Predictions for the binding of bisphenols to nuclear receptors AR and GR were investigated using two in silico tools: Endocrine Disruptome (Kolšek et al., 2014) and VirtualToxLab (Vedani et al., 2012; Vedani et al., 2009). Both of these tools are recommended for identification of EDCs by the EFSA and the European Chemicals Agency (ECA) (Andersson et al., 2018). Endocrine Disruptome uses molecular docking to predict the binding free energies of compounds to nuclear receptors, and it is available at http://endocrinedisruptome.ki.si/ (Kolšek et al., 2014). It uses Docking Interface for Target Systems (DoTS) for docking simulations and AutoDock Vina for docking calculations. Endocrine Disruptome differentiates between the agonist and antagonist conformations of AR and GR. The structures of each of these receptors were first validated using three databases (i.e., active compounds, agonists, antagonists), as can be downloaded as a pdbqt file (Kolšek et al., 2014). The results are returned as predicted binding free energy based on the sensitivity of each receptor structure: low probability (sensitivity >0.75), moderately low probability (0.5 < sensitivity <0.75), moderately high probability (0.25 < sensitivity <0.5) and high probability of binding (sensitivity <0.25) (Kolšek et al., 2014).

VirtualToxLab (version 5.8; Biographics Laboratory 3R, Basel, Switzerland) uses automated flexible molecular docking with Yeti/AutoDock to consider all compound orientations and conformations in the receptor binding site. This is combined with the multidimensional Quantitative Structure-Activity Relationships (mQSAR) software (Quasar) to define compound orientation, position, different solvation, protonation and conformation, for induced-fit models and the tautomeric state. This tool does not consider agonist or antagonist modes of action on AR and GR. The results are given as the predicted toxic potential (0–1, where 1 is most toxic) and the concentration at which the compound is expected to bind to the nuclear receptor (up to 100 µM considered) (Vedani et al., 2009; Vedani et al., 2012).

BPA (PubChem CID: 6623), BPAF (PubChem CID: 73864), BPS (PubChem CID: 6626), Mifepristone (RU486) (PubChem CID: 55245), Flutamide (PubChem CID: 3397) and esiRNA MISSION^® (EHU068001) for GRα were purchased from Sigma Aldrich (St Louis, MO, United States). Lipopolysaccharide (LPS) from Escherichia coli serotype 0127:B8 and the cell culture media and supplements were from Sigma Aldrich (St Louis, MO, United States). The mouse anti-human RACK1 monoclonal antibody (sc-17754) was from Santa Cruz Biotechnology (Dallas, TX, United States). The mouse monoclonal anti-β-tubulin antibody (T0198) was from Sigma Aldrich (St Louis, MO, United States). Electrophoresis reagents were from Bio-Rad (Richmond, CA, United States). Chemicals were diluted in dimethyl sulfoxide (DMSO) for 50 mM stock solutions, and sequentially properly diluted.

The THP-1 cell line was originally derived from an acute monocytic leukemia patient, and was from American Type Culture Collection (ATCC TIB-202; Manassas, VA, United States). For experiments, THP-1 cells, were diluted to 10⁶ cells/mL in RPMI 1640 without phenol red containing 2 mM L-glutamine, 0.1 mg/ml streptomycin, 100 IU/ml penicillin, gentamycin 10 μg/ml, 50 µM 2-mercaptoethanol, supplemented with 5% dextran-coated charcoal-treated fetal bovine serum (DCC-FBS) and cultured at 37°C in 5% CO₂ incubator. Preliminary experiments were conducted to identify non-cytotoxic concentrations [cell viability >80% (CV80)]. Cytotoxicity was assessed by propidium iodine staining and the CV80 determined for all compounds. For the different experiments, cells were then treated with increasing concentrations of BPA, BPAF and BPS (0.001–10 μM) or DMSO as vehicle control (final concentration of DMSO in culture medium <0.1%) as detailed in the legends.

Plasmids preparation, transient transfections and luciferase assays were performed as previously described (Buoso et al., 2019; Buoso et al., 2020b). The Δ1 reporter plasmid construct has been previously described (Del Vecchio et al., 2009). It was the longest construct available, 2105 nt long, which contains the RACK1 gene promoter region between nucleotides −1744 and +361, and includes the glucocorticoid responsive element (GRE) sequence. The pGL4.32 vector (E8491; Promega, Madison, WI, United States) luciferase-reporter construct plasmid DNA was also used (Fagiani et al., 2020). Plasmids for transfections were purified with the HiSpeed^® Plasmid Midi Kit (Qiagen, Valencia, CA, United States). DNA was quantified and assayed for purity using Quantus™ fluorometer (Promega, Madison, WI, United States). Transient transfections were performed in 24-well plates; for each well, 5 × 10⁵ cells were seeded in RPMI 1640 medium without phenol red, and with 5% DCC-FBS, 1% antibiotics and supplemented with 1% L-glutamine. Transfections were carried out using Lipofectamine® 2000 (Invitrogen Carlsbad, CA, United States), following the manufacturer instructions. Each luciferase reporter construct plasmid DNA was co-transfected with the pRL-TK renilla luciferase expressing vector to measure the transfection efficiency (Promega, Madison, WI, United States). During transfection, THP-1 cells were incubated at 37°C in 5% CO₂, and then treated with the selected compounds for the times and at concentrations specified in figure legends. Cells were then lysed (Passive Lysis Buffer, provided by the Dual-Luciferase Reporter Assay System; Promega, Madison, WI, United States), following the manufacturer specifications. Luminescence was measured with a 20/20n Luminometer (Turner Bio-Systems, Sunnyvale, CA, United States), with 10 s integration time.

Quantitative PCR (qPCR) was performed as previously described (Buoso et al., 2020b). Total RNA was extracted using RNeasy Plus Mini Kit (Qiagen, Valencia, CA, United States), following the manufacturer instructions. For the synthesis of cDNA, 2 µg total RNA was reverse transcribed using high-capacity cDNA archive kits (Applied Biosystems, Foster City, CA, United States), following the manufacturer instructions. RACK1 gene expression was determined by qPCR using the Taq-Man™ PCR technology. Primers were from Applied Biosystems. PCR reactions were performed in duplicate, according to the standard protocols suggested by the manufacturer. For each PCR reaction, 10 ng total RNA was used. 18S ribosomal RNA was used as an endogenous reference. Quantification of the transcripts was performed according to the 2^−ΔΔCT method.

Immunoblot analysis was performed as previously described (Buoso et al., 2020b). RACK1 expression was determined at the protein level in cell lysates by immunoblot analysis. After the treatments, the cells were harvested, washed with phosphate-buffered saline (PBS) 1X, centrifuged, lysed in 100 µl homogenization buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Triton X-100, protease inhibitor mix) and sonicated for 10 s. Protein content was measured using Bradford assay. Samples for Western blotting were prepared by mixing cell lysates with sample buffer (125 mM Tris-HCl pH 6, 8.4% sodium dodecyl sulfate (SDS), 20% glycerol, 6% β-mercaptoethanol, 0.1% bromophenol) and denatured at 95°C for 5 min. Equivalent amounts of extracted protein (10 μg) were electrophoresed into 10% SDS-PAGE under reducing conditions. The proteins were then transferred to polyvinylidene fluoride (PVDF) membranes (Amersham, Little Chalfont, United Kingdom), blocked in 5% (w/v) bovine serum albumin (BSA), 1X TBS, 0.1% Tween-20 for 1 h with gentle shaking. Proteins were visualized using primary antibodies diluted in 5% (w/v) BSA, 1X TBS, 0.1% Tween-20 for RACK1 (1:1000) and β-tubulin (1:1,000). In all of the experiments, the immunoreactivity was detected using host-specific secondary IgG peroxidase-conjugated antibodies (1:5,000), and developed using enhanced chemiluminescence (Clarity Western ECL blotting substrates, Bio-Rad). Immunoblot images were acquired with Molecular Imager Gel Doc XR (BioRad), and the optical density of the bands was determined using the ImageJ software (W. Rasband, Research Service Branch, National Institute of Mental Health, National Institutes of Health, Bethesda, MD, United States; and Laboratory for Optical and Computational Instrumentation, University of Wisconsin, WS, United States). The relative densities of the bands are expressed as arbitrary units, and are normalized to the control samples run under the same conditions.

Cytokine production was assessed using TNF-α (Invitrogen, Carlsbad, CA, United States) and IL-8 specific sandwich ELISAs (ImmunoTools, Friesoythe, Germany). To induce cytokine secretion, 10 ng/ml LPS was added with the 24 h bisphenol exposure, which was prolonged for 3 h for TNF-α and for 24 h for IL-8 assays. Cell-free supernatants obtained by centrifugation at 300 × g for 5 min were stored at − 80°C until measurement. Limits of detection were IL-8 2.6 pg/ml for IL 8, and 22 pg/ml for TNF-α. Results are expressed in pg/mL (Buoso et al., 2020b).

The three-dimensional (3D) structure of GPER, generated by homology modeling, was recovered from the web server GPCR-ITASSER (Zhang and Zhang, 2010; Zhang et al., 2015). The model_1 of the homology model HG ID: HG0714 (C score = −0.64; Estimated TM-score = 0.63 ± 0.13; Estimated RMSD = 8.1 ± 4.4Å) was employed for docking calculation. The three-dimensional structure was prepared for the subsequent structure-based calculations using the Protein Preparation Wizard utility of the Schrödinger suite (Sastry et al., 2013; Schrödinger LLC New York (USA), 2014d). Missing amino acid side chains were rebuilt, and potential atom types and bond connectivity issues into the homology model structure were fixed. Moreover, ionization and tautomerization states potentially present at physiological pH were also calculated with the Epik sub-routine (Shelley et al., 2007; Schrödinger LLC New York (USA), 2014d). Afterward, the pretreated structure was minimized according to the OPLS3e force field. The GPER receptor grid was generated at the center of the putative binding site (Rosano et al., 2016). The chemical structures of BPA and BPAF were designed in ChemDraw, imported into the Maestro software (Schrödinger, LLC, New York, NY (USA), 2014b), and prepared using the LigPrep utility available within the Schrodinger suite (Schrödinger, LLC, New York, NY (USA), 2014c). All of the potential states of ionization and tautomerism at a physiological pH of 7.4 ±0.2 were generated. The prepared bisphenols were subjected to 2,000 minimization steps with MacroModel, using the OPLS3e force field. The ligands thus prepared were finally docked to the protein. The Glide software was used for the docking calculation, using the default settings of the Standard Precision (SP) protocol (Schrödinger, LLC, New York, NY (USA), 2014a). The resulting ligand-protein complexes were ranked by docking score and visually inspected.

Data are expressed as means ± standard error of mean (SEM) of at least three independent experiments. Statistical analysis was performed using the InStat software, version 7.0 (GraphPad Software, La Jolla, CA, United States). Significant differences were determined using Student’s t-tests or analysis of variance (ANOVA), followed, when significant, by an appropriate post hoc test, as indicated in the Figure legends. In all of the reported statistical analysis effects were designated as significant if the p value was <0.05.

Prediction of the binding of BPA, BPAF, and BPS to the nuclear receptors AR and GR relevant for RACK1 expression was performed with two in silico programs: Endocrine Disruptome (Kolšek et al., 2014) and VirtualToxLab (Vedani et al., 2009; Vedani et al., 2012). These data are shown in Tables 1, 2. Both in silico tools returned interactions of the three bisphenols with AR. Endocrine Disruptome suggested a moderately high probability to the agonist conformation of AR for BPA and BPS, whereas a low binding probability for BPAF is indicated. Moreover, a high binding probability to the antagonist conformation of AR for BPA and BPAF, and a moderately high probability for BPS is suggested (Table 1). VirtualToxLab indicated binding concentrations to AR at 139 and 143 nM for BPA and BPAF, respectively, and at the higher concentration of 2.3 µM for BPS (Table 2). The same analysis on Endocrine Disruptome was also performed for GR, that returned a moderately low probability of binding to the agonist conformations for all of bisphenols, and low for the antagonist conformation (Table 1). VirtualToxLab indicated that BPA and BPAF can interact with GR at 341 and 363 nM respectively and BPS at 10.5 µM (Table 2).

Overall, based on the binding affinities indicated by these analyses, a role of BPA, BPAF and BPS in affecting RACK1 expression could be suggested, since they show an interaction profile with the nuclear receptors demonstrated to regulate RACK1 transcription, namely AR and GR (Racchi et al., 2017). However, since these softwares return a possible binding affinity for both the agonist and antagonist nuclear receptor structures for all bisphenols, this suggest that cellular context need to be considered to elucidate their actual mechanism of action. Noteworthy, a range of concentrations for bisphenols treatments were then selected based on this analysis in order to investigate potential non-monotonic dose-response effect commonly observed for hormones and hormone-active substances.

BPA is a diphenylmethane derivative that is used in the production of polycarbonate plastics and epoxy resins (Monneret, 2017) and is known to have hormone-like, xenoestrogen and estrogen-mimicking properties (Murata and Kang, 2018). Given its wide industrial use and these known endocrine-disrupting features, BPA exposure in humans has raised concerns, particularly for its possible influence on the immune system (Buoso et al., 2020c). Based on our previous data on the effects of EDCs on RACK1 expression and altered immune responses, we investigated the effects of BPA on RACK1 expression by reporter luciferase activity using the human RACK1 gene promoter, mRNA by qPCR, and protein levels by Western blotting. THP-1 cells were treated for 6, 16, 18 or 24 h with increasing concentrations of BPA (0.001–10 μM) or DMSO as vehicle control, according to literature data (Couleau et al., 2015). These timings were selected based on previous experiments as being optimal to investigate dehydroepiandrosterone (DHEA) and cortisol-induced RACK1 transcriptional activity, mRNA and protein expression (Buoso et al., 2011) as well as the effects of EDCs on RACK1 (Buoso et al., 2017a; Buoso et al., 2020b). While 6 h of BPA treatment did not affect RACK1 gene promoter activity at any of the concentrations tested here (Figure 1A), 10 μM BPA for 16 h induced significant RACK1 down-regulation (Figure 1B). Accordingly, 10 μM BPA also induced significant decreases in RACK1 mRNA (Figure 1C) and protein (Figure 1D) levels at 18 and 24 h, respectively. Hormone-related RACK1 expression is known to correlate with TNF-α and IL-8 cytokine secretion (Racchi et al., 2017). Upon 10 µM BPA treatment, we observed this correlation for both TNF-α (Figure 1E) and IL-8 (Figure 1F). It is important to note that BPA alone did not induce cytokine production (data not shown).

We previously reported that BPA induced a low but significant RACK1 down-regulation (Figure 1) and this could be due to its ability to bind AR in an antagonist conformation (Table 1). In line with this consideration, our previous data demonstrated that RACK1 basal expression is positively regulated by androgens and AR silencing induced a significant down-regulation of RACK1 expression (Racchi et al., 2017). However, since BPA and other bisphenols are known to bind and activate different nuclear receptors, including GRα (Prasanth et al., 2010; Zhang et al., 2017; Buoso et al., 2020c) and that GR binding to RACK1 promoter induces its down-regulation (Racchi et al., 2017), we cannot exclude GR involvement in the observed BPA-mediated effect on RACK1 expression. Therefore, to unmask AR-mediated BPA effects, we pre-treated THP-1 cells with mifepristone, a well-known GR inhibitor (Bertagna et al., 1984), in order to exclude GR involvement and highlight BPA antagonistic activity for AR. As shown in Figures 2A–C, in the cells treated with mifepristone and 10 μM BPA, there was significant up-regulation of RACK1 promoter activity (Figure 2A) and increased mRNA (Figure 2B) and protein (Figure 2C) levels compared to 10 μM BPA alone. The same unmasking effect on BPA was also observed with lower tested concentrations (Supplementary Figure S1A). In addition, these data were mirrored by 10 μM BPA treatment in GRα-silenced THP-1 cells (Figure 2D). Therefore, these data show that BPA effects on RACK1 expression are more likely correlated with its agonist profile for GRα rather than an antagonist profile for AR. This is in accordance with BPA moderately low affinity for GR agonist conformation (Table 1) that correlates with the slight BPA effect on RACK1.

BPAF is a bisphenol analog that has replaced BPA in several industrial settings. Increasing evidence has suggested, however, that BPAF is characterized by greater estrogenic properties compared to BPA. THP-1 cells were treated for 6, 16, 18 and 24 h with increasing concentrations of BPAF (0.001–10 μM) or DMSO as vehicle control. As shown in Figures 3A,B, 10 μM BPAF treatment induced a statistically significant decrease of RACK1 transcriptional activity at 6 and 16 h, respectively. These effects on RACK1 gene promoter activation were reflected by RACK1 mRNA (Figure 3C) but were not appreciable at protein level (Figure 3D). The slight reduction of RACK1 promoter activity appears to be sufficient to affect mRNA but not RACK1 protein and, although a slight decrease was observed, it did not reach statistical significance. This is most likely related to the turnover time of the protein in THP-1 cells (Corsini et al., 2002; Del Vecchio et al., 2009).

Since a strong correlation between RACK1 expression and TNF-α release has been demonstrated, in accordance with the slight but significant RACK1 down-regulation observed after BPAF treatment, a low decrease in LPS-induced TNF-α secretion is reported (Figure 3E). Accordingly, BPAF did not affect IL-8 secretion compared to LPS treatment alone (Figure 3F). It is important to note that BPAF alone did not induce cytokine production (data not shown).

Like BPA, BPAF can bind AR in an antagonist conformation (Table 1) and/or activate several other nuclear hormone receptors, including GRα (Kojima et al., 2019; Skledar et al., 2019). Therefore, to unmask the possible AR-mediated BPAF effects, we pre-treated THP-1 cells with mifepristone and, as for BPA, there was significant up-regulation of RACK1 promoter activity (Figure 3A) and increased mRNA (Figure 3B) and protein (Figure 3C) levels compared to 10 μM BPAF alone. The same unmasking effect on BPAF was also observed with lower tested concentrations (Supplementary Figure S1B). These data were mirrored by 10 μM BPAF treatment in the GRα-silenced THP-1 cells (Figure 4D) thus indicating that BPAF shows a mild agonist profile for GRα rather than an antagonist profile for AR, in line with its moderately low affinity for GR agonist conformation (Table 1).

Altogether, these data show that GRα inhibition or silencing can unmask another BPAF-related effect, which suggests that similar to BPA, BPAF can induce alterations in RACK1 levels via GRα signaling as well as through other cellular pathways.

Our data show that hampering BPA binding to GRα through either receptor inhibition or gene silencing, another BPA-related effect could be unmasked and its antagonist profile for AR can be excluded. In addition, BPA agonist activity on AR can also be ruled out since in immune context and in THP-1 cells, we demonstrated that not only RACK1 has a basal regulation controlled by androgens and AR that counteracts GR activity, but also that treatment with androgenic compounds results in a significant up-regulation of RACK1 (Buoso et al., 2017a; Buoso et al., 2017b; Buoso et al., 2020b; Buoso et al., 2020c; Racchi et al., 2017; Buoso et al., 2017c). Therefore, due to its known estrogenic-like features, as reported in the literature, BPA can bind and activate GPER (previously known as GPR30) (Périan and Vanacker, 2020). GPER activation has been reported to induce RACK1 increase by AR ligand independent activation (Buoso et al., 2020b). Therefore, with flutamide treatment, we investigated the possible mechanism that links the mifepristone-unmasked BPA effects on RACK1 up-regulation with hormone response element (HRE) identified in the RACK1 promoter region (Racchi et al., 2017). Cells were treated for 1 h with 50 μM flutamide, 30 min with 20 μM mifepristone and then for 16 h with 10 μM BPA or the DMSO vehicle control for luciferase activity. As shown in Figure 5A, BPA could still significantly induce RACK1 promoter activation in the presence of flutamide, suggesting that BPA masked positive effect on RACK1 is not AR-dependent. The same mechanism was confirmed with BPA lower tested concentrations (Supplementary Figure S1A). Since GPER is known to activate PI3K/Akt signaling cascade (Buoso et al., 2020b) and consequently NF-κB, we assessed BPA-mediated NF-κB activation due to the presence of c-Rel binding sites in RACK1 promoter region (Del Vecchio et al., 2009). To this purpose, we treated THP-1 cells with the IκB degradation inhibitor Bay 11–7085, along with mifepristone. As shown in Figure 5A, RACK1 promoter activity was completely and significantly prevented by Bay 11–7085 treatment, suggesting that BPA masked positive effect on RACK1 is related to the presence of c-Rel site according to our previous data (Buoso et al., 2013; Racchi et al., 2017; Masi et al., 2020). To confirm whether BPA can activate NF-κB, we transfected THP-1 cells with a NF-κB specific luciferase reporter vector (Fagiani et al., 2020). As shown in Figure 5B, BPA treatment resulted in a significant activation of the NF-κB-recognized vector, confirming that BPA unmasked effect on RACK1 could be NF-κB-mediated.

The same considerations were made for BPAF but, as shown in Figure 5C, flutamide completely blocked mifepristone-unmasked BPAF effect on RACK1 transcriptional activity, demonstrating the role of GPER-activated AR in its observed effects. The same mechanism was confirmed with BPAF lower tested concentrations (Supplementary Figure S1B).

To investigate BPA and BPAF different mechanism of action, molecular docking calculations were then performed to gain structural insight into their interactions with GPER. As the crystallographic structure of GPER is not available at present, its most recent structure that was generated by homology modeling (HG ID: HG0714) was downloaded from the web server GPCR-ITASSER (Zhang and Zhang, 2010; Zhang et al., 2015). GPCR-ITASSER is a computational suite derived from the parent I-TASSER package, and it is the most accurate homology modeling software customized for G-protein-coupled receptors (GPCRs). These two bisphenols were then docked into the putative ligand binding pocket of GPER, as suggested by a previous computational study (Rosano et al., 2016) (Figure 6A). Ten docking poses for each ligand were generated. The structures of the most favored docked GPER/ligand complexes are shown in Figures 6B–E. Interestingly, although BPA and BPAF have high structural similarity, they bind into two different clefts of the wide GPER binding site (Figures 6B,C). BPAF interacts with the most superficial portion of the receptor site and its binding pose is highly conserved among the first seven poses with the lowest docking score (in the range of −6.08 to −5.92). A highly negative docking score is associated with a more suitable receptor-ligand binding, and the narrow range of docking score is indicative of the robustness of the predicted binding mode. In the lowest energetic complex, one aromatic ring is allocated into a hydrophobic pocket delimited by Leu59, Tyr55 and Leu119, whereas the phenolic hydroxyl is involved in a hydrogen bond with the carbonyl backbone of Leu119. The second aromatic ring is instead involved in a face-to-edge π-π bond with Phe206. Lastly, the two polar trifluoromethyl (-CF3) groups are located within a polar cleft of the receptor binding site, partially exposed to the solvent, and delimited by Gln54, and His307 (Figure 6D).

Conversely, as depicted in Figure 6C, BPA is predicted to bind into a deeper and less exposed sub-pocket of GPER binding site and the predicted poses are well reproduced in all the GPER-BPA complexes. BPA is accommodated within a portion of the receptor binding site delimited by non-polar amino acids such as Met133, Leu137, Met141, Phe208, Ala209, and Phe314. This hydrophobic pocket is more favorable to host lipophilic molecules such BPA than BPAF, thus accounting for the different orientation observed for the two assessed ligands. In addition, both phenolic hydroxyls of BPA participate into a network of hydrogen bonds involving Tyr123, the backbone carbonyl of Cys207, Glu275, and Gln310, and that contribute to tight the binding of the BPA with the receptor (Figure 6E). These data support the experimental evidence reported above and suggest that, upon binding, BPA and BPAF trigger different GPER conformational changes that lead to the activation of different signaling pathways.

Finally, we investigated the effects of another BPA analogue, BPS, on RACK1 expression. As for the other bisphenols, THP-1 cells were treated for 6, 16, 18 and 24 h with increasing concentrations of BPS (0.001–10 μM) or DMSO as vehicle control. As shown in Figures 7A 6 h BPS treatment did not alter RACK1 promoter activity at any of the concentrations tested, while 16 h treatment induced significant RACK1 up-regulation only at 10 μM BPS (Figure 7B). In agreement with this, there was a statistically significant increase in RACK1 mRNA expression after 18 h of treatment with 10 μM BPS (Figure 7C), which was mirrored by increased RACK1 protein levels after 24 h of treatment (Figure 7D). Similar to BPA and BPAF, we also investigated TNF-α and IL-8 release for BPS-treated and LPS-stimulated THP-1 cells (Figures 7E,F). Consistent with the increase in RACK1 expression, there was a significant increase in TNF-α secretion for 10 μM BPS, although no increase in the release of IL-8 was reported. It is important to note that BPS alone did not induce cytokine production (data not shown).

Due to the BPS up-regulation of RACK1 expression, we investigated whether this increase was AR-mediated. THP-1 cells were, therefore, pre-treated with flutamide for 1 h. As shown in Figures 8 10 μM BPS alone induced a significant increase in RACK1 promoter activity (Figure 8A), and the consequent up-regulation of RACK1 at both the mRNA (Figure 8B) and protein (Figure 8C) levels. Then, in line with the previous data here, flutamide pre-treatment completely abolished BPS-mediated RACK1 up-regulation at the promoter (Figure 8A), and also at the mRNA (Figure 8B) and protein (Figure 8C) levels. To investigate a possible masked effect on BPS we performed the same experimental setting as for BPA and BPAF. However, mifepristone treatment did not unmask any other BPS effect on RACK1 expression. (Supplementary Figure S2). Hence, these data confirmed that these effects of BPS on RACK1 are mediated by AR and are in line with BPS moderately high AR agonist profile reported in Table 1.