Characterizing properties of non-estrogenic substituted bisphenol analogs using high throughput microscopy and image analysis

Animal studies have linked the estrogenic properties of bisphenol A (BPA) to adverse effects on the endocrine system. Because of concerns for similar effects in humans, there is a desire to replace BPA in consumer products, and a search for BPA replacements that lack endocrine-disrupting bioactivity is ongoing. We used multiple cell-based models, including an established multi-parametric, high throughput microscopy-based platform that incorporates engineered HeLa cell lines with visible ERα- or ERβ-regulated transcription loci, to discriminate the estrogen-like and androgen-like properties of previously uncharacterized substituted bisphenol derivatives and hydroquinone. As expected, BPA induced 70–80% of the estrogen-like activity via ERα and ERβ compared to E2 in the HeLa prolactin array cell line. 2,2’ BPA, Bisguaiacol F, CHDM 4-hydroxybuyl acrylate, hydroquinone, and TM modified variants of BPF showed very limited estrogen-like or androgen-like activity (< 10% of that observed with the control compounds). Interestingly, TM-BFP and CHDM 4-hydroxybuyl acrylate, but not their derivatives, demonstrated evidence of anti-estrogenic and anti-androgenic activity. Our findings indicate that Bisguaiacol F, TM-BFP-ER and TM-BPF-DGE demonstrate low potential for affecting estrogenic or androgenic endocrine activity. This suggest that the tested compounds could be suitable commercially viable alternatives to BPA.


Introduction
The impact of man-made chemicals on human health and the environment is a continuous global concern. Between the thousands of compounds synthesized, chemicals that impact the endocrine system took a central stage as they have the potential to affect whole live cycle of humans through central functions like reproduction and metabolism, while accumulating over time both in the human body and the environment. Probably the best studied endocrine disrupting chemicals (EDCs) are the ones with estrogenic or anti-estrogenic activity. Of these, Bisphenol A (BPA) has garnered attention due to its widespread detection in the human a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 with funding from the John S. Dunn Gulf Coast Consortium for Chemical Genomics, the Dan L. Duncan Cancer Center (NIH P30CA125123), and the NIH (HD007495, and DK56338). ATS is a K12 Scholar supported by NIH grant K12DK0083014, the multidisciplinary K12 Urologic Research (KURe) Career Development Program awarded to Dr. Dolores J. Lamb. DeepBio, Inc received funds and material support from the Valspar Corporation. Funders do not have scientific or editorial control over the design, conduct, analysis, or interpretation of the data, nor were they involved in the decision to publish or producing this manuscript. DeepBio, Inc provided support in the form of materials and consulting fees for A.T.S., F.S., and M.G.M. and provided salary for M.A.M. The specific roles of these authors are articulated in the 'author contributions' section. M.A.M participated in in study design, data interpretation, and final manuscript review, however, DeepBio, Inc did not have editorial control nor was involved in the decision to publish. epigenetic mark alterations, chromatin remodeling, and transcriptional regulation in response to ER ligands [13][14][15]. The PRL-HeLa cell line was further-adapted to high throughput microscopy-based screening by the generation of stable variants expressing either ERα or ERβ translationally-fused with green fluorescent protein (GFP-ERα:PRL-HeLa and GFP-ERβ: PRL-HeLa). When combined with our custom, automated image analysis platform [16], these cell lines were used to discriminate and classify the mechanistic effects of E2, ER antagonists, BPA and 18 closely related BPA analogs on ERα and ERβ, a dataset that has also been included in the EPA mathematical models [12,17,18]. In the present study, we use these ERα-and ERβexpressing PRL-HeLa, the widely utilized ER-positive MCF-7 breast cancer, and the AR-positive LNCaP prostate cancer cell lines in high content/high throughput assays to characterize and classify seven, less well studied, potential BPA substitutes identified by Valspar during an internal yeast-based in vitro assay (data not published). The molecules that we analyzed include: 2,2' BPA, bisguaiacol F (BGF), CHDM 4-hydroxybuyl acrylate (CHDM-4-HBA), hydroquinone (HydroQ), tetramethyl bisphenol F (TMBPF), tetramethyl bisphenol F diglycidyl ether (TMBPF-DGE), and TMBFP-ER (Fig 1).
For PRL-Hela or 2PB-mCherry-NLS:LnCaP experiments, cells were robotically seeded using a Titertek dispenser in 384-well optical Cyclo-Olefin-Polymer (COP) bottom plates (384-IQ, Aurora Biotechnologies) that have been empirically-shown to be free of estrogenic activity. Cells were allowed to adhere overnight prior to treatment with compounds alone or in combination with 10 nM E2 or DHT at concentrations ranging from 50 pM to 5 μM for either 1 hr or 24 hr.
For ERE-luciferase experiments, MCF-7 ERE-MAR cells (containing a stably integrated ERE-driven luciferase reporter gene) were manually plated into 24 well plates. Cells were allowed to adhere overnight prior to treatment with compounds alone at 2 μM and 5 μM. For combination experiments, cells were treated in the presence of 10 nM E2 for 24 hrs.

Immunofluorescence
Immunofluorescence was performed as previously described [17,19]. Cells were fixed in 4% EM-grade formaldehyde in PEM buffer (80 mM potassium PIPES [pH 6.8], 5 mM EGTA, and 2 mM MgCl 2 ) and quenched with 0.1 M ammonium chloride for 10 min. For samples in which no antibodies are used, cell membranes were permeabilized using 0.5% Triton X-100 for 10 min and DNA stained using DAPI (1 μg/ml) for 10 min. For samples with antibody labeling, permeabilization was with 0.5% Triton X-100 for 30 min. Cells were incubated at room temperature in blotto (5% milk in Tris-buffered saline/Tween 20) for 30 min, and then specific primary antibodies were added overnight at 4˚C prior to 1 hr of secondary antibody (Alexa 647 conjugates; Molecular Probes) and DAPI staining (1 μg/ml for 10 min). The primary antibodies used were: mouse anti-Ser5-phospho RNA polymerase II (Abcam, ab5401) and mouse anti-SRC-3 (BD Transduction Labs No. 611105).

High throughput microscopy and image analysis
Image data sets were collected using the IC-200 image cytometer (Vala Sciences, San Diego, CA) utilizing dual-step high-speed (50-100 ms) reflection-and image-based autofocusing, either a 40X/0.95 Nikon S-fluor or a 20X/0.75 Plan Apo objective, and a sCMOS 5.5-megapixel camera and LED illumination. Z-stacks were collected at 1 μm (20X) or 0.5 μm (40x) intervals and maximum-projected using the instrument-supplied Mexican Hat 13 x 13 synthetic focus algorithm. Cell, nucleus, array segmentation and signal quantification was performed using the myImageAnalysis web application powered by Pipeline Pilot software (Biovia) as previously described [16]. Aggregated cells, mitotic cells, and apoptotic cells were removed using filters based on nuclear size, nuclear shape, and nuclear intensity.

MCF-7 ERE-MAR luciferase assay
MCF-7 ERE-MAR (also known as MCF-7-XVA2-Luc) cells, a kind gift from Dr. Steffi Oesterreich (University of Pittsburgh), are a human breast cancer cell line that contains a stable integration of the firefly luciferase gene under control of a promoter containing the estrogen response element (ERE) and matrix attachment region (MAR) derived from the Xenopus vitellogenin A2 promoter and responds to estrogenic compounds with a 30-35-fold induction of firefly luciferase [20]. After 24 hr of compound treatment, culture medium was aspirated and plates were frozen at -20˚C until further processing. Cells were lysed with 1% Triton X-100, 10% glycerol, 2 mM EGTA, and 1 mM DTT. Luciferase was then measured using an automated microplate luminometer with the Promega Luciferase Assay System. Results from 4 replicates were collected and normalized to controls wells treated with E2 only.

Calculation of estrogenicity (%E2) and statistical analysis
The estrogenicity of compounds was calculated relative to the maximum response observed with 10 nM E2 treatment (%E2, a measure of response amplitude). Each compound at each concentration was measured in triplicate or quadruplicate. Basal activity of each assay was determined using "mock" control (MC) containing treatment media and vehicle controls (VC) containing 0.5% DMSO solvent in a minimum of eight samples on each assay plate. The MC was set to 0% resulting in a %E2 calculation (1): Typical values for VC were 0% ± 5% E2. Inclusion of a VC accounts for any extraneous residual estrogenicity that might exist in the media or derived from the materials used for sample preparation. A chemical is considered to be significantly different from VC controls if it produces a mean %E2 significantly greater (p < 0.05) than VC as determined by ANOVA and post-hoc pairwise comparisons made using the Tukey HSD method. A compound is considered to have inhibited the E2 induced response if, when added to 10nM E2, it produces a mean %E2 significantly less than E2 controls. As a general rule, we consider a compound to have significant estrogenic properties with a %E2 score greater than 10% when used alone or antiestrogenic properties with a %E2 change greater than 15%. %DHT calculations were done in an identical manner using 10 nM DHT at the control response.
Averaged results from a minimum of 4 independent experiments were graphed ± standard deviation (SD) using SigmaPlot software with EC 50 values determined by fitting a 4-parameter sigmoidal curve. All data used to generate summary data can be found in a supplemental data file (S1 Data). Clustering analysis was performed using Cluster 3.0 and visualized with Java Treeview.

Measure of estrogenic activity using ERα PRL-HeLa cells
As described in the introduction, we previously established the utility of an engineered PRL-HeLa cell line, stably-expressing either GFP-ERα or GFP-ERβ, for high content analysis of ER-mediated transcriptional regulation and classification of endocrine disruptor chemicals (S1 Fig) [13,17]. Fig 2A describes the overall workflow of the high throughput assays and analysis routines that were used in the present analysis.
We initially assessed the ability of the potential endocrine disrupting compounds (Fig 1 for abbreviations used, CID #, and structures) to induce GFP-ERα recruitment to the multi-copy ERE in the PRL array, as compared with E2 10nM, which is always used as a positive (agonist) control. Also, in all experiments, the selective estrogen receptor modulators (SERMs) 4HT and BZA are used as antagonist controls. ER recruitment to the PRL array reflects a combination of ligand binding, receptor dimerization and DNA binding and was assessed after 1hr of treatment. Fig 2B shows the percent of cells with visible GFP-ERα targeting the ERE-rich promoter locus. Consistent with our previous work, treatment with ER SERMs such as 10 nM 4HT or BZA produced a DNA binding response equivalent to that of E2 alone (%E2, 90% and 106% respectively), consistent with the understanding that ER bound to SERMs bind DNA but recruit co-repressor proteins rather than co-activator proteins. Treatment with 5 μM bisphenol A (BPA) also induced significant DNA binding, although with lower efficacy as indicated by a %E2 of 69%. With 1 hr treatment at 5 μM, we were unable to observe any significant induction of GFP-ERα DNA binding (%E2 < 1%) by any of the seven investigational compounds (BGF, 2,2' BPA, CHDM-4-HBA, HydroQ, TMBPF, TMBPF-DGE, or TMBFP-ER). Similar results were observed when treatment time was extended to 24 hr (Fig 2C) or concentrations ranging between 50 pM to 5 μM were tested ( Fig 2D). Concentrations beyond 5 μM resulted in significant cell toxicity and were not analyzed. As expected, treatment of cells with E2 and either ER SERM did not decrease the DNA binding as receptor bound to either the agonist or the SERM demonstrates high affinity for estrogen responsive promoter elements. Interestingly, two of the investigational compounds (CHDM-4-HBA and TMBPF) had significant effects when combined with E2 as they reduced the % of cells with visible arrays from 100% to~60% (TMBPF) and~20% (CHDM-4-HBA), potentially indicating a different mechanism of antagonism as compared with 4HT and BZA (Fig 2B, grey bars).
Next, we quantified the effect of compound treatment on the E2 induced recruitment of SRC-3, a central ER coregulator, to the PRL array in GFP-ERα:PRL-HeLa ( Fig 2E) after 1 hr of treatment. The p160 coactivator SRC-3 is a key coregulator for nuclear receptors and other transcription factors, and has been shown, along with SRC-1, to be significantly recruited to the PRL-array during E2-mediated transcriptional activity [13,15]. As expected, the ER antagonists 4HT and BZA significantly decreased SRC-3 recruitment with %E2 values reduced > 20%. BPA treatment resulted in %E2 values not significantly different from E2 treatment alone. The investigational compounds did not have a significant effect on E2 induced recruitment of SRC-3 to the PRL array, including TMBPF and CHDM-4-HBA. This suggests that while these compounds reduce the ability of ERα to bind DNA in the presence of E2, these compounds do not affect the ability of ERα to recruit SRC-3 once bound to the PRLarray.
We also observed and quantified PRL array targeting of phospho-serine 5 RNA polymerase II (pS5-RNAPII) following 1 hr of compound treatment in the presence of 10 nM E2 (Fig 2E). Phosphorylation at serine 5 identifies RNA polymerase actively transcribing genes, and increases at the PRL array in the presence of E2 [15]. As expected, the ER antagonist 4HT and BZA significantly reduced pS5-RNAPII recruitment by 22% and 38%. Importantly, the only investigational compound to have a significant effect was CHDM-4-HBA, with treatment resulting in the complete loss (100%) of pS5-RNAPII at the PRL array. This was observed despite maintaining SRC-3 recruitment to the PRL array, which is suggestive of a mechanism for an anti-estrogenic activity distinct from that observed with 4HT and BZA.
To investigate potential effects on ER-mediated transcription at the PRL array, we performed mRNA FISH using probes specific to the dsRED2 reporter protein that is regulated by the PRL promoter elements (S1 Fig). After 1 hour of treatment with 10 nM E2, there was a 4.4-fold increase in de novo dsRED2 mRNA production ( Fig 2F, black bars). Treatment with 5 μM BPA resulted in 60% of the activity observed with E2 treatment whereas none of the investigational compounds activity exceeded 7%. When cells were treated with 10 nM E2 in the presence of 5 μM of the investigational compounds, both CHDM-4-HBA and TMBPF were able to significantly reduce E2-induced de novo mRNA production by 84% and 87%, respectively (Fig 2F, gray bars). This suggests the anti-estrogenic-like effects observed with these compounds on ERα promoter recruitment and/or pS5-RNAPII recruitment leads to reduced ERα mediated transcriptional activity.
Finally, to ensure that results were not due to dramatic changes in GFP-ER expression levels, we quantified GFP-ERα protein level in samples treated with compounds for 24 hr ( Fig  2G). Protein levels were measured by quantifying the total GFP intensity per cell in only those cells analyzed for possible visible array formation. Previous studies demonstrated that E2 of maximal dose of compound. (F) Effects on de novo mRNA production at the PRL array after 1 hour of treatment with compound alone or with 10 nM E2. (G) Effects on GFP-ERα expression level after 24 hours of treatment with maximal dose of compound. All data represent mean of a minimum of 4 experiments ± standard deviation. (*) or (**) indicates a response significantly (p < 0.05) above or below control treatment.
https://doi.org/10.1371/journal.pone.0180141.g002 treatment results in proteasome-dependent degradation of ERα whereas 4HT treatment results in ER stabilization [21]. Consistent with these reports, E2 treatment decreased ERα level by 24% while 4HT and BZA significantly increased protein levels. BPA treatment increased the observed GFP-ERα level by 34%. Of the investigational compounds, both TMBPF-DGE and TMBPF-ER resulted in increased ERα with percent changes of 30% and 20%. The other compounds in this study did not significantly alter the observed GFP-ERα expression level.

Measure of estrogenic activity using GFP-ERβ:PRL-HeLa cells
To complement the GFP-ERα:PRL-HeLa cell line, we recently generated the GFP-ERβ: PRL-HeLa cell line stably expressing GFP-ERβ. We demonstrated that this cell line responds to E2 in a similar manner as the GFP-ERα:PRL-HeLa cell line and allows for a similar visualization and quantification of multiple aspects of the ERβ transcriptional response [17]. Using the pair of cell lines, we were able to demonstrate differential effects on ERα and ERβ of a panel of BPA analogs [17].
We treated the GFP-ERβ:PRL-HeLa cells with the compounds for 1 hr (Fig 3A) and determined the %E2 of cells with visible GFP-ERβ recruitment to the PRL array as an indicator of DNA binding. A similar pattern of responses to that observed in the GFP-ERα:PRL-HeLa cell line emerged. Both 10 nM 4HT and 10 nM BZA produced %E2 of 77% and 80%, indicating a strong ability to induce ERβ DNA binding. BPA treatment produced a %E2 score of 76%, similar to that observed with ERα. We observed a very weak but statistically significant ERβ response with 5 μM 2,2' BPA, CHDM-4-HBA, and TMBPF with %E2 scores ranging between 3% and 9%. The %E2 scores for BGF, HydroQ, TMBPF-DGE, and TMBPF-ER were < 1% and were not significantly different from vehicle. Similar results were observed when treatment time was extended to 24 hr (Fig 3B) or concentrations ranging between 50 pM to 5 μM were tested (Fig 3C).
Similar to ERα, we next determined if compound treatment could alter E2 induced ERβ binding to the PRL array by treating GFP-ERβ:PRL-HeLa with compounds in the presence of 10 nM E2 for 1 hr (Fig 3A, grey bars). As was observed for ERα, 4HT, BZA, and BPA did not inhibit E2 induced ERβ DNA binding. We did not observe the antagonistic effect of TMBPF and CHDM-4-HBA in the ERβ cell line as found in the ERα cell line, suggestive of receptor selectivity. Although CHDM-4-HBA treatment resulted in a %E2 decrease of 11%, this was not significantly different from E2 treatment alone and was dramatically different than the 81% decrease that was observed with ERα.
Although the investigational compounds did not alter GFP-ERβ recruitment to the PRL array, we performed dsRED2 mRNA FISH to determine if they altered ERβ transcriptional regulation at the PRL promoter. As expected, after 1 hour of treatment both E2 and BPA were able to increase de novo dsRED2 mRNA production with 10 nM E2 inducing a 3.4-fold increase and BPA treatment reaching 68% of the E2 response (Fig 3D, black bars). When cells were treated with 5 μM of the investigational compounds, no significant increase was observed. In contrast, when compounds were used in combination with 10 nM E2, CHDM-4-HBA, hydroQ, and TMBPF significantly decreased the E2-induced response between 24 to 30% (Fig 3D, gray bars). The anti-estrogenic-like effect of CHDM-4-HBA and TMBPF was markedly-less than what was observed in similar studies using ERα, further suggesting that the effects of these compounds is dependent on the ER isoform present.
Finally, we quantified the GFP-ERβ expression level in GFP-ERβ:PRL-HeLa samples treated with compounds for 24 hr (Fig 3E). In contrast to ERα, E2 and BPA treatment did not have a significant impact on ERβ expression levels. The stabilizing effect of 4HT and BZA was consistent across receptor isotypes, with treatment resulting in a percent change of 436% and 405%, respectively. None of the investigational compounds altered ERβ expression levels when cells were treated up to 5 μM.

Measure of estrogenic activity using an integrated ERE-luciferase reporter
Although the image based approach using the prolactin array cell line allows for an intensive and mechanistic interrogation of estrogenic activity at the level of individual cells, the HeLa cell line does not endogenously express ER and therefore is not inherently responsive to estrogenic compounds. To overcome this limitation using a routine population-based approach, we used an estrogen responsive MCF-7 line containing a genomic integration of an ERE-driven firefly luciferase reporter to characterize the compounds of interest.
Treatment of the MCF-7 ERE-MAR cell line with 10 nM E2 for 24 hours resulted in a~19-fold induction of luciferase activity (data not shown) whereas treatment with 2 μM BZA resulted in a %E2 luciferase activity score < 1% (Fig 4). BPA treatment resulted in a significant %E2 luciferase score of 57%. When tested at 5 μM, a majority of the investigational compounds did not result in a %E2 luciferase score significantly different from vehicle (Fig 4), which validates the results from the PRL array cell line. However, 2,2' BPA treatment resulted in a %E2 of 19%, which suggest the presence of weak estrogenic activity in this assay.
When MCF-7 ERE-MAR cells were treated with 10 nM E2 in the presence of the compounds to determine potential antagonistic activity, several were found to have an effect ( Fig  4). As expected, the ER antagonist BZA decreased E2 induced luciferase by > 95%. Treatment with 5 μM BPA did not significantly reduce the E2 response, as expected. With the investigational compounds, there was an observable decrease in cell number following treatment when directly examined, indicating a possible toxic effect at this dose in MCF-7 cells that was not observed in the HeLa lines. The compounds TMBPF-DGE, TMBPF-ER, HydroQ, and 2,2' BPA caused a significant inhibition with 41-61% reduction in luciferase activity. Further, TMBPF and CHDM-4-HBA also resulted in an observable decrease; however, these results were not statistically different from E2 alone. When concentrations of the investigational compounds were reduced to 2 μM, no evidence of cell loss was observed and there was no detectable anti-estrogenic-like activity; collectively, these data suggest that the previous results were likely due to non-specific toxicity effects.

Measure of androgenic activity using an integrated probasin-mCherry-NLS reporter
Computational structural analysis, animal studies, and in vitro assays have all generated evidence that BPA and BPA analogs are capable of altering androgenic signaling through direct interactions with the androgen receptor (AR) [22][23][24][25]. To assay potential androgenic-or antiandrogenic-like activity of the investigational compounds in this study, we used the androgen responsive LNCaP prostate cancer cell line with a stable integration of a probasin promoter element (2PB-mChrry-NLS:LNCAP) that contains multiple androgen response elements (AREs), and regulates expression of a mCherry-NLS-mCherry fluorescent protein. The parental LNCaP cell line contains a T877A mutation in the ligand binding domain of AR which is known to increase the receptor responsiveness to estrogens and the anti-androgen hydroxyflutamide.
As expected, DHT treatment was able to induce a dose dependent increase in the nuclear accumulation of the reporter protein, with a 4-fold increase observed after 48 h treatment with 30 nM DHT (Fig 5A). A strong response was also observed with E2 treatment due to the T877A mutation. We were unable to observe any androgenic-like activity from BPA or the investigational compounds after 48 hours of treatment. Next, cells were treated with increasing concentrations of DHT in the presence of 5 μM of reference or investigational compounds for 48 hours (Fig 5B). Again, we observed a strong androgenic-like effect of E2 in this cell line. Further, we were able to observe a weaker, but significant androgenic-like effect of BPA treatment. The known anti-androgens casodex (CDX, bicalutamide) and MDV3100 (enzalutamide) were able to decrease the DHT induced response by 88 and 76%. Importantly, the antiestrogens 4HT and BZA did not significantly alter the DHT response; however, the investigational compound TMBPF resulted in a significant 79% decrease in the DHT induced response indicating a strong anti-androgenic-like response.

Clustering analysis of estrogen-like activity across all assays
To gain a more comprehensive view of these disparate data sets, we compiled the above results and included previously published [17] data that were used to characterize a panel of BPA analogs (i.e., "BPX" compounds). We standardized all data to E2 treated controls. Because the antagonistic effects on E2 induced DNA binding, ER expression, and pS5-RNAPII recruitment were not examined in the published BPX work, these metrics were excluded from the cluster analysis. The resulting data set was range normalized and clustered using Euclidean distance Chemical compound effect on an integrated ERE-driven luciferase reporter in the ER responsive MCF-7 cell line. MCF-7-MAR-ERE-Luc were treated with indicated compounds at 5 μM, 5 μM + 10 nM E2, or 2 μM + 10 nM E2 for 24 hours before total luciferase activity determined. The %E2 calculated using control wells treated with E2 (10 nM) or DMSO. Dashed line indicates 10 nM E2 control response. All data represent mean of 3 experiments ± standard deviation. (*) or (**) indicates a response significantly (p < 0.05) above or below control treatment.
https://doi.org/10.1371/journal.pone.0180141.g004 (Fig 6). Analysis of the clustering reveals that E2 and mixed agonist/antagonist treatments (4HT, BZA) result in a distinct response pattern and suggest that neither the BPX compounds nor the investigational compounds in this study exhibit E2-or SERM-like activity. The second major cluster includes BPA and multiple BPA structural analogs, compounds that retain a significant degree of estrogenic activity. The investigational compounds 2,2' BPA, BGF, CHDM-4-HBA, TMBPF, TMBPF-DGE, and TMBPF-DGE cluster into a branch distant from that containing BPA. The investigational compounds are further subdivided into smaller branches based upon specific functional differences, such as minor differences in induced ERβ DNA binding and inhibition of E2 induced transcription.

Discussion
In response to concerns regarding BPA, there is an emphasis on manufacturing products that no longer use BPA. In some cases, BPA has been replaced with analogs such as BPS and BPF; consequently, human exposure to these analogs is now approaching the same levels as BPA [8]. It is becoming clear that BPA analogs can have similar effects as BPA in cell-based in vitro assays and in vivo animal models [8,17,24,26,27]. These results emphasize the need for rigorous studies to understand potential endocrine effects of compounds proposed as BPA alternatives before widespread adoption. To this end, we completed a set of cell-based studies that included quantitative and mechanistic metrics of estrogenic and androgenic activity on a set of compounds previously identified using a yeast-based in vitro screening approach as having minimal estrogen-like properties.
When utilizing the PRL-HeLa and MCF-7 cell models to characterize a set of reference compounds and seven investigational compounds (2',2 BPA, BGF, CHDM-4-HBA, HydroQ, TMBPF, TMBPF-DGE, and TMBPF-ER), the estrogenic effects of BPA were clear and  consistent with previous work. In PRL-HeLa cells expressing either GFP-ERα or GFP-ERβ, BPA exposure resulted in an extent of visible array formation comparable with E2 treatment. This indicates that at a sufficient concentration, BPA exposure results in the efficient binding of ER (α or β) to the estrogen response sequences located in the prolactin promoter, a crucial step in estrogen regulated gene expression. This result was confirmed by the significant induction of de novo mRNA production at the PRL array and ERE-driven reporter gene activity in the estrogen responsive MCF-7 breast cancer cell line.
In contrast, none of the seven investigational compounds were able to induce significant promoter binding in PRL-HeLa cells expressing either GFP-ERα or GFP-ERβ when tested up to micromolar concentrations at both short (1 hr) and long (24 hr) time points. Based upon hierarchical clustering using the assays presented, these 7 compounds populate a cluster distinct from BPA and 18 other previously-studied BPX compounds and are marked by a relative lack of estrogenic activity (Fig 5). Within in this cluster, 2,2' BPA is unique in that it is characterized by weak estrogenic activity in the MCF-7 cell line with little to no activity in the PRL-HeLa cell lines. This indicates that the altered position of the hydroxyl groups results in a receptor-ligand interaction that retains residual activity on the basic ERE element contained in the MCF-7 reporter cell line, but is insufficient to drive estrogenic activity in the context of the prolactin promoter elements used in the PRL-HeLa cell lines.
Unique within this cluster are the responses observed with CHDM-4-HBA and TMBPF. These are the first compounds we have observed that demonstrated any ability to alter the promoter binding induced by 10 nM E2, reducing ERα recruitment to the PRL array by~80% and~40% and significantly reducing de novo mRNA production. This is in contrast to wellstudied compounds with anti-estrogenic activity such as 4HT and BZA, which antagonize the E2 transcriptional response by the recruitment of corepressor complexes that inhibit transcription, but do not alter ERα recruitment to the PRL array [15,17]. The effect was greatest in the GFP-ERα:PRL-HeLa cell line and not observed or significantly reduced with the GFP-ERβ: PRL-HeLa and the MCF-7 cell lines. In addition, only TMBPF demonstrated an effect on the androgen induced transcriptional activity of the promiscuous AR receptor expressed in the LNCaP prostate cancer cell line. These results suggest that these effects on transcriptional activity are selective based upon receptor, receptor subtype preference, and/or cell background. When promoter binding was observed, neither compound altered the recruitment of the ER coregulator SRC-3 and only CHDM-4-HBA decreased levels of activated RNA Pol II at the integrated PRL array. One potential mechanism by which these compounds may be acting is through the activation of other type I nuclear receptors, as others have observed that activation of the glucocorticoid receptor GR can displace ER from DNA binding sites [28]. The PRL-HeLa cell line does express endogenous GR; however, we have shown that direct activation of endogenous GR using dexamethasone increases, not decreases, ERα recruitment to the PRL array which suggests this mechanism is not involved in the responses observed with CHDM-4-HBA and TMBPF [19]. Therefore, the mechanisms behind these observations remain unknown, and importantly, it has been subsequently shown that TMBPF and its related compounds do not show biological effects in either E-SCREEN or mouse uterotorphic in vivo assays [29].
The remaining four investigational compounds, including two TMBPF derivatives (BGF, TMBPF-DGE, TMBPF-ER, and HydroQ), demonstrated little activity in the in vitro assays for estrogenic, anti-estrogenic, androgenic, and anti-androgenic activity. Although a significant decrease in E2 induced reporter gene activity in the MCF-7 cell line was observed with 5 μM TMBPF-DGE, TMBPF-ER, and HydroQ, significant cell toxicity was also observed. When the concentration was reduced from 5 μM to 2 μM, cell toxicity was absent, and there was no observable effect on E2 induced reporter gene activity. This suggests that the results observed at the higher tested concentration were due to non-specific cell toxicity and not a specific antiestrogenic-like activity associated with the compounds. Although CHDM-4-HBA appeared to demonstrate anti-estrogenic-like activity in the PRL-HeLa assays, this result was not supported in the ERE-luciferase assays performed using the MCF-7 cell line that endogenously expresses ERα.
Considering these findings, it appears several of these investigational compounds, especially HydroQ, BGF and the TMBPF derivatives TMBPF-DGE and TMBPF-ER, demonstrate little of the estrogen-like activity shown by BPA and various BPA analogs. The TMBPF findings are consistent with a recent publication in which this compound was found to not have estrogenic-like properties in an in vivo uterotrophic assay [29]. HydroQ, a compound with evidence of weak anti-estrogen-like activity in the GFP-ERβ:PRL-HeLa cell line, is the only investigational compound that has been included in the EPA ToxCast program where it was found to have weak estrogenic activity at concentrations (EC 50 > 30 μM) exceeding those tested in this study [30]. There is little to no existing data describing the estrogenic or anti-estrogenic activity of BGF, TMBFP-DGE and TMBPF-ER, compounds that may represent useful alternatives to BPA. For example, the lack of activity by BGF is exciting as this is a "green" compound readily synthesized from plant material found in by-products from the paper industry [31]. Based upon the lack of activities demonstrated here and minimal pre-existing data, these compounds are worthwhile candidates for further investigation using an expanded panel of in vitro assays such as those sanctioned by the EPA and used in current models predicting estrogenic activity [12], and/or using in vivo models to further our understanding as suitable BPA alternatives. in the decision to publish or producing this manuscript. DeepBio, Inc provided support in the form of materials and consulting fees for A.T.S., F.S., and M.G.M. and provided salary for M. A.M. The specific roles of these authors are articulated in the 'author contributions' section. M.A.M participated in in study design, data interpretation, and final manuscript review, however, DeepBio, Inc did not have editorial control nor was involved in the decision to publish.