Improving Estrogenic Compound Screening Efficiency by Using Self-Modulating, Continuously Bioluminescent Human Cell Bioreporters Expressing a Synthetic Luciferase

Abstract

A synthetic bacterial luciferase-based autobioluminescent bioreporter, HEK293_ERE/Gal4-Lux, was developed in a human embryonic kidney (HEK293) cell line for the surveillance of chemicals displaying endocrine disrupting activity. Unlike alternative luminescent reporters, this bioreporter generates bioluminescence autonomously without requiring an external light-activating chemical substrate or cellular destruction. The bioreporter’s performance was validated against a library of 76 agonistic and antagonistic estrogenic endocrine disruptor chemicals and demonstrated reproducible half maximal effective concentration (EC₅₀) values meeting the U.S. Environmental Protection Agency (EPA) guidelines for Tier 1 endocrine disrupting chemical screening assays. For model compounds, such as the estrogen receptor (ER) agonist 17β-estradiol, HEK293_ERE/Gal4-Lux demonstrated an EC₅₀ value (7.9 × 10⁻¹² M) comparable to that of the current EPA-approved HeLa-9903 firefly luciferase-based estrogen receptor transcription assay (4.6 × 10⁻¹² M). Screening against an expanded array of common ER agonists likewise produced similar relative effect potencies as compared with existing assays. The self-initiated autobioluminescent signal of the bioreporter permitted facile monitoring of the effects of endocrine disrupting chemicals, which decreased the cost and hands-on time required to perform these assays. These characteristics make the HEK293_ERE/Gal4-Lux bioreporter potentially suitable as a high-throughput human cell-based assay for screening estrogenic activity.

A variety of chemicals, ranging from naturally occurring compounds like phytoestrogens to synthetics such as pesticides, personal care products, nutritional supplements, and plastics, have been classified as endocrine disrupting chemicals (EDCs) because of their ability to interfere with the proper balance and regulation of normal endocrine function (Bergman et al., 2012). Human exposure to EDCs can result in reproductive, developmental, and metabolic disorders, as well as breast, ovarian, testicular, and prostate cancers, early onset of puberty, genital tract abnormalities, reduced sperm counts, and various other health-related problems (Vogel, 2004). Environmental exposures, which predominantly affect wildlife (Kabir et al., 2015), can similarly result in the feminization or masculinization of sex organs, changes in sexual behavior, eggshell thinning, and immune system malfunction (Lyons, 2006).

These deleterious effects result in part from EDC interference with the normal functioning of the sex hormone estrogen (Cooper and Kavlock, 1997). Estrogen, produced in the ovaries in females and testes in males, is responsible for regulating reproductive health in humans. It also serves as an important regulatory molecule in the neuroendocrine, skeletal, adipogenic, and cardiovascular systems (Keyaerts et al., 2012). Estrogen signaling is achieved via binding of the compound to one of two nuclear hormone receptors (NRs), estrogen receptor α (ERα), or estrogen receptor β (ERβ), which leads to the formation and phosphorylation of receptor homo- or hetero-dimers that then relocate to the nucleus (Keyaerts et al., 2012; McDevitt et al., 2008). Following nuclear relocation, the DNA-binding domains of these activated estrogen receptors (ERs) bind to their target genes at conserved estrogen response element (ERE) sequences and regulate their expression (Park et al., 2011). ERs are expressed in a multitude of tissues. ERα is the predominant subtype in the uterus, prostate stroma, ovarian theca cells, testicular Leydig cells, the epididymis, the liver, and the breasts (Lane, 2008). ERβ is the predominant subtype in the testes, ovarian granulosa cells, prostate epithelium, bone marrow, and the brain (Weiser et al., 2008). Due to the prevalence of these ERs throughout the body, disruption of the estrogenic signaling pathway can have diverse and far-reaching health consequences.

Because of these potential health and environmental impacts, the U.S. Environmental Protection Agency (EPA) has established an aggressive program, EDSP21 (Endocrine Disruptor Screening Program for the 21st Century), designed to evaluate these compounds and how they affect the health and well-being of animal and human endocrine systems. EDSP21 encompasses a two-tiered approach. In Tier 1, in vitro assays are used to identify chemicals that have the potential to interact with the endocrine system. In Tier 2, those compounds that test positive are re-screened using in vivo assays to define their endocrine-related effects and obtain dosage-relevant information. There are currently over 87,000 chemicals awaiting screening under this program (Vogel, 2004). However, the present battery of Tier 1 assays are not well-suited to address this backlog because they use nonhuman cell lines that can obscure bioavailability data (Environmental Protection Agency, 2011a,c), require the use of radioactive materials that necessitate dedicated use areas and specially trained personnel (Environmental Protection Agency, 2011a,c,e), or rely on expensive analytical equipment (Environmental Protection Agency, 2011e,f). It is therefore imperative that more efficient, cost-effective, and higher-throughput assays be developed as second generation Tier 1 screens if the EDSP21 program is to be successful.

In this study, we describe the development of a human-cell-based bioreporter that can autonomously generate bioluminescence when exposed to estrogenic endocrine disruptor chemicals (EEDCs). A human embryonic kidney (HEK293) cell line was endowed with a synthetic version of the bacterial luciferase operon (lux) under the control of an ERE-based gene amplification circuit, which allows the cells to autonomously modulate bioluminescent signal generation without requiring cellular destruction or interfering with cellular metabolism (Xu et al., 2014). This approach reduces the number of requisite assay preparation steps and minimizes performance costs (Class et al., 2015). The functionality of this autobioluminescent estrogen-responsive bioreporter was validated against a suite of known estrogenic and potentially estrogenic compounds as outlined by the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM, 2003) and assay performance, efficacy, and reproducibility were validated experimentally and through comparisons with the published results of established Tier 1 screening assays.

MATERIALS AND METHODS

Cell types and culture conditions

All cell types were grown at 37°C in a humidified chamber under a 5% CO₂ atmosphere.

HEK293 cells (ATCC) were cultured in phenol red-free Dulbecco’s modified Eagle’s medium (DMEM/high modified) (Hyclone, GE Lifesciences) supplemented with 10% dextran-coated-charcoal-treated fetal bovine serum (DCC-FBS; Atlanta Biologicals) and 1% penicillin/streptomycin (Corning). The selective agents Zeocin (Thermo Fisher Scientific) and G418 (Calbiochem) were added at 50 μg/ml and 400 μg/ml, respectively, as noted for clonal selection of cell lines. Concentrations were reduced to 25 μg/ml and 100 μg/ml, respectively, for routine maintenance following selection. During compound testing, Zeocin and G418 were omitted from the growth medium.

The ER transcriptional activation assay was performed using the human cervical cancer hERα-HeLa-9903 cell line (Japanese Collection of Research Bioresources [JCRB] cell bank) maintained in phenol red-free DMEM supplemented with 10% DCC-FBS and 60 μg kanamycin/ml (Sigma).

Human adrenocortical carcinoma H295R cells (ATCC) were propagated in ATCC-formulated DMEM-F12 medium (Gibco) supplemented with 1× ITS+Premix (Corning; 6.25 μg/ml insulin, 6.25 μg/ml transferrin, 5.35 μg/ml linoleic acid, 1.25 mg/ml bovine serum albumin, 6.25 ng/ml selenium), and 2.5% Nu-Serum I (Corning). H295R cells were grown for five passages before they were used for steroidogenesis assays and/or frozen.

The MCF-7 human breast cancer cell line (ATCC) was cultured in phenol red-free DMEM supplemented with 10% DCC-FBS and 1% penicillin/streptomycin.

Chemicals

Reporter cells were assayed against the 76 chemical library recommended by the ICCVAM for the validation of in vitro ER test methods (ICCVAM, 2003), or subsets thereof. These chemicals and their sources are listed in Supplementary Table 1.

HEK293_ERE/Gal4-Lux and MCF7_ERE/Gal4-Lux molecular assembly

Human ERα cDNA was synthesized (GenScript) and cloned into the pcDNA3.1/Zeo vector (Life Technologies). The resulting vector, pcDNA3.1/Zeo/ERα, was transfected into HEK293 cells using Viafect transfection reagent (Promega) and stable transfectants were selected following growth on 50 μg Zeocin/ml. ERα expression was assessed for each clonal isolate using the PathScan Total Estrogen Receptor α Sandwich ELISA kit (Cell Signaling Technology) and the clone displaying the highest level of ERα expression, designated as HEK293_ERα, was selected for further development.

In order to achieve ERα-mediated activation of autobioluminescent production, a two-module signal amplification circuit was used to regulate the expression of the synthetic lux cassette (Figure 1). The first module used three tandem ERE repeats and a TATA minimal promoter to control the expression of a Gal4ff fusion gene consisting of the Gal4 DNA-binding domain and two transcriptional motifs of the herpes simplex virus VP16 transcription factor. The second module harbored the human-optimized lux cassette (Xu et al., 2014) under the regulation of five tandem repeats of the yeast upstream activating sequence (UAS) and a minimal promoter. This two-module circuit, designated as ERE-Gal4FF/UAS-Lux, was transfected into the HEK293_ERα cell line and selected with 400 μg G418/ml. Clonal isolates were grown in a 12-well tissue culture plate and each isolate was treated with 1 × 10⁻⁸ M 17β-estradiol (Sigma-Aldrich) dissolved in ethanol, or with only 0.1% ethanol as a vehicle control. Following 24 h of chemical exposure, all the cells from a single well of a 12-well plate were harvested in 200 μl of appropriate culture medium and then plated in a single well of a 96-well plate. Autobioluminescent outputs from each isolate were monitored at 60 min intervals for 24 h using a CLARIOstar plate reader (BMG Labtech) with an integration time of 60 s/well at 37°C under a 5% CO₂ atmosphere. At the end of this procedure, the clonal lineage demonstrating the highest up-regulation of bioluminescent production by 17β-estradiol was established as the HEK293_ERE/Gal4-Lux biosensor.

Figure 1.

Schematic representation of the genetic circuitry enabling self-initiated autobioluminescent EDC signaling. In this design, activation of the estrogen receptor by an estrogenic compound (ER/ligand complex) stimulates the 3×ERE/TATA promoter to initiate transcription of the Gal4FF gene. The Gal4FF activator then acts as an amplifier by binding to the downstream 5×UAS/TATA promoter, which initiates transcription of the human optimized lux cassette to generate continuous light output.

Biological triplicates of naturally ER-positive MCF7 cells were prepared for transient transfection by plating ∼3 × 10⁵ cells into individual wells of a 6-well plate 24 h before transfection. Following ERE-Gal4FF/UAS-Lux transfection, cells were recovered at 37°C and 5% CO₂ for 24 h, then dosed with test chemicals. After 24 h of exposure, cells were harvested and each biological replicate was transferred to technical triplicates in 100 μl volumes of assay medium in individual wells of a 96-well plate for autobioluminescence measurement.

HEK293_ERE/Gal4-Lux autobioluminescent assay

Approximately 2 × 10⁴ HEK293_ERE/Gal4-Lux cells/well were plated in white-bottom 96-well tissue culture plates (Corning) in 100 μl volumes of cell culture medium. The cells were allowed to settle and attach to the plate surface for 3 h after seeding, and were then treated with serial dilutions of test chemicals. Each dilution and vehicle control were tested in triplicate. The cells were incubated with the test compounds for 24 h at 37°C under a 5% CO₂ atmosphere before being transferred to a CLARIOstar plate reader (BMG) and assayed for bioluminescent production using a 60s/well integration time. EC₅₀ (concentration required to induce a half maximum effect) values and coefficients of variation (CV) were generated using GraphPad Prism 7. Relative effect potency (REP) was calculated as EC_{50(17β-estradiol)}/EC_50(compound). Z-factor was calculated according to Zhang et al. (1999).

HeLa-9903 estrogen receptor transcription assay

hERα-HeLa-9903 cells were plated at 1 × 10⁴ cells/100 μl of culture medium into each well of a white-bottom 96-well tissue culture plate. The cells were allowed to settle and attach to the plate surface for 3 h after seeding before being treated with serial dilutions of 17β-estradiol, 17α-estradiol, 17α-methyltestosterone, and corticosterone. An additional vehicle control treatment consisting of a 1:1000 dilution of solvent (DMSO or ethanol) in 100 μl culture medium was also applied. Each dilution and vehicle control were tested in triplicate. Cells were incubated with the chemicals for 24 h before harvesting. The firefly luciferase-based Steady-Glo Luciferase Assay system (Promega) was used to measure bioluminescence output according to the manufacturer’s guidelines. Briefly, for each well, 150 μl of medium was removed and 50 μl of Steady-Glo reagent mixture was added to the cells, mixed well, and incubated for 10 min at room temperature in the dark before luminescence was measured using a CLARIOstar plate reader with a 2s/well integration time.

H295R cell-based steroidogenesis assay

The H295R cell-based steroidogenesis assay was performed to evaluate if test compounds affected 17β-estradiol production. H295R cells were seeded in 24-well tissue culture plates at a density of 2 × 10⁵ cells/well in 1 ml volumes of medium. Cells were allowed to attach to the plate surface for 24 h prior to the addition of test chemicals. Triplicate replicates of 17α-estradiol, 17α-methyltestosterone, and corticosterone were tested at seven dilutions ranging from 10⁻¹² to 10⁻⁶ M, 10⁻¹¹ to 10⁻⁵ M, and 10⁻¹⁰ to 10⁻⁴ M, respectively, along with a 0.1% DMSO vehicle control. The known inducer forskolin and inhibitor prochloraz were included in triplicate as positive and negative controls, respectively. Cells were incubated with the test chemicals for 48 h at 37°C under a 5% CO₂ atmosphere. Following incubation, the medium was removed from each well and assayed for 17β-estradiol concentration using the 17β-estradiol ELISA kit (Abcam).

RESULTS

Development of the Estrogen-Responsive HEK293_ERE/Gal4-Lux Bioreporter

Transfection of the pcDNA3.1/Zeo/ERα vector into wild-type HEK293 cells generated a stable HEK293_ERα clone that displayed increased ER expression levels relative to both the negative control wild-type HEK293 cells and positive control MCF-7 breast cancer cells (Supplementary Figure 1). A clonal lineage of these isolates was subsequently transfected with the ERE-Gal4FF/UAS-Lux sequence to form a synthetic amplification circuit that placed transcription of the human optimized lux cassette (Xu et al., 2014) under the control of an exogenous yeast UAS rather than an endogenous EDC-responsive promoter (Figure 1). This allowed the Gal4FF transcriptional activator (Asakawa et al., 2008), to serve as a transcriptional amplifier following binding of an EDC to its upstream EREs without promoting leaky expression due to endogenous cellular processes. Individual clones expressing this full circuit (n = 6) were selected and tested for their autobioluminescent response to treatment with 1 × 10⁻⁸ M 17β-estradiol. The clone displaying the highest fold of induction compared with vehicle control was designated as HEK293_ERE/Gal4-Lux and used for all further experimentation. To validate HEK293_ERE/Gal4-Lux functionality and performance, multiple replicate assays were performed against the known strong inducer chemical, 17β-estradiol, using individual batches of biosensor cells prepared by different personnel. Each assay was run under identical conditions and the results were compared. The average intra-assay CV was determined to be 9.8 (±0.3)% (n = 30) and inter-assay CV was determined to be 15.7% (n = 63). The Z-factor of the assay was calculated to be 0.32 (±0.17).

HEK293_ERE/Gal4-Lux Validation Against the HeLa-9903 Estrogen Transcription Activation Assay

The HeLa-9903 ER transcription assay was selected as a comparative validation because both reporter systems utilize ER-mediated transcriptional luciferase activation to report EDC detection. However, because the HeLa-9903 ER transcription assay requires external chemical stimulation concurrent with sample destruction to initiate its luminescent signal, the two assays were run slightly differently. Each cohort of reporter cells was incubated with serial dilutions of the strong agonist 17β-estradiol (10⁻⁶ to 10⁻¹³ M), the weak agonist 17α-estradiol (10⁻⁵ to 10⁻¹² M), the very weak agonist 17α-methyltestosterone (10⁻⁴ to 10⁻¹⁰ M), and the negative control corticosterone (10⁻⁴ to 10⁻¹⁰ M) for 24 h in a 96-well plate format. However, although the plates containing autobioluminescent HEK293_ERE/Gal4-Lux cells were placed directly into the plate reader to obtain luminescence readings, the hERα-HeLa-9903 cells were first lysed and processed using the Steady-Glo luciferase assay reagent before luminescence measurement.

In both assays, the three positive ER agonists produced full sigmoidal dose-response curves and corticosterone tested negative (Figure 2). The HEK293_ERE/Gal4-Lux biosensor demonstrated an EC₅₀ value of 7.9 × 10⁻¹² M for the strong ER agonist 17β-estradiol, which was similar to the 4.6 × 10⁻¹² M EC₅₀ value obtained from the HeLa-9903 luciferase assay (Table 1). The HEK293_ERE/Gal4-Lux biosensor was less sensitive than the HeLa-9903 luciferase assay for detection of the weak agonist 17α-estradiol and the very weak agonist 17α-methyltestosterone, with EC₅₀ values of 2.9 × 10⁻¹⁰ M and 1.0 × 10⁻⁶ M compared to 1.5 × 10⁻¹¹ M and 8.2 × 10⁻⁸ M, respectively (Table 1). However, the relative effect potency (REP) values of these compounds as determined using the HEK293_ERE/Gal4-Lux biosensor were determined to be 0.03 and 0.000008, which is in agreement with previous literature reports (ICCVAM, 2003, 2006, 2011).

Figure 2.

Dose-response curves of exposure to the strong ER agonist 17β-estradiol, the weak agonist 17α-estradiol, and the very weak agonist 17α-methyltestestorone in the (a) HEK293ER_E/Gal4-Luxassay and (b) HeLa-9903 assay.

Table 1.

Comparison of the HEK293_ERE/Gal4-Lux Autobioluminescent Assay and the HeLa-9903 Estrogen Receptor Transcription Activation Assay

Chemical	HEK293ERE/Gal4-Lux EC50 (M)a	HEK293ERE/Gal4-Lux REPa	HeLa-9903EC50 (M)b	HeLa-9903 EC50 (M)a
17β-Estradiol	7.9 × 10−12	1	5.0 × 10−12 to 7.9 × 10−11	4.6 × 10−12
17α-Estradiol	2.9 × 10−10	0.03	2.5 × 10−10 to 4.0 × 10−9	1.5 × 10−11
17α-Methyltestosterone	1 × 10−6	0.000008	—	8.2 × 10−8
Corticosterone	Negative	Negative	Negative	Negative

^aData derived from this study.

^bData derived from EPA guideline (Environmental Protection Agency, 2011d).

HEK293_ERE/Gal4-Lux Validation Against the H295R Steroidogenesis Assay

In contrast to traditional in vitro cell culture-based systems, such as those like the HeLa-9903 assay that require cell lysis concurrent with each reading, ELISA-based screening methods provide a more economical and simplistic method for laboratories that do not have access to the specialized equipment required for automated cellular processing. Therefore, to determine if the autonomous nature of the HEK293_ERE/Gal4-Lux bioreporter would allow it to make in vitro cell culture-type assays more accessible to these laboratories, it was similarly validated against an ELISA-based steroidogenesis assay that utilizes the H295R human adrenocortical carcinoma cell line. Both the HEK293_ERE/Gal4-Lux and the H295R cell lines were analogously treated with serial dilutions of 17α-estradiol (10⁻⁶ to 10⁻¹² M), 17α-methyltestosterone (10⁻⁵ to 10⁻¹¹ M), and corticosterone (10⁻⁴ to 10⁻¹⁰ M) to allow for the comparison of results between both assay formats. The known inducer forskolin was also included as a positive control in the H295R steroidogenesis assay as suggested by EPA guidelines.

The H295R assay did not demonstrate 17β-estradiol production in response to corticosterone treatment but identified both 17α-estradiol and 17α-methyltestosterone as positive inducers. Furthermore, this assay format could not produce full sigmoidal dose-response curves using the tested concentrations. Therefore, PC₁₀ values, which are defined as the concentration required to induce 10% of maximum induction by the positive control forskolin, were instead calculated for each compound. The PC₁₀ concentration for the weak ER agonist 17α-estradiol was calculated to be 2.3 × 10⁻⁸ M, and the PC₁₀ of the very weak agonist 17α-methyltestosterone was calculated to be 1.7 × 10⁻⁵ M (Table 2). These data suggest that, according to the H295R ELISA assay, 17α-estradiol is approximately 739 times more potent than 17α-methyltestosterone. Despite their different testing formats, HEK293_ERE/Gal4-Lux identified similar relative estrogenic potency differences between the two positive agonists and was also nonresponsive to corticosterone. HEK293_ERE/Gal4-Lux estimated the PC₁₀ values of 17α-estradiol and 17α-methyltestosterone to be 7.5 × 10⁻¹¹ M and 7.2 × 10⁻⁸ M, respectively, representing an approximately 960-fold difference (Table 2).

Table 2.

Comparison of the HEK293_ERE/Gal4-Lux Autobioluminescent Assay and the H295R Steroidogenesis Assay

Chemical	HEK293ERE/Gal4-Lux PC10 (M)	H295R PC10 (M)
17α-Estradiol	7.5 × 10-11	2.3 × 10−8
17α-Methyltestosterone	7.2 × 10-8	1.7 × 10−5
Corticosterone	Negative	Negative

Performance Testing of the HEK293_ERE/Gal4-Lux Biosensor Against a Chemical Library

To succeed as a Tier 1 screening assay, HEK293_ERE/Gal4-Lux must be capable of rapidly and efficiently reporting the estrogenicity of a wide range of chemicals. To validate its ability to perform these types of screens, it was challenged with the 76 chemical ICCVAM in vitro ER test method validation library (ICCVAM, 2003) (Supplementary Table 1). These compounds encompass a wide range of agonist and antagonist effects, and oftentimes display conflicting results between previously validated assay formats (ICCVAM, 2003). Due to the variable responses of these compounds, autobioluminescent assays performed with HEK293_ERE/Gal4-Lux were considered successful if their qualitative response (positive/negative for EDC activity) was similar to the majority of responses in the published literature (ICCVAM, 2003).

Table 3 lists the EC₅₀ and REP values of representative examples of the test results. Similar to the natural estrogens 17β-estradiol and 17α-estradiol (Figure 2a), synthetic hormones with strong affinities for ER binding such as diethylstilbestrol and 17α-ethynylestradiol (Dickson and Eisenfeld, 1981; Okulicz and Johnson, 1987) were identified by the autobioluminescent assay as potent ER agonists. Both compounds induced full sigmoidal dose responses in the biosensor cells, with EC₅₀ values of 1.2 × 10⁻¹¹ M (Table 3 and Supplementary Figure 2). These compounds displayed the highest REP values (0.66 for each) among the test library in the autobioluminescent assay, which is in agreement with published reports that consistently rank them as highly estrogenic chemicals with potencies similar to 17β-estradiol (ICCVAM, 2003, 2006, 2011; Legler et al., 2002). For many of the tested compounds, the results obtained using the HEK293_ERE/Gal4-Lux assay were highly similar to those of alternative assays. For example, HEK293_ERE/Gal4-Lux identified the industrial plasticizer bisphenol A as having an EC₅₀ value of 4.6 × 10⁻⁷ M and a REP value of 0.000017, while the MCF7 cell-based luciferase reporter gene transcription activation assay reported an EC₅₀ value of 6.3 × 10⁻⁷ M and a REP of 0.000013 (Kitamura et al., 2005). HEK293_ERE/Gal4-Lux was also responsive to phytoestrogens such as daidzein and coumestrol, estimating their EC₅₀ values to be 4.7 × 10⁻⁷ M and 1.8 × 10⁻⁸ M, respectively (Table 3 and Supplementary Figure 2).

Table 3.

EC₅₀ Values of Common Endocrine Disruptor Chemicals Tested Using HEK293_ERE/Gal4-Lux Autobioluminescent Cells

Chemical	Note	EC50 (M)	REP
17α-Ethynylestradiol	Synthetic hormone	1.2 × 10−11	0.66
Diethylstilbestrol	Synthetic hormone	1.2 × 10−11	0.66
Bisphenol A	Industrial plasticizer	4.6 × 10−7	0.000017
Daidzein	Phytoestrogen	4.7 × 10−7	0.000017
Coumestrol	Phytoestrogen	1.8 × 10−8	0.00044

Across the full library, 68 of the 76 compounds (90%) were qualitatively identified as positive/negative for EDC activity in agreement with the majority of the published literature (Table 4). Only eight tested chemicals generated responses contradicting the majority-reported literature values. Morin, phenobarbital, and phorbol 12-myristate 13-acetate were identified as positive by HEK293_ERE/Gal4-Lux but as negative in the majority of the ICCVAM meta-analysis and 2,4,5-trichloro-phenoxyacetic acid, chlordecone/kepone, dexamethasone, dicofol, and raloxifene HCl demonstrated reciprocal results.

Table 4.

Qualitative ER Agonism Responses of the HEK293_ERE/Gal4-Lux Autobioluminescent Assay Relative to the Multi-Assay ICCVAM Meta-Analysis for All Tested Compounds

Chemical	Meta-Analysis Majority Responsea	HEK293ERE/Gal4-Lux Responseb
1,1-Dichloro-2,2-bis(4-chlorophenyl)ethene	Negative	Negative
17ß-Estradiol	Positive	Positive
17ß-Trenbolone	Negative	Negative
17α-Estradiol	Positive	Positive
17α-Ethynylestradiol	Positive	Positive
19-Nortestosterone/nandrolone	Positive	Positive
2-sec-Butylphenol	Negative	Negative
2,4,5-Trichloro-phenoxyacetic acidc	Positive	Negative
4-Androstenedione	Negative	Negative
4-Cumylphenol	Positive	Positive
4-Hydroxytamoxifen	Positive	Positive
4-Nonylphenol	Positive	Positive
4-tert-Octylphenol	Positive	Positive
5α-Dihydrotestosterone	Positive	Positive
Actinomycin D	Negative	Negative
Ammonium perchlorate	Negative	Negative
Apigenin	Positive	Positive
Apomorphine	Negative	Negative
Atrazine	Negative	Negative
Benzyl butyl phthalate	Positive	Positive
Bicalutamide	Negative	Negative
Bisphenol A	Positive	Positive
Bisphenol B	Positive	Positive
Chlordecone/Keponec	Positive	Negative
Chrysin	Positive	Positive
Clomiphene citrate	Positive	Positive
Corticosterone	Negative	Negative
Coumestrol	Positive	Positive
Cycloheximide	Negative	Negative
Cyproterone acetate	Negative	Negative
Daidzein	Positive	Positive
Dexamethasonec	Positive	Negative
Di-n-butyl phthalate	Positive	Positive
Dicofolc	Positive	Negative
Diethylhexyl phthalate	Negative	Negative
Diethylstilbestrol	Positive	Positive
Estrone	Positive	Positive
Ethyl paraben	Positive	Positive
Fenarimol	Positive	Positive
Finasteride	Negative	Negative
Flavone	Positive	Positive
Fluoranthene	Negative	Negative
Fluoxymestrone	Negative	Negative
Flutamide	Negative	Negative
Formestane	Negative	Negative
Genistein	Positive	Positive
Haloperidol	Negative	Negative
Hexestrol	Positive	Positive
4-Hydroxyflutamide	Negative	Negative
Kaempferol	Positive	Positive
Ketoconazole	Negative	Negative
l-Thyroxine	Negative	Negative
Linuron	Negative	Negative
Medroxyprogesterone acetate	Negative	Negative
17α-Methyltestosterone	Positive	Positive
Mifepristone	Negative	Negative
Morinc	Negative	Positive
Nilutamide	Negative	Negative
Norethynodrel	Positive	Positive
o, p′-DDT	Positive	Positive
p, p′-Methoxychlor	Positive	Positive
Phenobarbitalc	Negative	Positive
Phenolpthalin	Negative	Negative
Pimozide	Negative	Negative
Phorbol 12-myristate 13-acetatec	Negative	Positive
Procymidone	Negative	Negative
Progesterone	Negative	Negative
Propylthiouracil	Negative	Negative
Raloxifene HClc	Positive	Negative
Reserpine	Negative	Negative
Resveratrol	Positive	Positive
Sodium azide	Negative	Negative
Spironolactone	Negative	Negative
Tamoxifen	Positive	Positive
Testosterone	Negative	Negative
Vinclozolin	Negative	Negative

^aData derived from ER agonism meta-analyses from ICCVAM (2003, 2006, 2011).

^bData derived from this study.

^cChemicals displaying contradictory EDC response profiles.

Compound Assessment Using an Alternative MCF7 Cell Line

Although HEK293-based autobioluminescent screening provides a proof-in-principle demonstration of the systems capabilities, SERM effects and metabolic activity differences may limit the utility of using only a single host cell type for compound evaluation. The system was therefore recapitulated in the naturally ER-positive MCF7 cell line and challenged with 17β-estradiol (as a positive control) and raloxifene HCl, which was classified as positive in the ICCVAM meta-analysis, but tested negative in the HEK293-based autobioluminescent assay. Autobioluminescent MCF7_ERE/Gal4-Lux cells correctly identified both 17β-estradiol and raloxifene HCl as positive, demonstrating autobioluminescent inductions of 1.91 (± 0.23) and 1.41 (± 0.06)-fold, respectively, following compound treatment.

DISCUSSION

There are a wide variety of potential cell lines that can be selected as bioreporter development chassis. The choice of host cell is important because it will influence the sensitivity and performance of the assay due to innate differences in metabolic activity, selective estrogen receptor modulator (SERM) sensitivity, and robustness (Ball et al., 2009). Although the ideal assay would encompass reporter expression across a variety of cell lines to limit the influence of these variables, this work limited demonstration to the naturally ER-negative HEK293 cell line. HEK293 was chosen for several reasons. First, previous work has validated that expression of the autobioluminescent phenotype does not significantly alter its basal metabolic activity level (Close et al., 2010), which could potentially interfere with ER activation and/or reporter function. Additionally, SERM-based ER activation has been suggested to serve as a transcriptional activator in HEK293 cells relative to naturally ER-expressing cell lines (Leung et al., 2007), which should expand the number of compounds the assay is capable of identifying using the chosen reporter system. Furthermore, the lack of native EDC-mediated intracellular signaling pathways in HEK293 cells reduces background induction during assay performance to improve assay sensitivity. Human ERα was chosen because it is known to activate transcription when 17β-estradiol is present, although the 17β-estradiol/ERβ complex appears to inhibit activator protein (AP-1)-mediated transcription (Aranda and Pascual, 2001). Moreover, previous studies using animal models have shown that estrogenic effects induced by EDCs are mainly mediated through ERα (Dang et al., 2007). This provides the best chance that weakly active EDCs can be captured by the ER and subsequently trigger ER-mediated activation of the autobioluminescent amplification circuit.

In theory, this process for recognizing and reporting EDCs is identical to that of traditional bioluminescent EDC reporters. However, the use of an autobioluminescent signal requires the expression of multiple synthetic luciferase genes (Xu et al., 2014). This necessitates transcription of a relatively extended mRNA, which is not as efficient as transcription of a shorter, single gene sequence. It was therefore necessary to include the ERE-Gal4FF/UAS-Lux amplification circuit to achieve sufficient translational levels of each gene product to produce a visible signal without necessitating any external stimulation. However, by incorporating this amplification circuit to drive the expression of the full synthetic luciferase gene cassette, the reporter is capable of bypassing the need to perform any assay step associated with sample destruction or excitation. This provides several advantages relative to fluorescent or substrate-requiring assay systems that necessitate the application of a photonic activation signal or chemical substrate to initiate their signaling response. Fluorescent-based reporters, such as GFP, are somewhat similar in that they allow for near-continuous imaging, however, they are susceptible to photobleaching and often display poor signal-to-noise ratios due to autofluorescence from human cellular hosts. This limits their utility in human-relevant systems such as the one developed in this work. Conversely, firefly luciferase-based reporters generate extremely high signal-to-noise ratios due to the lack of endogenous background bioluminescence in host cells, but their reliance on the repetitive administration of a light activating chemical substrate (luciferin) that is expensive, sensitive, and yields only single time point data, is disruptive toward cost-effective, high-throughput screening. This use of autobioluminescence within a human cellular host simplifies the assay design, reduces performance costs by reducing regent usage and hands-on time, and allows for samples to be assayed at any relevant time point.

These advantages were evident through comparison with the HeLa-9903 ER transcription assay and the H295R steroidogenesis assay, both of which are approved as a part of the EPA’s EDSP (OCSPP Test Guideline 890.1550) Tier 1 screening battery for endocrine disruptors (Environmental Protection Agency, 2011b). Although both HEK293_ERE/Gal4-Lux and HeLa-9903 are in vitro cell culture-based systems that utilize luciferase to report ER-mediated transcriptional activation, the HeLa-9903 cells required an external chemical substrate application concurrent with cellular lysis to activate their luciferase. This process necessitates additional hands-on time or the use of automated injection equipment. These inefficiencies persisted following lysis and treatment, as the Steady-Glo reagent required to induce signal generation was found to have a half-life of greater than 5 h at 22°C, with an approximate 13% loss of luminescence per hour. This signal decay limited the number of plates/chemicals that could be processed per day. Also, after reconstitution, the reagent could only be subjected to a limited number of freeze-thaw cycles, thereby increasing the cost of the HeLa-9903 assay by requiring fresh reagent to be obtained during assays analyzing several hundred chemicals. In contrast, reagent stability was not an issue with the HEK293_ERE/Gal4-Lux bioreporter. Because these cells biosynthesize their own reagent and bioluminescence generation did not require sample destruction, they could be analyzed repeatedly or at any desired time.

Although not subject to these external substrate-dependent luciferase problems, the H295R steroidogenesis assay also presented several logistical challenges that were overcome by the autonomous nature of the HEK293_ERE/Gal4-Lux assay. Its ELISA-based format required a hands-on time of at least 3 h for sample preparation and necessitated that absorbance readings be performed within 30 min of adding the final reagent to reduce signal quenching. This renders the steroidogenesis assay unsuitable for high-throughput performance and significantly limits its serviceability for Tier 1 EDC screening. In contrast, HEK293_ERE/Gal4-Lux did not require any sample preparation time beyond plating and dosing the cells and reduced the cost of analysis substantially when compared with the fact that a single ELISA test kit (∼$500) could only assay a 96-well plate with four chemical treatments.

In fact, although all tests in this work were performed in 96-well plates to match the most widely reported assay format, the autobioluminescent assay format used by HEK293_ERE/Gal4-Lux has been demonstrated in microtiter plates up to 1536-well and can easily be transitioned to higher throughput parameters if required (Class et al., 2015). Furthermore, its lack of requirements for sample destruction or pre-processing prior to signal acquisition makes it highly amenable to automation. This is especially important given that ICCVAM has specifically called for the development of new assays capable of robotic integration (ICCVAM, 2003, 2006) to address the large chemical backlog of the EDSP21 program (Vogel, 2004). At present, there are few estrogen-responsive assays that meet this requirement. Of those that do, such as the MCF7:WS8 cell proliferation assay (Yang et al., 2014) and the BG1Luc reporter assay (Stoner et al., 2014), there are often other intrinsic hurdles that make them less attractive as Tier 1 screening platforms. The MCF7:WS8 assay, for instance, relies upon an indirect output to report ER-binding, which sacrifices sensitivity to allow for incorporation into robotic workflows. Similarly, the BG1Luc assay relies upon the addition of exogenous substrate. Although this step can be achieved using modern laboratory automation systems, it also increases assay costs similar to when performed manually. The HEK293_ERE/Gal4-Lux assay system, in contrast, can be scaled to allow for robotic integration where cell plating, dosing, incubation, and reading are automated. Because addition of exogenous substrate or sample manipulation post-treatment is not required, this system reduces assay complexity and facilitates rapid detection using automated systems.

It is hypothesized that this use of automation would also reduce assay variability by removing human variables. The primary weakness of the HEK293_ERE/Gal4-Lux assay system under its current incarnation is its variability. Although the 15.7% inter-assay CV observed in this work falls within the acceptable range for an assay of this type, it also suggests there is room for improvement in standardization (Reed et al., 2002). It is likely that the use of multiple cell lots and variations in the performance of cell preparation by different technicians contributed to the observed level of variability, as did the use of a 96-well format. In this instance, assay CV values are marginally higher than those from assays that have used autobioluminescent reporters in the 1536-well format and leveraged robotic loading of cells and test samples (Class et al., 2015). It is also possible that the increased variability is due to uneven internal luciferin production levels or promoter stimulation among the cells in the population. However, this demonstrates that, for repeated usage, observed variability should be maintained within acceptable levels.

Similarly, the 0.32 Z-factor was acceptable, but could also be improved to fall within the ideal range of >0.5 (Zhang et al., 1999). A potential limitation for this improvement may be the reliance of the autobioluminescent phenotype on intracellular substrate production. Unlike a traditional firefly luciferase system, in which the luciferin chemical is added at saturating concentrations following cellular lysis to expose the luciferase protein, the synthetic luciferase used in this work is necessarily limited to producing its luciferin at a level that does not impact cellular health or impose a detectable metabolic burden (Close et al., 2010). This results in a lower signal output maximum than a similar, externally stimulated luciferase reporter. The use of an inducible promoter system may also be a contributing factor, as metabolic activity assays using autobioluminescent reporters under the control of constitutive promoters have demonstrated Z-factors in the range of 0.58–0.82 (Class et al., 2015). This suggests that, although suitable in its present form, improvements such as optimizing the promoter sequences to improve fold change could improve the assay by enhancing its dynamic range.

Despite failing to achieve the highest tiers for Z-factor and CV, the HEK293_ERE/Gal4-Lux assay nonetheless performed well when used to evaluate a diverse chemical library. The 76 chemical ICCVAM in vitro ER test method validation library (ICCVAM, 2003) used in this work (Supplementary Table 1) encompasses a wide range of agonist and antagonist effects that oftentimes display conflicting results between previously validated assay formats (ICCVAM, 2003). However, despite these challenges, HEK293_ERE/Gal4-Lux identified 90% of the tested chemicals in agreement with the majority of the ICCVAM meta-analysis.

Interestingly, in all of the cases where the ICCVAM meta-analysis suggested a compound was negative, but HEK293_ERE/Gal4-Lux suggested it was positive (morin, phenobarbital and phorbol 12-myristate 13-acetate), the compounds previously had not been tested using transcriptional activation assays and had been identified as negative based on alternative mechanisms such as the activation of ligand independent cell division or thyroid hormone excretion. In the majority of the alternative instances, only one previous test had been recorded for the suspect compounds. However, in instances where multiple tests had been performed, such as for Chlordecone/Kepone and Dicofol, the ICCVAM had noted mixed results (4 of 6 tests positive for Chlordecone/Kepone and 3 of 4 tests positive for Dicofol). Only for raloxifene HCl did HEK293_ERE/Gal4-Lux contradict multiple previously positive tests (n = 3). In this case, the metabolic rate and capability of the HEK293 cells for xenobiotics, as well as its responsivity to SERM-based ER stimulation, likely played a role. This is especially true for compounds such as Chlordecone/Kepone, which are known to be metabolized slowly and often compete with estrogens or fail to trigger ER stimulation in ER-negative cell lines (Thomas and Dong, 2006). To test this hypothesis, the system was transiently transfected into the naturally ER-positive MCF7 cells and the assay was repeated. Using this alternative cell line, the autobioluminescent assay identified raloxifene HCl as positive in agreement with the ICCVAM meta-analysis. Although this test demonstrated a low overall fold of induction (1.41-fold), this is likely due to the use of transient transfection for reporter development, which necessarily resulted in less than 100% of the cells in the assay being capable of responding to compound treatment. However, even under transient transfection conditions, there was high confidence in these results given the low variation among both biological and technical assay replicates.

These results indicate that the detection of estrogenicity from xenobiotic compounds, such as o, p'-DDT, p, p'-DDE, methoxychlor, and atrazine, will be limited when using a single cell type-based assay system. Rather, a more comprehensive system would include alternative host cell types to act as secondary or co-screening tools. This will better identify the tissue-specific estrogenic potential of SERMs or compounds that induce negative metabolic activity dynamics at levels impacting reporter signal generation.

CONCLUSIONS

The adverse health effects of EDCs, and their ubiquity in food, water, soil, pesticides, and household products, demands that improved detection assays be developed to better control exposure. HEK293_ERE/Gal4-Lux provides a significant advantage over traditional luciferase-based detection systems that allows more chemicals to be screened for EDC activity by lowering testing costs, yet maintains the accuracy and specificity of estrogenic chemical detection criteria set forth by ICCVAM, EPA, and the international Organization for Economic Co-operation and Development (OECD). Because no exogenous luciferin chemical is required for the autobioluminescent reporter assay, the system can be expanded into additional host cell types to create a screening panel for tissue-specific EDC effects without incurring the accompanying cost increase that would be necessary for a similar luciferin-requiring assay. The expansion of this screening approach to additional host cells and the limitation of variability provided by the use of robotic systems can make it a suitable tool for improving EDC detection across a variety of tissue types.

CONFLICT OF INTEREST

S.R. and D.C. are board members for 490 BioTech, Inc. All other authors declare they have no actual or potential competing financial interests.

SUPPLEMENTARY DATA

Supplementary data are available at Toxicological Sciences online.

FUNDING

National Institutes of Health, National Institute of Environmental Health Sciences (5R44ES022567-03 and 1R15ES023979-01). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.