Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Jun 19.
Published in final edited form as: Chem Biol. 2014 May 22;21(6):743–753. doi: 10.1016/j.chembiol.2014.03.013

Defining Estrogenic Mechanisms of Bisphenol A Analogs through High Throughput Microscopy-based Contextual Assays

Fabio Stossi 1, Michael J Bolt 1, Felicity J Ashcroft 1, Jane E Lamerdin 2, Jonathan S Melnick 2, Reid T Powell 3, Radhika D Dandekar 1, Maureen G Mancini 1, Cheryl L Walker 3, John K Westwick 2, Michael A Mancini 1
PMCID: PMC4301571  NIHMSID: NIHMS600704  PMID: 24856822

Summary

Environmental exposures to chemically heterogeneous endocrine disrupting chemicals (EDCs) mimic or interfere with hormone actions, and negatively impact human health. Despite public interest and the prevalence of EDCs in the environment, methods to mechanistically classify these diverse chemicals in a high throughput (HT) manner have not been actively explored. Here, we describe the use of multi-parametric, HT microscopy-based platforms to examine how a prototypical EDC, Bisphenol A (BPA), and eighteen poorly studied analogs (BPXs), affect estrogen receptor (ER). We show that short exposure to BPA and most BPXs induce ERα and/or ERβ and change levels of target gene transcription. Many BPXs exhibit higher affinity for ERβ and act as ERβ antagonists, while they act largely as agonists or mixed agonists/antagonists on ERα. Finally, despite binding to ERs, some BPXs exhibit lower levels of activity. Our comprehensive view of BPXs activities allows their classification and evaluation of potential harmful effects. The strategy described here used on a large scale basis likely offers a faster, more cost-effective way to identify safer BPA alternatives.

Introduction

The impact of environmental endocrine disrupting chemicals (EDCs) on human health is widely discussed, but remains poorly understood. By definition, EDCs, usually binding to nuclear receptors (NRs), act by interfering with any aspect of hormone action (e.g., synthesis, activity, degradation, etc) thus altering hormone responsive cells and tissues (Zoeller et al., 2012, Myers et al., 2009). NRs are a large class of transcription factors that play significant roles in the pathophysiology of virtually every organ (Committee, 1999). The steroid receptor subfamily of NRs is the best studied in the context of EDCs, as it comprises the receptors for estrogens (ERα and ERβ), androgens (AR), progestins (PR) and glucocorticoids (GR). Not surprisingly, given widespread environmental exposure, EDCs have been consistently associated in epidemiological studies with increased incidence of a number of pathological conditions including hormone-dependent cancers (e.g., breast and prostate), metabolic (e.g., diabetes, obesity), fertility, neurological, behavioral (Weiss, 2012) and developmental defects (De Coster and van Larebeke, 2012, Rochester, 2013). The list of potential EDCs is comprised of a large and growing number of individual compounds or mixtures, and their metabolic/environmental derivatives. These compounds have diverse chemical structures and are introduced into the environment from both natural and industrial sources. As suggested by the Environmental Protection Agency (EPA), existing assays to interrogate known, or identify new, EDCs use single data points and low throughput assays (e.g., in vitro binding or reporter gene assays), even though recent efforts include HT in vitro assays using bulk populations of cells (Rotroff et al., 2013). Thus, there is an urgent need for multi-parametric, robust, and HT cell-based assay platforms that can rigorously investigate the complex mechanisms underlying the adverse effects of known EDCs, and identify new compounds with endocrine disrupting potential. Given the appreciation of cell-to-cell heterogeneity in tumors and cell cultures, single-cell-based approaches are receiving increased attention.

Bisphenol A (BPA) is an EDC of concern due to its ability to induce developmental reprogramming in animal models (Jirtle and Skinner, 2007, Susiarjo et al., 2013). BPA is in the top 2% of all high-production-volume chemicals and is frequently used in the manufacturing of polycarbonate plastics and epoxy resins. Collectively, there is widespread use of these polymers in manufacturing of milk and food containers, baby formula bottles, interior lining of food cans, paper receipts and dental resins (Brotons et al., 1995) (Olea et al., 1996), providing numerous sources for BPA exposure during key periods of development. BPA has been shown to leach in microgram amounts from polycarbonate plastics and epoxy resins into food and water supplies (Welshons et al., 2003), and exposure to BPA is nearly ubiquitous: urinary analysis reveals that BPA is detected in >93% of the population in the United States (Calafat et al., 2005, Calafat et al., 2008).

In part due to these health concerns, many BPA analogs have been synthesized and are now used as substitutes for the parent compound (e.g., BPB, DM BPA, TM BPA, BPF, BPS). Importantly, there is a paucity of information available on the effects of these and other BPA analogs on human health and the environment, as the vast majority of studies have focused largely on the parent compound. When examined, BPA analogs have been found to be potentially estrogenic and, in some cases, anti-androgenic (Kitamura et al., 2005, Rivas et al., 2002, Vinas and Watson, 2013). Thus, large data gaps exist in our understanding of how these analogs differ from the parent compound, and whether they present more or less of a hazard to human health.

To interrogate EDC activity and to provide platforms amenable to HT identification and characterization of potential new EDCs, we have developed a series of cell-based assay systems that incorporate multiple facets of ER biology (see Figure 1A for experimental workflow). These assays and platforms, based on high throughput microscopy and multi-parametric automated image analysis, have the ability to characterize EDC activity in the context of wide-ranging mechanistic depth and with great speed on a cell-by-cell basis; importantly, these assays are highly complementary with the ones used in the ToxCast program (Rotroff et al., 2013). In this study, we have examined eighteen BPA analogs (BPXs) using a HT multi-parametric “systems level” approach that provides data over a wide range of characteristics that underlie many facets of NRs mechanisms of action, including DNA- and protein-based interactions, chromatin remodeling, transcriptional output, and cell proliferation.

Figure 1. Comparison of GFP-ERβ:PRL-HeLa vs. GFP-ERα:PRL-HeLa stable cell lines.

Figure 1

Open in a new tab

A) Experimental workflow for BPX analysis. B) PRL-HeLa array cells (ERα in red and ERβ in blue) were treated with 10nM E2 for the indicated times and percent of cells with an array was measured. C-E) Six point dose response for the percent arrays measurement at the 30 min time point for E2 (C), 4-OHT (D) and Raloxifene (E). F) Six point dose response for the array area measurement (in pixels) after 30 min of E2 treatment. G-I) recruitment of SRC-1 (G), SRC-2 (H) and SRC-3 (I) to the array after 30 min of E2 and 4OHT treatment measured as loading (Bolt et al 2013). * p<0.05 between ERα and ERβ. J) Serine 5 phosphorylated RNA Polymerase II loading to the array after 30 minutes of E2 treatment (6 point dose-response). K) dsRED2 RNA FISH time course analysis after E2 treatment represented as intensity at the array normalized to vehicle treatment which was set as 1.

Using these approaches, we found that: 1) nearly all BPXs bound to ERα and/or ERβ, suggesting that the development of BPA analogs that do not act as ER ligands may be challenging; 2) the majority of BPXs showed higher affinity for ERβ, suggesting that screens focused exclusively on ERα could underestimate effects of potential toxicants; 3) BPXs were predominantly ERβ antagonists but ERα mixed agonists; and, 4) despite binding to ERs, some BPXs were inactive or had very low activity at the highest dose tested across multiple assays, indicating that ligand binding per se was insufficient to predict BPX activity.

Results

GFP-ERα:PRL-HeLa and GFP-ERβ:PRL-HeLa array cell lines as sensors of estrogen receptors DNA binding, coregulator recruitment, chromatin remodeling and gene transcription

We previously engineered a HeLa derived cell line with a multi-copy integration of the estrogen responsive unit of the prolactin gene (PRL-HeLa cells (Sharp et al., 2006)), which stably expresses GFP-tagged ERα (GFP-ERα:PRL-HeLa (Ashcroft et al., 2011)). This HT-amenable “Big Data” approach allows for direct visualization and quantification of multiple steps of ER-mediated transcriptional activation: DNA binding, coregulator recruitment, large scale chromatin modeling, and transcription. We previously used this model system to define and classify the effects of a small set of estrogenic compounds, including BPA (Ashcroft et al., 2011).

To complement the ERα-based system, we developed a similar GFP-tagged ERβ stable PRL array cell line (GFP-ERβ:PRL-HeLa) as described in Materials and Methods. Figure 1 shows the use of these two cell lines in assays that compare ERα and ERβ activity under agonist (17β-estradiol, E2) and antagonist (4-hydroxy-tamoxifen, 4OHT) treatments. First, we demonstrated that, following E2 treatment, the percent of cells exhibiting a visible fluorescent signal at the PRL promoter array increased rapidly (minutes; Figure 1B), and then decayed over time (hours) as previously described (Ashcroft et al., 2011, Bolt et al., 2013). This measurement represents ER binding to ERE-rich promoter/enhancer DNA, and both ERα and ERβ demonstrated similar response kinetics with E2; as such, the peak time point (30 minutes) was chosen for all subsequent experiments. Following these kinetic experiments, a six point E2 dose-response analysis was performed. We found that the logEC50 for ERα and ERβ localization to the array was similar (logEC50: −8.2M vs. -8.5M), and within the same range as previously observed with other in vitro assays (Figure 1C, (Kuiper et al., 1997)). The selective estrogen receptor modulator (SERM), 4OHT had a similar logEC50 for ERα and ERβ localization to the array, while the SERM Raloxifene showed a preference for ERα over ERβ (logEC50: −8.38M vs. −7.42M) (Figure 1D-E).

We next evaluated the extent of chromatin remodeling after array binding by measuring the increase in PRL array area in response to E2 (Figure 1F). We observed a much larger change in array size for ERα as compared to ERβ, and this difference correlated with the extent of mRNA FISH output (see below, Figure 1K). 4OHT, on the other hand, caused smaller, transcriptionally inactive arrays (Ashcroft et al., 2011). This suggests that ERα and ERβ recruit different chromatin remodeling complexes to the PRL array, and that this difference is detectable using these cell lines. We therefore further investigated the ability of ERα and ERβ to recruit the p160 family of coactivators (SRC-1, SRC-2 and SRC-3) to the PRL array when liganded with E2 or 4OHT. Coactivator recruitment was determined after 30 minutes of treatment, and measured as the ratio of array to nucleoplasm intensity (Bolt et al., 2013). SRC-1 was recruited by both ERα and ERβ in response to E2, with slight preference for ERβ (Figure 1G). SRC-2 was recruited equally by both ERs (Figure 1H). In contrast, SRC-3 was only recruited to the array by ERα (Figure 1I). As expected, 4OHT failed to recruit SRC-1, -2 or -3 to the PRL array with either receptor. Thus, the PRL-HeLa array model utilizes high throughput microscopy to visually detect differences in coregulator recruitment between ERα and ERβ, and between agonists (E2) and SERM (4OHT) ligands, which is similar to previous work by others using conventional in vitro chromatin immunoprecipitation (ChIP) assays (Monroe et al., 2003, Wong et al., 2001).

To correlate these observations with ER-mediated changes in transcription, we measured E2-induced mRNA from the integrated reporter gene at the PRL array locus (dsRED2) via mRNA FISH using fluorescently labeled oligonucleotide probes (Stellaris™) (Bolt et al., 2013). To measure immediate changes in transcription, as opposed to steady state mRNA levels, we specifically focused on the 30 minute time point after exposure to compound. Also, as a marker of transcription initiation, we immunolabeled cells with a monoclonal antibody to serine 5 phosphorylated RNA polymerase II (Figure 1J). The time course of E2-mediated dsRED2 transcription largely reflected what was observed for array occupancy (Figure 1B and 1K), with an early peak between 30 minutes and 2 hours before plateauing at ~50% max activity. Overall, ERα and ERβ showed similar kinetics for transcription in response to E2, albeit with a 2-3-fold difference in magnitude, with ERα eliciting higher levels of activity. Representative images for some of the data tabulated in Figure 1 are provided in Figure S1. The similarity in time course for the peak responses in array loading and gene transcription allowed us to utilize this early time point (30 minutes) across all experiments to compare the effects of BPXs on ERα and ERβ.

High throughput cell-based analyses of ERα vs. ERβ selectivity of BPXs using protein-fragment complementation (PCA) and PRL array platforms

Next, we further developed a suite of novel biosensor assays to characterize and quantify ER activity using HT, cell-based, automated microscopy platforms. The assay systems and experimental workflow, described in Figure 1A, combined previously validated protein-fragment complementation assays (PCA; (MacDonald et al., 2006, Michnick et al., 2006)) and the PRL array system (Ashcroft et al., 2011, Bolt et al., 2013, Sharp et al., 2006).

Engineered non-fluorescent fragments of the N- and C-terminal portion of the yellow fluorescent protein (YFP) were fused in-frame to sequences encoding the ligand binding domain (LBD) of ERα (amino acids 310-547) or ERβ (amino acids 263-489) (MacDonald et al., 2006). Pairs of fusion protein constructs were stably expressed in HEK293 cells to interrogate either ERα or ERβ LBD homodimerization upon treatment with estrogenic compounds. The fluorescent signal generated in cells is measurable by automated microscopy and image analysis as previously described (Figure 2; (MacDonald et al., 2006)). For each assay, changes in mean fluorescence intensity relative to the E2 control were quantified 8 hrs after treatment with the indicated compounds (comparable results were obtained after 24hrs).

Figure 2. HTM platforms define ERα vs. ERβ selectivity of BPXs.

Figure 2

Open in a new tab

A) log EC50 values were calculated after six point dose responses either using a protein fragment complementation assay (PCA) or the PRL array stable cell lines. In the heatmap red indicates high activity compounds while blue indicates low activity compounds. B-E) examples of dose-response curves for four BPXs comparing PCA and PRL array platforms.

For the PRL array system, treatment with E2 was for 30 minutes, as described in Figure 1, and percent array occupancy was measured off the GFP-ER channel as previously described (Ashcroft et al., 2011, Bolt et al., 2013). LogEC50 values were calculated for each tested compound after performing a minimum of three independent biological replicates in a six point dose-response (ranging from 100pM to10μM).

After validating the response and quantitation strategies of these four cell-by-cell assays using known ERα and ERβ ligands, we tested 18 BPXs (see Table S1 for names, chemical structures and CAS numbers) that are in commercial use; importantly, to date, the potential endocrine disrupting activity of these BPXs have only been partially assessed. As shown in Figure 2A, we determined the relative binding affinity of each BPX for ERα and ERβ in comparison with E2, 4OHT or Raloxifene using PCA assays, and determined array occupancy of liganded ERα and ERβ using PRL arrays.

In Figure 2A, BPXs logEC50 values across all four HT assays are presented in heatmap fashion, where red represents high and blue low activity on the assays. Figures S2 andS3 show full dose-response profiles for PRL array recruitment and the EC50 values. For most BPXs, ER binding and dimerization by PCA and array occupancy were similar, although differences were observed between ERα and ERβ. One of the most interesting observations from this initial analysis was a clear selectivity of multiple BPXs for ERβ as compared to ERα (Figure 2B-E). Some BPXs that are notable for their difference in potency between the two ERs include BPC, BPAF, BPB and BPZ. An example of a compound with a unique profile is BPS, a BPX that has recently been proposed as a possible substitute for BPA; however, it appears to disrupt E2 nongenomic activities (Vinas and Watson, 2013). BPS was unable to induce ERα binding to PRL arrays, very little ERα dimerization, and relative to other BPXs, exhibited only low levels of activity as a ligand for ERβ.

Interestingly, the ability of BPXs to induce ER dimerization did not necessarily correlate with PRL array occupancy, indicating that dimerization was not sufficient to induce DNA binding. BPF, MH-BPF, TM BPA, and MH-MM1, were able to induce receptor dimerization in the PCA assay at doses that were not cytotoxic while being largely unable to induce DNA binding to the ERE-rich PRL arrays.

The ability of BPXs to maximally induce receptor binding to DNA and chromatin remodeling relative to estrogen treatment was next determined by treating cells with 10μM of each BPX for 30 minutes (Figure 3A). Using this measure, most BPXs are capable of inducing ER-binding to the PRL array to a degree equal to or better than E2 in the ERβ containing cells, whereas a more heterogeneous response was observed with ERα. In ERβ containing cells, for example, some BPXs (MH-MM1, TM BPA) were capable of causing array formation in ~50% of the cells while none showed higher response than E2. In ERβ containing cells, on the other hand, the maximal tested dose of many BPXs caused greater responses than E2 (BPA, BPC, BPAF, BPAP, MM2, DMDMB BPA, DMB BPA, BPZ, BPB) while only few had an intermediate response (~50% of E2; BPS, MH-MM1).

Figure 3. Analysis of efficacy and chromatin remodeling potential of BPXs.

Figure 3

Open in a new tab

Percent of cells with an array (A) or array area (B) was measured at the maximal BPX dose (10μM) after 30min of treatment of GFP-ERα:PRL-HeLa and GFP-ERβ:PRL-HeLa cells. Heat maps were generated after normalizing the data to E2, which is set as 1.

In Figure 3B we were able to observe and quantify differences in the area of the PRL arrays elicited by the different agents, a reflection of large scale chromatin remodeling. A general observation was that, even for BPXs capable of forming arrays, the changes in array area were significantly smaller than when cells were treated with E2, and larger than with 4OHT, implicating an intermediate phenotype of BPXs between classic agonists and antagonists. This was particularly evident in ERα containing cells, where the PRL arrays are much larger in size after E2 treatment (see Figure 1F). However, in most cases the same trend was observed in ERβ containing cells, even though the size of the array is much smaller. These observations highlight the likely differential usage of chromatin remodeling complexes by ERα and ERβ, and the relative importance of conformational changes that some BPXs can elicit on ERβ vs. ERα. Further, these results also demonstrate the utility of these multi-parametric assay formats for dissecting differences between closely related chemical structures.

As shown in Figure 1K, the dsRED2 reporter linked to the prolactin promoter/enhancer regulatory unit allows the PRL array system to simultaneously measure transcriptional responses using mRNA FISH. ERα or ERβ expressing PRL-HeLa cells were treated with BPXs at a maximal dose of 10μM, alone (Figure 4A), or in combination with 10nM E2 for 30 minutes (Figure 4B). Following this short E2 treatment, de novo mRNA production increased ~4-8 fold in the ERα containing cells and ~2-4 fold in the ERβ containing cells. 4OHT had very little agonist activity itself, but it largely blocked E2 activity in both cell lines, and; for this reason, it was used as the control antagonist in these experiments.

Figure 4. Transcriptional response to BPXs in PRL array cell lines.

Figure 4

Open in a new tab

A-B) dsRED2 RNA FISH intensity at the array after 30 min of BPX 10μM (A) or BPX+E2 10nM (B) in GFP-ERα:PRL-HeLa (grey bars) or GFP-ERβ:PRL-HeLa (black bars) cells. C-D) Venn diagrams representing the different categories of responses by BPXs in ERα- (C) or ERβ-containing cells (D). See text for further description of the categories.

Following data mining of ERα-containing cells, BPXs could be divided into roughly four groups based on transcriptional activity (Figure 4C): A) largely inactive (<2 fold over vehicle or less than 50% reduction of E2 activity, e.g., BPF), B) partial agonists/antagonists (>2 fold over vehicle and >50% reversal of E2, e.g., BPZ), C) agonists (>2 fold over vehicle but no reversal of E2 activity, e.g., BPA), and, D) antagonists (<2 fold over vehicle and >50% reduction of E2, e.g., 4OHT, no BPX was a complete antagonist). From the RNA FISH analysis, it is clear that, on ERα, some BPXs are largely transcriptionally inactive even at 10μM (32%); however, a bigger group showed either agonist or partial agonist/antagonist activity (68%).

ERβ-containing PRL array cells revealed a distinctly different picture, with only two classes of responses being evident (Figure 4D). The BPXs were either inactive (42%, e.g., MH-BPF) or antagonistic (58%, e.g., BPAF). None of the compounds tested in the ERβ cell model led to mRNA accumulation at the dsRED2 transcriptional reporter gene locus. A very interesting and novel observation was that most of the BPXs appeared to antagonize E2 activation of ERβ even though they were categorized as agonist or mixed agonist/antagonist in ERα-containing cells. Future studies will address the structural basis of this phenomenon and determine if it occurs also in vivo.

Effect of BPXs on ERα activity in MCF-7 breast cancer cells

To evaluate the effect of BPXs on endogenous ER protein levels, we treated MCF-7 breast cancer cells either with BPXs alone (10μM), or with BPXs plus E2 (10nM) for 24 hrs. We used immunofluorescence (IF) to quantify the nuclear level of ERα protein (Figure 5A) throughout the population of cells. As expected, following 24 hours of treatment, E2 caused a 40-50% reduction of ERβ which is comparable to previous observations by us and others (Borras et al., 1996).

Figure 5. Effect of BPXs on endogenous ERα activity in MCF-7 breast cancer cells.

Figure 5

Open in a new tab

A) MCF-7 cells were treated for 24hrs with BPXs (10μM, grey bars) or BPX+E2 10nM (black bars) and then labeled with ERα antibody. Nuclear ER intensity was then quantified. Data are represented as relative to vehicle control. B) MCF-7 cells were treated for 24hrs with BPXs (10μM, grey bars) or BPX+E2 10nM (black bars) and then hybridized with GREB1 intron mRNA FISH probes. Data indicates the average number of transcriptionally active foci per cell after setting vehicle treatment at 1.

As shown in Figure 5A, a few BPXs reduced ERα levels to a degree similar to E2 (e.g., BPB, MM2, MonoMxy) while most of them showed little activity and none showed a 4OHT-like profile. In combination studies with E2, only one of the BPXs (DMB BPA) showed partial antagonism.

In order to test the effect of BPXs on endogenous ER target gene expression, we optimized a high throughput, low magnification (20×), mRNA FISH protocol for GREB1, a well characterized and highly E2-inducible mRNA. We used a set of specific fluorescently-labeled oligonucleotides that recognize GREB1 introns to identify active transcriptional loci in the nucleus (Raj and Tyagi, 2010) (see Figure S4 for representative images). This strategy allowed us to count the transcriptional active foci in a fast, accurate and reproducible way. We treated MCF-7 cells with 10nM E2 for 24 hours, a time point that was chosen to guarantee maximal E2-mediated induction of GREB1 mRNA in MCF-7 cells thus expanding the dynamic range of response.

Quantitative mRNA FISH reveals an E2-dependent increase in GREB1 transcriptional bursting from less than a half foci per cell to over 1.5 foci per cell (Figure 5B). As there are 4 copies of the GREB1 gene in MCF-7 cells (Kocanova et al., 2010), we saw a distribution of responses to E2 ranging from cells that have all 4 copies active to cells that do not show any transcriptional bursting.

We then treated MCF-7 cells with BPXs alone or with E2 for 24 hrs and measured the average number of GREB1 transcriptional foci per cell (Figure 5B). Using this analysis, most BPXs showed no or little activity when treated alone. However, some (e.g., DM DMB BPA, BPB, BPA) showed an increase in the number of bursts to ca. 1 per cell, while a few (e.g., BPZ, BPC) were capable of almost completely abrogating the agonistic effects of E2.

As a means of comparison between our HT data with more traditional assays used to measure EDC activity, we utilized a well-established MCF-7 cell proliferation assay (Soto et al., 1995). In brief, MCF-7 cells were treated with vehicle, 1nM E2 or BPXs in a six point dose response for six days. The effect of each BPX relative to 1nM E2 was calculated using two metrics (Table 1): RPE (relative proliferative effect) which represents the ratio between the highest cell number achieved with a given BPX and E2 (×100), and RPP (relative proliferative potency), which is the ratio between the minimum E2 concentration needed for maximal cell number and the minimal BPX concentration needed to achieve the same cell density.

Table 1.

BPXs effect on MCF-7 cell proliferation

Compound RPE, %1 RPP, %2
17β-estradiol 100 100
DMDMB BPA 91.93 0.033
BPC 77.97 0.0105
DMB BPA 110.16 0.0033
BPB 92.96 0.0032
MHMM1 73.82 0.00105
BPZ 70.64 0.00105
BPA 97.76 0.00033
BPAF 91.25 0.00033
MM2 85.54 0.00033
NSC17960 80.39 0.00033
BPF 91.64 0.000316
MH-BPF 67.3 0.000316
TM BPA 30.34 0.000316
BPS 97.54 0.000105
MonoMxy BPA 77.15 0.000033
TC BPA 142.21 0.0000316
BPAP * *
TB BPA * *
Raloxifene * *
Open in a new tab
1

RPE: Relative Proliferative Effect – highest cell yield with compound/highest cell yield with 17β-estradiol.

2

RPP: Relative Proliferative Potency – min [17β-Estradiol] needed for max cell yield/min [compound] needed to achieve similar effect

*

does not reach maximal cell density at any dose

In this proliferation assay, most BPXs showed an RPE>70% (14/19) indicating their ability to induce cell proliferation similarly to E2. Interestingly, two BPXs (BPAP and TBBPA) never reached a cell number similar to E2, indicating their lack of proliferative ability. Despite the ability of most BPXs to induce MCF-7 cell proliferation, their RPP was significantly lower than E2, similar to previous studies (Rivas et al., 2002). The most active in this assay was DM DMB BPA, which had 0.033% activity compared to E2.

Clustering analysis of BPX activity across all the HTM assays

In order to generate a more comprehensive view of BPXs activity, we compiled all the data (Figures 1-5, Table 1) into a cluster dendrogram (Figure 6). We first standardized the data from each experiment, including SRC-3 recruitment to the PRL array in ERβ containing cells (no ERβ data was included for this parameter since no recruitment of SRC-3 was measured after E2 treatment, see Figures 1I and S1), by range normalization and then we clustered them using Euclidean distance. Clusters highlight the fact that E2 and SERM treatments stand alone; indicating that none of the BPXs display pure estrogen- or SERM-like activity across all the assays performed. However, BPXs were divided into two major clusters: 1) a minimally-active subset (BPF, MH-BPF, TCBPA, TBBPA and BPS) and, 2) a BPA-centric group. Each of the BPX clusters can be further subdivided into smaller branches that highlight specific functional and mechanistic differences.

Figure 6. Clustering Analysis of BPXs activity across all the HTM assays performed.

Figure 6

Open in a new tab

Data in each assay row was range normalized before clustering analysis. Clustering was performed using Euclidean distance for both the compounds and the assays.

From the hierarchal clustering dendrogram, it is interesting to note that the assays are largely divided into three branches. In the first branch, the assays designed for dimerization and transcriptional complex formation (PCA and PRL array mRNA FISH) cluster together, and also with endogenous ER nuclear levels in MCF-7 cells. The second branch contains mostly functional and mechanistic measurements (e.g., array area, coregulator binding, efficacy in array formation). Lastly, a third branch is represented by transcriptional output and MCF-7 cell proliferation. Another interesting observation from the assay dendrogram is that there is no separation between ERα and ERβ assays; however the biggest distance between the two receptors occurs at the level of RNA FISH output when E2 is combined with BPXs and with the mechanistic end points (array area and efficacy in % arrays).

We also wanted to compare our assay platforms with chemical structure similarities in terms of classifying BPXs. In order to do this we calculated the Tanimoto distance between all the BPXs, 4OHT and E2 and generated a cluster diagram based upon their relative structural similarities (Figure S5). By both hierarchical clustering and multidimensional scaling it is clear that the BPXs are different from both E2 and 4OHT, as it is also evident from the clustering analysis shown in Figure 6. From this analysis it appears that the simple chemical structural analysis does not perfectly overlap with the estrogenic activity as determined by our multiple assays. This further highlight the relevance of our multi-endpoint mechanistic platforms to better characterize EDCs activity, which could be additive to more traditional SAR analyses currently based upon available data sets.

Discussion

The number of chemicals that are synthesized and dispersed in the environment has grown exponentially during recent years. It is now evident that many of them can have serious impact on the environment and on human health. However, with the increasing synthesis of new chemical compounds, there is a need for improved high throughput platforms to query and classify their effects as broadly as possible. One example of molecules with effects on human health is endocrine disrupting compounds (EDCs). These affect mostly the endocrine system altering basic physiological functions such as development, reproduction and metabolism. They do so by interfering with many aspects of hormonal regulation, which include acting as ligands for nuclear receptors, key sensors of hormonal and environmental stimuli. The best studied EDC is bisphenol A (BPA), whose widespread use in many common products has lead to its detection in over 80-90% of individuals living in developed countries (Rochester, 2013, Rubin, 2011). BPA levels have been linked to numerous developmental and reproductive defects in animal models (Rubin, 2011). For these reasons, plastic manufacturers have developed series of BPA derivatives (BPXs) at least in part to circumvent problems with BPA; however, rigorous and appropriate testing remains to be performed. The hope for some of these compounds would be to reduce the environmental and negative health footprint. Many of the BPXs are already used in the manufacturing of polycarbonate plastics and epoxy resins, yet, in some cases, a series of analytical approaches have been attempted, but they have not provided a deep understanding of their mechanism of action (Kitamura et al., 2005). A comprehensive analysis of these compounds is highly desirable, as it will enable an understanding of structure-function relationships that may facilitate rapid identification of potential EDCs, and help drive the manufacture and commercial use of non-toxic alternatives.

In recent years, the EPA and NIEHS started specific programs for the analysis of potentially toxic compounds which include EDCs (e.g., ToxCast and Tox21). The assays utilized in these programs are very robust and allow determination of a compound as a potential EDC. While these efforts are essential, ToxCast assays do not frequently query mechanistic aspects of these compounds, which become more important when series of analogs are synthesized.

Here, we developed a series of multi-endpoint, microscopy-based, high throughput platforms and queried the effects of eighteen BPXs on estrogen receptor-α and -β. These platforms have many advantages over more classic approaches (i.e., reporter gene assays) in that they readily offer a cost-effective, multi-parametric, mechanistic view of test compounds on a cell-by-cell basis that permits classification and identification of specific differences. As these platforms are amenable to high throughput analysis of known or potentially bioactive compounds that can directly bind to NRs or compete with endogenous hormones, they will help generate a complete cartography of EDCs. From this study we obtained several important insights on BPA analogs and identified some future directions that will be needed to fully generate a complete analysis of EDC action.

One interesting observation was that most BPXs (but not all) were capable of binding one or both estrogen receptors, generally with higher affinity for ERβ, a largely understudied receptor in EDC characterization. Moreover, BPXs acted mostly as ERβ antagonists in our model system, while they acted largely as agonists or mixed agonists/antagonists on ERα. Interestingly, despite being capable of binding to the estrogen receptors, some BPXs remained inactive in some functional end point assays (e.g., gene transcription, cell proliferation), indicating the importance of querying multiple parameters for each compound in order to more precisely classify their activity. A recognized limitation of our study is the lack of a stable and reliable in vitro model system to test the effect of BPXs on endogenous ERβ, hampering our ability to extend the observations obtained with our engineered ERβ model. Specifically, high quality imaging-compatible ERβ antibodies are being developed towards this goal.

This study also highlights the need for adding additional assays to query other important nuclear receptors and transcription factors that have the potential to bind EDCs (e.g., AhR, LXR, ERRs, PPARG etc.). We are moving toward this goal by generating multiple PCA assays that cover all nuclear receptors and by adapting the PRL array system to other NRs by the use of chimeric receptors (e.g., NRs with an ER DNA binding domain swapped in). The development of such tools will allow us to broaden the spectrum of assays for characterization of EDCs, their metabolites and mixtures. When evaluating potential EDCs, it is important to keep in mind the impossibility of covering all possible mechanisms of EDC action in one (or few) assays. For instance, although our HT platforms assess numerous mechanistic steps of EDC activities, they do yet not address hormone-induced non-genomic actions, hormone synthesis, metabolism etc. (Watson et al., 2013, Marino et al., 2012). For this reason, an inactive EDC by our assays does not guarantee complete inactivity in other cell culture or animal model systems, or in the context of environmental exposure. However, the platforms utilized in this study could definitely increase the knowledge of EDC actions and are highly complementary with the assays in the ToxCast and Tox21 programs.

In conclusion, at a technical level, the use of multi-parametric, high throughput microscopy-based platforms provides a remarkably complete view of EDC activity compared to existing technologies, and can be performed in a fast, quantitative and cost effective fashion. Further, the mechanistic data arising from these studies clearly indicate the need to more thoroughly understand EDC effects upon a much wider swath of gene regulators than, as has been the case for years, primarily being concerned with effects upon ERα. Lastly, the use of such platforms on a larger scale basis should provide a much-improved determination of (for example) plasticizers that are less likely to biologically active.

Significance

Endocrine Disruptors (EDCs) are a very large class of chemical structures that impinge on human health at multiple levels (i.e., development, reproduction, metabolism). Many of these compounds act through their direct binding to nuclear receptors, which include the estrogen receptors. One of the most studied EDCs is bisphenol A (BPA), which is produced and used as plasticizer in many day-to-day products. Because of the wealth of data on its impact on human health and the environment, many analogs (BPXs) have been synthesized as potential substitutes. These compounds and their effects have not been well studied. Here we developed a series of multi parametric, microscopy-based assays to generate a fine map of BPXs activity on estrogen receptor alpha and beta. We identified several important new characteristics of BPXs: most are capable of binding to either receptor, few are largely inactive and many act as antagonists on ERβ, a previously unappreciated mechanism of action. In conclusion, our study highlights the importance for a thorough analysis and classification of putative EDCs across contextual assay platforms to directly assess and classify the activity of chemicals on nuclear receptors.

Experimental Procedures

Cell lines, cell culture conditions and materials

The GFP-ERβ:PRL-HeLa cell line was created by stably expressing full length GFPERβ in PRL-HeLa parental cells (Sharp et al., 2006) using a blasticidin resistance gene. The cells are grown in phenol-red free DMEM containing 10% fetal bovine serum (FBS, Gemini Bioproducts), L-glutamine, sodium pyruvate, 0.8μg/ml blasticidin, 200μg/ml hygromycin and 10nM Raloxifene (Sigma). The GFP-ERα:PRL-HeLa cells are grown in the same media with 10nM 4OHT substituted for Raloxifene. Cells were grown in treatment media (phenol red-free DMEM containing 5% stripped/dialyzed FBS, L-glutamine, and sodium pyruvate, and no drugs) for at least 48 hours prior to treatment. MCF-7 cells were grown in phenol-red free DMEM containing 10% fetal bovine serum (FBS, Gemini Bioproducts), L-glutamine, and sodium pyruvate. Cells were robotically seeded using a Titertek dispenser in 384 well glass bottom plates (Greiner Bio-one) pre-coated with 5% charcoal stripped and dialyzed FBS for a minimum of 4 hours.

BPXs were obtained from Sigma or Santa Cruz Biotechnology and solubilized in DMSO. Estrogen, Raloxifene and 4-hydroxy-Tamoxifen (Sigma) were solubilized in ethanol.

Protein Fragment Complementation (PCA) assays

The ERα/α LBD and ERβ/β LBD stable cell lines were generated by cotransfecting fusion constructs encoding the LBD of ERα (amino acids 310-547) or ERβ (amino acids 263-489) fused to fragment 1 or fragment 2 of the Venus variant of YFP into HEK93T cells as previously described (MacDonald et al., 2006, Yu et al., 2003). The cells are grown in phenol-red free DMEM containing 10% fetal bovine serum (FBS, Gemini Bioproducts), L-glutamine, sodium pyruvate, 200μg/ml Zeocin (Life Technologies). Cells were seeded into 384-well poly-D-lysine coated plates (Perkin Elmer) in treatment media (phenol red-free DMEM containing 5% charcoal stripped and dialyzed FBS, L-glutamine, and sodium pyruvate) for 24-48 hours prior to drug treatment using a MultiDrop cell dispenser. Cells were treated with agent (in final DMSO concentration of 0.5%) in triplicate wells for 8 hours, fixed with 4% formaldehyde and stained with Draq5 to identify cells. Fluorescence data in the 488nm (YFP) and 635nm (Draq5) channels were collected on the Acumen eX3 laser scanning cytometer (TTP Labtech). For each well, the average PCA fluorescence intensity for all cells is normalized by the total nuclear area (normalized total intensity). Each data point for each experiment represents the average of three wells. LogEC50 values for each experiment were derived by normalizing each data point to the max E2 value on each plate (100nM). Data presented in all figures represent the average logEC50 value from at least three independent experiments, and all error bars represent the standard error of the mean across the independent experiments.

Cell proliferation assay

MCF-7 cells were maintained as described above. 48hr prior to drug treatment, cells were seeded into 384-well poly-D-lysine coated plates in treatment media. Cells were treated with vehicle, E2 or drug for 6 days as described for the E-Screen assay developed to identify estrogenic compounds (Soto et al., 1995) and the number of viable cells was assessed using Alamar Blue staining (Life Technologies). RPP (relative proliferative potency) and RPE (relative proliferative effect) of each test compound were calculated as described in (Soto et al., 1995).

RNA fluorescence in situ hybridization (FISH)

RNA FISH experiments were completed as described in (Bolt et al., 2013). Cells were fixed in 4% purified formaldehyde (Electron Microscopy Sciences) in RNase-free phosphate-buffered saline for 15 min and then permeabilized with 70% ethanol in RNase-free water at 4°C for 1 h. Cells were washed in 1 ml of wash buffer (2× SSC (Ambion) plus 10% formamide) followed by four hours of 37°C hybridization with mRNA FISH probes (dsRED2 or GREB1 Stellaris™ probes, Biosearch Technologies Inc.) in buffer (1 g of dextran sulfate, 1 ml of 20× SSC buffer, 1 ml of formamide and 8 ml of nuclease-free water) followed by one wash in wash buffer for 30 min at 37°C, then followed by DNA staining with DAPI for 10 min at 37°C. Finally, cells were washed in 2× SCC buffer, and imaged.

Immunofluorescence

Immunofluorescence experiments were completed as described in (Bolt et al., 2013). Briefly, cells were fixed in 4% formaldehyde in PEM buffer (80 mM potassium PIPES, pH 6.8, 5 mM EGTA and 2 mM MgCl2), quenched with 0.1 M ammonium chloride for 10 min and permeabilized with 0.5% Triton X-100 for 30 minutes. Cells were incubated at room temperature in blotto (5% milk in 1× TBS–Tween 20) for 1 h, and then specific antibodies were added overnight at 4°C prior to 30 min of secondary antibody (Alexa conjugates, Molecular Probes) and DAPI staining (1μg/ml for ~1 min). The primary antibodies used were: mouse anti-Ser5-phospho RNA polymerase II (Abcam, ab5401), mouse anti-SRC-1 (BD Transduction Labs # 612378), mouse anti-SRC-2 (BD Transduction Labs # 610985), mouse anti-SRC-3 (BD Transduction Labs # 611105), and rabbit anti-ERα (Millipore 04-820).

High throughput microscopy and image analysis

Automated imaging was carried out using the IC-200 high throughput microscope (Vala Sciences). Dual step high speed (50-100ms) reflection- and image-based autofocused image fields were acquired with a sCMOS 5.5MPixel camera through a Nikon S Fluor 40×/0.90 NA objective. Z-stacks were imaged at 1 μm intervals at 1 × 1 binning. Nuclear array segmentation and signal quantification were performed using PipelinePilot image analysis software (Accelrys) as previously described (Ashcroft et al., 2011, Bolt et al., 2013). Aggregated cells, mitotic cells and apoptotic cells were removed using filters based on nuclear size, nuclear shape and nuclear intensity.

Statistical Data analysis

Every experiment was standardized based on estrogenic activity (E=U-Umin/Umax-Umin). Heatmaps were generated using the Matrix2png program; clustering was performed using Cluster 3.0 and visualized with Java Treeview; Tanimoto distance and MDS plots were calculated using ChemMine tools (Backman et al., 2011, de Hoon et al., 2004, Saldanha, 2004, Pavlidis and Noble, 2003). GraphPad Prism 5.0 was used for making graphs, dose-response curves, calculate EC50, performing t-tests, etc. Every experiment was repeated between three and six independent times. In the 384 well plate experiments, every condition was tested in quadruplicate.

Supplementary Material

01
02

Highlights.

  • Bisphenol A analogs (BPXs) show preference in binding ERβ over ERα

  • Some BPXs have very low activity despite binding to ERs

  • BPXs act mostly as antagonists on ERβ

  • Microscopy-based high throughput multiparametric BPX classification

Acknowledgments

We gratefully acknowledge Z.D. Sharp for aid in the development of the PRL-HeLa cell line, Hans Johansson and Arturo Orjalo (Biosearch Technologies) for help with mRNA FISH, and I. Mikic, T.J. Moran, A.T. Szafran, and J.Y. Newberg for help developing automated high content analysis tools used in this study. We acknowledge imaging and automation resource support from the John S. Dunn Gulf Coast for Chemical Genomics (P.J. Davies and M.A.M.); NIEHS Grand Opportunity Grant 1RC2ES018789-01 (M.A.M., C.L.W., S.-M. Ho), NIEHS R01 (1R01ES023206-01; M.A.M., C.W.L., B.W. O'Malley and M.T. Bedford), NIEHS P30 (ES023512-01; Center of Excellence in Environmental Health), and the Diana Helis Henry Medical Research Foundation (M.A.M) through its direct engagement in the continuous active conduct of medical research in conjunction with Baylor College of Medicine and the Cancer Program. This project was supported by the Integrated Microscopy Core at Baylor College of Medicine, with funding from the National Institutes of Health (grants HD007495, DK56338, and CA125123) and the Dan L. Duncan Cancer Center.

M.A.M. has and continues to do beta-testing/consulting for Biosearch Technologies, a vendor that was used to acquire mRNA FISH probes used in the some of the assays in the paper. J.K.W., J.E.L., and J.S.M. previously worked at Odyssey Thera, where they performed some of the assays in the manuscript. In late 2013, the company dissolved. M.A.M. also previously consulted for Odyssey Thera, ending in June 2013.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Ashcroft FJ, Newberg JY, Jones ED, Mikic I, Mancini MA. High content imaging-based assay to classify estrogen receptor-alpha ligands based on defined mechanistic outcomes. Gene. 2011;477:42–52. doi: 10.1016/j.gene.2011.01.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Backman TW, Cao Y, Girke T. ChemMine tools: an online service for analyzing and clustering small molecules. Nucleic Acids Res. 2011;39:W486–91. doi: 10.1093/nar/gkr320. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bolt MJ, Stossi F, Newberg JY, Orjalo A, Johansson HE, Mancini MA. Coactivators enable glucocorticoid receptor recruitment to fine-tune estrogen receptor transcriptional responses. Nucleic Acids Res. 2013;41:4036–48. doi: 10.1093/nar/gkt100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Borras M, Laios I, El Khissiin A, Seo HS, Lempereur F, Legros N, Leclercq G. Estrogenic and antiestrogenic regulation of the half-life of covalently labeled estrogen receptor in MCF-7 breast cancer cells. J Steroid Biochem Mol Biol. 1996;57:203–13. doi: 10.1016/0960-0760(95)00272-3. [DOI] [PubMed] [Google Scholar]
  5. Brotons JA, Olea-Serrano MF, Villalobos M, Pedraza V, Olea N. Xenoestrogens released from lacquer coatings in food cans. Environ Health Perspect. 1995;103:608–12. doi: 10.1289/ehp.95103608. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Calafat AM, Kuklenyik Z, Reidy JA, Caudill SP, Ekong J, Needham LL. Urinary concentrations of bisphenol A and 4-nonylphenol in a human reference population. Environ Health Perspect. 2005;113:391–5. doi: 10.1289/ehp.7534. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Calafat AM, Ye X, Wong LY, Reidy JA, Needham LL. Exposure of the U.S. population to bisphenol A and 4-tertiary-octylphenol: 2003-2004. Environ Health Perspect. 2008;116:39–44. doi: 10.1289/ehp.10753. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Committee NR. A unified nomenclature system for the nuclear receptor superfamily. Cell. 1999;97:161–3. doi: 10.1016/s0092-8674(00)80726-6. [DOI] [PubMed] [Google Scholar]
  9. De Coster S, Van Larebeke N. Endocrine-disrupting chemicals: associated disorders and mechanisms of action. J Environ Public Health. 2012;2012:713696. doi: 10.1155/2012/713696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. De Hoon MJ, Imoto S, Nolan J, Miyano S. Open source clustering software. Bioinformatics. 2004;20:1453–4. doi: 10.1093/bioinformatics/bth078. [DOI] [PubMed] [Google Scholar]
  11. Jirtle RL, Skinner MK. Environmental epigenomics and disease susceptibility. Nat Rev Genet. 2007;8:253–62. doi: 10.1038/nrg2045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Kitamura S, Suzuki T, Sanoh S, Kohta R, Jinno N, Sugihara K, Yoshihara S, Fujimoto N, Watanabe H, Ohta S. Comparative study of the endocrine-disrupting activity of bisphenol A and 19 related compounds. Toxicol Sci. 2005;84:249–59. doi: 10.1093/toxsci/kfi074. [DOI] [PubMed] [Google Scholar]
  13. Kocanova S, Kerr EA, Rafique S, Boyle S, Katz E, Caze-Subra S, Bickmore WA, Bystricky K. Activation of estrogen-responsive genes does not require their nuclear co-localization. PLoS Genet. 2010;6:e1000922. doi: 10.1371/journal.pgen.1000922. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Kuiper GG, Carlsson B, Grandien K, Enmark E, Haggblad J, Nilsson S, Gustafsson JA. Comparison of the ligand binding specificity and transcript tissue distribution of estrogen receptors alpha and beta. Endocrinology. 1997;138:863–70. doi: 10.1210/endo.138.3.4979. [DOI] [PubMed] [Google Scholar]
  15. Macdonald ML, Lamerdin J, Owens S, Keon BH, Bilter GK, Shang Z, Huang Z, Yu H, Dias J, Minami T, Michnick SW, Westwick JK. Identifying off-target effects and hidden phenotypes of drugs in human cells. Nat Chem Biol. 2006;2:329–37. doi: 10.1038/nchembio790. [DOI] [PubMed] [Google Scholar]
  16. Marino M, Pellegrini M, La Rosa P, Acconcia F. Susceptibility of estrogen receptor rapid responses to xenoestrogens: Physiological outcomes. Steroids. 2012;77:910–7. doi: 10.1016/j.steroids.2012.02.019. [DOI] [PubMed] [Google Scholar]
  17. Michnick SW, Macdonald ML, Westwick JK. Chemical genetic strategies to delineate MAP kinase signaling pathways using protein-fragment complementation assays (PCA). Methods. 2006;40:287–93. doi: 10.1016/j.ymeth.2006.07.016. [DOI] [PubMed] [Google Scholar]
  18. Monroe DG, Johnsen SA, Subramaniam M, Getz BJ, Khosla S, Riggs BL, Spelsberg TC. Mutual antagonism of estrogen receptors alpha and beta and their preferred interactions with steroid receptor coactivators in human osteoblastic cell lines. J Endocrinol. 2003;176:349–57. doi: 10.1677/joe.0.1760349. [DOI] [PubMed] [Google Scholar]
  19. Myers JP, Zoeller RT, Vom Saal FS. A clash of old and new scientific concepts in toxicity, with important implications for public health. Environ Health Perspect. 2009;117:1652–5. doi: 10.1289/ehp.0900887. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Olea N, Pulgar R, Perez P, Olea-Serrano F, Rivas A, Novillo-Fertrell A, Pedraza V, Soto AM, Sonnenschein C. Estrogenicity of resin-based composites and sealants used in dentistry. Environ Health Perspect. 1996;104:298–305. doi: 10.1289/ehp.96104298. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Pavlidis P, Noble WS. Matrix2png: a utility for visualizing matrix data. Bioinformatics. 2003;19:295–6. doi: 10.1093/bioinformatics/19.2.295. [DOI] [PubMed] [Google Scholar]
  22. Raj A, Tyagi S. Detection of individual endogenous RNA transcripts in situ using multiple singly labeled probes. Methods Enzymol. 2010;472:365–86. doi: 10.1016/S0076-6879(10)72004-8. [DOI] [PubMed] [Google Scholar]
  23. Rivas A, Lacroix M, Olea-Serrano F, Laios I, Leclercq G, Olea N. Estrogenic effect of a series of bisphenol analogues on gene and protein expression in MCF-7 breast cancer cells. J Steroid Biochem Mol Biol. 2002;82:45–53. doi: 10.1016/s0960-0760(02)00146-2. [DOI] [PubMed] [Google Scholar]
  24. Rochester JR. Bisphenol A and human health: A review of the literature. Reprod Toxicol. 2013;42C:132–155. doi: 10.1016/j.reprotox.2013.08.008. [DOI] [PubMed] [Google Scholar]
  25. Rotroff DM, Dix DJ, Houck KA, Knudsen TB, Martin MT, Mclaurin KW, Reif DM, Crofton KM, Singh AV, Xia M, Huang R, Judson RS. Using in vitro high throughput screening assays to identify potential endocrine-disrupting chemicals. Environ Health Perspect. 2013;121:7–14. doi: 10.1289/ehp.1205065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Rubin BS. Bisphenol A: an endocrine disruptor with widespread exposure and multiple effects. J Steroid Biochem Mol Biol. 2011;127:27–34. doi: 10.1016/j.jsbmb.2011.05.002. [DOI] [PubMed] [Google Scholar]
  27. Saldanha AJ. Java Treeview--extensible visualization of microarray data. Bioinformatics. 2004;20:3246–8. doi: 10.1093/bioinformatics/bth349. [DOI] [PubMed] [Google Scholar]
  28. Sharp ZD, Mancini MG, Hinojos CA, Dai F, Berno V, Szafran AT, Smith KP, Lele TP, Ingber DE, Mancini MA. Estrogen-receptor-alpha exchange and chromatin dynamics are ligand- and domain-dependent. J Cell Sci. 2006;119:4101–16. doi: 10.1242/jcs.03161. [DOI] [PubMed] [Google Scholar]
  29. Soto AM, Sonnenschein C, Chung KL, Fernandez MF, Olea N, Serrano FO. The E-SCREEN assay as a tool to identify estrogens: an update on estrogenic environmental pollutants. Environ Health Perspect. 1995;103(Suppl 7):113–22. doi: 10.1289/ehp.95103s7113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Susiarjo M, Sasson I, Mesaros C, Bartolomei MS. Bisphenol a exposure disrupts genomic imprinting in the mouse. PLoS Genet. 2013;9:e1003401. doi: 10.1371/journal.pgen.1003401. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Vinas R, Watson CS. Bisphenol S disrupts estradiol-induced nongenomic signaling in a rat pituitary cell line: effects on cell functions. Environ Health Perspect. 2013;121:352–8. doi: 10.1289/ehp.1205826. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Watson CS, Hu G, Paulucci-Holthauzen AA. Rapid actions of xenoestrogens disrupt normal estrogenic signaling. Steroids. 2014;81:36–42. doi: 10.1016/j.steroids.2013.11.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Weiss B. The intersection of neurotoxicology and endocrine disruption. Neurotoxicology. 2012;33:1410–9. doi: 10.1016/j.neuro.2012.05.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Welshons WV, Thayer KA, Judy BM, Taylor JA, Curran EM, Vom Saal FS. Large effects from small exposures. I. Mechanisms for endocrine- disrupting chemicals with estrogenic activity. Environ Health Perspect. 2003;111:994–1006. doi: 10.1289/ehp.5494. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Wong CW, Komm B, Cheskis BJ. Structure-function evaluation of ER alpha and beta interplay with SRC family coactivators. ER selective ligands. Biochemistry. 2001;40:6756–65. doi: 10.1021/bi010379h. [DOI] [PubMed] [Google Scholar]
  36. Yu H, West M, Keon BH, Bilter GK, Owens S, Lamerdin J, Westwick JK. Measuring drug action in the cellular context using protein-fragment complementation assays. Assay Drug Dev Technol. 2003;1:811–22. doi: 10.1089/154065803772613444. [DOI] [PubMed] [Google Scholar]
  37. Zoeller RT, Brown TR, Doan LL, Gore AC, Skakkebaek NE, Soto AM, Woodruff TJ, Vom Saal FS. Endocrine-disrupting chemicals and public health protection: a statement of principles from The Endocrine Society. Endocrinology. 2012;153:4097–110. doi: 10.1210/en.2012-1422. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01
NIHMS600704-supplement-01.pdf (498.2KB, pdf)
02
NIHMS600704-supplement-02.xlsx (13.1KB, xlsx)

RESOURCES