A Mechanistic High Content Analysis Assay using Chimeric Androgen Receptor that Rapidly Characterizes Androgenic Chemicals

Abstract

Human health is at risk due to environmental exposures to a wide range of chemical toxicants and endocrine disrupting chemicals (EDCs). As part of understanding this risk, the US EPA has been pursuing new high throughput in vitro assays and computational models to characterize EDCs. EPA models have incorporated our high content analysis (HCA)-based GFP-ER:PRL-HeLa assay that allows direct visualization of estrogen receptor binding to DNA regulatory elements. Here, we characterize a modified functional assay based upon the stable expression of a chimeric androgen receptor (ARER), wherein a region containing the native AR DNA binding domain (DBD) was replaced with the ERα DBD (amino acids 183–254). We demonstrate that the AR agonist dihydrotestosterone (DHT) induces GFP-ARER nuclear translocation, PRL promoter binding, and transcriptional activity at physiologically relevant concentrations (<1 nM). In contrast, the AR antagonist bicalutamide induces only nuclear translocation of the GFP-ARER receptor (at μM concentrations). Estradiol also fails to induce visible chromatin binding, indicating androgen specificity. In a screen of reference chemicals from the EPA and the ATSDR, the GFP-ARER cell model identified and mechanistically grouped activity by known (anti-)androgens based on the ability to induce nuclear translocation and/or chromatin binding. Finally, the cell model was used to identify potential (anti-)androgens in environmental samples in collaboration with the Houston Ship Channel/Galveston Bay TAMU EPA Superfund Research Program. Based on this data, the chromatin-binding, in vitro assay-based GFP-ARER model represents a selective tool for rapidly identifying androgenic activity associated with drugs, chemicals, and environmental samples.

Introduction

The androgen receptor (AR) is a nuclear receptor (NR), a family of 48 transcription factors in humans that regulate gene expression in response to circulating ligands. In response to endogenous androgens, the AR regulates a gene network important in multiple physiological responses, including those that result in the differentiation and maintenance of the male sexual phenotype. In common with other nuclear receptors, AR has a modular domain structure comprising of a relatively large N-terminal activation function (AF-1) domain, a DNA-binding domain (DBD), and a C-terminal ligand binding domain (LBD) that also harbors a ligand-dependent transcriptional activation function (AF-2). In the absence of ligand, AR usually resides in the cytoplasm in complex with chaperone proteins. Androgen binding results in a conformational change of AR, chaperone shedding, and translocation into the nucleus. Once in the nucleus, the receptor interacts with DNA androgen response elements (AREs) and protein coactivators to initiate gene transcription.¹ This is a carefully regulated process and disruptions can result in proliferative, reproductive, and metabolic disorders, including birth defects, hormone-dependent cancers and infertility.^1,2

Endocrine disrupting chemicals (EDCs) are exogenous chemicals that interfere with normal hormone action.^2,3 Although impact upon estrogen receptor signaling has received the most focus, EDCs have been shown to likely impact human health through multiple nuclear receptors including the estrogen-related receptor (ERR), pregnane X receptor (PXR), thyroid hormone receptors (TR), and AR.^2-4 Societal concerns about the impact of EDC exposures is reflected in governmental regulations and international guidelines that support research seeking to understand potential exposure, the effects of exposure, and how to minimize future exposure risks.^2,5-9 These regulations and guidelines often described a tiered approach to understanding EDC activity using a combination of in vitro and in vivo based assays.^10,11 More recently, there has been an increased focus on deploying high throughput in vitro screening approaches to better capture the diverse landscape of chemicals currently in use and allow the development of computational models to better predict in vivo EDC activity.^12,13

A considerable number of chemicals components, including pesticides, flame retardants, packaging chemicals, and industrial chemicals have been identified as androgenic or anti-androgenic.^14-16 Although in vivo assays (rat pubertal, Hershberger assay) remain the gold standard for determining androgenic activity, multiple in vitro cell models for detecting effects on AR transcriptional activity have been developed.^3,5,10,11 These reporter systems generally consist of either rodent (CHO) or human (22RV1, HeLa, MDA-MB-453, PC3, U2OS) cell lines engineered to express human AR (hAR) and containing an MMTV/ARE-luciferase reporter. These cell models have been used extensively to characterize EDC activity, however, they are vulnerable to poor selectively due to endogenous progesterone (PR), mineralocorticoid (MR) , and glucocorticoid receptor (GR) expression, nuclear receptors known to bind and activate the MMTV/ARE-luciferase reporter gene.^3,11

To overcome potential limitations association with reporter cell models, we have used a novel cell model that has been developed based upon direct visualization of NR-DNA binding, coregulator accumulation, and transcriptional activity. The estrogen responsive biosensor cell model (GFP-ERα:PRL-HeLa) is based on a multicopy integration of prolactin promoter/enhancer elements, GFP fusion technology, and automated imaging and analysis.^17,18 This cell model has been used to characterize mechanisms affecting ERα and ERβ activity, including ligand specificity, receptor mobility on promoter elements, functional importance of receptor domains, and regulation by ubiquitin ligase.^18-21 The systems-biology nature of the cell model has also allowed it to be used as a tool to characterize EDC potential of bisphenol analogs and BPA alternatives; it is important to note that the actions of the BPA alternatives, which were almost all active, were quantified and classified by the robust mechanistic/phenotypic HCA data that are derived from the PRL-Hela models.^22,23

Recently, the modular nature of NRs that allows the generation of functional chimeras has been exploited to allow the use of the PRL-HeLa cell model to understand protein kinase regulation of the progesterone receptor (PR).^24-28 The GFP-PRER:PRL-HeLa cell model, in which the PR DBD has been swapped for the ERα DBD, demonstrated ligand specificity for progestins and anti-progestins and allowed the direct visualization of protein kinase effects on PR nuclear translocation, DNA binding, coactivator recruitment, chromatin remodeling, and transcriptional activity.²⁸ Here, we report the generation of a stable PRL-HeLa cell model that expresses a functional chimeric AR (ARER) based upon standard in vitro transcription assays possessing the ERα DBD inserted in place of the native AR DBD. We demonstrate the new cell model possesses quantifiable mechanistic steps associated with androgen-induced transcriptional activity and altered mechanisms associated androgenic/anti-androgenic exposures. The speed, sensitivity and multiple mechanistic readouts of the new model were tested for ligand specificity via the use of a focused library of EDC reference chemicals. Evaluation of this approach was performed in the context of real-life environmental samples collected in conjunction with a US EPA Superfund Research Program involving water samples collected from Galveston Bay and the Houston Ship Channel. Based on our observations, the GFP-ARER:PRL-HeLa cell model allows the rapid detection of androgenic/anti-androgenic EDC activity with potentially distinct mechanistic phenotypes.

Materials and Methods

Materials

Cell culture tetracycline-free fetal bovine serum (FBS) was obtained from Gemini Bio-products (Sacramento, CA); trypsin-EDTA was obtained from Sigma (St. Louis, MO); cell culture media, G418, hygromycin, and L-Glutamine was obtained from Corning Cellgro (Corning, NY); and blasticidin was obtained from Invitrogen (Carlsbad, CA). The dihydrotestosterone (DHT), bicalutamide, and 17β-estradiol (E2) were obtained from Sigma. A set of 43 estrogenic EDC reference chemicals with well documented in vivo activity were provided in DMSO by agreement from Dr. Keith Houck (US EPA, Research Triangle Park, NC). ATSDR reference chemicals and 16 complex mixtures were provided in DMSO by the lab of Dr. Ivan Rusyn (Texas A&M, College Station, TX) as part of the Superfund Research Program. Further information about chemicals used can be found in Supplemental Table 1. Information about complex mixtures provided upon request.

As part of a Superfund Research Program collaboration (PI, Ivan Rusyn), environmental samples (sediments) were collected by the lab of Dr. Anthony Knap (Texas A&M, College Station, TX) and extracted by the Rusyn lab using previously established protocols.^29,30 Specifically, 1 gram of freeze-dried sample was added to a 15 mL conical-bottom disposable plastic tube (Corning, Vernon Hills, IL) and mixed with 2 mL of cyclohexane (HPLC grade, Fisher Scientific, Waltham, MA) and 2 mL of DMSO (≥99%, Santa Cruz Biotechnology, Santa Cruz, CA) pre-equilibrated for 24 hours with cyclohexane at 10:1 ratio. Tubes were vortexed for 1 minute and centrifuged for 5 min at 4,700 rpm. Next, 1 mL of DMSO layer was removed into a clean glass vial (Lab Products, Inc. Houston, TX). Additional 2 mL of DMSO was added to the tube with the sample and vortexing/centrifugation steps were repeated. The DMSO layer (2 mL) was removed and combined with the DMSO fraction from the first step of extraction. Samples were labeled by core (1, 2, 3, etc.); where multiple samples were extracted from a single core corresponding to differing sediment depths, the sample label contains an additional alpha code (a, b, c, etc.).

Cell Model Generation

A pSG5-flag backbone vector with human AR N-terminal domain (1-E540), ERα DNA-binding domain (S178-K252), and AR ligand binding domain (G627-919) were used to create and functionally test the chimeric receptor in standard transcription assays in vitro (data not shown). DNA binding to EREs was confirmed in electrophoretic mobility shift assays (data not shown). The ARER coding region was subcloned using XmaI and BamHI sites into an enhanced green fluorescent protein (eGFP) expression vector, pEGFP-C1-AR 108, that contains a truncated version of full-length AR, placing the GFP tag on the N-terminus of the chimeric protein. Next, using HindIII & BamHI sites, the ARER coding region was cloned into pENTR-EGFP-C1, a vector we prepared from pENTR (Invitrogen) that contains the EGFP and C1 reading frame MCS from pEGFP-C1. Using the recombinase method (LR Clonase II, Invitrogen), the EGFP-C1 ARER coding region was cloned into pINDUCER20, containing a tetracycline-inducible promoter, a gift from Dr. Trey Westbrook (Baylor College of Medicine, Houston, TX). High titer Lentivirus was packaged by System Biosciences LLC (Palo Alto, CA) despite eGFP-ARER exceeding the recommended packaging size. The resulting lentivirus particles were used to transduce the PRL-HeLa cell line. PRL-HeLa cells stably expressing GFP-ARER were enriched using geneticin (G418) drug selection, flow cytometry, and single-cell cloning using limiting dilution. The final population of cells is >95% GFP positive following doxycycline induction. The GFP-ARER:PRL-HeLa cell line is maintained in phenol red-free Dulbecco’s modified Eagle’s medium (DMEM-HG) supplemented with 5% FBS, 200 μg/ml hygromycin, and 400 μg/ml G418.

Cell Culture

GFP-ER:PRL-HeLa cells were maintained in phenol red-free DMEM containing 5% FBS, 0.8 μg/ml blasticidin, 200 μg/ml hygromycin, and 10 nM 4-hydroxytamoxifen as previously described.^14,18,19 GFP-ARER and GFP-ER PRL-HeLa cells were seeded in DMEM with 5% charcoal-stripped/dialyzed FBS without selection agents or inhibitors on 96-well (Greiner SensoplatePlus) or 384-well plates (384-IQ, Aurora Biotechnologies) at a target density of 8,000 cells/well (96) or 3,000 cells/well (384). Cells were allowed to adhere overnight and then the GFP-ARER cells were treated with doxycycline (0.8 μg/ml) for 18-20 hours before media was exchanged prior to treatment.

C4-2 and 22RV1 cells are genotype verified cell lines obtained from the American Type Culture Collection (ATCC) and were maintained according to recommended conditions in DMEM/F12 media with 10% FBS. Cells were seeded on multi-well plates in phenol red-free DMEM with 5% SD-FBS 24 hours prior to use.

Treatment with Chemicals, Mixtures, and Environmental Samples

The EPA and ATSDR reference chemicals and mixtures were provided in pre-formatted multi-well plates in DMSO at a known concentration. In vitro screening in GFP-ARER:PRL-HeLa cells was performed at concentrations ranging from 10 nM to 10 μM (EPA), 0.1 nM to 10 μM (ATSDR), and 200-fold to 200,000-fold dilution (mixtures, samples) and a treatment time of 2 hours. For EPA chemicals and environmental samples, stock chemical solutions were directly transferred to assay plates using an acoustic transfer device (Labcyte Echo 550). ATSDR chemicals and mixtures, immediately prior to compound treatment, the stock compound solutions were transferred to 96-well working dilution plates in phenol red-free DMEM with 5% SD-FBS. Working dilutions of compounds were then added in quadruplicate to assay plates containing prepared cells. For all experiments, negative control wells containing 0.5% DMSO and positive control wells containing DHT ranging from 0.01 nM to 1000 nM were included in the plate layout.

Immunofluorescence Labeling

After incubation period was completed, cells were fixed using 4% EM-grade formaldehyde in PEM buffer (80 mM potassium PIPES [pH 6.8], 5 mM EGTA, and 2 mM MgCl2) and quenched with 0.1 M ammonium chloride for 10 min. Cell membranes were disrupted by incubating samples with a 0.5% Triton X-100 solution for 10 minutes. Nuclei were stained using DAPI (1 μg/ml) for 10 minutes. For samples in which antibody labeling was used, incubation with 0.5% Triton-X was extended to 30 minutes and cells were incubated in blotto (5% milk in Tris-buffered saline/Tween 20) for 30 minutes prior to addition of primary antibody solution (AR 441, 1 μg/ml, gift from Dr. Dean Edwards; SRC-1, 0.25 μg/ml, BD Transduction Labs; SRC-3, 1 μg/ml, gift from Dr. Bert O’Malley; BRG1, 0.9 μg/ml, Bethyl; Ser5-phospho RNA polymerase II, 0.5 μg/mL, Abcam). Primary antibody solution was incubated overnight at 4C prior to incubation with secondary antibody (Alexa 647 conjugated Anti-Ms IgG, Molecular Probes) for 1 hour at room temperature. Nuclei were stained using DAPI (1 μg/ml) for 10 minutes. All samples were stored in PBS buffer containing calcium, magnesium, and sodium azide at 4C prior to imaging.

Fluorescence in situ hybridization

GFP-ARER:PRL-HeLa were fixed in 4% formaldehyde in ribonuclease-free phosphate-buffered saline for 30 minutes, washed in PBS, and then permeabilized with 70% ethanol in ribonuclease-free water for a minimum of 1 hour at 4°C. Cells were washed in 1 mL of wash buffer A (LGC Biosearch Technologies, Novato, CA) containing 10% v/v formamide followed by hybridization with custom made dsRED2 RNA probes^18,19,24 (Stellaris® probes; LGC Biosearch Technologies) diluted 1:500 (50 nM) in hybridization buffer (1 g dextran sulfate, 1 mL 20X saline sodium citrate (SSC) buffer and 1 mL formamide in 8 mL of nuclease-free water) overnight at 37C. After hybridization, cells were washed twice (15 min at 37C) with wash buffer A then with wash buffer B (LGC Biosearch Technologies) for 30 minutes at 37C and then stained with DAPI for 10 minutes at 37C. After DAPI labeling the cells were washed x1 with Dulbecco’s PBS, then stored and imaged in Dulbecco’s PBS + 0.02% Sodium azide.

Imaging and Quantitation

Automated imaging was carried out using either a Vala Sciences IC-200 (San Diego, CA) or a Molecular Devices ImageXpress Micro (Downingtown, PA) image cytometer. Image acquisition was performed with a 20x/0.75NA objective and a scientific CMOS camera. Z-stacks were imaged at 1 μm intervals (for a total of 5 μm ) at 1 × 1 binning. Nuclear array segmentation and automated image analyses were performed using custom measurement and reporting routines incorporated into myImageAnalysis/Pipeline Pilot software.³¹ Cell, nucleus, and array (if present) were identified and metrics describing shape and intensity features quantified. Aggregated cells, mitotic cells, and apoptotic cells were removed using filters based on nuclear size, nuclear shape, and nuclear intensity. Degree of nuclear translocation of the GFP signal was determined by measuring the ratio of nuclear mean intensity to cytoplasmic mean intensity. Loading on the PRL-array was determined by calculating the ratio of array mean intensity to nuclear mean intensity. GFP expression per cell was determined by integrating the total GFP pixel intensity per cell.

Data Analysis

The chemical responses were analyzed using a standardized analysis pipeline generated using Pipeline Pilot (Biovia). This pipeline performed all necessary baseline correction, normalization, curve-fitting, and hit-calling. All screening data were normalized to a fractional range with DMSO and 100 nM DHT treated controls set at 0 and 1. All dose response data was fit to either a constant model or a hill model using least squares curve fitting. The activity observed for each chemical on measured endpoints is summarized by integrating the response over the tested concentration range (Area Under Curve, AUC). AUC values are normalized to DHT and labeled as significant if greater than 0.1. This threshold was selected based on EPA computational models.²⁶

Statistical Analysis

Data presented were acquired from a minimum of 3 independent experiments. Statistical significance was determined using one-way ANOVA with Tukey’s post hoc test for comparisons across multiple samples. Tests were performed using Minitab 19 software, and p < 0.05 was considered statistically significant. All graphs were generated using SigmaPlot 10 software.

Results

Regulation of AR activity is a complex process involving expression, protein interactions, and subcellular compartmentation of the receptor. Direct analysis of AR transcriptional activity using reporter assays is routinely complicated by bulk cell analysis and the coexpression of other nuclear receptors that may generate nonspecific results; furthermore, single end-point assays may not capture global effects on receptor activity.^11,26 To address these issues, we mirrored our previous ER studies to establish a androgenic cell model capable of measuring multiple mechanistic steps of AR signaling by high throughput microscopy and high content image analysis.²⁸ As described previously, the PRL-HeLa parental cell line contains a multi-copy (~100 copies) integration of a cassette derived from the prolactin promoter and multimerized synergy element (repeats containing a Pit-1 and ERE binding sites; see Methods) linked to a dsRED2 fluorescent protein reporter.¹⁷ When GFP-ERα (or ERβ) is expressed, agonist treatment results in receptor binding at the ERE-rich locus (‘promoter array’) forming a visible intranuclear GFP spot. Binding at the promoter array is associated with coregulator recruitment (visualized using immunofluorescence) and de novo transcription of the dsRED2 reporter (detected using mRNA FISH).^18,19,22,23 The versatility of the systems is well-exemplified by the ability to utilize chimeric receptors in single cell assays within the context of additional endocrine models, as described in our progesterone receptor/estrogen receptor chimera model.²⁸

To enable the study of AR in the PRL-HeLa cell line, a chimeric AR was generated by swapping the DBD sequence of AR with the sequence containing the DBD from ERα (e.g., ARER) (Figure 1A). The chimeric protein also includes an GFP protein fusion attached to the N-terminus (GFP-ARER). GFP-ARER was next cloned into a vector that included a tetracycline-inducible promoter. After stable transfection into the PRL-HeLa cell line, single cell cloning was used to generate the final GFP-ARER:PRL-HeLa cell line that allows for culture maintenance in the absence of AR expression. Without doxycycline-induction, GFP-ARER expression is negligible, and background levels of dsRED2 mRNA are observed (Supplemental Figure 1). Doxycycline induction generates ARER expression, elevated relative to AR expression in prostate cancer cell lines (Supplemental Figure 2). Treatment for 2 hours with the androgen 5α-dihydrotestosterone (DHT) results in GFP-ARER nuclear translocation, visible nuclear PRL array spot associated with chromatin binding, stabilized GFP-ARER expression, and a 4.7-fold increase in dsRED2 reporter mRNA levels over no ligand conditions, indicating an active transcriptional response (Figure 1B, Supplemental Figure 2, Supplemental Video 1).

Figure 1.

Generation and characterization of stable cell lines expressing inducible chimeric AR. (A) Schematic illustrating the generation of the chimeric AR by swapping a region in AR containing the DNA binding domain (amino acids 541–626) for a region of ER containing the ER DNA binding domain (amino acids 178–252). Diagram numbers indicate NR domains. Slanted lines indicate region of AR N-terminal domain not shown. (B) A stable cell line expressing inducible GFP-ARER was treated with 200 ng/ mL of doxycycline for 24 hours, followed by treatment with vehicle or 100nM DHT for 120 minutes. Cells were imaged at 20X/0.75 magnification. Representative contrast enhanced GFP image is shown with computer generated nuclear segmentation in red. Scale bar indicates 10 μM. (C) GFP-ARER nuclear translocation (N/C Ratio), percentage of cells with arrays (% Array), GFP expression (Expression), and dsRED2skl mRNA signal (FISH Intensity) were quantified from >500 cells per sample at the indicated ligand concentrations. Results normalized to untreated control samples. Graph represents the mean ± SE from 3 independent samples.

To further explore GFP-ARER ligand dependent responses, GFP-ARER:PRL-HeLa cells were treated with increasing concentrations of DHT, the anti-androgen bicalutamide (i.e., Casodex™), or 17β-estradiol (E2), and analyzed by high content analysis (Figure 1B). DHT treatment results in a dose dependent nuclear translocation, visible array/chromatin binding, stabilization of receptor expression, and dsRED2 mRNA synthesis. These responses occurred with EC₅₀ values ranging between 0.15 and 0.24 nM (Table 1), concentrations consistent with known wild type AR sensitivity to DHT and previously observed results using GFP-AR.^32-34 In contrast, bicalutamide induced only nuclear translocation of GFP-ARER at an EC₅₀ concentration greater than 200 nM (Figure 1C, Supplemental Video 2). E2, the obligate ligand for the GFP-ER, induced nuclear translocation of GFP-ARER with an EC₅₀ value of 2.2 nM, only partial stabilization at concentrations exceeding 3 μM, and failed to induce visible chromatin binding (Figure 1C). The ability of E2 to induce nuclear translocation of AR has been shown previously.³² These responses are consistent with the modular nature of nuclear receptors and the ability of chimeric receptors to retain molecular functions associated with the constituent AR and ER domains.

Table 1.

Measured EC₅₀ values of observed responses

Response	DHT (nM)	Bicalutamide (nM)	E2 (nM)
GFP-AR/ER:PRL-HeLa
Nuclear Translocation	0.23 ± 0.02	230 ± 2	2.2 ± 0.2
Chromatin Binding	0.21 ± 0.01	n.d.	n.d.
AR/ER Stablization	0.15 ± 0.01	n.d.	3100 ± 100*
Transcriptional Activity	0.24 ± 0.03	n.d.	n.d.
GFP-ER:PRL-HeLa**
Chromatin Binding	1600 ± 40	n.t.	0.22 ± 0.05
GFP-AR:HeLa31
Nuclear Translocation	11 ± 2	600 ± 10	n.t.
Transcriptional Activity	44 ± 6	n.d.	n.t.

^*Estimated from incomplete dose response curve.

^**Data from screen of EPA reference compounds, manuscript in preparation.

n.d. – EC₅₀ could not be determined due to a lack of response.

n.t. – Chemical was not tested in the indicated cell model.

To understand the molecular mechanisms associated with GFP-ARER mediated transcriptional activity, we used ‘visual ChIP’ to measure coregulator protein recruitment to the integrated PRL array in the presence of DHT. Specifically, we used immunofluorescence to measure recruitment of the p160 coactivators SRC-1 and SRC-3, the chromatin remodeler BRG1, and activated RNA Pol II (serine 5 phosphorylated, S5P) in the GFP-ARER cell model compared to the GFP-ER cell model (Figure 2). We have previously used relative loading, defined as average array signal divided by average nuclear signal, to quantify coregulator recruitment in PRL-HeLa cell lines.^18,22,23,28 In the presence of DHT, GFP-ARER cells recruited all proteins to the PRL array, however, the relative loading (array signal intensity relative to nuclear signal intensity) of the p160 coactivators SRC-1 and SRC-3 was significantly lower compared to E2 induced recruitment in GFP-ER cells despite similar recruitment of BRG1 and S5P (Figure 2A). Further investigation found that this is likely a function of GFP-ARER cells having higher nuclear signal intensity of SRC-1 and SRC-3 compared to GFP-ER cells (Figure 2B). This indicates the GFP-ARER chimera can recruit SRC-1 and SRC-3 to the PRL array in a similar manner as GFP-ER, consistent with known AR interactions with the p160 coactivators and components of the transcriptional machinery.³⁵

Figure 2.

Quantification of agonist induced coregulator and transcriptional machinery recruitment to the PRL-array. GFP-ER:PRL-HeLa and GFP-ARER:PRL-HeLa cells were prepared and treated with either 10 nM E2 (GFP-ER) or 10 nM DHT (GFP-ARER) for 120 minutes. Cells were processed for immunofluorescence with antibodies against SRC-1, SRC-2, BRG1, and phosphorylated RNA Pol II (serine 5). (A) The ratio of array to nuclear intensity (Relative Loading) was determined. Dashed line reference line indicates result if no relative loading was observed. (B) Nuclear intensity was also quantified and normalized to GFP-ER. Results indicate mean ± SE from 3 independent samples shown with >500 cells per sample measured. (*) p < 0.05.

Assessment of GFP-ARER:PRL-HeLa detection of androgenic bioactivity relied upon chemical libraries defined by EPA or the Agency for Toxic Substances and Disease Registry (ATSDR). The EPA reference chemicals consists of 43 chemicals with defined in vivo activity; the ATSDR reference chemicals consists of 42 pesticides, herbicides, heavy metals, or otherwise hazardous contaminants found in the environment. ATSDR individual chemicals were also combined into defined complex mixtures to mimic the complexity of environmental pollution. Leaning further towards environmental situations, we sought to also explore responses of the GFP-ARER model to endocrine disruption-relevant bisphenol A derivatives (BPXs) that have previously been exhaustively-characterized by HCA using the GFP-ER:PRL-HeLa cell model.²² Further information about the chemicals and mixtures used can be found in Supplemental Table 1.

Within the EPA reference set of chemicals, we detected nineteen that induced nuclear translocation of GFP-ARER (Figure 3). Among the strongest responders were known androgens (5alpha-DHT, 17-Methyltestosterone), anti-androgens (hydroxyflutamide, promycidone), and strong agonists for other nuclear receptors (estradiols, estrone, progesterone, spironolactone, corticosterone). However, only 6 of the 19 chemicals were able to induce a visible spot indicating chromatin binding by GFP-ARER. In addition to the known androgens, these included spironolactone, estrone, 17alpha-estradiol, and 17beta-estradiol, compounds with reported weak androgenic activity.^36,37 However, the AUC of chromatin binding of these other chemicals was lower than that for DHT and 17-methyltestosterone indicating that induction of chromatin binding required higher concentrations.

Figure 3.

GFP-ARER as biosensor to query EPA reference chemicals. GFP-ARER:PRL-HeLa cells were seeded onto 384-well plates and induced with 200 ng/ml doxycycline for 24 hours prior to treatment with 10 nM to 10 μM of DHT or 43 reference chemicals for 2 hours. Calculated AUC scores for (A) nuclear translocation or (B) chromatin binding are listed along the x-axis. All scores above 0.1 were used to define a chemical as active (solid markers). Results represents mean SE of 3 samples. Dashed gray reference line indicates the 0.1 threshold of activity. Solid gray reference line indicates DHT reference value.

The performance of the supplemented ATSDR chemicals and mixtures further confirmed the selective nature of GFP-ARER chromatin binding (Supplemental Figure 3). Of the 53 individual chemicals, only bisphenol C (BPC) induced a strong nuclear translocation response (65% of DHT response). Other chemicals (bisphenol F, tetramethyl-bisphenol A) and mixtures (AC50 mx, CAT APP) were able to induce a weak nuclear translocation response (10-12% of DHT response). None of the individual chemicals or mixtures were able to induce chromatin binding, suggestive of minimal androgenic activity in the supplemented ATSDR reference set of individual chemicals or complex mixtures.

To understand the relationships between the observed responses, we compiled nuclear translocation, chromatin binding, and receptor stabilization metric AUC data from active chemicals/mixtures. All data is standardized to a range based on vehicle and DHT controls and clustered using a k-means algorithm (Figure 4). Cluster 1 contains chemicals that induced a high AUC score for nuclear translocation, chromatin binding, and receptor stabilization. It contains 5alpha-dihydrotesterone and 17-methyltestosterone, the only two strong/moderate AR agonist in the chemicals tested (Supplemental Table 1).

Figure 4.

Unsupervised clustering of GFP-ARER endpoints of active chemicals. Normalized AUC scores from active chemicals identified in either the EPA or ATSDR reference sets were clustered using a K-means based algorithm. All inactive scores (< 0.1) were set to zero to improve clustering performance and visualization. Both the chemicals and endpoints were clustered.

Cluster 2 contains 3 chemicals that induced a low/moderate AUC score for all three metrics. Spironolactone has demonstrated mixed in vitro and in vivo (anti-)androgenic activity in previous studies.^36,37 Estrone has previously demonstrated in vitro androgenic activity.³⁷ Literature review classifies 17alpha-estradiol as inactive, however, an EPA computational model predicting androgenic activity based on multiple in vitro assays identified this chemical as a moderate/weak androgen.³⁶ This suggests cluster 2 compounds are those with mixed/weak androgenic activity in in vitro assays.

Cluster 3 features chemicals that predominantly have a moderate to weak nuclear translocation AUC score. Those chemicals with stronger AUC scores include known anti-androgens (bisphenol C, procymidone, hydroxyflutamide, 17alpha-ethynilestradiol).^36-38 E2 (17beta-estradiol) also demonstrates a moderate AUC score despite being generally considered inactive with regard to (anti-)androgenic activity.^36,37 Chemicals with weak translocation AUC scores include several that have demonstrated anti-androgenic activity (progesterone, corticosterone, benzyl butyl phthalate), but largely consist of chemicals with unknown (anti-)androgenic potential. A summary of chemical information, previously reported activity, and cluster assignment can be found in Supplemental Table 1.

As a proof-of-concept demonstration of how the GFP-ARER:PRL-HeLa cell model may be used as an androgen biosensor, we screened a panel of 44 environmental samples collected at 24 unique sites of a major shipping channel/industrialized area in Texas as part of an EPA-funded Houston Shipping Channel/Galveston Bay TAMU Superfund Research Program. Samples were prepared and concentrated from sediment samples using a cyclohexane-DMSO extraction method (see Methods) and tested at 1:200 to 1:200,000 dilutions. Cells were treated for 2 hours and activity was compared to measured AUC response from DHT at concentrations ranging between 0.01 nM to 10 nM. K-means cluster analysis of results identified three samples with moderate/weak induction of GFP-ARER nuclear translocation, chromatin binding, and receptor stabilization (Figure 5A-B). These samples demonstrated 9-33% of DHT activity. An additional 13 samples induced nuclear translocation only achieving 10-29% of the DHT response. Active samples were dispersed throughout the sampled region (Figure 5C), however, further analysis will need to be done using orthogonal approaches to verify activity. Importantly, total processing time of samples (cell treatment, imaging, and data analysis) was less than 12 hours, indicating that the cell model may be applied to rapidly measure potential EDC activity and identify areas of concern.

Figure 5.

GFP-ARER used to detect potential (anti-)androgenic activity in sediment samples. GFP-ARER:PRL-HeLa cells were seeded onto 384-well plates and induced with 200 ng/ml doxycycline for 24 hours prior to treatment. Cells were treated with 1:200 to 1:200,000 dilutions of cyclohexane-DMSO extracted samples collected from 24 sites around the Houston, TX shipping channel. Cells were then processed for fluorescence and imaged using the Molecular Devices ImageXpress at 20X magnification. (A) AUC scores for GFP-ARER endpoints were range normalized to DHT and visualized using K-means clustering. All inactive scores (< 0.1) were set to zero to improve clustering performance and visualization. (B) Representative contrast enhanced GFP images of active samples shown with computer generated nuclear segmentation shown in red. (C) Geo-mapping of inactive (blue), partially active (red), and active (green) samples generated using geographic coordinates recorded at sample collection and Google Maps. Numbers correspond to sites with detected activity.

Discussion

We have taken the established image-based high throughput GFP-ER:PRL-HeLa assay for estrogenic activity and adapted it to detect (anti-)androgenic activity by exploiting the modular nature of nuclear receptors. The GFP-ARER:PRL-HeLa cell model allows the direct visualization of multiple mechanistic steps of AR genomic signaling at the single cell level. By swapping the AR DBD for that of ER, the GFP-ARER based model allowed for the single cell quantification of (anti-)androgen-induced nuclear translocation, chromatin binding at an ERE rich locus, coregulator recruitment, and transcriptional activity. We were also able to observe androgen-induced increases in total GFP signal per cell (AR stabilization), consistent with established agonist induced decrease in degradation rate of AR.³⁹ Although a fraction of the total increase GFP intensity may be due to consolidation of a diffuse cytoplasmic signal into the nuclear compartment as nuclear translocation occurs, we did not see similar increases in total GFP signal with the bicalutamide antagonist which also induces nuclear translocation.

These are similar endpoints as used in the GFP-ER:PRL-HeLa model and recapitulate those measured using multiple biochemical and molecular methods, generally at the population level. Importantly, the observed responses occur rapidly with nuclear translocation and chromatin binding occurring in less than 60 min and complete response to DHT observed within 120 minutes (Supplemental Video 1 and 2). Interestingly, previous studies using either ER or chimeric PRER have shown changes in regulated transcriptional activity is routinely detected at timepoints less than 60 min.^18,28 This allows for analysis of factors directly affecting AR signaling with a low potential for secondary effects due to prolonged treatment times (i.e., altered coregulator protein expression, cell cycle synchronization, cytotoxicity).⁴⁰

One key difference observed between the GFP-ARER cell model and the GFP-ER cell model is the ability of antagonist to induce visible receptor chromatin binding. In the GFP-ER model, selective estrogen modulators (SERMs) and pure ER antagonists such as Fulvestrant (ICI 182 780) induce small visible arrays, indicative of chromatin binding with minimal chromatin remodeling.^13,14,18,19 In contrast, AR antagonist bicalutamide and hydroxyflutamide induced GFP-ARER nuclear translocation but undetectable evidence of chromatin binding at the ERE rich PRL locus. Antagonist induced nuclear translocation of endogenous wild type AR in prostate cancer cells is well established.⁴¹ In addition, genome-wide ChIP-seq approaches have confirmed that both antagonist-bound ER and AR can bind some, but not all chromatin binding sites associated with agonist induced transcriptional regulation.^42,43 This suggest that the ERE-rich PRL array resembles a promoter region recognized by both agonist- and antagonist-bound receptor in the GFP-ER cell model, whereas the PRL array resembles a region recognized thus far only by androgen-bound receptor in the GFP-ARER cell model. Interestingly in the GFP-PRER:PRL-HeLa cell model, the PR antagonists RU486 induces visible chromatin binding whereas ZK98299 does not, suggesting that further screening may yet identify AR anti-androgens that result in stable chromatin binding by GFP-ARER at the PRL-array.²⁸

The ligand-dependent array binding of AR highlights how the GFP-ARER:PRL-HeLa cell model may represent a beneficial compromise between rapid transcriptional reporter assays and slower androgen dependent growth assays. While reporter assays are fundamental for the high throughput analysis of potential EDC activity, they measure AR activity within the context of a single promoter (either endogenous or synthetic) driving reporter gene expression. There is yet no defined minimal set of AR-responsive genes known to faithfully predict in vivo (anti-)androgenic EDC activity and these assays are vulnerable to non-specific results due to co-expression of other nuclear receptors.¹¹ The GFP-ARER:PRL-HeLa cell model is also reliant upon the single PRL promoter construct integrated into the PRL-HeLa cell line as a promoter array, however, the direct visualization of GFP-ARER chromatin binding (and transcriptional activity via mRNA FISH) minimizes the risk of non-specific results. Further, the multiplex nature of the assay also allows the quantification of global effects on AR signaling such as nuclear translocation, subnuclear organization and expression levels. This is similar to our earlier efforts using GFP-AR expression in HeLa cells and others have advanced image-based AR bioassays to efficiently and inexpensively monitor androgenic activity from multiple types of samples, including urine samples in the detection of anabolic steroid abuse.^32,40,44,45

Although the GFP-ARER:PRL-HeLa cell model has advantages over traditional methods for detecting androgenic activity, challenges remain. Reliable detection of chromatin binding by GFP-ARER is dependent on the quality of sample preparation, requires high resolution automated imaging platforms, and, as with most high content assays, is dependent on custom image analysis and data analysis algorithms. Although the pattern of nuclear translocation with negligible specific chromatin binding suggests anti-androgen-like activity, additional experiments are required to more directly detect inhibition of DHT-induced responses. To avoid further complexities of downstream effects from ligand exposure and fit the assay within a high throughput context, short term experiments were the logical focus, which naturally precluded observations of responses requiring more time. Cell proliferation assays for the detection of estrogenic EDC activity are well established, simple, and sensitive. Because cell proliferation assays typically last for days/weeks, they potentially allow the detection of active metabolites or other important secondary changes that may not be detected in a HT assay focusing on early (~1-2hr) mechanistic steps in receptor activation. However, relatively few androgen-dependent cell proliferation assays have been developed and tend to rely upon cell lines that express AR’s harboring point mutations known to alter ligand specificity.⁴⁵

Relatively little is known about the potential (anti-)androgenic activity of the many thousands of man-made chemicals in the environment. There is growing concern that a number pesticides, flame retardants, packaging, and industrial chemicals may harbor AR bioactivity that could potentially impact human health.^14,16,36,37 Through our participation in the Texas A&M Superfund Research Program, we were able to successfully detect androgenic activity in sediment samples collected from the Houston Ship Channel and Galveston Bay. The sediment samples provide a depth-dependent historical record of environmental contamination of this highly urbanized region containing markedly-busy seaports and numerous permitted industrial sites located throughout the watershed. In this analysis, we did not detect androgenic activity in samples from the Galveston Bay, sampling sites furthest from the watershed. This suggests that higher concentrations of androgenic chemicals are rapidly dispersed with distance. Identifying the androgenic chemicals and their potential sources is an ongoing project within the Superfund Research Program. In addition, we are in the process of developing an anti-androgenic protocol using the GFP-ARER cell model to improve the ability us to distinguish between weak androgens and anti-androgens.

It is clear that adoption of high throughput in vitro screening assays have enabled the development of computational models capable of predicting in vivo activity of endocrine disruptors.^36,46 A recent androgenic model relies on 11 different assays spanning three assay types (hormone binding, coactivator protein complementation, transcriptional reporter) in order to achieve 95.2 to 97.5% predictive accuracy.³⁶ The GFP-ARER:PRL-HeLa cell model rapidly measures three additional mechanistic steps of AR signaling (nuclear translocation, specific chromatin binding, receptor stabilization) in a single assay that would complement those already being used in computational models. Continued studies are warranted to further explore assay performance and classification models using additional EPA-validated reference chemicals, and additional real-world samples (e.g., water samples from flooded areas). A meritorious aspect of expanding the use of the PRL-HeLa model with chimeric nuclear receptors is the fact that functional post-translational modification sites are retained, thus allowing further exploration of activation/inhibition of cell signaling pathways that could alter AR signaling in cancer research. In conclusion, we have developed the GFP-ARER:PRL-HeLa cell model that allows for the rapid quantification of multiple mechanistic aspects of AR signaling. This adds to our library of PRL-HeLa cell models based upon ERα, ERβ, and PRER expression, expanding the versatility of the assay platform, and enhancing the ability to identify molecules and cell signaling pathways that alter nuclear receptor function, and potentially protecting and improving human health.