin in the E2 group further suggests a role for ESR1 in regulating cytoskeletal architecture (Fig. 5).

STRING analysis also resulted in a number of interesting observations in the xenoestrogen exposed groups. Both E2 and BPA (but not genistein) recruited Cadherin-1 (CDH1), which exhibited a primary relationship with ESR1 and is involved in maintaining the epithelial phenotype of cells and its expression is regulated by ligand-independent actions of ESR1 (Cardamone et al., 2009). The identification of CDH1 suggests that E2 and BPA may influence cellular phenotype via inducing an interaction between ESR1 and this protein. CDH1 was also identified in the DMSO group in our analysis; however, this is not surprising as previous studies have indicated a ligand-independent role for ESR1 in maintaining the epithelial phenotype through modulation of CDH1 expression that is repressed upon ligand-binding thereby promoting an exchange of coactivators for corepressors (Cardamone et al., 2009).

Of note is the BPA-specific recruitment of connective tissue growth factor (CTGF), a matricellular protein involved in numerous biological processes include wound healing, proliferation, and cell–cell adhesion and migration (Hall-Glenn and Lyons, 2011). Previous studies have indicated that BPA is capable of increasing cell proliferation through up-regulation of CTGF expression via G protein-coupled estrogen receptor (GPER) in breast cancer fibroblasts (Pupo et al., 2012). Our results suggest another possible mechanism of BPA-induced regulation of CTGF activity through interactions with ESR1. Interestingly, E2, BPA, and genistein but not vehicle control induced an interaction between ESR1 and transmembrane protein 205 which has been implicated in resistance to the chemotherapeutic cisplatin in vitro (Shen et al., 2010). Future studies should investigate a role for ESR1 in modulating cisplatin resistance.

Through a subnetwork enrichment analysis of proteins identified in each treatment group we identified severl functional relationships. Not surprisingly, a number of the proteins recruited to ESR1 in response to E2 and xenoestrogen binding were involved in apoptosis, cell growth, and cell cycle (Fig. 6). ESR1 is known to influence these processes in MCF-7 cells (Brünner et al., 1989; Wang and Phang, 1995; Mawson et al., 2005) and may be influenced by contaminant exposure (Hsieh et al., 1998; Diel et al., 2002). E2 has also been shown to influence the differentiation of other cell types such as endothelial progenitor cells (Imanishi et al., 2005) and osteoblasts (Qu et al., 1998). While it is not surprising that these processes were found to be enriched with proteins identified in our analysis, our assay was not quantitative thus we cannot determine whether these processes were up or down-regulated.

We also identified a number of xenoestrogen exposure-specific effects. For example, BPA and genistein but not E2 recruited a large percentage of proteins involved in oxidative stress which represents a possible target of xenoestrogen action caused by differential co-regulatory protein recruitment. This is in accordance with previous studies indicating an association between BPA exposure and oxidative stress in postmenopausal women (Yang et al., 2009), and that GEN up-regulated the expression of genes involved in oxidative stress in MCF-7 cells (Borrás et al., 2006). Another striking observation of the subnetwork enrichment analysis was the highly discrepant enrichment of cellular processes by genistein compared to E2 and BPA. Genistein specifically induced recruitment of proteins involved in mitochondrial damage, fatty acid oxidation, lipid transport, and DNA damage checkpoint (Fig. 6).

We expect our co-immunoprecipitation method to be useful in investigating the recruitment of co-regulatory proteins to ligand-activated ESR1 in other species. Because ER signaling is fairly conserved across species (Lam et al., 2011), a comparative approach would increase our understanding of the complex mechanisms of action of endocrine disrupting contaminants in both human and aquatic targets. The observation that BPA and genistein are able to bind to ERs, albeit with lower affinity than E2, and influence reproductive endpoints in fish species (Scholz and Mayer, 2008) further highlights the importance of studying their interactions with the ERs. In fact, it has been found that various fish species respond differently to exposure to environmental estrogens (Miyagawa et al., 2014) which suggests that variable expression of co-regulatory proteins may play a role in such responses in addition to amino acid substitutions in ER sequences across species. Most studies to date have focused on a small suite of co-regulatory proteins such as TIF2 (Tan et al., 2005) and COUP-TFI and COUP-TFII (Métivier et al., 2000), but no reports have utilized a global approach to examine the influence of differential co-regulatory protein recruitment on molecular responses in fish exposed to xenoestrogens.

As a first-pass screen of the translatability of our results to an aquatic species, we performed an in silico analysis of interactions between putative co-regulatory proteins identified in our functional proteomic analysis and zebrafish esr1. We converted the 76 proteins in our analysis that were recruited to ESR1 in response to E2 binding to their zebrafish orthologs using ZFIN and constructed plausible interaction networks using STRING 10 (Fig. 7). In total, 64 homologous zebrafish proteins were successfully identified and 15.6% exhibited known or inferred interaction with zebrafish esr1 while 57.8% of human orthologs shared an interaction (Table 4). These results are not surprising due to the lack of data regarding co-regulatory protein recruitment of fish ERs, differences in experimental methods, and variable ligand affinity and co-regulatory protein expression across species. It is well accepted that the ligand affinities of various xenoestrogens for ESR1 differ among species such as medaka (Oryzias latipes), stickleback, bluegill (Lepomis macrochirus), and guppy (Poecilia reticulata), and one group showed that constructing esr1 chimeras in which the AF-containing domains were swapped between the fish species changed the ligand binding affinities and responsiveness of that species to the xenoestrogens (Miyagawa et al., 2014). These data suggest that the structure of the AF-domain can alter ligand binding which has implications for the recruitment and interaction of co-regulatory proteins with the ERs and downstream signaling responses. To be sure, other factors may also contribute to differential ER signaling responses not limited to species-specific amino acid sequences and the presence of multiple ER isoforms. Our co-immunoprecipitation method would be perfectly suited to study differences in recruitment of co-regulatory proteins to esr1 between fish species thereby increasing our basic understanding of ER signaling in fish.

Overall, development of relevant high-throughput screening assays affords us the ability to test and generate ER-interacting ligand and protein profiles that may be predictive of their activity in cell and tissue environments in humans and fish. As transcriptional complexes require intricate protein–protein and protein–DNA interactions, the need for identifying these complexes in a more comprehensive manner and assessing alternate promoter sites is increasingly acute in understanding ER signaling. Herein we have shown that various ESR1 agonists and antagonists exhibit variable binding affinities and that these binding affinities do not necessarily correlate with co-regulatory protein recruitment profiles, an observation that has puzzled investigators and highlights the complexity of hormonal signaling (Kuiper et al., 1998a; Chang et al., 1999b; O'Malley, 2005). Data presented here offer possible mechanisms in support of this observation as it relates to both known (SRC-3) and unknown ER co-regulatory proteins and xenoestrogens. These data also suggest that investigating recruitment of select peptides derived from known co-regulatory proteins is not an accurate predictor of the recruitment of intact, endogenous proteins to ESR1 by xenoestrogens as genistein exhibited the greatest recruitment of SRC-3 peptides, excluding E2, but recruited the fewest intact, endogenous proteins to ESR1. Further, the select suite of intact, endogenous proteins recruited to ESR1 in response to each ligand increases our understanding of the cellular consequences of xenoestrogen exposure that may have relevance for both mammalian and fish species.